From 7d766dda2eaee781b8a08b9c1debad73565dffbf Mon Sep 17 00:00:00 2001
From: Mickael Ide <mide@nvidia.com>
Date: Wed, 17 Jan 2024 21:10:37 +0100
Subject: [PATCH] Remove unused sources and headers

Signed-off-by: Mickael Ide <mide@nvidia.com>
---
 build.sh                                      |    6 +-
 cpp/CMakeLists.txt                            |  161 ---
 cpp/bench/ann/src/raft/raft_cagra_wrapper.h   |   10 +-
 cpp/bench/micro/neighbors/cagra_bench.cuh     |    4 +-
 cpp/include/cuvs/neighbors/ball_cover-ext.cuh |  124 --
 cpp/include/cuvs/neighbors/ball_cover-inl.cuh |  395 ------
 cpp/include/cuvs/neighbors/ball_cover.cuh     |   24 -
 .../cuvs/neighbors/ball_cover_types.hpp       |  169 ---
 .../cuvs/neighbors/brute_force-ext.cuh        |  149 ---
 .../cuvs/neighbors/brute_force-inl.cuh        |  355 ------
 cpp/include/cuvs/neighbors/brute_force.cuh    |   92 --
 .../cuvs/neighbors/brute_force_types.hpp      |  283 -----
 cpp/include/cuvs/neighbors/cagra.hpp          |  109 +-
 cpp/include/cuvs/neighbors/cagra_types.hpp    |  363 ------
 .../cuvs/neighbors/epsilon_neighborhood.cuh   |  123 --
 cpp/include/cuvs/neighbors/ivf_flat-ext.cuh   |  206 ----
 cpp/include/cuvs/neighbors/ivf_flat-inl.cuh   |  602 ---------
 cpp/include/cuvs/neighbors/ivf_flat.cuh       |   24 -
 .../cuvs/neighbors/ivf_flat_codepacker.hpp    |   90 --
 .../cuvs/neighbors/ivf_flat_helpers.cuh       |  147 ---
 .../cuvs/neighbors/ivf_flat_serialize.cuh     |  154 ---
 cpp/include/cuvs/neighbors/ivf_flat_types.hpp |  406 ------
 cpp/include/cuvs/neighbors/ivf_list.hpp       |  196 ---
 cpp/include/cuvs/neighbors/ivf_list_types.hpp |   79 --
 cpp/include/cuvs/neighbors/ivf_pq-ext.cuh     |  227 ----
 cpp/include/cuvs/neighbors/ivf_pq-inl.cuh     |  529 --------
 cpp/include/cuvs/neighbors/ivf_pq.cuh         |   24 -
 cpp/include/cuvs/neighbors/ivf_pq_helpers.cuh |  798 ------------
 .../cuvs/neighbors/ivf_pq_serialize.cuh       |  146 ---
 cpp/include/cuvs/neighbors/ivf_pq_types.hpp   |  580 ---------
 .../cuvs/neighbors/neighbors_types.hpp        |   63 -
 cpp/include/cuvs/neighbors/nn_descent.cuh     |  181 ---
 .../cuvs/neighbors/nn_descent_types.hpp       |  147 ---
 cpp/include/cuvs/neighbors/refine-ext.cuh     |   78 --
 cpp/include/cuvs/neighbors/refine-inl.cuh     |  104 --
 cpp/include/cuvs/neighbors/refine.cuh         |   24 -
 cpp/include/cuvs/neighbors/sample_filter.cuh  |   49 -
 .../cuvs/neighbors/sample_filter_types.hpp    |  175 ---
 .../cuvs/neighbors/specializations.cuh        |   22 -
 cpp/include/cuvs/spatial/knn/ann.cuh          |   83 --
 cpp/include/cuvs/spatial/knn/ann_common.h     |  103 --
 cpp/include/cuvs/spatial/knn/ann_types.hpp    |   45 -
 cpp/include/cuvs/spatial/knn/ball_cover.cuh   |   70 --
 .../cuvs/spatial/knn/ball_cover_types.hpp     |   37 -
 cpp/include/cuvs/spatial/knn/common.hpp       |   23 -
 .../cuvs/spatial/knn/detail/ann_quantized.cuh |  147 ---
 .../cuvs/spatial/knn/detail/ann_utils.cuh     |  576 ---------
 .../cuvs/spatial/knn/detail/ball_cover.cuh    |  549 ---------
 .../spatial/knn/detail/ball_cover/common.cuh  |   73 --
 .../knn/detail/ball_cover/registers-ext.cuh   |  129 --
 .../knn/detail/ball_cover/registers-inl.cuh   |  794 ------------
 .../knn/detail/ball_cover/registers.cuh       |   24 -
 .../knn/detail/ball_cover/registers_types.cuh |   66 -
 .../knn/detail/epsilon_neighborhood.cuh       |  241 ----
 .../spatial/knn/detail/fused_l2_knn-ext.cuh   |   74 --
 .../spatial/knn/detail/fused_l2_knn-inl.cuh   | 1062 ----------------
 .../cuvs/spatial/knn/detail/fused_l2_knn.cuh  |   24 -
 .../spatial/knn/detail/haversine_distance.cuh |  143 ---
 .../cuvs/spatial/knn/detail/processing.cuh    |  189 ---
 .../cuvs/spatial/knn/detail/processing.hpp    |   45 -
 .../cuvs/spatial/knn/epsilon_neighborhood.cuh |   38 -
 cpp/include/cuvs/spatial/knn/ivf_flat.cuh     |   39 -
 .../cuvs/spatial/knn/ivf_flat_types.hpp       |   40 -
 cpp/include/cuvs/spatial/knn/ivf_pq.cuh       |   39 -
 cpp/include/cuvs/spatial/knn/ivf_pq_types.hpp |   40 -
 cpp/include/cuvs/spatial/knn/knn.cuh          |  231 ----
 .../cuvs/spatial/knn/specializations.cuh      |   22 -
 .../cuvs/spatial/knn/specializations/knn.cuh  |   22 -
 cpp/include/cuvs_runtime/cluster/kmeans.hpp   |   97 --
 .../cuvs_runtime/distance/fused_l2_nn.hpp     |   65 -
 .../distance/pairwise_distance.hpp            |   50 -
 cpp/include/cuvs_runtime/matrix/select_k.hpp  |   32 -
 .../cuvs_runtime/neighbors/brute_force.hpp    |   38 -
 .../cuvs_runtime/neighbors/ivf_flat.hpp       |   83 --
 cpp/include/cuvs_runtime/neighbors/ivf_pq.hpp |   96 --
 cpp/include/cuvs_runtime/neighbors/refine.hpp |   48 -
 cpp/src/cuvs_runtime/cluster/cluster_cost.cuh |   87 --
 .../cluster/cluster_cost_double.cu            |   33 -
 .../cluster/cluster_cost_float.cu             |   33 -
 .../cuvs_runtime/cluster/kmeans_fit_double.cu |   33 -
 .../cuvs_runtime/cluster/kmeans_fit_float.cu  |   33 -
 .../cluster/kmeans_init_plus_plus_double.cu   |   31 -
 .../cluster/kmeans_init_plus_plus_float.cu    |   31 -
 .../cuvs_runtime/cluster/update_centroids.cuh |   72 --
 .../cluster/update_centroids_double.cu        |   46 -
 .../cluster/update_centroids_float.cu         |   46 -
 .../cuvs_runtime/distance/fused_l2_min_arg.cu |  105 --
 .../distance/pairwise_distance.cu             |   52 -
 .../matrix/select_k_float_int64_t.cu          |   36 -
 .../brute_force_knn_int64_t_float.cu          |   47 -
 .../cuvs_runtime/neighbors/ivf_flat_build.cu  |   62 -
 .../cuvs_runtime/neighbors/ivf_flat_search.cu |   40 -
 .../neighbors/ivf_flat_serialize.cu           |   65 -
 cpp/src/cuvs_runtime/neighbors/ivfpq_build.cu |   59 -
 .../neighbors/ivfpq_deserialize.cu            |   31 -
 .../neighbors/ivfpq_search_float_int64_t.cu   |   38 -
 .../neighbors/ivfpq_search_int8_t_int64_t.cu  |   38 -
 .../neighbors/ivfpq_search_uint8_t_int64_t.cu |   38 -
 .../cuvs_runtime/neighbors/ivfpq_serialize.cu |   31 -
 .../neighbors/refine_d_int64_t_float.cu       |   33 -
 .../neighbors/refine_d_int64_t_int8_t.cu      |   33 -
 .../neighbors/refine_d_int64_t_uint8_t.cu     |   33 -
 .../neighbors/refine_h_int64_t_float.cu       |   34 -
 .../neighbors/refine_h_int64_t_int8_t.cu      |   33 -
 .../neighbors/refine_h_int64_t_uint8_t.cu     |   33 -
 cpp/src/cuvs_runtime/random/common.cuh        |   41 -
 ...rmat_rectangular_generator_int64_double.cu |   23 -
 .../rmat_rectangular_generator_int64_float.cu |   23 -
 .../rmat_rectangular_generator_int_double.cu  |   23 -
 .../rmat_rectangular_generator_int_float.cu   |   23 -
 .../pairwise_matrix/dispatch_00_generate.py   |  194 ---
 ...patch_canberra_double_double_double_int.cu |   55 -
 ...dispatch_canberra_float_float_float_int.cu |   50 -
 ...ch_correlation_double_double_double_int.cu |   55 -
 ...patch_correlation_float_float_float_int.cu |   55 -
 ...ispatch_cosine_double_double_double_int.cu |   51 -
 .../dispatch_cosine_float_float_float_int.cu  |   51 -
 ...ing_unexpanded_double_double_double_int.cu |   50 -
 ...amming_unexpanded_float_float_float_int.cu |   50 -
 ...inger_expanded_double_double_double_int.cu |   55 -
 ...ellinger_expanded_float_float_float_int.cu |   50 -
 ...jensen_shannon_double_double_double_int.cu |   55 -
 ...ch_jensen_shannon_float_float_float_int.cu |   55 -
 ..._kl_divergence_double_double_double_int.cu |   50 -
 ...tch_kl_divergence_float_float_float_int.cu |   50 -
 .../dispatch_l1_double_double_double_int.cu   |   50 -
 .../dispatch_l1_float_float_float_int.cu      |   50 -
 ...ch_l2_expanded_double_double_double_int.cu |   51 -
 ...patch_l2_expanded_float_float_float_int.cu |   51 -
 ..._l2_unexpanded_double_double_double_int.cu |   55 -
 ...tch_l2_unexpanded_float_float_float_int.cu |   50 -
 ...dispatch_l_inf_double_double_double_int.cu |   50 -
 .../dispatch_l_inf_float_float_float_int.cu   |   50 -
 ..._lp_unexpanded_double_double_double_int.cu |   55 -
 ...tch_lp_unexpanded_float_float_float_int.cu |   50 -
 .../detail/pairwise_matrix/dispatch_rbf.cu    |   64 -
 ...tch_russel_rao_double_double_double_int.cu |   55 -
 ...spatch_russel_rao_float_float_float_int.cu |   50 -
 cpp/src/distance/distance.cu                  |  934 --------------
 cpp/src/distance/fused_l2_nn.cu               |   54 -
 .../matrix/detail/select_k_double_int64_t.cu  |   34 -
 .../matrix/detail/select_k_double_uint32_t.cu |   35 -
 cpp/src/matrix/detail/select_k_float_int32.cu |   34 -
 .../matrix/detail/select_k_float_int64_t.cu   |   34 -
 .../matrix/detail/select_k_float_uint32_t.cu  |   34 -
 .../matrix/detail/select_k_half_int64_t.cu    |   34 -
 .../matrix/detail/select_k_half_uint32_t.cu   |   34 -
 cpp/src/neighbors/ball_cover.cu               |   66 -
 cpp/src/neighbors/brute_force_00_generate.py  |  106 --
 .../brute_force_fused_l2_knn_float_int64_t.cu |   45 -
 .../neighbors/brute_force_knn_index_float.cu  |   39 -
 .../brute_force_knn_int64_t_float_int64_t.cu  |   47 -
 .../brute_force_knn_int64_t_float_uint32_t.cu |   47 -
 .../brute_force_knn_int_float_int.cu          |   47 -
 ...brute_force_knn_uint32_t_float_uint32_t.cu |   47 -
 cpp/src/neighbors/cagra_build_float.cpp       |   68 +-
 cpp/src/neighbors/cagra_build_int8.cpp        |   68 +-
 cpp/src/neighbors/cagra_build_uint8.cpp       |   68 +-
 cpp/src/neighbors/cagra_optimize.cpp          |   22 +-
 cpp/src/neighbors/cagra_search_float.cpp      |   20 +-
 cpp/src/neighbors/cagra_search_int8.cpp       |   20 +-
 cpp/src/neighbors/cagra_search_uint8.cpp      |   20 +-
 .../cagra/search_multi_cta_00_generate.py     |  108 --
 ...arch_multi_cta_float_uint32_dim1024_t32.cu |   66 -
 ...search_multi_cta_float_uint32_dim128_t8.cu |   66 -
 ...earch_multi_cta_float_uint32_dim256_t16.cu |   66 -
 ...earch_multi_cta_float_uint32_dim512_t32.cu |   66 -
 ...arch_multi_cta_float_uint64_dim1024_t32.cu |   66 -
 ...search_multi_cta_float_uint64_dim128_t8.cu |   66 -
 ...earch_multi_cta_float_uint64_dim256_t16.cu |   66 -
 ...earch_multi_cta_float_uint64_dim512_t32.cu |   66 -
 ...earch_multi_cta_int8_uint32_dim1024_t32.cu |   66 -
 .../search_multi_cta_int8_uint32_dim128_t8.cu |   66 -
 ...search_multi_cta_int8_uint32_dim256_t16.cu |   66 -
 ...search_multi_cta_int8_uint32_dim512_t32.cu |   66 -
 ...arch_multi_cta_uint8_uint32_dim1024_t32.cu |   66 -
 ...search_multi_cta_uint8_uint32_dim128_t8.cu |   66 -
 ...earch_multi_cta_uint8_uint32_dim256_t16.cu |   66 -
 ...earch_multi_cta_uint8_uint32_dim512_t32.cu |   66 -
 .../cagra/search_single_cta_00_generate.py    |  113 --
 ...rch_single_cta_float_uint32_dim1024_t32.cu |   67 -
 ...earch_single_cta_float_uint32_dim128_t8.cu |   67 -
 ...arch_single_cta_float_uint32_dim256_t16.cu |   67 -
 ...arch_single_cta_float_uint32_dim512_t32.cu |   67 -
 ...rch_single_cta_float_uint64_dim1024_t32.cu |   67 -
 ...earch_single_cta_float_uint64_dim128_t8.cu |   67 -
 ...arch_single_cta_float_uint64_dim256_t16.cu |   67 -
 ...arch_single_cta_float_uint64_dim512_t32.cu |   67 -
 ...arch_single_cta_int8_uint32_dim1024_t32.cu |   67 -
 ...search_single_cta_int8_uint32_dim128_t8.cu |   67 -
 ...earch_single_cta_int8_uint32_dim256_t16.cu |   67 -
 ...earch_single_cta_int8_uint32_dim512_t32.cu |   67 -
 ...rch_single_cta_uint8_uint32_dim1024_t32.cu |   67 -
 ...earch_single_cta_uint8_uint32_dim128_t8.cu |   67 -
 ...arch_single_cta_uint8_uint32_dim256_t16.cu |   67 -
 ...arch_single_cta_uint8_uint32_dim512_t32.cu |   67 -
 ...at_interleaved_scan_float_float_int64_t.cu |   42 -
 ...interleaved_scan_int8_t_int32_t_int64_t.cu |   42 -
 ...terleaved_scan_uint8_t_uint32_t_int64_t.cu |   42 -
 cpp/src/neighbors/detail/ivf_flat_search.cu   |   40 -
 .../ivf_pq_compute_similarity_00_generate.py  |  108 --
 .../ivf_pq_compute_similarity_float_float.cu  |   81 --
 ...f_pq_compute_similarity_float_fp8_false.cu |   81 --
 ...vf_pq_compute_similarity_float_fp8_true.cu |   81 --
 .../ivf_pq_compute_similarity_float_half.cu   |   81 --
 ...vf_pq_compute_similarity_half_fp8_false.cu |   81 --
 ...ivf_pq_compute_similarity_half_fp8_true.cu |   81 --
 .../ivf_pq_compute_similarity_half_half.cu    |   81 --
 .../detail/refine_host_float_float.cpp        |   29 -
 .../detail/refine_host_int8_t_float.cpp       |   29 -
 .../detail/refine_host_uint8_t_float.cpp      |   30 -
 .../detail/selection_faiss_00_generate.py     |   79 --
 .../detail/selection_faiss_int32_t_float.cu   |   44 -
 .../detail/selection_faiss_int64_t_double.cu  |   44 -
 .../detail/selection_faiss_int64_t_half.cu    |   44 -
 .../detail/selection_faiss_int_double.cu      |   44 -
 .../detail/selection_faiss_long_float.cu      |   44 -
 .../detail/selection_faiss_size_t_double.cu   |   44 -
 .../detail/selection_faiss_size_t_float.cu    |   44 -
 .../detail/selection_faiss_uint32_t_double.cu |   44 -
 .../detail/selection_faiss_uint32_t_float.cu  |   44 -
 .../detail/selection_faiss_uint32_t_half.cu   |   44 -
 cpp/src/neighbors/ivf_flat_00_generate.py     |  148 ---
 .../neighbors/ivf_flat_build_float_int64_t.cu |   50 -
 .../ivf_flat_build_int8_t_int64_t.cu          |   50 -
 .../ivf_flat_build_uint8_t_int64_t.cu         |   50 -
 .../ivf_flat_extend_float_int64_t.cu          |   58 -
 .../ivf_flat_extend_int8_t_int64_t.cu         |   58 -
 .../ivf_flat_extend_uint8_t_int64_t.cu        |   58 -
 .../ivf_flat_search_float_int64_t.cu          |   49 -
 .../ivf_flat_search_int8_t_int64_t.cu         |   49 -
 .../ivf_flat_search_uint8_t_int64_t.cu        |   49 -
 .../neighbors/ivfpq_build_float_int64_t.cu    |   36 -
 .../neighbors/ivfpq_build_int8_t_int64_t.cu   |   36 -
 .../neighbors/ivfpq_build_uint8_t_int64_t.cu  |   36 -
 .../neighbors/ivfpq_extend_float_int64_t.cu   |   50 -
 .../neighbors/ivfpq_extend_int8_t_int64_t.cu  |   50 -
 .../neighbors/ivfpq_extend_uint8_t_int64_t.cu |   50 -
 .../neighbors/ivfpq_search_float_int64_t.cu   |   42 -
 .../neighbors/ivfpq_search_int8_t_int64_t.cu  |   42 -
 .../neighbors/ivfpq_search_uint8_t_int64_t.cu |   42 -
 cpp/src/neighbors/refine_00_generate.py       |   78 --
 cpp/src/neighbors/refine_float_float.cu       |   50 -
 cpp/src/neighbors/refine_int8_t_float.cu      |   50 -
 cpp/src/neighbors/refine_uint8_t_float.cu     |   50 -
 .../knn/detail/ball_cover/registers.cu        |   60 -
 .../ball_cover/registers_00_generate.py       |  112 --
 .../ball_cover/registers_pass_one_2d_dist.cu  |   48 -
 .../registers_pass_one_2d_euclidean.cu        |   48 -
 .../registers_pass_one_2d_haversine.cu        |   48 -
 .../ball_cover/registers_pass_one_3d_dist.cu  |   48 -
 .../registers_pass_one_3d_euclidean.cu        |   48 -
 .../registers_pass_one_3d_haversine.cu        |   48 -
 .../ball_cover/registers_pass_two_2d_dist.cu  |   48 -
 .../registers_pass_two_2d_euclidean.cu        |   48 -
 .../registers_pass_two_2d_haversine.cu        |   48 -
 .../ball_cover/registers_pass_two_3d_dist.cu  |   48 -
 .../registers_pass_two_3d_euclidean.cu        |   48 -
 .../registers_pass_two_3d_haversine.cu        |   48 -
 .../knn/detail/fused_l2_knn_int32_t_float.cu  |   42 -
 .../knn/detail/fused_l2_knn_int64_t_float.cu  |   42 -
 .../knn/detail/fused_l2_knn_uint32_t_float.cu |   43 -
 cpp/template/src/cagra_example.cu             |    2 +-
 cpp/test/CMakeLists.txt                       |  124 --
 cpp/test/cluster/cluster_solvers.cu           |  103 --
 cpp/test/cluster/kmeans.cu                    |  359 ------
 cpp/test/cluster/kmeans_balanced.cu           |  236 ----
 cpp/test/cluster/kmeans_find_k.cu             |  140 ---
 cpp/test/cluster/linkage.cu                   |  675 ----------
 cpp/test/distance/dist_adj.cu                 |  194 ---
 cpp/test/distance/dist_adj.cuh                |   71 --
 .../distance/dist_adj_distance_instance.cu    |   63 -
 cpp/test/distance/dist_adj_threshold.cuh      |   36 -
 cpp/test/distance/dist_canberra.cu            |   70 --
 cpp/test/distance/dist_correlation.cu         |   94 --
 cpp/test/distance/dist_cos.cu                 |  110 --
 cpp/test/distance/dist_hamming.cu             |   71 --
 cpp/test/distance/dist_hellinger.cu           |   71 --
 cpp/test/distance/dist_inner_product.cu       |   74 --
 cpp/test/distance/dist_jensen_shannon.cu      |   71 --
 cpp/test/distance/dist_kl_divergence.cu       |   71 --
 cpp/test/distance/dist_l1.cu                  |   70 --
 cpp/test/distance/dist_l2_exp.cu              |  113 --
 cpp/test/distance/dist_l2_sqrt_exp.cu         |   74 --
 cpp/test/distance/dist_l2_unexp.cu            |   71 --
 cpp/test/distance/dist_l_inf.cu               |   70 --
 cpp/test/distance/dist_lp_unexp.cu            |   71 --
 cpp/test/distance/dist_russell_rao.cu         |   71 --
 cpp/test/distance/distance_base.cuh           |  673 ----------
 cpp/test/distance/fused_l2_nn.cu              |  436 -------
 cpp/test/distance/gram.cu                     |  170 ---
 cpp/test/distance/gram_base.cuh               |   88 --
 cpp/test/distance/masked_nn.cu                |  435 -------
 .../distance/masked_nn_compress_to_bits.cu    |  217 ----
 cpp/test/ext_headers/00_generate.py           |   79 --
 cpp/test/ext_headers/raft_core_logger.cpp     |   27 -
 ...istance_detail_pairwise_matrix_dispatch.cu |   27 -
 .../ext_headers/raft_distance_distance.cu     |   27 -
 .../ext_headers/raft_distance_fused_l2_nn.cu  |   27 -
 .../raft_linalg_detail_coalesced_reduction.cu |   27 -
 .../raft_matrix_detail_select_k.cu            |   27 -
 .../ext_headers/raft_neighbors_ball_cover.cu  |   27 -
 .../ext_headers/raft_neighbors_brute_force.cu |   27 -
 ...ghbors_detail_ivf_flat_interleaved_scan.cu |   27 -
 .../raft_neighbors_detail_ivf_flat_search.cu  |   27 -
 ...ghbors_detail_ivf_pq_compute_similarity.cu |   27 -
 .../raft_neighbors_detail_selection_faiss.cu  |   27 -
 .../ext_headers/raft_neighbors_ivf_flat.cu    |   27 -
 cpp/test/ext_headers/raft_neighbors_ivf_pq.cu |   27 -
 cpp/test/ext_headers/raft_neighbors_refine.cu |   27 -
 ...spatial_knn_detail_ball_cover_registers.cu |   27 -
 .../raft_spatial_knn_detail_fused_l2_knn.cu   |   27 -
 .../ext_headers/raft_util_memory_pool.cpp     |   27 -
 cpp/test/neighbors/ann_cagra.cuh              |   22 +-
 .../ann_cagra/test_float_uint32_t.cu          |   13 -
 .../ann_cagra/test_int8_t_uint32_t.cu         |   14 +-
 .../ann_cagra/test_uint8_t_uint32_t.cu        |   14 +-
 cpp/test/neighbors/ann_ivf_flat.cuh           |  615 ---------
 .../ann_ivf_flat/test_filter_float_int64_t.cu |   29 -
 .../ann_ivf_flat/test_float_int64_t.cu        |   32 -
 .../ann_ivf_flat/test_int8_t_int64_t.cu       |   28 -
 .../ann_ivf_flat/test_uint8_t_int64_t.cu      |   28 -
 cpp/test/neighbors/ann_ivf_pq.cuh             | 1095 -----------------
 .../ann_ivf_pq/test_filter_float_int64_t.cu   |   26 -
 .../ann_ivf_pq/test_filter_int8_t_int64_t.cu  |   27 -
 .../ann_ivf_pq/test_float_int64_t.cu          |   27 -
 .../ann_ivf_pq/test_float_uint32_t.cu         |   37 -
 .../ann_ivf_pq/test_int8_t_int64_t.cu         |   27 -
 .../ann_ivf_pq/test_uint8_t_int64_t.cu        |   27 -
 cpp/test/neighbors/ann_nn_descent.cuh         |  156 ---
 .../ann_nn_descent/test_float_uint32_t.cu     |   28 -
 .../ann_nn_descent/test_int8_t_uint32_t.cu    |   28 -
 .../ann_nn_descent/test_uint8_t_uint32_t.cu   |   28 -
 cpp/test/neighbors/ball_cover.cu              |  372 ------
 cpp/test/neighbors/epsilon_neighborhood.cu    |  121 --
 cpp/test/neighbors/fused_l2_knn.cu            |  175 ---
 cpp/test/neighbors/haversine.cu               |  133 --
 cpp/test/neighbors/knn.cu                     |  197 ---
 cpp/test/neighbors/knn_utils.cuh              |   94 --
 cpp/test/neighbors/refine.cu                  |  129 --
 cpp/test/neighbors/selection.cu               |  499 --------
 cpp/test/neighbors/spatial_data.h             |   38 -
 cpp/test/neighbors/tiled_knn.cu               |  354 ------
 cpp/test/sparse/dist_coo_spmv.cu              |  697 -----------
 cpp/test/sparse/distance.cu                   |  853 -------------
 cpp/test/sparse/gram.cu                       |  328 -----
 cpp/test/sparse/neighbors/brute_force.cu      |  179 ---
 .../sparse/neighbors/cross_component_nn.cu    | 1036 ----------------
 cpp/test/sparse/neighbors/knn_graph.cu        |  127 --
 cpp/test/sparse/spectral_matrix.cu            |   83 --
 cpp/test/stats/accuracy.cu                    |  106 --
 cpp/test/stats/adjusted_rand_index.cu         |  207 ----
 cpp/test/stats/completeness_score.cu          |  136 --
 cpp/test/stats/contingencyMatrix.cu           |  167 ---
 cpp/test/stats/cov.cu                         |  193 ---
 cpp/test/stats/dispersion.cu                  |  132 --
 cpp/test/stats/entropy.cu                     |  123 --
 cpp/test/stats/histogram.cu                   |  318 -----
 cpp/test/stats/homogeneity_score.cu           |  134 --
 cpp/test/stats/information_criterion.cu       |  151 ---
 cpp/test/stats/kl_divergence.cu               |  107 --
 cpp/test/stats/mean.cu                        |  149 ---
 cpp/test/stats/meanvar.cu                     |  161 ---
 cpp/test/stats/minmax.cu                      |  208 ----
 cpp/test/stats/mutual_info_score.cu           |  162 ---
 cpp/test/stats/neighborhood_recall.cu         |  178 ---
 cpp/test/stats/r2_score.cu                    |  114 --
 cpp/test/stats/rand_index.cu                  |  129 --
 cpp/test/stats/regression_metrics.cu          |  146 ---
 cpp/test/stats/silhouette_score.cu            |  227 ----
 cpp/test/stats/stddev.cu                      |  196 ---
 cpp/test/stats/sum.cu                         |  111 --
 cpp/test/stats/trustworthiness.cu             |  352 ------
 cpp/test/stats/v_measure.cu                   |  139 ---
 cpp/test/stats/weighted_mean.cu               |  339 -----
 375 files changed, 215 insertions(+), 41426 deletions(-)
 delete mode 100644 cpp/include/cuvs/neighbors/ball_cover-ext.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ball_cover-inl.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ball_cover.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ball_cover_types.hpp
 delete mode 100644 cpp/include/cuvs/neighbors/brute_force-ext.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/brute_force-inl.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/brute_force.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/brute_force_types.hpp
 delete mode 100644 cpp/include/cuvs/neighbors/cagra_types.hpp
 delete mode 100644 cpp/include/cuvs/neighbors/epsilon_neighborhood.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_flat-ext.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_flat-inl.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_flat.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_flat_codepacker.hpp
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_flat_helpers.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_flat_serialize.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_flat_types.hpp
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_list.hpp
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_list_types.hpp
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_pq-ext.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_pq-inl.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_pq.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_pq_helpers.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_pq_serialize.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/ivf_pq_types.hpp
 delete mode 100644 cpp/include/cuvs/neighbors/neighbors_types.hpp
 delete mode 100644 cpp/include/cuvs/neighbors/nn_descent.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/nn_descent_types.hpp
 delete mode 100644 cpp/include/cuvs/neighbors/refine-ext.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/refine-inl.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/refine.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/sample_filter.cuh
 delete mode 100644 cpp/include/cuvs/neighbors/sample_filter_types.hpp
 delete mode 100644 cpp/include/cuvs/neighbors/specializations.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/ann.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/ann_common.h
 delete mode 100644 cpp/include/cuvs/spatial/knn/ann_types.hpp
 delete mode 100644 cpp/include/cuvs/spatial/knn/ball_cover.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/ball_cover_types.hpp
 delete mode 100644 cpp/include/cuvs/spatial/knn/common.hpp
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/ann_quantized.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/ann_utils.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/ball_cover.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/ball_cover/common.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/ball_cover/registers-ext.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/ball_cover/registers.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/ball_cover/registers_types.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/epsilon_neighborhood.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/fused_l2_knn-ext.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/fused_l2_knn-inl.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/fused_l2_knn.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/haversine_distance.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/processing.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/detail/processing.hpp
 delete mode 100644 cpp/include/cuvs/spatial/knn/epsilon_neighborhood.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/ivf_flat.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/ivf_flat_types.hpp
 delete mode 100644 cpp/include/cuvs/spatial/knn/ivf_pq.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/ivf_pq_types.hpp
 delete mode 100644 cpp/include/cuvs/spatial/knn/knn.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/specializations.cuh
 delete mode 100644 cpp/include/cuvs/spatial/knn/specializations/knn.cuh
 delete mode 100644 cpp/include/cuvs_runtime/cluster/kmeans.hpp
 delete mode 100644 cpp/include/cuvs_runtime/distance/fused_l2_nn.hpp
 delete mode 100644 cpp/include/cuvs_runtime/distance/pairwise_distance.hpp
 delete mode 100644 cpp/include/cuvs_runtime/matrix/select_k.hpp
 delete mode 100644 cpp/include/cuvs_runtime/neighbors/brute_force.hpp
 delete mode 100644 cpp/include/cuvs_runtime/neighbors/ivf_flat.hpp
 delete mode 100644 cpp/include/cuvs_runtime/neighbors/ivf_pq.hpp
 delete mode 100644 cpp/include/cuvs_runtime/neighbors/refine.hpp
 delete mode 100644 cpp/src/cuvs_runtime/cluster/cluster_cost.cuh
 delete mode 100644 cpp/src/cuvs_runtime/cluster/cluster_cost_double.cu
 delete mode 100644 cpp/src/cuvs_runtime/cluster/cluster_cost_float.cu
 delete mode 100644 cpp/src/cuvs_runtime/cluster/kmeans_fit_double.cu
 delete mode 100644 cpp/src/cuvs_runtime/cluster/kmeans_fit_float.cu
 delete mode 100644 cpp/src/cuvs_runtime/cluster/kmeans_init_plus_plus_double.cu
 delete mode 100644 cpp/src/cuvs_runtime/cluster/kmeans_init_plus_plus_float.cu
 delete mode 100644 cpp/src/cuvs_runtime/cluster/update_centroids.cuh
 delete mode 100644 cpp/src/cuvs_runtime/cluster/update_centroids_double.cu
 delete mode 100644 cpp/src/cuvs_runtime/cluster/update_centroids_float.cu
 delete mode 100644 cpp/src/cuvs_runtime/distance/fused_l2_min_arg.cu
 delete mode 100644 cpp/src/cuvs_runtime/distance/pairwise_distance.cu
 delete mode 100644 cpp/src/cuvs_runtime/matrix/select_k_float_int64_t.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/brute_force_knn_int64_t_float.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/ivf_flat_build.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/ivf_flat_search.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/ivf_flat_serialize.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/ivfpq_build.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/ivfpq_deserialize.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/ivfpq_search_float_int64_t.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/ivfpq_search_int8_t_int64_t.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/ivfpq_search_uint8_t_int64_t.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/ivfpq_serialize.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_float.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_int8_t.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_uint8_t.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_float.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_int8_t.cu
 delete mode 100644 cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_uint8_t.cu
 delete mode 100644 cpp/src/cuvs_runtime/random/common.cuh
 delete mode 100644 cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int64_double.cu
 delete mode 100644 cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int64_float.cu
 delete mode 100644 cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int_double.cu
 delete mode 100644 cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int_float.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_00_generate.py
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_canberra_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_canberra_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_correlation_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_correlation_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_cosine_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_cosine_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_hellinger_expanded_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_hellinger_expanded_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_jensen_shannon_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_jensen_shannon_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_kl_divergence_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_kl_divergence_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_l1_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_l1_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_l2_expanded_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_l2_expanded_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_l2_unexpanded_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_l2_unexpanded_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_l_inf_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_l_inf_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_lp_unexpanded_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_lp_unexpanded_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_rbf.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_russel_rao_double_double_double_int.cu
 delete mode 100644 cpp/src/distance/detail/pairwise_matrix/dispatch_russel_rao_float_float_float_int.cu
 delete mode 100644 cpp/src/distance/distance.cu
 delete mode 100644 cpp/src/distance/fused_l2_nn.cu
 delete mode 100644 cpp/src/matrix/detail/select_k_double_int64_t.cu
 delete mode 100644 cpp/src/matrix/detail/select_k_double_uint32_t.cu
 delete mode 100644 cpp/src/matrix/detail/select_k_float_int32.cu
 delete mode 100644 cpp/src/matrix/detail/select_k_float_int64_t.cu
 delete mode 100644 cpp/src/matrix/detail/select_k_float_uint32_t.cu
 delete mode 100644 cpp/src/matrix/detail/select_k_half_int64_t.cu
 delete mode 100644 cpp/src/matrix/detail/select_k_half_uint32_t.cu
 delete mode 100644 cpp/src/neighbors/ball_cover.cu
 delete mode 100644 cpp/src/neighbors/brute_force_00_generate.py
 delete mode 100644 cpp/src/neighbors/brute_force_fused_l2_knn_float_int64_t.cu
 delete mode 100644 cpp/src/neighbors/brute_force_knn_index_float.cu
 delete mode 100644 cpp/src/neighbors/brute_force_knn_int64_t_float_int64_t.cu
 delete mode 100644 cpp/src/neighbors/brute_force_knn_int64_t_float_uint32_t.cu
 delete mode 100644 cpp/src/neighbors/brute_force_knn_int_float_int.cu
 delete mode 100644 cpp/src/neighbors/brute_force_knn_uint32_t_float_uint32_t.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_00_generate.py
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim1024_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim128_t8.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim256_t16.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim512_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim1024_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim128_t8.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim256_t16.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim512_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim1024_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim128_t8.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim256_t16.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim512_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim1024_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim128_t8.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim256_t16.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim512_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_00_generate.py
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim1024_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim128_t8.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim256_t16.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim512_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim1024_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim128_t8.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim256_t16.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim512_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim1024_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim128_t8.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim256_t16.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim512_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim1024_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim128_t8.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim256_t16.cu
 delete mode 100644 cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim512_t32.cu
 delete mode 100644 cpp/src/neighbors/detail/ivf_flat_interleaved_scan_float_float_int64_t.cu
 delete mode 100644 cpp/src/neighbors/detail/ivf_flat_interleaved_scan_int8_t_int32_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/detail/ivf_flat_interleaved_scan_uint8_t_uint32_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/detail/ivf_flat_search.cu
 delete mode 100644 cpp/src/neighbors/detail/ivf_pq_compute_similarity_00_generate.py
 delete mode 100644 cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_float.cu
 delete mode 100644 cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_fp8_false.cu
 delete mode 100644 cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_fp8_true.cu
 delete mode 100644 cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_half.cu
 delete mode 100644 cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_fp8_false.cu
 delete mode 100644 cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_fp8_true.cu
 delete mode 100644 cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_half.cu
 delete mode 100644 cpp/src/neighbors/detail/refine_host_float_float.cpp
 delete mode 100644 cpp/src/neighbors/detail/refine_host_int8_t_float.cpp
 delete mode 100644 cpp/src/neighbors/detail/refine_host_uint8_t_float.cpp
 delete mode 100644 cpp/src/neighbors/detail/selection_faiss_00_generate.py
 delete mode 100644 cpp/src/neighbors/detail/selection_faiss_int32_t_float.cu
 delete mode 100644 cpp/src/neighbors/detail/selection_faiss_int64_t_double.cu
 delete mode 100644 cpp/src/neighbors/detail/selection_faiss_int64_t_half.cu
 delete mode 100644 cpp/src/neighbors/detail/selection_faiss_int_double.cu
 delete mode 100644 cpp/src/neighbors/detail/selection_faiss_long_float.cu
 delete mode 100644 cpp/src/neighbors/detail/selection_faiss_size_t_double.cu
 delete mode 100644 cpp/src/neighbors/detail/selection_faiss_size_t_float.cu
 delete mode 100644 cpp/src/neighbors/detail/selection_faiss_uint32_t_double.cu
 delete mode 100644 cpp/src/neighbors/detail/selection_faiss_uint32_t_float.cu
 delete mode 100644 cpp/src/neighbors/detail/selection_faiss_uint32_t_half.cu
 delete mode 100644 cpp/src/neighbors/ivf_flat_00_generate.py
 delete mode 100644 cpp/src/neighbors/ivf_flat_build_float_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivf_flat_build_int8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivf_flat_build_uint8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivf_flat_extend_float_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivf_flat_extend_int8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivf_flat_extend_uint8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivf_flat_search_float_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivf_flat_search_int8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivf_flat_search_uint8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivfpq_build_float_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivfpq_build_int8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivfpq_build_uint8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivfpq_extend_float_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivfpq_extend_int8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivfpq_extend_uint8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivfpq_search_float_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivfpq_search_int8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/ivfpq_search_uint8_t_int64_t.cu
 delete mode 100644 cpp/src/neighbors/refine_00_generate.py
 delete mode 100644 cpp/src/neighbors/refine_float_float.cu
 delete mode 100644 cpp/src/neighbors/refine_int8_t_float.cu
 delete mode 100644 cpp/src/neighbors/refine_uint8_t_float.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_00_generate.py
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_dist.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_euclidean.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_haversine.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_dist.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_euclidean.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_haversine.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_dist.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_euclidean.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_haversine.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_dist.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_euclidean.cu
 delete mode 100644 cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_haversine.cu
 delete mode 100644 cpp/src/spatial/knn/detail/fused_l2_knn_int32_t_float.cu
 delete mode 100644 cpp/src/spatial/knn/detail/fused_l2_knn_int64_t_float.cu
 delete mode 100644 cpp/src/spatial/knn/detail/fused_l2_knn_uint32_t_float.cu
 delete mode 100644 cpp/test/cluster/cluster_solvers.cu
 delete mode 100644 cpp/test/cluster/kmeans.cu
 delete mode 100644 cpp/test/cluster/kmeans_balanced.cu
 delete mode 100644 cpp/test/cluster/kmeans_find_k.cu
 delete mode 100644 cpp/test/cluster/linkage.cu
 delete mode 100644 cpp/test/distance/dist_adj.cu
 delete mode 100644 cpp/test/distance/dist_adj.cuh
 delete mode 100644 cpp/test/distance/dist_adj_distance_instance.cu
 delete mode 100644 cpp/test/distance/dist_adj_threshold.cuh
 delete mode 100644 cpp/test/distance/dist_canberra.cu
 delete mode 100644 cpp/test/distance/dist_correlation.cu
 delete mode 100644 cpp/test/distance/dist_cos.cu
 delete mode 100644 cpp/test/distance/dist_hamming.cu
 delete mode 100644 cpp/test/distance/dist_hellinger.cu
 delete mode 100644 cpp/test/distance/dist_inner_product.cu
 delete mode 100644 cpp/test/distance/dist_jensen_shannon.cu
 delete mode 100644 cpp/test/distance/dist_kl_divergence.cu
 delete mode 100644 cpp/test/distance/dist_l1.cu
 delete mode 100644 cpp/test/distance/dist_l2_exp.cu
 delete mode 100644 cpp/test/distance/dist_l2_sqrt_exp.cu
 delete mode 100644 cpp/test/distance/dist_l2_unexp.cu
 delete mode 100644 cpp/test/distance/dist_l_inf.cu
 delete mode 100644 cpp/test/distance/dist_lp_unexp.cu
 delete mode 100644 cpp/test/distance/dist_russell_rao.cu
 delete mode 100644 cpp/test/distance/distance_base.cuh
 delete mode 100644 cpp/test/distance/fused_l2_nn.cu
 delete mode 100644 cpp/test/distance/gram.cu
 delete mode 100644 cpp/test/distance/gram_base.cuh
 delete mode 100644 cpp/test/distance/masked_nn.cu
 delete mode 100644 cpp/test/distance/masked_nn_compress_to_bits.cu
 delete mode 100644 cpp/test/ext_headers/00_generate.py
 delete mode 100644 cpp/test/ext_headers/raft_core_logger.cpp
 delete mode 100644 cpp/test/ext_headers/raft_distance_detail_pairwise_matrix_dispatch.cu
 delete mode 100644 cpp/test/ext_headers/raft_distance_distance.cu
 delete mode 100644 cpp/test/ext_headers/raft_distance_fused_l2_nn.cu
 delete mode 100644 cpp/test/ext_headers/raft_linalg_detail_coalesced_reduction.cu
 delete mode 100644 cpp/test/ext_headers/raft_matrix_detail_select_k.cu
 delete mode 100644 cpp/test/ext_headers/raft_neighbors_ball_cover.cu
 delete mode 100644 cpp/test/ext_headers/raft_neighbors_brute_force.cu
 delete mode 100644 cpp/test/ext_headers/raft_neighbors_detail_ivf_flat_interleaved_scan.cu
 delete mode 100644 cpp/test/ext_headers/raft_neighbors_detail_ivf_flat_search.cu
 delete mode 100644 cpp/test/ext_headers/raft_neighbors_detail_ivf_pq_compute_similarity.cu
 delete mode 100644 cpp/test/ext_headers/raft_neighbors_detail_selection_faiss.cu
 delete mode 100644 cpp/test/ext_headers/raft_neighbors_ivf_flat.cu
 delete mode 100644 cpp/test/ext_headers/raft_neighbors_ivf_pq.cu
 delete mode 100644 cpp/test/ext_headers/raft_neighbors_refine.cu
 delete mode 100644 cpp/test/ext_headers/raft_spatial_knn_detail_ball_cover_registers.cu
 delete mode 100644 cpp/test/ext_headers/raft_spatial_knn_detail_fused_l2_knn.cu
 delete mode 100644 cpp/test/ext_headers/raft_util_memory_pool.cpp
 delete mode 100644 cpp/test/neighbors/ann_ivf_flat.cuh
 delete mode 100644 cpp/test/neighbors/ann_ivf_flat/test_filter_float_int64_t.cu
 delete mode 100644 cpp/test/neighbors/ann_ivf_flat/test_float_int64_t.cu
 delete mode 100644 cpp/test/neighbors/ann_ivf_flat/test_int8_t_int64_t.cu
 delete mode 100644 cpp/test/neighbors/ann_ivf_flat/test_uint8_t_int64_t.cu
 delete mode 100644 cpp/test/neighbors/ann_ivf_pq.cuh
 delete mode 100644 cpp/test/neighbors/ann_ivf_pq/test_filter_float_int64_t.cu
 delete mode 100644 cpp/test/neighbors/ann_ivf_pq/test_filter_int8_t_int64_t.cu
 delete mode 100644 cpp/test/neighbors/ann_ivf_pq/test_float_int64_t.cu
 delete mode 100644 cpp/test/neighbors/ann_ivf_pq/test_float_uint32_t.cu
 delete mode 100644 cpp/test/neighbors/ann_ivf_pq/test_int8_t_int64_t.cu
 delete mode 100644 cpp/test/neighbors/ann_ivf_pq/test_uint8_t_int64_t.cu
 delete mode 100644 cpp/test/neighbors/ann_nn_descent.cuh
 delete mode 100644 cpp/test/neighbors/ann_nn_descent/test_float_uint32_t.cu
 delete mode 100644 cpp/test/neighbors/ann_nn_descent/test_int8_t_uint32_t.cu
 delete mode 100644 cpp/test/neighbors/ann_nn_descent/test_uint8_t_uint32_t.cu
 delete mode 100644 cpp/test/neighbors/ball_cover.cu
 delete mode 100644 cpp/test/neighbors/epsilon_neighborhood.cu
 delete mode 100644 cpp/test/neighbors/fused_l2_knn.cu
 delete mode 100644 cpp/test/neighbors/haversine.cu
 delete mode 100644 cpp/test/neighbors/knn.cu
 delete mode 100644 cpp/test/neighbors/knn_utils.cuh
 delete mode 100644 cpp/test/neighbors/refine.cu
 delete mode 100644 cpp/test/neighbors/selection.cu
 delete mode 100644 cpp/test/neighbors/spatial_data.h
 delete mode 100644 cpp/test/neighbors/tiled_knn.cu
 delete mode 100644 cpp/test/sparse/dist_coo_spmv.cu
 delete mode 100644 cpp/test/sparse/distance.cu
 delete mode 100644 cpp/test/sparse/gram.cu
 delete mode 100644 cpp/test/sparse/neighbors/brute_force.cu
 delete mode 100644 cpp/test/sparse/neighbors/cross_component_nn.cu
 delete mode 100644 cpp/test/sparse/neighbors/knn_graph.cu
 delete mode 100644 cpp/test/sparse/spectral_matrix.cu
 delete mode 100644 cpp/test/stats/accuracy.cu
 delete mode 100644 cpp/test/stats/adjusted_rand_index.cu
 delete mode 100644 cpp/test/stats/completeness_score.cu
 delete mode 100644 cpp/test/stats/contingencyMatrix.cu
 delete mode 100644 cpp/test/stats/cov.cu
 delete mode 100644 cpp/test/stats/dispersion.cu
 delete mode 100644 cpp/test/stats/entropy.cu
 delete mode 100644 cpp/test/stats/histogram.cu
 delete mode 100644 cpp/test/stats/homogeneity_score.cu
 delete mode 100644 cpp/test/stats/information_criterion.cu
 delete mode 100644 cpp/test/stats/kl_divergence.cu
 delete mode 100644 cpp/test/stats/mean.cu
 delete mode 100644 cpp/test/stats/meanvar.cu
 delete mode 100644 cpp/test/stats/minmax.cu
 delete mode 100644 cpp/test/stats/mutual_info_score.cu
 delete mode 100644 cpp/test/stats/neighborhood_recall.cu
 delete mode 100644 cpp/test/stats/r2_score.cu
 delete mode 100644 cpp/test/stats/rand_index.cu
 delete mode 100644 cpp/test/stats/regression_metrics.cu
 delete mode 100644 cpp/test/stats/silhouette_score.cu
 delete mode 100644 cpp/test/stats/stddev.cu
 delete mode 100644 cpp/test/stats/sum.cu
 delete mode 100644 cpp/test/stats/trustworthiness.cu
 delete mode 100644 cpp/test/stats/v_measure.cu
 delete mode 100644 cpp/test/stats/weighted_mean.cu

diff --git a/build.sh b/build.sh
index eb360ff32..c4b7a7bf7 100755
--- a/build.sh
+++ b/build.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 
-# Copyright (c) 2020-2023, NVIDIA CORPORATION.
+# Copyright (c) 2020-2024, NVIDIA CORPORATION.
 
 # cuvs build scripts
 
@@ -77,8 +77,8 @@ INSTALL_TARGET=install
 BUILD_REPORT_METRICS=""
 BUILD_REPORT_INCL_CACHE_STATS=OFF
 
-TEST_TARGETS="CLUSTER_TEST;DISTANCE_TEST;NEIGHBORS_TEST;NEIGHBORS_ANN_CAGRA_TEST;NEIGHBORS_ANN_NN_DESCENT_TEST;NEIGHBORS_ANN_IVF_TEST"
-BENCH_TARGETS="CLUSTER_BENCH;NEIGHBORS_BENCH;DISTANCE_BENCH"
+TEST_TARGETS="NEIGHBORS_ANN_CAGRA_TEST"
+BENCH_TARGETS="NEIGHBORS_BENCH"
 
 CACHE_ARGS=""
 NVTX=ON
diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
index 8f914227a..2239a7e15 100644
--- a/cpp/CMakeLists.txt
+++ b/cpp/CMakeLists.txt
@@ -191,167 +191,6 @@ include(cmake/thirdparty/get_cutlass.cmake)
 
 add_library(
   cuvs SHARED
-  # src/distance/detail/pairwise_matrix/dispatch_canberra_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_canberra_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_correlation_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_correlation_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_cosine_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_cosine_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_hellinger_expanded_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_hellinger_expanded_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_jensen_shannon_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_jensen_shannon_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_kl_divergence_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_kl_divergence_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_l1_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_l1_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_l2_expanded_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_l2_expanded_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_l2_unexpanded_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_l2_unexpanded_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_l_inf_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_l_inf_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_lp_unexpanded_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_lp_unexpanded_float_float_float_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_rbf.cu
-  # src/distance/detail/pairwise_matrix/dispatch_russel_rao_double_double_double_int.cu
-  # src/distance/detail/pairwise_matrix/dispatch_russel_rao_float_float_float_int.cu
-  # src/distance/distance.cu
-  # src/distance/fused_l2_nn.cu
-  # src/matrix/detail/select_k_double_int64_t.cu
-  # src/matrix/detail/select_k_double_uint32_t.cu
-  # src/matrix/detail/select_k_float_int64_t.cu
-  # src/matrix/detail/select_k_float_uint32_t.cu
-  # src/matrix/detail/select_k_float_int32.cu
-  # src/matrix/detail/select_k_half_int64_t.cu
-  # src/matrix/detail/select_k_half_uint32_t.cu
-  # src/neighbors/ball_cover.cu
-  # src/neighbors/brute_force_fused_l2_knn_float_int64_t.cu
-  # src/neighbors/brute_force_knn_int64_t_float_int64_t.cu
-  # src/neighbors/brute_force_knn_int64_t_float_uint32_t.cu
-  # src/neighbors/brute_force_knn_int_float_int.cu
-  # src/neighbors/brute_force_knn_uint32_t_float_uint32_t.cu
-  # src/neighbors/brute_force_knn_index_float.cu
-  # src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim128_t8.cu
-  # src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim256_t16.cu
-  # src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim512_t32.cu
-  # src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim1024_t32.cu
-  # src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim128_t8.cu
-  # src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim256_t16.cu
-  # src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim512_t32.cu
-  # src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim1024_t32.cu
-  # src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim128_t8.cu
-  # src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim256_t16.cu
-  # src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim512_t32.cu
-  # src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim1024_t32.cu
-  # src/neighbors/detail/cagra/search_single_cta_float_uint32_dim128_t8.cu
-  # src/neighbors/detail/cagra/search_single_cta_float_uint32_dim256_t16.cu
-  # src/neighbors/detail/cagra/search_single_cta_float_uint32_dim512_t32.cu
-  # src/neighbors/detail/cagra/search_single_cta_float_uint32_dim1024_t32.cu
-  # src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim128_t8.cu
-  # src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim256_t16.cu
-  # src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim512_t32.cu
-  # src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim1024_t32.cu
-  # src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim128_t8.cu
-  # src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim256_t16.cu
-  # src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim512_t32.cu
-  # src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim1024_t32.cu
-  # src/neighbors/detail/ivf_flat_interleaved_scan_float_float_int64_t.cu
-  # src/neighbors/detail/ivf_flat_interleaved_scan_int8_t_int32_t_int64_t.cu
-  # src/neighbors/detail/ivf_flat_interleaved_scan_uint8_t_uint32_t_int64_t.cu
-  # src/neighbors/detail/ivf_flat_search.cu
-  # src/neighbors/detail/ivf_pq_compute_similarity_float_float.cu
-  # src/neighbors/detail/ivf_pq_compute_similarity_float_fp8_false.cu
-  # src/neighbors/detail/ivf_pq_compute_similarity_float_fp8_true.cu
-  # src/neighbors/detail/ivf_pq_compute_similarity_float_half.cu
-  # src/neighbors/detail/ivf_pq_compute_similarity_half_fp8_false.cu
-  # src/neighbors/detail/ivf_pq_compute_similarity_half_fp8_true.cu
-  # src/neighbors/detail/ivf_pq_compute_similarity_half_half.cu
-  # src/neighbors/detail/refine_host_float_float.cpp
-  # src/neighbors/detail/refine_host_int8_t_float.cpp
-  # src/neighbors/detail/refine_host_uint8_t_float.cpp
-  # src/neighbors/detail/selection_faiss_int32_t_float.cu
-  # src/neighbors/detail/selection_faiss_int_double.cu
-  # src/neighbors/detail/selection_faiss_long_float.cu
-  # src/neighbors/detail/selection_faiss_size_t_double.cu
-  # src/neighbors/detail/selection_faiss_size_t_float.cu
-  # src/neighbors/detail/selection_faiss_uint32_t_float.cu
-  # src/neighbors/detail/selection_faiss_int64_t_double.cu
-  # src/neighbors/detail/selection_faiss_int64_t_half.cu
-  # src/neighbors/detail/selection_faiss_uint32_t_double.cu
-  # src/neighbors/detail/selection_faiss_uint32_t_half.cu
-  # src/neighbors/ivf_flat_build_float_int64_t.cu
-  # src/neighbors/ivf_flat_build_int8_t_int64_t.cu
-  # src/neighbors/ivf_flat_build_uint8_t_int64_t.cu
-  # src/neighbors/ivf_flat_extend_float_int64_t.cu
-  # src/neighbors/ivf_flat_extend_int8_t_int64_t.cu
-  # src/neighbors/ivf_flat_extend_uint8_t_int64_t.cu
-  # src/neighbors/ivf_flat_search_float_int64_t.cu
-  # src/neighbors/ivf_flat_search_int8_t_int64_t.cu
-  # src/neighbors/ivf_flat_search_uint8_t_int64_t.cu
-  # src/neighbors/ivfpq_build_float_int64_t.cu
-  # src/neighbors/ivfpq_build_int8_t_int64_t.cu
-  # src/neighbors/ivfpq_build_uint8_t_int64_t.cu
-  # src/neighbors/ivfpq_extend_float_int64_t.cu
-  # src/neighbors/ivfpq_extend_int8_t_int64_t.cu
-  # src/neighbors/ivfpq_extend_uint8_t_int64_t.cu
-  # src/neighbors/ivfpq_search_float_int64_t.cu
-  # src/neighbors/ivfpq_search_int8_t_int64_t.cu
-  # src/neighbors/ivfpq_search_uint8_t_int64_t.cu
-  # src/neighbors/refine_float_float.cu
-  # src/neighbors/refine_int8_t_float.cu
-  # src/neighbors/refine_uint8_t_float.cu
-  # src/cuvs_runtime/cluster/cluster_cost.cuh
-  # src/cuvs_runtime/cluster/cluster_cost_double.cu
-  # src/cuvs_runtime/cluster/cluster_cost_float.cu
-  # src/cuvs_runtime/cluster/kmeans_fit_double.cu
-  # src/cuvs_runtime/cluster/kmeans_fit_float.cu
-  # src/cuvs_runtime/cluster/kmeans_init_plus_plus_double.cu
-  # src/cuvs_runtime/cluster/kmeans_init_plus_plus_float.cu
-  # src/cuvs_runtime/cluster/update_centroids.cuh
-  # src/cuvs_runtime/cluster/update_centroids_double.cu
-  # src/cuvs_runtime/cluster/update_centroids_float.cu
-  # src/cuvs_runtime/distance/fused_l2_min_arg.cu
-  # src/cuvs_runtime/distance/pairwise_distance.cu
-  # src/cuvs_runtime/matrix/select_k_float_int64_t.cu
-  # src/cuvs_runtime/neighbors/brute_force_knn_int64_t_float.cu
-  # src/cuvs_runtime/neighbors/ivf_flat_build.cu
-  # src/cuvs_runtime/neighbors/ivf_flat_search.cu
-  # src/cuvs_runtime/neighbors/ivf_flat_serialize.cu
-  # src/cuvs_runtime/neighbors/ivfpq_build.cu
-  # src/cuvs_runtime/neighbors/ivfpq_deserialize.cu
-  # src/cuvs_runtime/neighbors/ivfpq_search_float_int64_t.cu
-  # src/cuvs_runtime/neighbors/ivfpq_search_int8_t_int64_t.cu
-  # src/cuvs_runtime/neighbors/ivfpq_search_uint8_t_int64_t.cu
-  # src/cuvs_runtime/neighbors/ivfpq_serialize.cu
-  # src/cuvs_runtime/neighbors/refine_d_int64_t_float.cu
-  # src/cuvs_runtime/neighbors/refine_d_int64_t_int8_t.cu
-  # src/cuvs_runtime/neighbors/refine_d_int64_t_uint8_t.cu
-  # src/cuvs_runtime/neighbors/refine_h_int64_t_float.cu
-  # src/cuvs_runtime/neighbors/refine_h_int64_t_int8_t.cu
-  # src/cuvs_runtime/neighbors/refine_h_int64_t_uint8_t.cu
-  # src/cuvs_runtime/random/rmat_rectangular_generator_int64_double.cu
-  # src/cuvs_runtime/random/rmat_rectangular_generator_int64_float.cu
-  # src/cuvs_runtime/random/rmat_rectangular_generator_int_double.cu
-  # src/cuvs_runtime/random/rmat_rectangular_generator_int_float.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_one_2d_dist.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_one_2d_euclidean.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_one_2d_haversine.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_one_3d_dist.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_one_3d_euclidean.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_one_3d_haversine.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_two_2d_dist.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_two_2d_euclidean.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_two_2d_haversine.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_two_3d_dist.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_two_3d_euclidean.cu
-  # src/spatial/knn/detail/ball_cover/registers_pass_two_3d_haversine.cu
-  # src/spatial/knn/detail/fused_l2_knn_int32_t_float.cu
-  # src/spatial/knn/detail/fused_l2_knn_int64_t_float.cu
-  # src/spatial/knn/detail/fused_l2_knn_uint32_t_float.cu
-
   src/neighbors/cagra_build_float.cpp
   src/neighbors/cagra_build_int8.cpp
   src/neighbors/cagra_build_uint8.cpp
diff --git a/cpp/bench/ann/src/raft/raft_cagra_wrapper.h b/cpp/bench/ann/src/raft/raft_cagra_wrapper.h
index 90c4218c5..c846416a4 100644
--- a/cpp/bench/ann/src/raft/raft_cagra_wrapper.h
+++ b/cpp/bench/ann/src/raft/raft_cagra_wrapper.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2023, NVIDIA CORPORATION.
+ * Copyright (c) 2023-2024, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -17,9 +17,7 @@
 
 #include <cassert>
 #include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/cagra.cuh>
-#include <cuvs/neighbors/cagra_serialize.cuh>
-#include <cuvs/neighbors/cagra_types.hpp>
+#include <cuvs/neighbors/cagra.hpp>
 #include <cuvs/neighbors/detail/cagra/cagra_build.cuh>
 #include <cuvs/neighbors/ivf_pq_types.hpp>
 #include <cuvs/neighbors/nn_descent_types.hpp>
@@ -57,7 +55,7 @@ class RaftCagra : public ANN<T> {
   using typename ANN<T>::AnnSearchParam;
 
   struct SearchParam : public AnnSearchParam {
-    cuvs::neighbors::experimental::cagra::search_params p;
+    cuvs::neighbors::cagra::search_params p;
     AllocatorType graph_mem   = AllocatorType::Device;
     AllocatorType dataset_mem = AllocatorType::Device;
     auto needs_dataset() const -> bool override { return true; }
@@ -209,7 +207,7 @@ void RaftCagra<T, IdxT>::set_search_param(const AnnSearchParam& param)
                   allocator_to_string(dataset_mem_).c_str());
 
     auto mr = get_mr(dataset_mem_);
-    cuvs::neighbors::cagra::detail::copy_with_padding(handle_, dataset_, input_dataset_v_, mr);
+    raft::neighbors::cagra::detail::copy_with_padding(handle_, dataset_, input_dataset_v_, mr);
 
     index_->update_dataset(handle_, make_const_mdspan(dataset_.view()));
 
diff --git a/cpp/bench/micro/neighbors/cagra_bench.cuh b/cpp/bench/micro/neighbors/cagra_bench.cuh
index 3be664db8..0cc8c9578 100644
--- a/cpp/bench/micro/neighbors/cagra_bench.cuh
+++ b/cpp/bench/micro/neighbors/cagra_bench.cuh
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2023, NVIDIA CORPORATION.
+ * Copyright (c) 2023-2024, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -17,7 +17,7 @@
 #pragma once
 
 #include <common/benchmark.hpp>
-#include <cuvs/neighbors/cagra.cuh>
+#include <cuvs/neighbors/cagra.hpp>
 #include <cuvs/neighbors/sample_filter.cuh>
 #include <raft/random/rng.cuh>
 #include <raft/util/itertools.hpp>
diff --git a/cpp/include/cuvs/neighbors/ball_cover-ext.cuh b/cpp/include/cuvs/neighbors/ball_cover-ext.cuh
deleted file mode 100644
index b1cd2b4ed..000000000
--- a/cpp/include/cuvs/neighbors/ball_cover-ext.cuh
+++ /dev/null
@@ -1,124 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#include <cstdint>                              // uint32_t
-#include <cuvs/distance/distance_types.hpp>     // cuvs::distance::DistanceType
-#include <cuvs/neighbors/ball_cover_types.hpp>  // BallCoverIndex
-#include <raft/util/raft_explicit.hpp>          // RAFT_EXPLICIT
-
-#ifdef RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-namespace cuvs::neighbors::ball_cover {
-
-template <typename idx_t, typename value_t, typename int_t, typename matrix_idx_t>
-void build_index(raft::resources const& handle,
-                 BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index) RAFT_EXPLICIT;
-
-template <typename idx_t, typename value_t, typename int_t, typename matrix_idx_t>
-void all_knn_query(raft::resources const& handle,
-                   BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index,
-                   int_t k,
-                   idx_t* inds,
-                   value_t* dists,
-                   bool perform_post_filtering = true,
-                   float weight                = 1.0) RAFT_EXPLICIT;
-
-template <typename idx_t, typename value_t, typename int_t, typename matrix_idx_t>
-void all_knn_query(raft::resources const& handle,
-                   BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index,
-                   raft::device_matrix_view<idx_t, matrix_idx_t, raft::row_major> inds,
-                   raft::device_matrix_view<value_t, matrix_idx_t, raft::row_major> dists,
-                   int_t k,
-                   bool perform_post_filtering = true,
-                   float weight                = 1.0) RAFT_EXPLICIT;
-
-template <typename idx_t, typename value_t, typename int_t>
-void knn_query(raft::resources const& handle,
-               const BallCoverIndex<idx_t, value_t, int_t>& index,
-               int_t k,
-               const value_t* query,
-               int_t n_query_pts,
-               idx_t* inds,
-               value_t* dists,
-               bool perform_post_filtering = true,
-               float weight                = 1.0) RAFT_EXPLICIT;
-
-template <typename idx_t, typename value_t, typename int_t, typename matrix_idx_t>
-void knn_query(raft::resources const& handle,
-               const BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index,
-               raft::device_matrix_view<const value_t, matrix_idx_t, raft::row_major> query,
-               raft::device_matrix_view<idx_t, matrix_idx_t, raft::row_major> inds,
-               raft::device_matrix_view<value_t, matrix_idx_t, raft::row_major> dists,
-               int_t k,
-               bool perform_post_filtering = true,
-               float weight                = 1.0) RAFT_EXPLICIT;
-
-}  // namespace cuvs::neighbors::ball_cover
-
-#endif  // RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-#define instantiate_raft_neighbors_ball_cover(idx_t, value_t, int_t, matrix_idx_t)                 \
-  extern template void                                                                             \
-  cuvs::neighbors::ball_cover::build_index<idx_t, value_t, int_t, matrix_idx_t>(                   \
-    raft::resources const& handle,                                                                 \
-    cuvs::neighbors::ball_cover::BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index);      \
-                                                                                                   \
-  extern template void                                                                             \
-  cuvs::neighbors::ball_cover::all_knn_query<idx_t, value_t, int_t, matrix_idx_t>(                 \
-    raft::resources const& handle,                                                                 \
-    cuvs::neighbors::ball_cover::BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index,       \
-    int_t k,                                                                                       \
-    idx_t* inds,                                                                                   \
-    value_t* dists,                                                                                \
-    bool perform_post_filtering,                                                                   \
-    float weight);                                                                                 \
-                                                                                                   \
-  extern template void                                                                             \
-  cuvs::neighbors::ball_cover::all_knn_query<idx_t, value_t, int_t, matrix_idx_t>(                 \
-    raft::resources const& handle,                                                                 \
-    cuvs::neighbors::ball_cover::BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index,       \
-    raft::device_matrix_view<idx_t, matrix_idx_t, raft::row_major> inds,                           \
-    raft::device_matrix_view<value_t, matrix_idx_t, raft::row_major> dists,                        \
-    int_t k,                                                                                       \
-    bool perform_post_filtering,                                                                   \
-    float weight);                                                                                 \
-                                                                                                   \
-  extern template void cuvs::neighbors::ball_cover::knn_query<idx_t, value_t, int_t>(              \
-    raft::resources const& handle,                                                                 \
-    const cuvs::neighbors::ball_cover::BallCoverIndex<idx_t, value_t, int_t>& index,               \
-    int_t k,                                                                                       \
-    const value_t* query,                                                                          \
-    int_t n_query_pts,                                                                             \
-    idx_t* inds,                                                                                   \
-    value_t* dists,                                                                                \
-    bool perform_post_filtering,                                                                   \
-    float weight);                                                                                 \
-                                                                                                   \
-  extern template void                                                                             \
-  cuvs::neighbors::ball_cover::knn_query<idx_t, value_t, int_t, matrix_idx_t>(                     \
-    raft::resources const& handle,                                                                 \
-    const cuvs::neighbors::ball_cover::BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index, \
-    raft::device_matrix_view<const value_t, matrix_idx_t, raft::row_major> query,                  \
-    raft::device_matrix_view<idx_t, matrix_idx_t, raft::row_major> inds,                           \
-    raft::device_matrix_view<value_t, matrix_idx_t, raft::row_major> dists,                        \
-    int_t k,                                                                                       \
-    bool perform_post_filtering,                                                                   \
-    float weight);
-
-instantiate_raft_neighbors_ball_cover(int64_t, float, uint32_t, uint32_t);
-
-#undef instantiate_raft_neighbors_ball_cover
diff --git a/cpp/include/cuvs/neighbors/ball_cover-inl.cuh b/cpp/include/cuvs/neighbors/ball_cover-inl.cuh
deleted file mode 100644
index 4d0f170df..000000000
--- a/cpp/include/cuvs/neighbors/ball_cover-inl.cuh
+++ /dev/null
@@ -1,395 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#ifndef __BALL_COVER_H
-#define __BALL_COVER_H
-
-#pragma once
-
-#include <cstdint>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/ball_cover_types.hpp>
-#include <cuvs/spatial/knn/detail/ball_cover.cuh>
-#include <cuvs/spatial/knn/detail/ball_cover/common.cuh>
-#include <thrust/transform.h>
-
-namespace cuvs::neighbors::ball_cover {
-
-/**
- * @defgroup random_ball_cover Random Ball Cover algorithm
- * @{
- */
-
-/**
- * Builds and populates a previously unbuilt BallCoverIndex
- *
- * Usage example:
- * @code{.cpp}
- *
- *  #include <raft/core/resources.hpp>
- *  #include <cuvs/neighbors/ball_cover.cuh>
- *  #include <cuvs/distance/distance_types.hpp>
- *  using namespace cuvs::neighbors;
- *
- *  raft::resources handle;
- *  ...
- *  auto metric = cuvs::distance::DistanceType::L2Expanded;
- *  BallCoverIndex index(handle, X, metric);
- *
- *  ball_cover::build_index(handle, index);
- * @endcode
- *
- * @tparam idx_t knn index type
- * @tparam value_t knn value type
- * @tparam int_t integral type for knn params
- * @tparam matrix_idx_t matrix indexing type
- * @param[in] handle library resource management handle
- * @param[inout] index an empty (and not previous built) instance of BallCoverIndex
- */
-template <typename idx_t, typename value_t, typename int_t, typename matrix_idx_t>
-void build_index(raft::resources const& handle,
-                 BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index)
-{
-  ASSERT(index.n <= 3, "only 2d and 3d vectors are supported in current implementation");
-  if (index.metric == cuvs::distance::DistanceType::Haversine) {
-    cuvs::spatial::knn::detail::rbc_build_index(
-      handle, index, spatial::knn::detail::HaversineFunc<value_t, int_t>());
-  } else if (index.metric == cuvs::distance::DistanceType::L2SqrtExpanded ||
-             index.metric == cuvs::distance::DistanceType::L2SqrtUnexpanded) {
-    cuvs::spatial::knn::detail::rbc_build_index(
-      handle, index, spatial::knn::detail::EuclideanFunc<value_t, int_t>());
-  } else {
-    RAFT_FAIL("Metric not support");
-  }
-
-  index.set_index_trained();
-}
-
-/** @} */  // end group random_ball_cover
-
-/**
- * Performs a faster exact knn in metric spaces using the triangle
- * inequality with a number of landmark points to reduce the
- * number of distance computations from O(n^2) to O(sqrt(n)). This
- * performs an all neighbors knn, which can reuse memory when
- * the index and query are the same array. This function will
- * build the index and assumes rbc_build_index() has not already
- * been called.
- * @tparam idx_t knn index type
- * @tparam value_t knn distance type
- * @tparam int_t type for integers, such as number of rows/cols
- * @param[in] handle raft handle for resource management
- * @param[inout] index ball cover index which has not yet been built
- * @param[in] k number of nearest neighbors to find
- * @param[in] perform_post_filtering if this is false, only the closest k landmarks
- *                               are considered (which will return approximate
- *                               results).
- * @param[out] inds output knn indices
- * @param[out] dists output knn distances
- * @param[in] weight a weight for overlap between the closest landmark and
- *               the radius of other landmarks when pruning distances.
- *               Setting this value below 1 can effectively turn off
- *               computing distances against many other balls, enabling
- *               approximate nearest neighbors. Recall can be adjusted
- *               based on how many relevant balls are ignored. Note that
- *               many datasets can still have great recall even by only
- *               looking in the closest landmark.
- */
-template <typename idx_t, typename value_t, typename int_t, typename matrix_idx_t>
-void all_knn_query(raft::resources const& handle,
-                   BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index,
-                   int_t k,
-                   idx_t* inds,
-                   value_t* dists,
-                   bool perform_post_filtering = true,
-                   float weight                = 1.0)
-{
-  ASSERT(index.n <= 3, "only 2d and 3d vectors are supported in current implementation");
-  if (index.metric == cuvs::distance::DistanceType::Haversine) {
-    cuvs::spatial::knn::detail::rbc_all_knn_query(
-      handle,
-      index,
-      k,
-      inds,
-      dists,
-      spatial::knn::detail::HaversineFunc<value_t, int_t>(),
-      perform_post_filtering,
-      weight);
-  } else if (index.metric == cuvs::distance::DistanceType::L2SqrtExpanded ||
-             index.metric == cuvs::distance::DistanceType::L2SqrtUnexpanded) {
-    cuvs::spatial::knn::detail::rbc_all_knn_query(
-      handle,
-      index,
-      k,
-      inds,
-      dists,
-      spatial::knn::detail::EuclideanFunc<value_t, int_t>(),
-      perform_post_filtering,
-      weight);
-  } else {
-    RAFT_FAIL("Metric not supported");
-  }
-
-  index.set_index_trained();
-}
-
-/**
- * @ingroup random_ball_cover
- * @{
- */
-
-/**
- * Performs a faster exact knn in metric spaces using the triangle
- * inequality with a number of landmark points to reduce the
- * number of distance computations from O(n^2) to O(sqrt(n)). This
- * performs an all neighbors knn, which can reuse memory when
- * the index and query are the same array. This function will
- * build the index and assumes rbc_build_index() has not already
- * been called.
- *
- * Usage example:
- * @code{.cpp}
- *
- *  #include <raft/core/resources.hpp>
- *  #include <cuvs/neighbors/ball_cover.cuh>
- *  #include <cuvs/distance/distance_types.hpp>
- *  using namespace cuvs::neighbors;
- *
- *  raft::resources handle;
- *  ...
- *  auto metric = cuvs::distance::DistanceType::L2Expanded;
- *
- *  // Construct a ball cover index
- *  BallCoverIndex index(handle, X, metric);
- *
- *  // Perform all neighbors knn query
- *  ball_cover::all_knn_query(handle, index, inds, dists, k);
- * @endcode
- *
- * @tparam idx_t knn index type
- * @tparam value_t knn distance type
- * @tparam int_t type for integers, such as number of rows/cols
- * @tparam matrix_idx_t matrix indexing type
- *
- * @param[in] handle raft handle for resource management
- * @param[in] index ball cover index which has not yet been built
- * @param[out] inds output knn indices
- * @param[out] dists output knn distances
- * @param[in] k number of nearest neighbors to find
- * @param[in] perform_post_filtering if this is false, only the closest k landmarks
- *                               are considered (which will return approximate
- *                               results).
- * @param[in] weight a weight for overlap between the closest landmark and
- *               the radius of other landmarks when pruning distances.
- *               Setting this value below 1 can effectively turn off
- *               computing distances against many other balls, enabling
- *               approximate nearest neighbors. Recall can be adjusted
- *               based on how many relevant balls are ignored. Note that
- *               many datasets can still have great recall even by only
- *               looking in the closest landmark.
- */
-template <typename idx_t, typename value_t, typename int_t, typename matrix_idx_t>
-void all_knn_query(raft::resources const& handle,
-                   BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index,
-                   raft::device_matrix_view<idx_t, matrix_idx_t, raft::row_major> inds,
-                   raft::device_matrix_view<value_t, matrix_idx_t, raft::row_major> dists,
-                   int_t k,
-                   bool perform_post_filtering = true,
-                   float weight                = 1.0)
-{
-  RAFT_EXPECTS(index.n <= 3, "only 2d and 3d vectors are supported in current implementation");
-  RAFT_EXPECTS(k <= index.m,
-               "k must be less than or equal to the number of data points in the index");
-  RAFT_EXPECTS(inds.extent(1) == dists.extent(1) && dists.extent(1) == static_cast<matrix_idx_t>(k),
-               "Number of columns in output indices and distances matrices must be equal to k");
-
-  RAFT_EXPECTS(inds.extent(0) == dists.extent(0) && dists.extent(0) == index.get_X().extent(0),
-               "Number of rows in output indices and distances matrices must equal number of rows "
-               "in index matrix.");
-
-  all_knn_query(
-    handle, index, k, inds.data_handle(), dists.data_handle(), perform_post_filtering, weight);
-}
-
-/** @} */
-
-/**
- * Performs a faster exact knn in metric spaces using the triangle
- * inequality with a number of landmark points to reduce the
- * number of distance computations from O(n^2) to O(sqrt(n)). This
- * function does not build the index and assumes rbc_build_index() has
- * already been called. Use this function when the index and
- * query arrays are different, otherwise use rbc_all_knn_query().
- * @tparam idx_t index type
- * @tparam value_t distances type
- * @tparam int_t integer type for size info
- * @param[in] handle raft handle for resource management
- * @param[inout] index ball cover index which has not yet been built
- * @param[in] k number of nearest neighbors to find
- * @param[in] query the
- * @param[in] perform_post_filtering if this is false, only the closest k landmarks
- *                               are considered (which will return approximate
- *                               results).
- * @param[out] inds output knn indices
- * @param[out] dists output knn distances
- * @param[in] weight a weight for overlap between the closest landmark and
- *               the radius of other landmarks when pruning distances.
- *               Setting this value below 1 can effectively turn off
- *               computing distances against many other balls, enabling
- *               approximate nearest neighbors. Recall can be adjusted
- *               based on how many relevant balls are ignored. Note that
- *               many datasets can still have great recall even by only
- *               looking in the closest landmark.
- * @param[in] n_query_pts number of query points
- */
-template <typename idx_t, typename value_t, typename int_t>
-void knn_query(raft::resources const& handle,
-               const BallCoverIndex<idx_t, value_t, int_t>& index,
-               int_t k,
-               const value_t* query,
-               int_t n_query_pts,
-               idx_t* inds,
-               value_t* dists,
-               bool perform_post_filtering = true,
-               float weight                = 1.0)
-{
-  ASSERT(index.n <= 3, "only 2d and 3d vectors are supported in current implementation");
-  if (index.metric == cuvs::distance::DistanceType::Haversine) {
-    cuvs::spatial::knn::detail::rbc_knn_query(handle,
-                                              index,
-                                              k,
-                                              query,
-                                              n_query_pts,
-                                              inds,
-                                              dists,
-                                              spatial::knn::detail::HaversineFunc<value_t, int_t>(),
-                                              perform_post_filtering,
-                                              weight);
-  } else if (index.metric == cuvs::distance::DistanceType::L2SqrtExpanded ||
-             index.metric == cuvs::distance::DistanceType::L2SqrtUnexpanded) {
-    cuvs::spatial::knn::detail::rbc_knn_query(handle,
-                                              index,
-                                              k,
-                                              query,
-                                              n_query_pts,
-                                              inds,
-                                              dists,
-                                              spatial::knn::detail::EuclideanFunc<value_t, int_t>(),
-                                              perform_post_filtering,
-                                              weight);
-  } else {
-    RAFT_FAIL("Metric not supported");
-  }
-}
-
-/**
- * @ingroup random_ball_cover
- * @{
- */
-
-/**
- * Performs a faster exact knn in metric spaces using the triangle
- * inequality with a number of landmark points to reduce the
- * number of distance computations from O(n^2) to O(sqrt(n)). This
- * function does not build the index and assumes rbc_build_index() has
- * already been called. Use this function when the index and
- * query arrays are different, otherwise use rbc_all_knn_query().
- *
- * Usage example:
- * @code{.cpp}
- *
- *  #include <raft/core/resources.hpp>
- *  #include <cuvs/neighbors/ball_cover.cuh>
- *  #include <cuvs/distance/distance_types.hpp>
- *  using namespace cuvs::neighbors;
- *
- *  raft::resources handle;
- *  ...
- *  auto metric = cuvs::distance::DistanceType::L2Expanded;
- *
- *  // Build a ball cover index
- *  BallCoverIndex index(handle, X, metric);
- *  ball_cover::build_index(handle, index);
- *
- *  // Perform all neighbors knn query
- *  ball_cover::knn_query(handle, index, inds, dists, k);
- * @endcode
-
- *
- * @tparam idx_t index type
- * @tparam value_t distances type
- * @tparam int_t integer type for size info
- * @tparam matrix_idx_t
- * @param[in] handle raft handle for resource management
- * @param[in] index ball cover index which has not yet been built
- * @param[in] query device matrix containing query data points
- * @param[out] inds output knn indices
- * @param[out] dists output knn distances
- * @param[in] k number of nearest neighbors to find
- * @param[in] perform_post_filtering if this is false, only the closest k landmarks
- *                               are considered (which will return approximate
- *                               results).
- * @param[in] weight a weight for overlap between the closest landmark and
- *               the radius of other landmarks when pruning distances.
- *               Setting this value below 1 can effectively turn off
- *               computing distances against many other balls, enabling
- *               approximate nearest neighbors. Recall can be adjusted
- *               based on how many relevant balls are ignored. Note that
- *               many datasets can still have great recall even by only
- *               looking in the closest landmark.
- */
-template <typename idx_t, typename value_t, typename int_t, typename matrix_idx_t>
-void knn_query(raft::resources const& handle,
-               const BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index,
-               raft::device_matrix_view<const value_t, matrix_idx_t, raft::row_major> query,
-               raft::device_matrix_view<idx_t, matrix_idx_t, raft::row_major> inds,
-               raft::device_matrix_view<value_t, matrix_idx_t, raft::row_major> dists,
-               int_t k,
-               bool perform_post_filtering = true,
-               float weight                = 1.0)
-{
-  RAFT_EXPECTS(k <= index.m,
-               "k must be less than or equal to the number of data points in the index");
-  RAFT_EXPECTS(inds.extent(1) == dists.extent(1) && dists.extent(1) == static_cast<idx_t>(k),
-               "Number of columns in output indices and distances matrices must be equal to k");
-
-  RAFT_EXPECTS(inds.extent(0) == dists.extent(0) && dists.extent(0) == query.extent(0),
-               "Number of rows in output indices and distances matrices must equal number of rows "
-               "in search matrix.");
-
-  RAFT_EXPECTS(query.extent(1) == index.get_X().extent(1),
-               "Number of columns in query and index matrices must match.");
-
-  knn_query(handle,
-            index,
-            k,
-            query.data_handle(),
-            query.extent(0),
-            inds.data_handle(),
-            dists.data_handle(),
-            perform_post_filtering,
-            weight);
-}
-
-/** @} */
-
-// TODO: implement functions for:
-//  4. rbc_eps_neigh() - given a populated index, perform query against different query array
-//  5. rbc_all_eps_neigh() - populate a BallCoverIndex and query against training data
-
-}  // namespace cuvs::neighbors::ball_cover
-
-#endif
diff --git a/cpp/include/cuvs/neighbors/ball_cover.cuh b/cpp/include/cuvs/neighbors/ball_cover.cuh
deleted file mode 100644
index 41c5d0310..000000000
--- a/cpp/include/cuvs/neighbors/ball_cover.cuh
+++ /dev/null
@@ -1,24 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#ifndef RAFT_EXPLICIT_INSTANTIATE_ONLY
-#include "ball_cover-inl.cuh"
-#endif
-
-#ifdef RAFT_COMPILED
-#include "ball_cover-ext.cuh"
-#endif
diff --git a/cpp/include/cuvs/neighbors/ball_cover_types.hpp b/cpp/include/cuvs/neighbors/ball_cover_types.hpp
deleted file mode 100644
index c6e9fab2c..000000000
--- a/cpp/include/cuvs/neighbors/ball_cover_types.hpp
+++ /dev/null
@@ -1,169 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cstdint>
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resources.hpp>
-#include <rmm/device_uvector.hpp>
-
-namespace cuvs::neighbors::ball_cover {
-
-/**
- * @ingroup random_ball_cover
- * @{
- */
-
-/**
- * Stores raw index data points, sampled landmarks, the 1-nns of index points
- * to their closest landmarks, and the ball radii of each landmark. This
- * class is intended to be constructed once and reused across subsequent
- * queries.
- * @tparam value_idx
- * @tparam value_t
- * @tparam value_int
- */
-template <typename value_idx,
-          typename value_t,
-          typename value_int  = std::uint32_t,
-          typename matrix_idx = std::uint32_t>
-class BallCoverIndex {
- public:
-  explicit BallCoverIndex(raft::resources const& handle_,
-                          const value_t* X_,
-                          value_int m_,
-                          value_int n_,
-                          cuvs::distance::DistanceType metric_)
-    : handle(handle_),
-      X(raft::make_device_matrix_view<const value_t, matrix_idx>(X_, m_, n_)),
-      m(m_),
-      n(n_),
-      metric(metric_),
-      /**
-       * the sqrt() here makes the sqrt(m)^2 a linear-time lower bound
-       *
-       * Total memory footprint of index: (2 * sqrt(m)) + (n * sqrt(m)) + (2 * m)
-       */
-      n_landmarks(sqrt(m_)),
-      R_indptr(raft::make_device_vector<value_idx, matrix_idx>(handle, sqrt(m_) + 1)),
-      R_1nn_cols(raft::make_device_vector<value_idx, matrix_idx>(handle, m_)),
-      R_1nn_dists(raft::make_device_vector<value_t, matrix_idx>(handle, m_)),
-      R_closest_landmark_dists(raft::make_device_vector<value_t, matrix_idx>(handle, m_)),
-      R(raft::make_device_matrix<value_t, matrix_idx>(handle, sqrt(m_), n_)),
-      R_radius(raft::make_device_vector<value_t, matrix_idx>(handle, sqrt(m_))),
-      index_trained(false)
-  {
-  }
-
-  explicit BallCoverIndex(raft::resources const& handle_,
-                          raft::device_matrix_view<const value_t, matrix_idx, raft::row_major> X_,
-                          cuvs::distance::DistanceType metric_)
-    : handle(handle_),
-      X(X_),
-      m(X_.extent(0)),
-      n(X_.extent(1)),
-      metric(metric_),
-      /**
-       * the sqrt() here makes the sqrt(m)^2 a linear-time lower bound
-       *
-       * Total memory footprint of index: (2 * sqrt(m)) + (n * sqrt(m)) + (2 * m)
-       */
-      n_landmarks(sqrt(X_.extent(0))),
-      R_indptr(raft::make_device_vector<value_idx, matrix_idx>(handle, sqrt(X_.extent(0)) + 1)),
-      R_1nn_cols(raft::make_device_vector<value_idx, matrix_idx>(handle, X_.extent(0))),
-      R_1nn_dists(raft::make_device_vector<value_t, matrix_idx>(handle, X_.extent(0))),
-      R_closest_landmark_dists(raft::make_device_vector<value_t, matrix_idx>(handle, X_.extent(0))),
-      R(raft::make_device_matrix<value_t, matrix_idx>(handle, sqrt(X_.extent(0)), X_.extent(1))),
-      R_radius(raft::make_device_vector<value_t, matrix_idx>(handle, sqrt(X_.extent(0)))),
-      index_trained(false)
-  {
-  }
-
-  auto get_R_indptr() const -> raft::device_vector_view<const value_idx, matrix_idx>
-  {
-    return R_indptr.view();
-  }
-  auto get_R_1nn_cols() const -> raft::device_vector_view<const value_idx, matrix_idx>
-  {
-    return R_1nn_cols.view();
-  }
-  auto get_R_1nn_dists() const -> raft::device_vector_view<const value_t, matrix_idx>
-  {
-    return R_1nn_dists.view();
-  }
-  auto get_R_radius() const -> raft::device_vector_view<const value_t, matrix_idx>
-  {
-    return R_radius.view();
-  }
-  auto get_R() const -> raft::device_matrix_view<const value_t, matrix_idx, raft::row_major>
-  {
-    return R.view();
-  }
-  auto get_R_closest_landmark_dists() const -> raft::device_vector_view<const value_t, matrix_idx>
-  {
-    return R_closest_landmark_dists.view();
-  }
-
-  raft::device_vector_view<value_idx, matrix_idx> get_R_indptr() { return R_indptr.view(); }
-  raft::device_vector_view<value_idx, matrix_idx> get_R_1nn_cols() { return R_1nn_cols.view(); }
-  raft::device_vector_view<value_t, matrix_idx> get_R_1nn_dists() { return R_1nn_dists.view(); }
-  raft::device_vector_view<value_t, matrix_idx> get_R_radius() { return R_radius.view(); }
-  raft::device_matrix_view<value_t, matrix_idx, raft::row_major> get_R() { return R.view(); }
-  raft::device_vector_view<value_t, matrix_idx> get_R_closest_landmark_dists()
-  {
-    return R_closest_landmark_dists.view();
-  }
-  raft::device_matrix_view<const value_t, matrix_idx, raft::row_major> get_X() const { return X; }
-
-  cuvs::distance::DistanceType get_metric() const { return metric; }
-
-  value_int get_n_landmarks() const { return n_landmarks; }
-  bool is_index_trained() const { return index_trained; };
-
-  // This should only be set by internal functions
-  void set_index_trained() { index_trained = true; }
-
-  raft::resources const& handle;
-
-  value_int m;
-  value_int n;
-  value_int n_landmarks;
-
-  raft::device_matrix_view<const value_t, matrix_idx, raft::row_major> X;
-
-  cuvs::distance::DistanceType metric;
-
- private:
-  // CSR storing the neighborhoods for each data point
-  raft::device_vector<value_idx, matrix_idx> R_indptr;
-  raft::device_vector<value_idx, matrix_idx> R_1nn_cols;
-  raft::device_vector<value_t, matrix_idx> R_1nn_dists;
-  raft::device_vector<value_t, matrix_idx> R_closest_landmark_dists;
-
-  raft::device_vector<value_t, matrix_idx> R_radius;
-
-  raft::device_matrix<value_t, matrix_idx, raft::row_major> R;
-
- protected:
-  bool index_trained;
-};
-
-/** @} */
-
-}  // namespace cuvs::neighbors::ball_cover
diff --git a/cpp/include/cuvs/neighbors/brute_force-ext.cuh b/cpp/include/cuvs/neighbors/brute_force-ext.cuh
deleted file mode 100644
index bc4773513..000000000
--- a/cpp/include/cuvs/neighbors/brute_force-ext.cuh
+++ /dev/null
@@ -1,149 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <optional>
-
-#include <cuvs/distance/distance_types.hpp>  // cuvs::distance::DistanceType
-#include <cuvs/neighbors/brute_force_types.hpp>
-#include <raft/core/device_mdspan.hpp>  // raft::device_matrix_view
-#include <raft/core/operators.hpp>      // raft::identity_op
-#include <raft/core/resources.hpp>      // raft::resources
-#include <raft/util/raft_explicit.hpp>  // RAFT_EXPLICIT
-
-#ifdef RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-namespace cuvs::neighbors::brute_force {
-
-template <typename value_t, typename idx_t>
-inline void knn_merge_parts(
-  raft::resources const& handle,
-  raft::device_matrix_view<const value_t, idx_t, raft::row_major> in_keys,
-  raft::device_matrix_view<const idx_t, idx_t, raft::row_major> in_values,
-  raft::device_matrix_view<value_t, idx_t, raft::row_major> out_keys,
-  raft::device_matrix_view<idx_t, idx_t, raft::row_major> out_values,
-  size_t n_samples,
-  std::optional<raft::device_vector_view<idx_t, idx_t>> translations = std::nullopt) RAFT_EXPLICIT;
-
-template <typename T, typename Accessor>
-index<T> build(
-  raft::resources const& res,
-  raft::mdspan<const T, raft::matrix_extent<int64_t>, raft::row_major, Accessor> dataset,
-  cuvs::distance::DistanceType metric = distance::DistanceType::L2Unexpanded,
-  T metric_arg                        = 0.0) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-void search(raft::resources const& res,
-            const index<T>& idx,
-            raft::device_matrix_view<const T, int64_t, raft::row_major> queries,
-            raft::device_matrix_view<IdxT, int64_t, raft::row_major> neighbors,
-            raft::device_matrix_view<T, int64_t, raft::row_major> distances) RAFT_EXPLICIT;
-
-template <typename idx_t,
-          typename value_t,
-          typename matrix_idx,
-          typename index_layout,
-          typename search_layout,
-          typename epilogue_op = raft::identity_op>
-void knn(raft::resources const& handle,
-         std::vector<raft::device_matrix_view<const value_t, matrix_idx, index_layout>> index,
-         raft::device_matrix_view<const value_t, matrix_idx, search_layout> search,
-         raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,
-         raft::device_matrix_view<value_t, matrix_idx, raft::row_major> distances,
-         distance::DistanceType metric         = distance::DistanceType::L2Unexpanded,
-         std::optional<float> metric_arg       = std::make_optional<float>(2.0f),
-         std::optional<idx_t> global_id_offset = std::nullopt,
-         epilogue_op distance_epilogue         = raft::identity_op()) RAFT_EXPLICIT;
-
-template <typename value_t, typename idx_t, typename idx_layout, typename query_layout>
-void fused_l2_knn(raft::resources const& handle,
-                  raft::device_matrix_view<const value_t, idx_t, idx_layout> index,
-                  raft::device_matrix_view<const value_t, idx_t, query_layout> query,
-                  raft::device_matrix_view<idx_t, idx_t, raft::row_major> out_inds,
-                  raft::device_matrix_view<value_t, idx_t, raft::row_major> out_dists,
-                  cuvs::distance::DistanceType metric) RAFT_EXPLICIT;
-
-}  // namespace cuvs::neighbors::brute_force
-
-#endif  // RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-// No extern template for cuvs::neighbors::brute_force::knn_merge_parts
-
-#define instantiate_raft_neighbors_brute_force_knn(                                         \
-  idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op)                     \
-  extern template void cuvs::neighbors::brute_force::                                       \
-    knn<idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op>(              \
-      raft::resources const& handle,                                                        \
-      std::vector<raft::device_matrix_view<const value_t, matrix_idx, index_layout>> index, \
-      raft::device_matrix_view<const value_t, matrix_idx, search_layout> search,            \
-      raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                 \
-      raft::device_matrix_view<value_t, matrix_idx, raft::row_major> distances,             \
-      cuvs::distance::DistanceType metric,                                                  \
-      std::optional<float> metric_arg,                                                      \
-      std::optional<idx_t> global_id_offset,                                                \
-      epilogue_op distance_epilogue);
-
-instantiate_raft_neighbors_brute_force_knn(
-  int64_t, float, uint32_t, raft::row_major, raft::row_major, raft::identity_op);
-instantiate_raft_neighbors_brute_force_knn(
-  int64_t, float, int64_t, raft::row_major, raft::row_major, raft::identity_op);
-instantiate_raft_neighbors_brute_force_knn(
-  int, float, int, raft::row_major, raft::row_major, raft::identity_op);
-instantiate_raft_neighbors_brute_force_knn(
-  uint32_t, float, uint32_t, raft::row_major, raft::row_major, raft::identity_op);
-
-#undef instantiate_raft_neighbors_brute_force_knn
-
-namespace cuvs::neighbors::brute_force {
-
-extern template void search<float, int>(
-  raft::resources const& res,
-  const cuvs::neighbors::brute_force::index<float>& idx,
-  raft::device_matrix_view<const float, int64_t, raft::row_major> queries,
-  raft::device_matrix_view<int, int64_t, raft::row_major> neighbors,
-  raft::device_matrix_view<float, int64_t, raft::row_major> distances);
-
-extern template void search<float, int64_t>(
-  raft::resources const& res,
-  const cuvs::neighbors::brute_force::index<float>& idx,
-  raft::device_matrix_view<const float, int64_t, raft::row_major> queries,
-  raft::device_matrix_view<int64_t, int64_t, raft::row_major> neighbors,
-  raft::device_matrix_view<float, int64_t, raft::row_major> distances);
-
-extern template cuvs::neighbors::brute_force::index<float> build<float>(
-  raft::resources const& res,
-  raft::device_matrix_view<const float, int64_t, raft::row_major> dataset,
-  cuvs::distance::DistanceType metric,
-  float metric_arg);
-}  // namespace cuvs::neighbors::brute_force
-
-#define instantiate_raft_neighbors_brute_force_fused_l2_knn(             \
-  value_t, idx_t, idx_layout, query_layout)                              \
-  extern template void cuvs::neighbors::brute_force::fused_l2_knn(       \
-    raft::resources const& handle,                                       \
-    raft::device_matrix_view<const value_t, idx_t, idx_layout> index,    \
-    raft::device_matrix_view<const value_t, idx_t, query_layout> query,  \
-    raft::device_matrix_view<idx_t, idx_t, raft::row_major> out_inds,    \
-    raft::device_matrix_view<value_t, idx_t, raft::row_major> out_dists, \
-    cuvs::distance::DistanceType metric);
-
-instantiate_raft_neighbors_brute_force_fused_l2_knn(float,
-                                                    int64_t,
-                                                    raft::row_major,
-                                                    raft::row_major)
-
-#undef instantiate_raft_neighbors_brute_force_fused_l2_knn
diff --git a/cpp/include/cuvs/neighbors/brute_force-inl.cuh b/cpp/include/cuvs/neighbors/brute_force-inl.cuh
deleted file mode 100644
index 3d5c449a9..000000000
--- a/cpp/include/cuvs/neighbors/brute_force-inl.cuh
+++ /dev/null
@@ -1,355 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/brute_force_types.hpp>
-#include <cuvs/neighbors/detail/knn_brute_force.cuh>
-#include <cuvs/spatial/knn/detail/fused_l2_knn.cuh>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-
-namespace cuvs::neighbors::brute_force {
-
-/**
- * @defgroup brute_force_knn Brute-force K-Nearest Neighbors
- * @{
- */
-
-/**
- * @brief Performs a k-select across several (contiguous) row-partitioned index/distance
- * matrices formatted like the following:
- *
- *     part1row1: k0, k1, k2, k3
- *     part1row2: k0, k1, k2, k3
- *     part1row3: k0, k1, k2, k3
- *     part2row1: k0, k1, k2, k3
- *     part2row2: k0, k1, k2, k3
- *     part2row3: k0, k1, k2, k3
- *     etc...
- *
- * The example above shows what an aggregated index/distance matrix
- * would look like with two partitions when n_samples=3 and k=4.
- *
- * When working with extremely large data sets that have been broken
- * over multiple indexes, such as when computing over multiple GPUs,
- * the ids will often start at 0 for each local knn index but the
- * global ids need to be used when merging them together. An optional
- * translations vector can be supplied to map the starting id of
- * each partition to its global id so that the final merged knn
- * is based on the global ids.
- *
- * Usage example:
- * @code{.cpp}
- *  #include <raft/core/resources.hpp>
- *  #include <cuvs/neighbors/brute_force.cuh>
- *  using namespace cuvs::neighbors;
- *
- *  raft::resources handle;
- *  ...
- *  compute multiple knn graphs and aggregate row-wise
- *  (see detailed description above)
- *  ...
- *  brute_force::knn_merge_parts(handle, in_keys, in_values, out_keys, out_values, n_samples);
- * @endcode
- *
- * @tparam idx_t
- * @tparam value_t
- *
- * @param[in] handle
- * @param[in] in_keys matrix of input keys (size n_samples * n_parts * k)
- * @param[in] in_values matrix of input values (size n_samples * n_parts * k)
- * @param[out] out_keys matrix of output keys (size n_samples * k)
- * @param[out] out_values matrix of output values (size n_samples * k)
- * @param[in] n_samples number of rows in each partition
- * @param[in] translations optional vector of starting global id mappings for each local partition
- */
-template <typename value_t, typename idx_t>
-inline void knn_merge_parts(
-  raft::resources const& handle,
-  raft::device_matrix_view<const value_t, idx_t, raft::row_major> in_keys,
-  raft::device_matrix_view<const idx_t, idx_t, raft::row_major> in_values,
-  raft::device_matrix_view<value_t, idx_t, raft::row_major> out_keys,
-  raft::device_matrix_view<idx_t, idx_t, raft::row_major> out_values,
-  size_t n_samples,
-  std::optional<raft::device_vector_view<idx_t, idx_t>> translations = std::nullopt)
-{
-  RAFT_EXPECTS(in_keys.extent(1) == in_values.extent(1) && in_keys.extent(0) == in_values.extent(0),
-               "in_keys and in_values must have the same shape.");
-  RAFT_EXPECTS(
-    out_keys.extent(0) == out_values.extent(0) && out_keys.extent(0) == n_samples,
-    "Number of rows in output keys and val matrices must equal number of rows in search matrix.");
-  RAFT_EXPECTS(
-    out_keys.extent(1) == out_values.extent(1) && out_keys.extent(1) == in_keys.extent(1),
-    "Number of columns in output indices and distances matrices must be equal to k");
-
-  idx_t* translations_ptr = nullptr;
-  if (translations.has_value()) { translations_ptr = translations.value().data_handle(); }
-
-  auto n_parts = in_keys.extent(0) / n_samples;
-  detail::knn_merge_parts(in_keys.data_handle(),
-                          in_values.data_handle(),
-                          out_keys.data_handle(),
-                          out_values.data_handle(),
-                          n_samples,
-                          n_parts,
-                          in_keys.extent(1),
-                          resource::get_cuda_stream(handle),
-                          translations_ptr);
-}
-
-/**
- * @brief Flat C++ API function to perform a brute force knn on
- * a series of input arrays and combine the results into a single
- * output array for indexes and distances. Inputs can be either
- * row- or column-major but the output matrices will always be in
- * row-major format.
- *
- * Usage example:
- * @code{.cpp}
- *  #include <raft/core/resources.hpp>
- *  #include <cuvs/neighbors/brute_force.cuh>
- *  #include <cuvs/distance/distance_types.hpp>
- *  using namespace cuvs::neighbors;
- *
- *  raft::resources handle;
- *  ...
- *  auto metric = cuvs::distance::DistanceType::L2SqrtExpanded;
- *  brute_force::knn(handle, index, search, indices, distances, metric);
- * @endcode
- *
- * @param[in] handle: the cuml handle to use
- * @param[in] index: vector of device matrices (each size m_i*d) to be used as the knn index
- * @param[in] search: matrix (size n*d) to be used for searching the index
- * @param[out] indices: matrix (size n*k) to store output knn indices
- * @param[out] distances: matrix (size n*k) to store the output knn distance
- * @param[in] metric: distance metric to use. Euclidean (L2) is used by default
- * @param[in] metric_arg: the value of `p` for Minkowski (l-p) distances. This
- * 					 is ignored if the metric_type is not Minkowski.
- * @param[in] global_id_offset: optional starting global id mapping for the local partition
- *                              (assumes the index contains contiguous ids in the global id space)
- * @param[in] distance_epilogue: optional epilogue function to run after computing distances. This
-                                 function takes a triple of the (value, rowid, colid) for each
-                                 element in the pairwise distances and returns a transformed value
-                                 back.
- */
-template <typename idx_t,
-          typename value_t,
-          typename matrix_idx,
-          typename index_layout,
-          typename search_layout,
-          typename epilogue_op = raft::identity_op>
-void knn(raft::resources const& handle,
-         std::vector<raft::device_matrix_view<const value_t, matrix_idx, index_layout>> index,
-         raft::device_matrix_view<const value_t, matrix_idx, search_layout> search,
-         raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,
-         raft::device_matrix_view<value_t, matrix_idx, raft::row_major> distances,
-         distance::DistanceType metric         = distance::DistanceType::L2Unexpanded,
-         std::optional<float> metric_arg       = std::make_optional<float>(2.0f),
-         std::optional<idx_t> global_id_offset = std::nullopt,
-         epilogue_op distance_epilogue         = raft::identity_op())
-{
-  RAFT_EXPECTS(index[0].extent(1) == search.extent(1),
-               "Number of dimensions for both index and search matrices must be equal");
-
-  RAFT_EXPECTS(indices.extent(0) == distances.extent(0) && distances.extent(0) == search.extent(0),
-               "Number of rows in output indices and distances matrices must equal number of rows "
-               "in search matrix.");
-  RAFT_EXPECTS(indices.extent(1) == distances.extent(1) && distances.extent(1),
-               "Number of columns in output indices and distances matrices must the same");
-
-  bool rowMajorIndex = std::is_same_v<index_layout, raft::layout_c_contiguous>;
-  bool rowMajorQuery = std::is_same_v<search_layout, raft::layout_c_contiguous>;
-
-  std::vector<value_t*> inputs;
-  std::vector<matrix_idx> sizes;
-  for (std::size_t i = 0; i < index.size(); ++i) {
-    inputs.push_back(const_cast<value_t*>(index[i].data_handle()));
-    sizes.push_back(index[i].extent(0));
-  }
-
-  std::vector<idx_t> trans;
-  if (global_id_offset.has_value()) { trans.push_back(global_id_offset.value()); }
-
-  std::vector<idx_t>* trans_arg = global_id_offset.has_value() ? &trans : nullptr;
-
-  cuvs::neighbors::detail::brute_force_knn_impl(handle,
-                                                inputs,
-                                                sizes,
-                                                index[0].extent(1),
-                                                // TODO: This is unfortunate. Need to fix.
-                                                const_cast<value_t*>(search.data_handle()),
-                                                search.extent(0),
-                                                indices.data_handle(),
-                                                distances.data_handle(),
-                                                indices.extent(1),
-                                                rowMajorIndex,
-                                                rowMajorQuery,
-                                                trans_arg,
-                                                metric,
-                                                metric_arg.value_or(2.0f),
-                                                distance_epilogue);
-}
-
-/**
- * @brief Compute the k-nearest neighbors using L2 expanded/unexpanded distance.
- *
- * This is a specialized function for fusing the k-selection with the distance
- * computation when k < 64. The value of k will be inferred from the number
- * of columns in the output matrices.
- *
- * Usage example:
- * @code{.cpp}
- *  #include <raft/core/resources.hpp>
- *  #include <cuvs/neighbors/brute_force.cuh>
- *  #include <cuvs/distance/distance_types.hpp>
- *  using namespace cuvs::neighbors;
- *
- *  raft::resources handle;
- *  ...
- *  auto metric = cuvs::distance::DistanceType::L2SqrtExpanded;
- *  brute_force::fused_l2_knn(handle, index, search, indices, distances, metric);
- * @endcode
-
- * @tparam value_t type of values
- * @tparam idx_t type of indices
- * @tparam idx_layout layout type of index matrix
- * @tparam query_layout layout type of query matrix
- * @param[in] handle raft handle for sharing expensive resources
- * @param[in] index input index array on device (size m * d)
- * @param[in] query input query array on device (size n * d)
- * @param[out] out_inds output indices array on device (size n * k)
- * @param[out] out_dists output dists array on device (size n * k)
- * @param[in] metric type of distance computation to perform (must be a variant of L2)
- */
-template <typename value_t, typename idx_t, typename idx_layout, typename query_layout>
-void fused_l2_knn(raft::resources const& handle,
-                  raft::device_matrix_view<const value_t, idx_t, idx_layout> index,
-                  raft::device_matrix_view<const value_t, idx_t, query_layout> query,
-                  raft::device_matrix_view<idx_t, idx_t, raft::row_major> out_inds,
-                  raft::device_matrix_view<value_t, idx_t, raft::row_major> out_dists,
-                  cuvs::distance::DistanceType metric)
-{
-  int k = static_cast<int>(out_inds.extent(1));
-
-  RAFT_EXPECTS(k <= 64, "For fused k-selection, k must be < 64");
-  RAFT_EXPECTS(out_inds.extent(1) == out_dists.extent(1), "Value of k must match for outputs");
-  RAFT_EXPECTS(index.extent(1) == query.extent(1),
-               "Number of columns in input matrices must be the same.");
-
-  RAFT_EXPECTS(metric == distance::DistanceType::L2Expanded ||
-                 metric == distance::DistanceType::L2Unexpanded ||
-                 metric == distance::DistanceType::L2SqrtUnexpanded ||
-                 metric == distance::DistanceType::L2SqrtExpanded,
-               "Distance metric must be L2");
-
-  size_t n_index_rows = index.extent(0);
-  size_t n_query_rows = query.extent(0);
-  size_t D            = index.extent(1);
-
-  RAFT_EXPECTS(raft::is_row_or_column_major(index), "Index must be row or column major layout");
-  RAFT_EXPECTS(raft::is_row_or_column_major(query), "Query must be row or column major layout");
-
-  const bool rowMajorIndex = raft::is_row_major(index);
-  const bool rowMajorQuery = raft::is_row_major(query);
-
-  cuvs::spatial::knn::detail::fusedL2Knn(D,
-                                         out_inds.data_handle(),
-                                         out_dists.data_handle(),
-                                         index.data_handle(),
-                                         query.data_handle(),
-                                         n_index_rows,
-                                         n_query_rows,
-                                         k,
-                                         rowMajorIndex,
-                                         rowMajorQuery,
-                                         raft::resource::get_cuda_stream(handle),
-                                         metric);
-}
-
-/**
- * @brief Build the index from the dataset for efficient search.
- *
- * @tparam T data element type
- *
- * @param[in] res
- * @param[in] dataset a matrix view (host or device) to a row-major matrix [n_rows, dim]
- * @param[in] metric: distance metric to use. Euclidean (L2) is used by default
- * @param[in] metric_arg: the value of `p` for Minkowski (l-p) distances. This
- *           is ignored if the metric_type is not Minkowski.
- *
- * @return the constructed brute force index
- */
-template <typename T, typename Accessor>
-index<T> build(
-  raft::resources const& res,
-  raft::mdspan<const T, raft::matrix_extent<int64_t>, raft::row_major, Accessor> dataset,
-  cuvs::distance::DistanceType metric = distance::DistanceType::L2Unexpanded,
-  T metric_arg                        = 0.0)
-{
-  // certain distance metrics can benefit by pre-calculating the norms for the index dataset
-  // which lets us avoid calculating these at query time
-  std::optional<raft::device_vector<T, int64_t>> norms;
-  if (metric == cuvs::distance::DistanceType::L2Expanded ||
-      metric == cuvs::distance::DistanceType::L2SqrtExpanded ||
-      metric == cuvs::distance::DistanceType::CosineExpanded) {
-    norms = raft::make_device_vector<T, int64_t>(res, dataset.extent(0));
-    // cosine needs the l2norm, where as l2 distances needs the squared norm
-    if (metric == cuvs::distance::DistanceType::CosineExpanded) {
-      raft::linalg::norm(res,
-                         dataset,
-                         norms->view(),
-                         raft::linalg::NormType::L2Norm,
-                         raft::linalg::Apply::ALONG_ROWS,
-                         raft::sqrt_op{});
-    } else {
-      raft::linalg::norm(res,
-                         dataset,
-                         norms->view(),
-                         raft::linalg::NormType::L2Norm,
-                         raft::linalg::Apply::ALONG_ROWS);
-    }
-  }
-
-  return index<T>(res, dataset, std::move(norms), metric, metric_arg);
-}
-
-/**
- * @brief Brute Force search using the constructed index.
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- *
- * @param[in] res raft resources
- * @param[in] idx brute force index
- * @param[in] queries a device matrix view to a row-major matrix [n_queries, index->dim()]
- * @param[out] neighbors a device matrix view to the indices of the neighbors in the source dataset
- * [n_queries, k]
- * @param[out] distances a device matrix view to the distances to the selected neighbors [n_queries,
- * k]
- */
-template <typename T, typename IdxT>
-void search(raft::resources const& res,
-            const index<T>& idx,
-            raft::device_matrix_view<const T, int64_t, raft::row_major> queries,
-            raft::device_matrix_view<IdxT, int64_t, raft::row_major> neighbors,
-            raft::device_matrix_view<T, int64_t, raft::row_major> distances)
-{
-  cuvs::neighbors::detail::brute_force_search<T, IdxT>(res, idx, queries, neighbors, distances);
-}
-/** @} */  // end group brute_force_knn
-}  // namespace cuvs::neighbors::brute_force
diff --git a/cpp/include/cuvs/neighbors/brute_force.cuh b/cpp/include/cuvs/neighbors/brute_force.cuh
deleted file mode 100644
index 91065d35f..000000000
--- a/cpp/include/cuvs/neighbors/brute_force.cuh
+++ /dev/null
@@ -1,92 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-#include <memory>
-
-#ifndef RAFT_EXPLICIT_INSTANTIATE_ONLY
-#include "brute_force-inl.cuh"
-#endif
-
-#ifdef RAFT_COMPILED
-#include "brute_force-ext.cuh"
-#endif
-
-#include <cuvs/neighbors/detail/knn_brute_force_batch_k_query.cuh>
-
-namespace cuvs::neighbors::brute_force {
-/**
- * @brief Make a brute force query over batches of k
- *
- * This lets you query for batches of k. For example, you can get
- * the first 100 neighbors, then the next 100 neighbors etc.
- *
- * Example usage:
- * @code{.cpp}
- * #include <cuvs/neighbors/brute_force.cuh>
- * #include <raft/core/device_mdarray.hpp>
- * #include <raft/random/make_blobs.cuh>
-
- * // create a random dataset
- * int n_rows = 10000;
- * int n_cols = 10000;
-
- * raft::device_resources res;
- * auto dataset = raft::make_device_matrix<float, int>(res, n_rows, n_cols);
- * auto labels = raft::make_device_vector<float, int>(res, n_rows);
-
- * raft::make_blobs(res, dataset.view(), labels.view());
- *
- * // create a brute_force knn index from the dataset
- * auto index = cuvs::neighbors::brute_force::build(res,
- *                                                  raft::make_const_mdspan(dataset.view()));
- *
- * // search the index in batches of 128 nearest neighbors
- * auto search = raft::make_const_mdspan(dataset.view());
- * auto query = make_batch_k_query<float, int>(res, index, search, 128);
- * for (auto & batch: *query) {
- *  // batch.indices() and batch.distances() contain the information on the current batch
- * }
- *
- * // we can also support variable sized batches - loaded up a different number
- * // of neighbors at each iteration through the ::advance method
- * int64_t batch_size = 128;
- * query = make_batch_k_query<float, int>(res, index, search, batch_size);
- * for (auto it = query->begin(); it != query->end(); it.advance(batch_size)) {
- *  // batch.indices() and batch.distances() contain the information on the current batch
- *
- *  batch_size += 16; // load up an extra 16 items in the next batch
- * }
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- * @param[in] res
- * @param[in] index The index to query
- * @param[in] query A device matrix view to query for [n_queries, index->dim()]
- * @param[in] batch_size The size of each batch
- */
-
-template <typename T, typename IdxT>
-std::shared_ptr<batch_k_query<T, IdxT>> make_batch_k_query(
-  const raft::resources& res,
-  const cuvs::neighbors::brute_force::index<T>& index,
-  raft::device_matrix_view<const T, int64_t, raft::row_major> query,
-  int64_t batch_size)
-{
-  return std::shared_ptr<batch_k_query<T, IdxT>>(
-    new detail::gpu_batch_k_query<T, IdxT>(res, index, query, batch_size));
-}
-}  // namespace cuvs::neighbors::brute_force
diff --git a/cpp/include/cuvs/neighbors/brute_force_types.hpp b/cpp/include/cuvs/neighbors/brute_force_types.hpp
deleted file mode 100644
index 0d3252d71..000000000
--- a/cpp/include/cuvs/neighbors/brute_force_types.hpp
+++ /dev/null
@@ -1,283 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "ann_types.hpp"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/neighbors_types.hpp>
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/error.hpp>
-#include <raft/core/host_mdarray.hpp>
-#include <raft/core/mdspan_types.hpp>
-#include <raft/core/resources.hpp>
-
-#include <raft/core/logger.hpp>
-
-namespace cuvs::neighbors::brute_force {
-/**
- * @addtogroup brute_force_knn
- * @{
- */
-
-/**
- * @brief Brute Force index.
- *
- * The index stores the dataset and norms for the dataset in device memory.
- *
- * @tparam T data element type
- */
-template <typename T>
-struct index : ann::index {
- public:
-  /** Distance metric used for retrieval */
-  [[nodiscard]] constexpr inline cuvs::distance::DistanceType metric() const noexcept
-  {
-    return metric_;
-  }
-
-  /** Total length of the index (number of vectors). */
-  [[nodiscard]] constexpr inline int64_t size() const noexcept { return dataset_view_.extent(0); }
-
-  /** Dimensionality of the data. */
-  [[nodiscard]] constexpr inline uint32_t dim() const noexcept { return dataset_view_.extent(1); }
-
-  /** Dataset [size, dim] */
-  [[nodiscard]] inline auto dataset() const noexcept
-    -> raft::device_matrix_view<const T, int64_t, raft::row_major>
-  {
-    return dataset_view_;
-  }
-
-  /** Dataset norms */
-  [[nodiscard]] inline auto norms() const
-    -> raft::device_vector_view<const T, int64_t, raft::row_major>
-  {
-    return norms_view_.value();
-  }
-
-  /** Whether or not this index has dataset norms */
-  [[nodiscard]] inline bool has_norms() const noexcept { return norms_view_.has_value(); }
-
-  [[nodiscard]] inline T metric_arg() const noexcept { return metric_arg_; }
-
-  // Don't allow copying the index for performance reasons (try avoiding copying data)
-  index(const index&)                    = delete;
-  index(index&&)                         = default;
-  auto operator=(const index&) -> index& = delete;
-  auto operator=(index&&) -> index&      = default;
-  ~index()                               = default;
-
-  /** Construct a brute force index from dataset
-   *
-   * Constructs a brute force index from a dataset. This lets us precompute norms for
-   * the dataset, providing a speed benefit over doing this at query time.
-
-   * If the dataset is already in GPU memory, then this class stores a non-owning reference to
-   * the dataset. If the dataset is in host memory, it will be copied to the device and the
-   * index will own the device memory.
-   */
-  template <typename data_accessor>
-  index(raft::resources const& res,
-        raft::mdspan<const T, raft::matrix_extent<int64_t>, raft::row_major, data_accessor> dataset,
-        std::optional<raft::device_vector<T, int64_t>>&& norms,
-        cuvs::distance::DistanceType metric,
-        T metric_arg = 0.0)
-    : ann::index(),
-      metric_(metric),
-      dataset_(raft::make_device_matrix<T, int64_t>(res, 0, 0)),
-      norms_(std::move(norms)),
-      metric_arg_(metric_arg)
-  {
-    if (norms_) { norms_view_ = raft::make_const_mdspan(norms_.value().view()); }
-    update_dataset(res, dataset);
-    raft::resource::sync_stream(res);
-  }
-
-  /** Construct a brute force index from dataset
-   *
-   * This class stores a non-owning reference to the dataset and norms here.
-   * Having precomputed norms gives us a performance advantage at query time.
-   */
-  index(raft::resources const& res,
-        raft::device_matrix_view<const T, int64_t, raft::row_major> dataset_view,
-        std::optional<raft::device_vector_view<const T, int64_t>> norms_view,
-        cuvs::distance::DistanceType metric,
-        T metric_arg = 0.0)
-    : ann::index(),
-      metric_(metric),
-      dataset_(raft::make_device_matrix<T, int64_t>(res, 0, 0)),
-      dataset_view_(dataset_view),
-      norms_view_(norms_view),
-      metric_arg_(metric_arg)
-  {
-  }
-
- private:
-  /**
-   * Replace the dataset with a new dataset.
-   */
-  void update_dataset(raft::resources const& res,
-                      raft::device_matrix_view<const T, int64_t, raft::row_major> dataset)
-  {
-    dataset_view_ = dataset;
-  }
-
-  /**
-   * Replace the dataset with a new dataset.
-   *
-   * We create a copy of the dataset on the device. The index manages the lifetime of this copy.
-   */
-  void update_dataset(raft::resources const& res,
-                      raft::host_matrix_view<const T, int64_t, raft::row_major> dataset)
-  {
-    dataset_ = raft::make_device_matrix<T, int64_t>(dataset.extents(0), dataset.extents(1));
-    raft::copy(dataset_.data_handle(),
-               dataset.data_handle(),
-               dataset.size(),
-               resource::get_cuda_stream(res));
-    dataset_view_ = raft::make_const_mdspan(dataset_.view());
-  }
-
-  cuvs::distance::DistanceType metric_;
-  raft::device_matrix<T, int64_t, raft::row_major> dataset_;
-  std::optional<raft::device_vector<T, int64_t>> norms_;
-  std::optional<raft::device_vector_view<const T, int64_t>> norms_view_;
-  raft::device_matrix_view<const T, int64_t, raft::row_major> dataset_view_;
-  T metric_arg_;
-};
-
-/**
- * @brief Interface for performing queries over values of k
- *
- * This interface lets you iterate over batches of k from a brute_force::index.
- * This lets you do things like retrieve the first 100 neighbors for a query,
- * apply post processing to remove any unwanted items and then if needed get the
- * next 100 closest neighbors for the query.
- *
- * This query interface exposes C++ iterators through the ::begin and ::end, and
- * is compatible with range based for loops.
- *
- * Note that this class is an abstract class without any cuda dependencies, meaning
- * that it doesn't require a cuda compiler to use - but also means it can't be directly
- * instantiated.  See the cuvs::neighbors::brute_force::make_batch_k_query
- * function for usage examples.
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- */
-template <typename T, typename IdxT = int64_t>
-class batch_k_query {
- public:
-  batch_k_query(const raft::resources& res,
-                int64_t index_size,
-                int64_t query_size,
-                int64_t batch_size)
-    : res(res), index_size(index_size), query_size(query_size), batch_size(batch_size)
-  {
-  }
-  virtual ~batch_k_query() {}
-
-  using value_type = cuvs::neighbors::batch<T, IdxT>;
-
-  class iterator {
-   public:
-    using value_type = cuvs::neighbors::batch<T, IdxT>;
-    using reference  = const value_type&;
-    using pointer    = const value_type*;
-
-    iterator(const batch_k_query<T, IdxT>* query, int64_t offset = 0)
-      : current(query->res, 0, 0), batches(query->res, 0, 0), query(query), offset(offset)
-    {
-      query->load_batch(offset, query->batch_size, &batches);
-      query->slice_batch(batches, offset, query->batch_size, &current);
-    }
-
-    reference operator*() const { return current; }
-
-    pointer operator->() const { return &current; }
-
-    iterator& operator++()
-    {
-      advance(query->batch_size);
-      return *this;
-    }
-
-    iterator operator++(int)
-    {
-      iterator previous(*this);
-      operator++();
-      return previous;
-    }
-
-    /**
-     * @brief Advance the iterator, using a custom size for the next batch
-     *
-     * Using operator++ means that we will load up the same batch_size for each
-     * batch. This method allows us to get around this restriction, and load up
-     * arbitrary batch sizes on each iteration.
-     * See cuvs::neighbors::brute_force::make_batch_k_query for a usage example.
-     *
-     * @param[in] next_batch_size: size of the next batch to load up
-     */
-    void advance(int64_t next_batch_size)
-    {
-      offset = std::min(offset + current.batch_size(), query->index_size);
-      if (offset + next_batch_size > batches.batch_size()) {
-        query->load_batch(offset, next_batch_size, &batches);
-      }
-      query->slice_batch(batches, offset, next_batch_size, &current);
-    }
-
-    friend bool operator==(const iterator& lhs, const iterator& rhs)
-    {
-      return (lhs.query == rhs.query) && (lhs.offset == rhs.offset);
-    };
-    friend bool operator!=(const iterator& lhs, const iterator& rhs) { return !(lhs == rhs); };
-
-   protected:
-    // the current batch of data
-    value_type current;
-
-    // the currently loaded group of data (containing multiple batches of data that we can iterate
-    // through)
-    value_type batches;
-
-    const batch_k_query<T, IdxT>* query;
-    int64_t offset, current_batch_size;
-  };
-
-  iterator begin() const { return iterator(this); }
-  iterator end() const { return iterator(this, index_size); }
-
- protected:
-  // these two methods need cuda code, and are implemented in the subclass
-  virtual void load_batch(int64_t offset,
-                          int64_t next_batch_size,
-                          batch<T, IdxT>* output) const = 0;
-  virtual void slice_batch(const value_type& input,
-                           int64_t offset,
-                           int64_t batch_size,
-                           value_type* output) const    = 0;
-
-  const raft::resources& res;
-  int64_t index_size, query_size, batch_size;
-};
-/** @} */
-
-}  // namespace cuvs::neighbors::brute_force
diff --git a/cpp/include/cuvs/neighbors/cagra.hpp b/cpp/include/cuvs/neighbors/cagra.hpp
index 8a4a8f017..3a0c60b78 100644
--- a/cpp/include/cuvs/neighbors/cagra.hpp
+++ b/cpp/include/cuvs/neighbors/cagra.hpp
@@ -18,11 +18,11 @@
 
 #include "ann_types.hpp"
 #include <cuvs/distance/distance_types.hpp>
-#include <raft/neighbors/cagra_types.hpp>
 #include <raft/core/device_mdspan.hpp>
 #include <raft/core/host_device_accessor.hpp>
 #include <raft/core/mdspan.hpp>
 #include <raft/core/resources.hpp>
+#include <raft/neighbors/cagra_types.hpp>
 #include <rmm/cuda_stream_view.hpp>
 
 namespace cuvs::neighbors::cagra {
@@ -53,16 +53,17 @@ struct index_params : ann::index_params {
   size_t nn_descent_niter = 20;
 
   /** Build a raft CAGRA index params from an existing cuvs CAGRA index params. */
-  operator raft::neighbors::cagra::index_params() const {
+  operator raft::neighbors::cagra::index_params() const
+  {
     return raft::neighbors::cagra::index_params{
       {
-        .metric = static_cast<raft::distance::DistanceType>((int)this->metric),
-        .metric_arg = this->metric_arg,
+        .metric            = static_cast<raft::distance::DistanceType>((int)this->metric),
+        .metric_arg        = this->metric_arg,
         .add_data_on_build = this->add_data_on_build,
       },
       .intermediate_graph_degree = intermediate_graph_degree,
-      .graph_degree = graph_degree,
-      .build_algo = static_cast<raft::neighbors::cagra::graph_build_algo>((int)build_algo),
+      .graph_degree              = graph_degree,
+      .build_algo       = static_cast<raft::neighbors::cagra::graph_build_algo>((int)build_algo),
       .nn_descent_niter = nn_descent_niter};
   }
 };
@@ -122,7 +123,8 @@ struct search_params : ann::search_params {
   uint64_t rand_xor_mask = 0x128394;
 
   /** Build a raft CAGRA search params from an existing cuvs CAGRA search params. */
-  operator raft::neighbors::cagra::search_params() const {
+  operator raft::neighbors::cagra::search_params() const
+  {
     raft::neighbors::cagra::search_params result = {
       {},
       max_queries,
@@ -156,7 +158,6 @@ static_assert(std::is_aggregate_v<search_params>);
  */
 template <typename T, typename IdxT>
 struct index : ann::index {
-
   /** Build a cuvs CAGRA index from an existing RAFT CAGRA index. */
   index(raft::neighbors::cagra::index<T, IdxT>&& raft_idx)
     : ann::index(),
@@ -174,10 +175,7 @@ struct index : ann::index {
   }
 
   /** Total length of the index (number of vectors). */
-  [[nodiscard]] constexpr inline auto size() const noexcept -> IdxT
-  {
-    return raft_index_->size();
-  }
+  [[nodiscard]] constexpr inline auto size() const noexcept -> IdxT { return raft_index_->size(); }
 
   /** Dimensionality of the data. */
   [[nodiscard]] constexpr inline auto dim() const noexcept -> uint32_t
@@ -215,7 +213,8 @@ struct index : ann::index {
   index(raft::resources const& res,
         cuvs::distance::DistanceType metric = cuvs::distance::DistanceType::L2Expanded)
     : ann::index(),
-      raft_index_(std::make_unique<raft::neighbors::cagra::index<T, IdxT>>(res, static_cast<raft::distance::DistanceType>((int)metric)))
+      raft_index_(std::make_unique<raft::neighbors::cagra::index<T, IdxT>>(
+        res, static_cast<raft::distance::DistanceType>((int)metric)))
   {
   }
   /** Construct an index from dataset and knn_graph arrays
@@ -339,57 +338,55 @@ struct index : ann::index {
   {
     return raft_index_.get();
   }
-  auto get_raft_index() -> raft::neighbors::cagra::index<T, IdxT>*
-  {
-    return raft_index_.get();
-  }
+  auto get_raft_index() -> raft::neighbors::cagra::index<T, IdxT>* { return raft_index_.get(); }
+
  private:
   std::unique_ptr<raft::neighbors::cagra::index<T, IdxT>> raft_index_;
 };
 
 // Using device and host_matrix_view avoids needing to typedef multiple mdspans based on accessors
-#define CUVS_INST_CAGRA_FUNCS(T, IdxT)                                             \
-  auto build(raft::resources const& handle,                                        \
-             const cuvs::neighbors::cagra::index_params& params,                   \
+#define CUVS_INST_CAGRA_FUNCS(T, IdxT)                                                   \
+  auto build(raft::resources const& handle,                                              \
+             const cuvs::neighbors::cagra::index_params& params,                         \
              raft::device_matrix_view<const T, int64_t, raft::row_major> dataset)        \
-    ->cuvs::neighbors::cagra::index<T, IdxT>;                                      \
-                                                                                   \
-  auto build(raft::resources const& handle,                                        \
-             const cuvs::neighbors::cagra::index_params& params,                   \
+    ->cuvs::neighbors::cagra::index<T, IdxT>;                                            \
+                                                                                         \
+  auto build(raft::resources const& handle,                                              \
+             const cuvs::neighbors::cagra::index_params& params,                         \
              raft::host_matrix_view<const T, int64_t, raft::row_major> dataset)          \
-    ->cuvs::neighbors::cagra::index<T, IdxT>;                                      \
-                                                                                   \
-  void build_device(raft::resources const& handle,                                 \
-                    const cuvs::neighbors::cagra::index_params& params,            \
+    ->cuvs::neighbors::cagra::index<T, IdxT>;                                            \
+                                                                                         \
+  void build_device(raft::resources const& handle,                                       \
+                    const cuvs::neighbors::cagra::index_params& params,                  \
                     raft::device_matrix_view<const T, int64_t, raft::row_major> dataset, \
-                    cuvs::neighbors::cagra::index<T, IdxT>& idx);                  \
-                                                                                   \
-  void build_host(raft::resources const& handle,                                   \
-                  const cuvs::neighbors::cagra::index_params& params,              \
+                    cuvs::neighbors::cagra::index<T, IdxT>& idx);                        \
+                                                                                         \
+  void build_host(raft::resources const& handle,                                         \
+                  const cuvs::neighbors::cagra::index_params& params,                    \
                   raft::host_matrix_view<const T, int64_t, raft::row_major> dataset,     \
-                  cuvs::neighbors::cagra::index<T, IdxT>& idx);                    \
-                                                                                   \
-  void search(raft::resources const& handle,                                       \
-              cuvs::neighbors::cagra::search_params const& params,                 \
-              const cuvs::neighbors::cagra::index<T, IdxT>& index,                 \
+                  cuvs::neighbors::cagra::index<T, IdxT>& idx);                          \
+                                                                                         \
+  void search(raft::resources const& handle,                                             \
+              cuvs::neighbors::cagra::search_params const& params,                       \
+              const cuvs::neighbors::cagra::index<T, IdxT>& index,                       \
               raft::device_matrix_view<const T, int64_t, raft::row_major> queries,       \
               raft::device_matrix_view<IdxT, int64_t, raft::row_major> neighbors,        \
               raft::device_matrix_view<float, int64_t, raft::row_major> distances);      \
-  void serialize_file(raft::resources const& handle,                               \
-                      const std::string& filename,                                 \
-                      const cuvs::neighbors::cagra::index<T, IdxT>& index,         \
-                      bool include_dataset = true);                                \
-                                                                                   \
-  void deserialize_file(raft::resources const& handle,                             \
-                        const std::string& filename,                               \
-                        cuvs::neighbors::cagra::index<T, IdxT>* index);            \
-  void serialize(raft::resources const& handle,                                    \
-                 std::string& str,                                                 \
-                 const cuvs::neighbors::cagra::index<T, IdxT>& index,              \
-                 bool include_dataset = true);                                     \
-                                                                                   \
-  void deserialize(raft::resources const& handle,                                  \
-                   const std::string& str,                                         \
+  void serialize_file(raft::resources const& handle,                                     \
+                      const std::string& filename,                                       \
+                      const cuvs::neighbors::cagra::index<T, IdxT>& index,               \
+                      bool include_dataset = true);                                      \
+                                                                                         \
+  void deserialize_file(raft::resources const& handle,                                   \
+                        const std::string& filename,                                     \
+                        cuvs::neighbors::cagra::index<T, IdxT>* index);                  \
+  void serialize(raft::resources const& handle,                                          \
+                 std::string& str,                                                       \
+                 const cuvs::neighbors::cagra::index<T, IdxT>& index,                    \
+                 bool include_dataset = true);                                           \
+                                                                                         \
+  void deserialize(raft::resources const& handle,                                        \
+                   const std::string& str,                                               \
                    cuvs::neighbors::cagra::index<T, IdxT>* index);
 
 CUVS_INST_CAGRA_FUNCS(float, uint32_t);
@@ -398,12 +395,12 @@ CUVS_INST_CAGRA_FUNCS(uint8_t, uint32_t);
 
 #undef CUVS_INST_CAGRA_FUNCS
 
-#define CUVS_INST_CAGRA_OPTIMIZE(IdxT)                                               \
-  void optimize_device(raft::resources const& res,                                   \
+#define CUVS_INST_CAGRA_OPTIMIZE(IdxT)                                                     \
+  void optimize_device(raft::resources const& res,                                         \
                        raft::device_matrix_view<IdxT, int64_t, raft::row_major> knn_graph, \
                        raft::host_matrix_view<IdxT, int64_t, raft::row_major> new_graph);  \
-                                                                                     \
-  void optimize_host(raft::resources const& res,                                     \
+                                                                                           \
+  void optimize_host(raft::resources const& res,                                           \
                      raft::host_matrix_view<IdxT, int64_t, raft::row_major> knn_graph,     \
                      raft::host_matrix_view<IdxT, int64_t, raft::row_major> new_graph);
 
diff --git a/cpp/include/cuvs/neighbors/cagra_types.hpp b/cpp/include/cuvs/neighbors/cagra_types.hpp
deleted file mode 100644
index 546279de1..000000000
--- a/cpp/include/cuvs/neighbors/cagra_types.hpp
+++ /dev/null
@@ -1,363 +0,0 @@
-/*
- * Copyright (c) 2023-2024, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "ann_types.hpp"
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/neighbors/cagra_types.hpp>
-
-#include <cuvs/distance/distance_types.hpp>
-//#include <cuvs/neighbors/detail/cagra/utils.hpp>
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/error.hpp>
-#include <raft/core/host_mdarray.hpp>
-#include <raft/core/mdspan_types.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/util/integer_utils.hpp>
-
-#include <memory>
-#include <optional>
-#include <string>
-#include <thrust/fill.h>
-#include <type_traits>
-
-#include <raft/core/logger.hpp>
-namespace cuvs::neighbors::cagra {
-/**
- * @addtogroup cagra
- * @{
- */
-
-/**
- * @brief ANN algorithm used by CAGRA to build knn graph
- *
- */
-enum class graph_build_algo {
-  /* Use IVF-PQ to build all-neighbors knn graph */
-  IVF_PQ,
-  /* Experimental, use NN-Descent to build all-neighbors knn graph */
-  NN_DESCENT
-};
-
-struct index_params : ann::index_params {
-  /** Degree of input graph for pruning. */
-  size_t intermediate_graph_degree = 128;
-  /** Degree of output graph. */
-  size_t graph_degree = 64;
-  /** ANN algorithm to build knn graph. */
-  graph_build_algo build_algo = graph_build_algo::IVF_PQ;
-  /** Number of Iterations to run if building with NN_DESCENT */
-  size_t nn_descent_niter = 20;
-
-  /** Build a raft CAGRA index params from an existing cuvs CAGRA index params. */
-  operator raft::neighbors::cagra::index_params() const {
-    return raft::neighbors::cagra::index_params{
-      {
-        .metric = static_cast<raft::distance::DistanceType>((int)this->metric),
-        .metric_arg = this->metric_arg,
-        .add_data_on_build = this->add_data_on_build,
-      },
-      .intermediate_graph_degree = intermediate_graph_degree,
-      .graph_degree = graph_degree,
-      .build_algo = static_cast<raft::neighbors::cagra::graph_build_algo>((int)build_algo),
-      .nn_descent_niter = nn_descent_niter};
-  }
-};
-
-enum class search_algo {
-  /** For large batch sizes. */
-  SINGLE_CTA,
-  /** For small batch sizes. */
-  MULTI_CTA,
-  MULTI_KERNEL,
-  AUTO
-};
-
-enum class hash_mode { HASH, SMALL, AUTO };
-
-struct search_params : ann::search_params {
-  /** Maximum number of queries to search at the same time (batch size). Auto select when 0.*/
-  size_t max_queries = 0;
-
-  /** Number of intermediate search results retained during the search.
-   *
-   *  This is the main knob to adjust trade off between accuracy and search speed.
-   *  Higher values improve the search accuracy.
-   */
-  size_t itopk_size = 64;
-
-  /** Upper limit of search iterations. Auto select when 0.*/
-  size_t max_iterations = 0;
-
-  // In the following we list additional search parameters for fine tuning.
-  // Reasonable default values are automatically chosen.
-
-  /** Which search implementation to use. */
-  search_algo algo = search_algo::AUTO;
-
-  /** Number of threads used to calculate a single distance. 4, 8, 16, or 32. */
-  size_t team_size = 0;
-
-  /** Number of graph nodes to select as the starting point for the search in each iteration. aka
-   * search width?*/
-  size_t search_width = 1;
-  /** Lower limit of search iterations. */
-  size_t min_iterations = 0;
-
-  /** Thread block size. 0, 64, 128, 256, 512, 1024. Auto selection when 0. */
-  size_t thread_block_size = 0;
-  /** Hashmap type. Auto selection when AUTO. */
-  hash_mode hashmap_mode = hash_mode::AUTO;
-  /** Lower limit of hashmap bit length. More than 8. */
-  size_t hashmap_min_bitlen = 0;
-  /** Upper limit of hashmap fill rate. More than 0.1, less than 0.9.*/
-  float hashmap_max_fill_rate = 0.5;
-
-  /** Number of iterations of initial random seed node selection. 1 or more. */
-  uint32_t num_random_samplings = 1;
-  /** Bit mask used for initial random seed node selection. */
-  uint64_t rand_xor_mask = 0x128394;
-
-  /** Build a raft CAGRA search params from an existing cuvs CAGRA search params. */
-  operator raft::neighbors::cagra::search_params() const {
-    raft::neighbors::cagra::search_params result = {
-      {},
-      max_queries,
-      itopk_size,
-      max_iterations,
-      static_cast<raft::neighbors::cagra::search_algo>((int)algo),
-      team_size,
-      search_width,
-      min_iterations,
-      thread_block_size,
-      static_cast<raft::neighbors::cagra::hash_mode>((int)hashmap_mode),
-      hashmap_min_bitlen,
-      hashmap_max_fill_rate,
-      num_random_samplings,
-      rand_xor_mask};
-    return result;
-  }
-};
-
-static_assert(std::is_aggregate_v<index_params>);
-static_assert(std::is_aggregate_v<search_params>);
-
-/**
- * @brief CAGRA index.
- *
- * The index stores the dataset and a kNN graph in device memory.
- *
- * @tparam T data element type
- * @tparam IdxT type of the vector indices (represent dataset.extent(0))
- *
- */
-template <typename T, typename IdxT>
-struct index : ann::index {
-
-  /** Build a cuvs CAGRA index from an existing RAFT CAGRA index. */
-  index(raft::neighbors::cagra::index<T, IdxT>&& raft_idx)
-    : ann::index(),
-      raft_index_{std::make_unique<raft::neighbors::cagra::index<T, IdxT>>(std::move(raft_idx))}
-  {
-  }
-  static_assert(!raft::is_narrowing_v<uint32_t, IdxT>,
-                "IdxT must be able to represent all values of uint32_t");
-
- public:
-  /** Distance metric used for clustering. */
-  [[nodiscard]] constexpr inline auto metric() const noexcept -> cuvs::distance::DistanceType
-  {
-    return static_cast<cuvs::distance::DistanceType>((int)raft_index_->metric());
-  }
-
-  /** Total length of the index (number of vectors). */
-  [[nodiscard]] constexpr inline auto size() const noexcept -> IdxT
-  {
-    return raft_index_->size();
-  }
-
-  /** Dimensionality of the data. */
-  [[nodiscard]] constexpr inline auto dim() const noexcept -> uint32_t
-  {
-    return raft_index_->dim();
-  }
-  /** Graph degree */
-  [[nodiscard]] constexpr inline auto graph_degree() const noexcept -> uint32_t
-  {
-    return raft_index_->graph_degree();
-  }
-
-  /** Dataset [size, dim] */
-  [[nodiscard]] inline auto dataset() const noexcept
-    -> raft::device_matrix_view<const T, int64_t, raft::layout_stride>
-  {
-    return raft_index_->dataset();
-  }
-
-  /** neighborhood graph [size, graph-degree] */
-  [[nodiscard]] inline auto graph() const noexcept
-    -> raft::device_matrix_view<const IdxT, int64_t, raft::row_major>
-  {
-    return raft_index_->graph();
-  }
-
-  // Don't allow copying the index for performance reasons (try avoiding copying data)
-  index(const index&)                    = delete;
-  index(index&&)                         = default;
-  auto operator=(const index&) -> index& = delete;
-  auto operator=(index&&) -> index&      = default;
-  ~index()                               = default;
-
-  /** Construct an empty index. */
-  index(raft::resources const& res,
-        cuvs::distance::DistanceType metric = cuvs::distance::DistanceType::L2Expanded)
-    : ann::index(),
-      raft_index_(std::make_unique<raft::neighbors::cagra::index<T, IdxT>>(res, static_cast<raft::distance::DistanceType>((int)metric)))
-  {
-  }
-  /** Construct an index from dataset and knn_graph arrays
-   *
-   * If the dataset and graph is already in GPU memory, then the index is just a thin wrapper around
-   * these that stores a non-owning a reference to the arrays.
-   *
-   * The constructor also accepts host arrays. In that case they are copied to the device, and the
-   * device arrays will be owned by the index.
-   *
-   * In case the dasates rows are not 16 bytes aligned, then we create a padded copy in device
-   * memory to ensure alignment for vectorized load.
-   *
-   * Usage examples:
-   *
-   * - Cagra index is normally created by the cagra::build
-   * @code{.cpp}
-   *   using namespace cuvs::neighbors::experimental;
-   *   auto dataset = raft::make_host_matrix<float, int64_t>(n_rows, n_cols);
-   *   load_dataset(dataset.view());
-   *   // use default index parameters
-   *   cagra::index_params index_params;
-   *   // create and fill the index from a [N, D] dataset
-   *   auto index = cagra::build(res, index_params, dataset);
-   *   // use default search parameters
-   *   cagra::search_params search_params;
-   *   // search K nearest neighbours
-   *   auto neighbors = raft::make_device_matrix<uint32_t, int64_t>(res, n_queries, k);
-   *   auto distances = raft::make_device_matrix<float, int64_t>(res, n_queries, k);
-   *   cagra::search(res, search_params, index, queries, neighbors, distances);
-   * @endcode
-   *   In the above example, we have passed a host dataset to build. The returned index will own a
-   * device copy of the dataset and the knn_graph. In contrast, if we pass the dataset as a
-   * raft::device_mdspan to build, then it will only store a reference to it.
-   *
-   * - Constructing index using existing knn-graph
-   * @code{.cpp}
-   *   using namespace cuvs::neighbors::experimental;
-   *
-   *   auto dataset = raft::make_device_matrix<float, int64_t>(res, n_rows, n_cols);
-   *   auto knn_graph = raft::make_device_matrix<uint32_n, int64_t>(res, n_rows, graph_degree);
-   *
-   *   // custom loading and graph creation
-   *   // load_dataset(dataset.view());
-   *   // create_knn_graph(knn_graph.view());
-   *
-   *   // Wrap the existing device arrays into an index structure
-   *   cagra::index<T, IdxT> index(res, metric, raft::make_const_mdspan(dataset.view()),
-   *                               raft::make_const_mdspan(knn_graph.view()));
-   *
-   *   // Both knn_graph and dataset objects have to be in scope while the index is used because
-   *   // the index only stores a reference to these.
-   *   cagra::search(res, search_params, index, queries, neighbors, distances);
-   * @endcode
-   *
-   */
-  template <typename data_accessor, typename graph_accessor>
-  index(raft::resources const& res,
-        cuvs::distance::DistanceType metric,
-        raft::mdspan<const T, raft::matrix_extent<int64_t>, raft::row_major, data_accessor> dataset,
-        raft::mdspan<const IdxT, raft::matrix_extent<int64_t>, raft::row_major, graph_accessor>
-          knn_graph)
-    : ann::index(),
-      raft_index_(std::make_unique<raft::neighbors::cagra::index<T, IdxT>>(
-        res, static_cast<raft::distance::DistanceType>((int)metric), dataset, knn_graph))
-  {
-    RAFT_EXPECTS(dataset.extent(0) == knn_graph.extent(0),
-                 "Dataset and knn_graph must have equal number of rows");
-    update_dataset(res, dataset);
-    update_graph(res, knn_graph);
-    raft::resource::sync_stream(res);
-  }
-
-  /**
-   * Replace the dataset with a new dataset.
-   *
-   * If the new dataset rows are aligned on 16 bytes, then only a reference is stored to the
-   * dataset. It is the caller's responsibility to ensure that dataset stays alive as long as the
-   * index.
-   */
-  void update_dataset(raft::resources const& res,
-                      raft::device_matrix_view<const T, int64_t, raft::row_major> dataset)
-  {
-    raft_index_->update_dataset(res, dataset);
-  }
-  /**
-   * Replace the dataset with a new dataset.
-   *
-   * We create a copy of the dataset on the device. The index manages the lifetime of this copy.
-   */
-  void update_dataset(raft::resources const& res,
-                      raft::host_matrix_view<const T, int64_t, raft::row_major> dataset)
-  {
-    raft_index_->update_dataset(res, dataset);
-  }
-
-  /**
-   * Replace the graph with a new graph.
-   *
-   * Since the new graph is a device array, we store a reference to that, and it is
-   * the caller's responsibility to ensure that knn_graph stays alive as long as the index.
-   */
-  void update_graph(raft::resources const& res,
-                    raft::device_matrix_view<const IdxT, int64_t, raft::row_major> knn_graph)
-  {
-    raft_index_->update_graph(res, knn_graph);
-  }
-
-  /**
-   * Replace the graph with a new graph.
-   *
-   * We create a copy of the graph on the device. The index manages the lifetime of this copy.
-   */
-  void update_graph(raft::resources const& res,
-                    raft::host_matrix_view<const IdxT, int64_t, raft::row_major> knn_graph)
-  {
-    raft_index_->update_graph(res, knn_graph);
-  }
-
-  auto get_raft_index() const -> const raft::neighbors::cagra::index<T, IdxT>*
-  {
-    return raft_index_.get();
-  }
-  auto get_raft_index() -> raft::neighbors::cagra::index<T, IdxT>*
-  {
-    return raft_index_.get();
-  }
- private:
-  std::unique_ptr<raft::neighbors::cagra::index<T, IdxT>> raft_index_;
-};
-
-/** @} */
-
-}  // namespace cuvs::neighbors::cagra
diff --git a/cpp/include/cuvs/neighbors/epsilon_neighborhood.cuh b/cpp/include/cuvs/neighbors/epsilon_neighborhood.cuh
deleted file mode 100644
index dfa300c22..000000000
--- a/cpp/include/cuvs/neighbors/epsilon_neighborhood.cuh
+++ /dev/null
@@ -1,123 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#ifndef __EPSILON_NEIGH_H
-#define __EPSILON_NEIGH_H
-
-#pragma once
-
-#include <cuvs/spatial/knn/detail/epsilon_neighborhood.cuh>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::neighbors::epsilon_neighborhood {
-
-/**
- * @brief Computes epsilon neighborhood for the L2-Squared distance metric
- *
- * @tparam value_t   IO and math type
- * @tparam idx_t    Index type
- *
- * @param[out] adj    adjacency matrix [row-major] [on device] [dim = m x n]
- * @param[out] vd     vertex degree array [on device] [len = m + 1]
- *                    `vd + m` stores the total number of edges in the adjacency
- *                    matrix. Pass a nullptr if you don't need this info.
- * @param[in]  x      first matrix [row-major] [on device] [dim = m x k]
- * @param[in]  y      second matrix [row-major] [on device] [dim = n x k]
- * @param[in]  m      number of rows in x
- * @param[in]  n      number of rows in y
- * @param[in]  k      number of columns in x and k
- * @param[in]  eps    defines epsilon neighborhood radius (should be passed as
- *                    squared as we compute L2-squared distance in this method)
- * @param[in]  stream cuda stream
- */
-template <typename value_t, typename idx_t>
-void epsUnexpL2SqNeighborhood(bool* adj,
-                              idx_t* vd,
-                              const value_t* x,
-                              const value_t* y,
-                              idx_t m,
-                              idx_t n,
-                              idx_t k,
-                              value_t eps,
-                              cudaStream_t stream)
-{
-  spatial::knn::detail::epsUnexpL2SqNeighborhood<value_t, idx_t>(
-    adj, vd, x, y, m, n, k, eps, stream);
-}
-
-/**
- * @defgroup epsilon_neighbors Epislon Neighborhood Operations
- * @{
- */
-
-/**
- * @brief Computes epsilon neighborhood for the L2-Squared distance metric and given ball size.
- * The epsilon neighbors is represented by a dense boolean adjacency matrix of size m * n and
- * an array of degrees for each vertex, which can be used as a compressed sparse row (CSR)
- * indptr array.
- *
- * @code{.cpp}
- *  #include <cuvs/neighbors/epsilon_neighborhood.cuh>
- *  #include <raft/core/resources.hpp>
- *  #include <raft/core/device_mdarray.hpp>
- *  using namespace cuvs::neighbors;
- *  raft::raft::resources handle;
- *  ...
- *  auto adj = raft::make_device_matrix<bool>(handle, m * n);
- *  auto vd = raft::make_device_vector<int>(handle, m+1);
- *  epsilon_neighborhood::eps_neighbors_l2sq(handle, x, y, adj.view(), vd.view(), eps);
- * @endcode
- *
- * @tparam value_t   IO and math type
- * @tparam idx_t    Index type
- * @tparam matrix_idx_t matrix indexing type
- *
- * @param[in]  handle raft handle to manage library resources
- * @param[in]  x      first matrix [row-major] [on device] [dim = m x k]
- * @param[in]  y      second matrix [row-major] [on device] [dim = n x k]
- * @param[out] adj    adjacency matrix [row-major] [on device] [dim = m x n]
- * @param[out] vd     vertex degree array [on device] [len = m + 1]
- *                    `vd + m` stores the total number of edges in the adjacency
- *                    matrix. Pass a nullptr if you don't need this info.
- * @param[in]  eps    defines epsilon neighborhood radius (should be passed as
- *                    squared as we compute L2-squared distance in this method)
- */
-template <typename value_t, typename idx_t, typename matrix_idx_t>
-void eps_neighbors_l2sq(raft::resources const& handle,
-                        raft::device_matrix_view<const value_t, matrix_idx_t, raft::row_major> x,
-                        raft::device_matrix_view<const value_t, matrix_idx_t, raft::row_major> y,
-                        raft::device_matrix_view<bool, matrix_idx_t, raft::row_major> adj,
-                        raft::device_vector_view<idx_t, matrix_idx_t> vd,
-                        value_t eps)
-{
-  epsUnexpL2SqNeighborhood<value_t, idx_t>(adj.data_handle(),
-                                           vd.data_handle(),
-                                           x.data_handle(),
-                                           y.data_handle(),
-                                           x.extent(0),
-                                           y.extent(0),
-                                           x.extent(1),
-                                           eps,
-                                           resource::get_cuda_stream(handle));
-}
-
-/** @} */  // end group epsilon_neighbors
-
-}  // namespace cuvs::neighbors::epsilon_neighborhood
-
-#endif
\ No newline at end of file
diff --git a/cpp/include/cuvs/neighbors/ivf_flat-ext.cuh b/cpp/include/cuvs/neighbors/ivf_flat-ext.cuh
deleted file mode 100644
index 3b66a589b..000000000
--- a/cpp/include/cuvs/neighbors/ivf_flat-ext.cuh
+++ /dev/null
@@ -1,206 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cstdint>  // int64_t
-
-#include <cuvs/neighbors/ivf_flat_serialize.cuh>
-#include <cuvs/neighbors/ivf_flat_types.hpp>      // cuvs::neighbors::ivf_flat::index
-#include <raft/core/device_mdspan.hpp>            // raft::device_matrix_view
-#include <raft/core/resources.hpp>                // raft::resources
-#include <raft/util/raft_explicit.hpp>            // RAFT_EXPLICIT
-#include <rmm/mr/device/per_device_resource.hpp>  // rmm::mr::device_memory_resource
-
-#ifdef RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-namespace cuvs::neighbors::ivf_flat {
-
-template <typename T, typename IdxT>
-auto build(raft::resources const& handle,
-           const index_params& params,
-           const T* dataset,
-           IdxT n_rows,
-           uint32_t dim) -> index<T, IdxT> RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-auto build(raft::resources const& handle,
-           const index_params& params,
-           raft::device_matrix_view<const T, IdxT, raft::row_major> dataset)
-  -> index<T, IdxT> RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-void build(raft::resources const& handle,
-           const index_params& params,
-           raft::device_matrix_view<const T, IdxT, raft::row_major> dataset,
-           cuvs::neighbors::ivf_flat::index<T, IdxT>& idx) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-auto extend(raft::resources const& handle,
-            const index<T, IdxT>& orig_index,
-            const T* new_vectors,
-            const IdxT* new_indices,
-            IdxT n_rows) -> index<T, IdxT> RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-auto extend(raft::resources const& handle,
-            raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,
-            std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices,
-            const index<T, IdxT>& orig_index) -> index<T, IdxT> RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-void extend(raft::resources const& handle,
-            index<T, IdxT>* index,
-            const T* new_vectors,
-            const IdxT* new_indices,
-            IdxT n_rows) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-void extend(raft::resources const& handle,
-            raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,
-            std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices,
-            index<T, IdxT>* index) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT, typename IvfSampleFilterT>
-void search_with_filtering(raft::resources const& handle,
-                           const search_params& params,
-                           const index<T, IdxT>& index,
-                           const T* queries,
-                           uint32_t n_queries,
-                           uint32_t k,
-                           IdxT* neighbors,
-                           float* distances,
-                           rmm::mr::device_memory_resource* mr = nullptr,
-                           IvfSampleFilterT sample_filter      = IvfSampleFilterT()) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-void search(raft::resources const& handle,
-            const search_params& params,
-            const index<T, IdxT>& index,
-            const T* queries,
-            uint32_t n_queries,
-            uint32_t k,
-            IdxT* neighbors,
-            float* distances,
-            rmm::mr::device_memory_resource* mr = nullptr) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT, typename IvfSampleFilterT>
-void search_with_filtering(raft::resources const& handle,
-                           const search_params& params,
-                           const index<T, IdxT>& index,
-                           raft::device_matrix_view<const T, IdxT, raft::row_major> queries,
-                           raft::device_matrix_view<IdxT, IdxT, raft::row_major> neighbors,
-                           raft::device_matrix_view<float, IdxT, raft::row_major> distances,
-                           IvfSampleFilterT sample_filter = IvfSampleFilterT()) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-void search(raft::resources const& handle,
-            const search_params& params,
-            const index<T, IdxT>& index,
-            raft::device_matrix_view<const T, IdxT, raft::row_major> queries,
-            raft::device_matrix_view<IdxT, IdxT, raft::row_major> neighbors,
-            raft::device_matrix_view<float, IdxT, raft::row_major> distances) RAFT_EXPLICIT;
-
-}  // namespace cuvs::neighbors::ivf_flat
-
-#endif  // RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-#define instantiate_raft_neighbors_ivf_flat_build(T, IdxT)            \
-  extern template auto cuvs::neighbors::ivf_flat::build<T, IdxT>(     \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    const T* dataset,                                                 \
-    IdxT n_rows,                                                      \
-    uint32_t dim)                                                     \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                      \
-                                                                      \
-  extern template auto cuvs::neighbors::ivf_flat::build<T, IdxT>(     \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset) \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                      \
-                                                                      \
-  extern template void cuvs::neighbors::ivf_flat::build<T, IdxT>(     \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset, \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>& idx);
-
-instantiate_raft_neighbors_ivf_flat_build(float, int64_t);
-instantiate_raft_neighbors_ivf_flat_build(int8_t, int64_t);
-instantiate_raft_neighbors_ivf_flat_build(uint8_t, int64_t);
-#undef instantiate_raft_neighbors_ivf_flat_build
-
-#define instantiate_raft_neighbors_ivf_flat_extend(T, IdxT)                \
-  extern template auto cuvs::neighbors::ivf_flat::extend<T, IdxT>(         \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index,           \
-    const T* new_vectors,                                                  \
-    const IdxT* new_indices,                                               \
-    IdxT n_rows)                                                           \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                           \
-                                                                           \
-  extern template auto cuvs::neighbors::ivf_flat::extend<T, IdxT>(         \
-    raft::resources const& handle,                                         \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,  \
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index)           \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                           \
-                                                                           \
-  extern template void cuvs::neighbors::ivf_flat::extend<T, IdxT>(         \
-    raft::resources const& handle,                                         \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>* index,                      \
-    const T* new_vectors,                                                  \
-    const IdxT* new_indices,                                               \
-    IdxT n_rows);                                                          \
-                                                                           \
-  extern template void cuvs::neighbors::ivf_flat::extend<T, IdxT>(         \
-    raft::resources const& handle,                                         \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,  \
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>* index);
-
-instantiate_raft_neighbors_ivf_flat_extend(float, int64_t);
-instantiate_raft_neighbors_ivf_flat_extend(int8_t, int64_t);
-instantiate_raft_neighbors_ivf_flat_extend(uint8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_flat_extend
-
-#define instantiate_raft_neighbors_ivf_flat_search(T, IdxT)           \
-  extern template void cuvs::neighbors::ivf_flat::search<T, IdxT>(    \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::search_params& params,           \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,           \
-    const T* queries,                                                 \
-    uint32_t n_queries,                                               \
-    uint32_t k,                                                       \
-    IdxT* neighbors,                                                  \
-    float* distances,                                                 \
-    rmm::mr::device_memory_resource* mr);                             \
-                                                                      \
-  extern template void cuvs::neighbors::ivf_flat::search<T, IdxT>(    \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::search_params& params,           \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,           \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> queries, \
-    raft::device_matrix_view<IdxT, IdxT, raft::row_major> neighbors,  \
-    raft::device_matrix_view<float, IdxT, raft::row_major> distances);
-
-instantiate_raft_neighbors_ivf_flat_search(float, int64_t);
-instantiate_raft_neighbors_ivf_flat_search(int8_t, int64_t);
-instantiate_raft_neighbors_ivf_flat_search(uint8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_flat_search
diff --git a/cpp/include/cuvs/neighbors/ivf_flat-inl.cuh b/cpp/include/cuvs/neighbors/ivf_flat-inl.cuh
deleted file mode 100644
index d956f060c..000000000
--- a/cpp/include/cuvs/neighbors/ivf_flat-inl.cuh
+++ /dev/null
@@ -1,602 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/neighbors/detail/ivf_flat_build.cuh>
-#include <cuvs/neighbors/detail/ivf_flat_search.cuh>
-#include <cuvs/neighbors/ivf_flat_serialize.cuh>
-#include <cuvs/neighbors/ivf_flat_types.hpp>
-
-#include <raft/core/resources.hpp>
-
-#include <raft/core/device_mdspan.hpp>
-#include <rmm/cuda_stream_view.hpp>
-#include <rmm/mr/device/per_device_resource.hpp>
-
-namespace cuvs::neighbors::ivf_flat {
-
-/**
- * @brief Build the index from the dataset for efficient search.
- *
- * NB: Currently, the following distance metrics are supported:
- * - L2Expanded
- * - L2Unexpanded
- * - InnerProduct
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   // use default index parameters
- *   ivf_flat::index_params index_params;
- *   // create and fill the index from a [N, D] dataset
- *   auto index = ivf_flat::build(handle, index_params, dataset, N, D);
- *   // use default search parameters
- *   ivf_flat::search_params search_params;
- *   // search K nearest neighbours for each of the N queries
- *   ivf_flat::search(handle, search_params, index, queries, N, K, out_inds, out_dists);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[in] params configure the index building
- * @param[in] dataset a device pointer to a row-major matrix [n_rows, dim]
- * @param[in] n_rows the number of samples
- * @param[in] dim the dimensionality of the data
- *
- * @return the constructed ivf-flat index
- */
-template <typename T, typename IdxT>
-auto build(raft::resources const& handle,
-           const index_params& params,
-           const T* dataset,
-           IdxT n_rows,
-           uint32_t dim) -> index<T, IdxT>
-{
-  return cuvs::neighbors::ivf_flat::detail::build(handle, params, dataset, n_rows, dim);
-}
-
-/**
- * @defgroup ivf_flat IVF Flat Algorithm
- * @{
- */
-
-/**
- * @brief Build the index from the dataset for efficient search.
- *
- * NB: Currently, the following distance metrics are supported:
- * - L2Expanded
- * - L2Unexpanded
- * - InnerProduct
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   // use default index parameters
- *   ivf_flat::index_params index_params;
- *   // create and fill the index from a [N, D] dataset
- *   auto index = ivf_flat::build(handle, dataset, index_params);
- *   // use default search parameters
- *   ivf_flat::search_params search_params;
- *   // search K nearest neighbours for each of the N queries
- *   ivf_flat::search(handle, search_params, index, queries, out_inds, out_dists);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[in] params configure the index building
- * @param[in] dataset a device pointer to a row-major matrix [n_rows, dim]
- *
- * @return the constructed ivf-flat index
- */
-template <typename T, typename IdxT>
-auto build(raft::resources const& handle,
-           const index_params& params,
-           raft::device_matrix_view<const T, IdxT, raft::row_major> dataset) -> index<T, IdxT>
-{
-  return cuvs::neighbors::ivf_flat::detail::build(handle,
-                                                  params,
-                                                  dataset.data_handle(),
-                                                  static_cast<IdxT>(dataset.extent(0)),
-                                                  static_cast<IdxT>(dataset.extent(1)));
-}
-
-/**
- * @brief Build the index from the dataset for efficient search.
- *
- * NB: Currently, the following distance metrics are supported:
- * - L2Expanded
- * - L2Unexpanded
- * - InnerProduct
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   // use default index parameters
- *   ivf_flat::index_params index_params;
- *   // create and fill the index from a [N, D] dataset
- *   ivf_flat::index<decltype(dataset::value_type), decltype(dataset::index_type)> index;
- *   ivf_flat::build(handle, dataset, index_params, index);
- *   // use default search parameters
- *   ivf_flat::search_params search_params;
- *   // search K nearest neighbours for each of the N queries
- *   ivf_flat::search(handle, search_params, index, queries, out_inds, out_dists);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[in] params configure the index building
- * @param[in] dataset raft::device_matrix_view to a row-major matrix [n_rows, dim]
- * @param[out] idx reference to ivf_flat::index
- *
- */
-template <typename T, typename IdxT>
-void build(raft::resources const& handle,
-           const index_params& params,
-           raft::device_matrix_view<const T, IdxT, raft::row_major> dataset,
-           cuvs::neighbors::ivf_flat::index<T, IdxT>& idx)
-{
-  idx = cuvs::neighbors::ivf_flat::detail::build(handle,
-                                                 params,
-                                                 dataset.data_handle(),
-                                                 static_cast<IdxT>(dataset.extent(0)),
-                                                 static_cast<IdxT>(dataset.extent(1)));
-}
-
-/** @} */
-
-/**
- * @brief Build a new index containing the data of the original plus new extra vectors.
- *
- * Implementation note:
- *    The new data is clustered according to existing kmeans clusters, then the cluster
- *    centers are adjusted to match the newly labeled data.
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   ivf_flat::index_params index_params;
- *   index_params.add_data_on_build = false;      // don't populate index on build
- *   index_params.kmeans_trainset_fraction = 1.0; // use whole dataset for kmeans training
- *   // train the index from a [N, D] dataset
- *   auto index_empty = ivf_flat::build(handle, index_params, dataset, N, D);
- *   // fill the index with the data
- *   auto index = ivf_flat::extend(handle, index_empty, dataset, nullptr, N);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[in] orig_index original index
- * @param[in] new_vectors a device pointer to a row-major matrix [n_rows, index.dim()]
- * @param[in] new_indices a device pointer to a vector of indices [n_rows].
- *    If the original index is empty (`orig_index.size() == 0`), you can pass `nullptr`
- *    here to imply a continuous range `[0...n_rows)`.
- * @param[in] n_rows number of rows in `new_vectors`
- *
- * @return the constructed extended ivf-flat index
- */
-template <typename T, typename IdxT>
-auto extend(raft::resources const& handle,
-            const index<T, IdxT>& orig_index,
-            const T* new_vectors,
-            const IdxT* new_indices,
-            IdxT n_rows) -> index<T, IdxT>
-{
-  return cuvs::neighbors::ivf_flat::detail::extend(
-    handle, orig_index, new_vectors, new_indices, n_rows);
-}
-
-/**
- * @ingroup ivf_flat
- * @{
- */
-
-/**
- * @brief Build a new index containing the data of the original plus new extra vectors.
- *
- * Implementation note:
- *    The new data is clustered according to existing kmeans clusters, then the cluster
- *    centers are adjusted to match the newly labeled data.
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   ivf_flat::index_params index_params;
- *   index_params.add_data_on_build = false;      // don't populate index on build
- *   index_params.kmeans_trainset_fraction = 1.0; // use whole dataset for kmeans training
- *   // train the index from a [N, D] dataset
- *   auto index_empty = ivf_flat::build(handle, dataset, index_params, dataset);
- *   // fill the index with the data
- *   std::optional<raft::device_vector_view<const IdxT, IdxT>> no_op = std::nullopt;
- *   auto index = ivf_flat::extend(handle, index_empty, no_op, dataset);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[in] new_vectors raft::device_matrix_view to a row-major matrix [n_rows, index.dim()]
- * @param[in] new_indices optional raft::device_vector_view to a vector of indices [n_rows].
- *    If the original index is empty (`orig_index.size() == 0`), you can pass `std::nullopt`
- *    here to imply a continuous range `[0...n_rows)`.
- * @param[in] orig_index original index
- *
- * @return the constructed extended ivf-flat index
- */
-template <typename T, typename IdxT>
-auto extend(raft::resources const& handle,
-            raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,
-            std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices,
-            const index<T, IdxT>& orig_index) -> index<T, IdxT>
-{
-  return extend<T, IdxT>(handle,
-                         orig_index,
-                         new_vectors.data_handle(),
-                         new_indices.has_value() ? new_indices.value().data_handle() : nullptr,
-                         new_vectors.extent(0));
-}
-
-/** @} */
-
-/**
- * @brief Extend the index in-place with the new data.
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   ivf_flat::index_params index_params;
- *   index_params.add_data_on_build = false;      // don't populate index on build
- *   index_params.kmeans_trainset_fraction = 1.0; // use whole dataset for kmeans training
- *   // train the index from a [N, D] dataset
- *   auto index_empty = ivf_flat::build(handle, index_params, dataset, N, D);
- *   // fill the index with the data
- *   ivf_flat::extend(handle, index_empty, dataset, nullptr, N);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param handle
- * @param[inout] index
- * @param[in] new_vectors a device pointer to a row-major matrix [n_rows, index.dim()]
- * @param[in] new_indices a device pointer to a vector of indices [n_rows].
- *    If the original index is empty (`orig_index.size() == 0`), you can pass `nullptr`
- *    here to imply a continuous range `[0...n_rows)`.
- * @param[in] n_rows the number of samples
- */
-template <typename T, typename IdxT>
-void extend(raft::resources const& handle,
-            index<T, IdxT>* index,
-            const T* new_vectors,
-            const IdxT* new_indices,
-            IdxT n_rows)
-{
-  cuvs::neighbors::ivf_flat::detail::extend(handle, index, new_vectors, new_indices, n_rows);
-}
-
-/**
- * @ingroup ivf_flat
- * @{
- */
-
-/**
- * @brief Extend the index in-place with the new data.
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   ivf_flat::index_params index_params;
- *   index_params.add_data_on_build = false;      // don't populate index on build
- *   index_params.kmeans_trainset_fraction = 1.0; // use whole dataset for kmeans training
- *   // train the index from a [N, D] dataset
- *   auto index_empty = ivf_flat::build(handle, index_params, dataset);
- *   // fill the index with the data
- *   std::optional<raft::device_vector_view<const IdxT, IdxT>> no_op = std::nullopt;
- *   ivf_flat::extend(handle, dataset, no_opt, &index_empty);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[in] new_vectors raft::device_matrix_view to a row-major matrix [n_rows, index.dim()]
- * @param[in] new_indices optional raft::device_vector_view to a vector of indices [n_rows].
- *    If the original index is empty (`orig_index.size() == 0`), you can pass `std::nullopt`
- *    here to imply a continuous range `[0...n_rows)`.
- * @param[inout] index pointer to index, to be overwritten in-place
- */
-template <typename T, typename IdxT>
-void extend(raft::resources const& handle,
-            raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,
-            std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices,
-            index<T, IdxT>* index)
-{
-  extend(handle,
-         index,
-         new_vectors.data_handle(),
-         new_indices.has_value() ? new_indices.value().data_handle() : nullptr,
-         static_cast<IdxT>(new_vectors.extent(0)));
-}
-
-/** @} */
-
-/**
- * @brief Search ANN using the constructed index with the given filter.
- *
- * See the [ivf_flat::build](#ivf_flat::build) documentation for a usage example.
- *
- * Note, this function requires a temporary buffer to store intermediate results between cuda kernel
- * calls, which may lead to undesirable allocations and slowdown. To alleviate the problem, you can
- * pass a pool memory resource or a large enough pre-allocated memory resource to reduce or
- * eliminate entirely allocations happening within `search`:
- * @code{.cpp}
- *   ...
- *   // Create a pooling memory resource with a pre-defined initial size.
- *   rmm::mr::pool_memory_resource<rmm::mr::device_memory_resource> mr(
- *     rmm::mr::get_current_device_resource(), 1024 * 1024);
- *   // use default search parameters
- *   ivf_flat::search_params search_params;
- *   filtering::none_ivf_sample_filter filter;
- *   // Use the same allocator across multiple searches to reduce the number of
- *   // cuda memory allocations
- *   ivf_flat::search_with_filtering(
- *     handle, search_params, index, queries1, N1, K, out_inds1, out_dists1, &mr, filter);
- *   ivf_flat::search_with_filtering(
- *     handle, search_params, index, queries2, N2, K, out_inds2, out_dists2, &mr, filter);
- *   ivf_flat::search_with_filtering(
- *     handle, search_params, index, queries3, N3, K, out_inds3, out_dists3, &mr, filter);
- *   ...
- * @endcode
- * The exact size of the temporary buffer depends on multiple factors and is an implementation
- * detail. However, you can safely specify a small initial size for the memory pool, so that only a
- * few allocations happen to grow it during the first invocations of the `search`.
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- * @tparam IvfSampleFilterT Device filter function, with the signature
- *         `(uint32_t query_ix, uint32 cluster_ix, uint32_t sample_ix) -> bool` or
- *         `(uint32_t query_ix, uint32 sample_ix) -> bool`
- *
- * @param[in] handle
- * @param[in] params configure the search
- * @param[in] index ivf-flat constructed index
- * @param[in] queries a device pointer to a row-major matrix [n_queries, index->dim()]
- * @param[in] n_queries the batch size
- * @param[in] k the number of neighbors to find for each query.
- * @param[out] neighbors a device pointer to the indices of the neighbors in the source dataset
- * [n_queries, k]
- * @param[out] distances a device pointer to the distances to the selected neighbors [n_queries, k]
- * @param[in] mr an optional memory resource to use across the searches (you can provide a large
- * enough memory pool here to avoid memory allocations within search).
- * @param[in] sample_filter a device filter function that greenlights samples for a given query
- */
-template <typename T, typename IdxT, typename IvfSampleFilterT>
-void search_with_filtering(raft::resources const& handle,
-                           const search_params& params,
-                           const index<T, IdxT>& index,
-                           const T* queries,
-                           uint32_t n_queries,
-                           uint32_t k,
-                           IdxT* neighbors,
-                           float* distances,
-                           rmm::mr::device_memory_resource* mr = nullptr,
-                           IvfSampleFilterT sample_filter      = IvfSampleFilterT())
-{
-  cuvs::neighbors::ivf_flat::detail::search(
-    handle, params, index, queries, n_queries, k, neighbors, distances, mr, sample_filter);
-}
-
-/**
- * @brief Search ANN using the constructed index using the given filter.
- *
- * See the [ivf_flat::build](#ivf_flat::build) documentation for a usage example.
- *
- * Note, this function requires a temporary buffer to store intermediate results between cuda kernel
- * calls, which may lead to undesirable allocations and slowdown. To alleviate the problem, you can
- * pass a pool memory resource or a large enough pre-allocated memory resource to reduce or
- * eliminate entirely allocations happening within `search`:
- * @code{.cpp}
- *   ...
- *   // Create a pooling memory resource with a pre-defined initial size.
- *   rmm::mr::pool_memory_resource<rmm::mr::device_memory_resource> mr(
- *     rmm::mr::get_current_device_resource(), 1024 * 1024);
- *   // use default search parameters
- *   ivf_flat::search_params search_params;
- *   // Use the same allocator across multiple searches to reduce the number of
- *   // cuda memory allocations
- *   ivf_flat::search(handle, search_params, index, queries1, N1, K, out_inds1, out_dists1, &mr);
- *   ivf_flat::search(handle, search_params, index, queries2, N2, K, out_inds2, out_dists2, &mr);
- *   ivf_flat::search(handle, search_params, index, queries3, N3, K, out_inds3, out_dists3, &mr);
- *   ...
- * @endcode
- * The exact size of the temporary buffer depends on multiple factors and is an implementation
- * detail. However, you can safely specify a small initial size for the memory pool, so that only a
- * few allocations happen to grow it during the first invocations of the `search`.
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- *
- * @param[in] handle
- * @param[in] params configure the search
- * @param[in] index ivf-flat constructed index
- * @param[in] queries a device pointer to a row-major matrix [n_queries, index->dim()]
- * @param[in] n_queries the batch size
- * @param[in] k the number of neighbors to find for each query.
- * @param[out] neighbors a device pointer to the indices of the neighbors in the source dataset
- * [n_queries, k]
- * @param[out] distances a device pointer to the distances to the selected neighbors [n_queries, k]
- * @param[in] mr an optional memory resource to use across the searches (you can provide a large
- * enough memory pool here to avoid memory allocations within search).
- */
-template <typename T, typename IdxT>
-void search(raft::resources const& handle,
-            const search_params& params,
-            const index<T, IdxT>& index,
-            const T* queries,
-            uint32_t n_queries,
-            uint32_t k,
-            IdxT* neighbors,
-            float* distances,
-            rmm::mr::device_memory_resource* mr = nullptr)
-{
-  cuvs::neighbors::ivf_flat::detail::search(handle,
-                                            params,
-                                            index,
-                                            queries,
-                                            n_queries,
-                                            k,
-                                            neighbors,
-                                            distances,
-                                            mr,
-                                            cuvs::neighbors::filtering::none_ivf_sample_filter());
-}
-
-/**
- * @ingroup ivf_flat
- * @{
- */
-
-/**
- * @brief Search ANN using the constructed index with the given filter.
- *
- * See the [ivf_flat::build](#ivf_flat::build) documentation for a usage example.
- *
- * Note, this function requires a temporary buffer to store intermediate results between cuda kernel
- * calls, which may lead to undesirable allocations and slowdown. To alleviate the problem, you can
- * pass a pool memory resource or a large enough pre-allocated memory resource to reduce or
- * eliminate entirely allocations happening within `search`:
- * @code{.cpp}
- *   ...
- *   // use default search parameters
- *   ivf_flat::search_params search_params;
- *   filtering::none_ivf_sample_filter filter;
- *   // Use the same allocator across multiple searches to reduce the number of
- *   // cuda memory allocations
- *   ivf_flat::search_with_filtering(
- *     handle, search_params, index, queries1, out_inds1, out_dists1, filter);
- *   ivf_flat::search_with_filtering(
- *     handle, search_params, index, queries2, out_inds2, out_dists2, filter);
- *   ivf_flat::search_with_filtering(
- *     handle, search_params, index, queries3, out_inds3, out_dists3, filter);
- *   ...
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- * @tparam IvfSampleFilterT Device filter function, with the signature
- *         `(uint32_t query_ix, uint32 cluster_ix, uint32_t sample_ix) -> bool` or
- *         `(uint32_t query_ix, uint32 sample_ix) -> bool`
- *
- * @param[in] handle
- * @param[in] params configure the search
- * @param[in] index ivf-flat constructed index
- * @param[in] queries a device pointer to a row-major matrix [n_queries, index->dim()]
- * @param[out] neighbors a device pointer to the indices of the neighbors in the source dataset
- * [n_queries, k]
- * @param[out] distances a device pointer to the distances to the selected neighbors [n_queries, k]
- * @param[in] sample_filter a device filter function that greenlights samples for a given query
- */
-template <typename T, typename IdxT, typename IvfSampleFilterT>
-void search_with_filtering(raft::resources const& handle,
-                           const search_params& params,
-                           const index<T, IdxT>& index,
-                           raft::device_matrix_view<const T, IdxT, raft::row_major> queries,
-                           raft::device_matrix_view<IdxT, IdxT, raft::row_major> neighbors,
-                           raft::device_matrix_view<float, IdxT, raft::row_major> distances,
-                           IvfSampleFilterT sample_filter = IvfSampleFilterT())
-{
-  RAFT_EXPECTS(
-    queries.extent(0) == neighbors.extent(0) && queries.extent(0) == distances.extent(0),
-    "Number of rows in output neighbors and distances matrices must equal the number of queries.");
-
-  RAFT_EXPECTS(neighbors.extent(1) == distances.extent(1),
-               "Number of columns in output neighbors and distances matrices must be equal");
-
-  RAFT_EXPECTS(queries.extent(1) == index.dim(),
-               "Number of query dimensions should equal number of dimensions in the index.");
-
-  search_with_filtering(handle,
-                        params,
-                        index,
-                        queries.data_handle(),
-                        static_cast<std::uint32_t>(queries.extent(0)),
-                        static_cast<std::uint32_t>(neighbors.extent(1)),
-                        neighbors.data_handle(),
-                        distances.data_handle(),
-                        resource::get_workspace_resource(handle),
-                        sample_filter);
-}
-
-/**
- * @brief Search ANN using the constructed index.
- *
- * See the [ivf_flat::build](#ivf_flat::build) documentation for a usage example.
- *
- * Note, this function requires a temporary buffer to store intermediate results between cuda kernel
- * calls, which may lead to undesirable allocations and slowdown. To alleviate the problem, you can
- * pass a pool memory resource or a large enough pre-allocated memory resource to reduce or
- * eliminate entirely allocations happening within `search`:
- * @code{.cpp}
- *   ...
- *   // use default search parameters
- *   ivf_flat::search_params search_params;
- *   // Use the same allocator across multiple searches to reduce the number of
- *   // cuda memory allocations
- *   ivf_flat::search(handle, search_params, index, queries1, out_inds1, out_dists1);
- *   ivf_flat::search(handle, search_params, index, queries2, out_inds2, out_dists2);
- *   ivf_flat::search(handle, search_params, index, queries3, out_inds3, out_dists3);
- *   ...
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- *
- * @param[in] handle
- * @param[in] params configure the search
- * @param[in] index ivf-flat constructed index
- * @param[in] queries a device pointer to a row-major matrix [n_queries, index->dim()]
- * @param[out] neighbors a device pointer to the indices of the neighbors in the source dataset
- * [n_queries, k]
- * @param[out] distances a device pointer to the distances to the selected neighbors [n_queries, k]
- */
-template <typename T, typename IdxT>
-void search(raft::resources const& handle,
-            const search_params& params,
-            const index<T, IdxT>& index,
-            raft::device_matrix_view<const T, IdxT, raft::row_major> queries,
-            raft::device_matrix_view<IdxT, IdxT, raft::row_major> neighbors,
-            raft::device_matrix_view<float, IdxT, raft::row_major> distances)
-{
-  search_with_filtering(handle,
-                        params,
-                        index,
-                        queries,
-                        neighbors,
-                        distances,
-                        cuvs::neighbors::filtering::none_ivf_sample_filter());
-}
-
-/** @} */
-
-}  // namespace cuvs::neighbors::ivf_flat
diff --git a/cpp/include/cuvs/neighbors/ivf_flat.cuh b/cpp/include/cuvs/neighbors/ivf_flat.cuh
deleted file mode 100644
index 8fd9628a4..000000000
--- a/cpp/include/cuvs/neighbors/ivf_flat.cuh
+++ /dev/null
@@ -1,24 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#ifndef RAFT_EXPLICIT_INSTANTIATE_ONLY
-#include "ivf_flat-inl.cuh"
-#endif
-
-#ifdef RAFT_COMPILED
-#include "ivf_flat-ext.cuh"
-#endif
diff --git a/cpp/include/cuvs/neighbors/ivf_flat_codepacker.hpp b/cpp/include/cuvs/neighbors/ivf_flat_codepacker.hpp
deleted file mode 100644
index 9f1b43380..000000000
--- a/cpp/include/cuvs/neighbors/ivf_flat_codepacker.hpp
+++ /dev/null
@@ -1,90 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/neighbors/detail/div_utils.hpp>
-#include <cuvs/neighbors/ivf_flat_types.hpp>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::neighbors::ivf_flat::codepacker {
-
-/**
- * Write one flat code into a block by the given offset. The offset indicates the id of the record
- * in the list. This function interleaves the code and is intended to later copy the interleaved
- * codes over to the IVF list on device. NB: no memory allocation happens here; the block must fit
- * the record (offset + 1).
- *
- * @tparam T
- *
- * @param[in] flat_code input flat code
- * @param[out] block block of memory to write interleaved codes to
- * @param[in] dim dimension of the flat code
- * @param[in] veclen size of interleaved data chunks
- * @param[in] offset how many records to skip before writing the data into the list
- */
-template <typename T>
-_RAFT_HOST_DEVICE void pack_1(
-  const T* flat_code, T* block, uint32_t dim, uint32_t veclen, uint32_t offset)
-{
-  // The data is written in interleaved groups of `index::kGroupSize` vectors
-  using interleaved_group = neighbors::detail::div_utils<kIndexGroupSize>;
-
-  // Interleave dimensions of the source vector while recording it.
-  // NB: such `veclen` is selected, that `dim % veclen == 0`
-  auto group_offset = interleaved_group::roundDown(offset);
-  auto ingroup_id   = interleaved_group::mod(offset) * veclen;
-
-  for (uint32_t l = 0; l < dim; l += veclen) {
-    for (uint32_t j = 0; j < veclen; j++) {
-      block[group_offset * dim + l * kIndexGroupSize + ingroup_id + j] = flat_code[l + j];
-    }
-  }
-}
-
-/**
- * Unpack 1 record of a single list (cluster) in the index to fetch the flat code. The offset
- * indicates the id of the record. This function fetches one flat code from an interleaved code.
- *
- * @tparam T
- *
- * @param[in] block interleaved block. The block can be thought of as the whole inverted list in
- * interleaved format.
- * @param[out] flat_code output flat code
- * @param[in] dim dimension of the flat code
- * @param[in] veclen size of interleaved data chunks
- * @param[in] offset fetch the flat code by the given offset
- */
-template <typename T>
-_RAFT_HOST_DEVICE void unpack_1(
-  const T* block, T* flat_code, uint32_t dim, uint32_t veclen, uint32_t offset)
-{
-  // The data is written in interleaved groups of `index::kGroupSize` vectors
-  using interleaved_group = neighbors::detail::div_utils<kIndexGroupSize>;
-
-  // NB: such `veclen` is selected, that `dim % veclen == 0`
-  auto group_offset = interleaved_group::roundDown(offset);
-  auto ingroup_id   = interleaved_group::mod(offset) * veclen;
-
-  for (uint32_t l = 0; l < dim; l += veclen) {
-    for (uint32_t j = 0; j < veclen; j++) {
-      flat_code[l + j] = block[group_offset * dim + l * kIndexGroupSize + ingroup_id + j];
-    }
-  }
-}
-}  // namespace cuvs::neighbors::ivf_flat::codepacker
\ No newline at end of file
diff --git a/cpp/include/cuvs/neighbors/ivf_flat_helpers.cuh b/cpp/include/cuvs/neighbors/ivf_flat_helpers.cuh
deleted file mode 100644
index cca83cea0..000000000
--- a/cpp/include/cuvs/neighbors/ivf_flat_helpers.cuh
+++ /dev/null
@@ -1,147 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/neighbors/detail/ivf_flat_build.cuh>
-#include <cuvs/neighbors/ivf_flat_types.hpp>
-
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resources.hpp>
-
-#include <cuvs/spatial/knn/detail/ann_utils.cuh>
-
-namespace cuvs::neighbors::ivf_flat::helpers {
-using namespace cuvs::spatial::knn::detail;  // NOLINT
-/**
- * @defgroup ivf_flat_helpers Helper functions for manipulationg IVF Flat Index
- * @{
- */
-
-namespace codepacker {
-
-/**
- * Write flat codes into an existing list by the given offset.
- *
- * NB: no memory allocation happens here; the list must fit the data (offset + n_vec).
- *
- * Usage example:
- * @code{.cpp}
- *   auto list_data  = index.lists()[label]->data.view();
- *   // allocate the buffer for the input codes
- *   auto codes = raft::make_device_matrix<T>(res, n_vec, index.dim());
- *   ... prepare n_vecs to pack into the list in codes ...
- *   // write codes into the list starting from the 42nd position
- *   ivf_pq::helpers::codepacker::pack(
- *       res, make_const_mdspan(codes.view()), index.veclen(), 42, list_data);
- * @endcode
- *
- * @tparam T
- * @tparam IdxT
- *
- * @param[in] res
- * @param[in] codes flat codes [n_vec, dim]
- * @param[in] veclen size of interleaved data chunks
- * @param[in] offset how many records to skip before writing the data into the list
- * @param[inout] list_data block to write into
- */
-template <typename T, typename IdxT>
-void pack(
-  raft::resources const& res,
-  raft::device_matrix_view<const T, uint32_t, raft::row_major> codes,
-  uint32_t veclen,
-  uint32_t offset,
-  raft::device_mdspan<T, typename list_spec<uint32_t, T, IdxT>::list_extents, raft::row_major>
-    list_data)
-{
-  cuvs::neighbors::ivf_flat::detail::pack_list_data<T, IdxT>(res, codes, veclen, offset, list_data);
-}
-
-/**
- * @brief Unpack `n_take` consecutive records of a single list (cluster) in the compressed index
- * starting at given `offset`.
- *
- * Usage example:
- * @code{.cpp}
- *   auto list_data = index.lists()[label]->data.view();
- *   // allocate the buffer for the output
- *   uint32_t n_take = 4;
- *   auto codes = raft::make_device_matrix<T>(res, n_take, index.dim());
- *   uint32_t offset = 0;
- *   // unpack n_take elements from the list
- *   ivf_pq::helpers::codepacker::unpack(res, list_data, index.veclen(), offset, codes.view());
- * @endcode
- *
- * @tparam T
- * @tparam IdxT
- *
- * @param[in] res raft resource
- * @param[in] list_data block to read from
- * @param[in] veclen size of interleaved data chunks
- * @param[in] offset
- *   How many records in the list to skip.
- * @param[inout] codes
- *   the destination buffer [n_take, index.dim()].
- *   The length `n_take` defines how many records to unpack,
- *   it must be <= the list size.
- */
-template <typename T, typename IdxT>
-void unpack(
-  raft::resources const& res,
-  raft::device_mdspan<const T, typename list_spec<uint32_t, T, IdxT>::list_extents, raft::row_major>
-    list_data,
-  uint32_t veclen,
-  uint32_t offset,
-  raft::device_matrix_view<T, uint32_t, raft::row_major> codes)
-{
-  cuvs::neighbors::ivf_flat::detail::unpack_list_data<T, IdxT>(
-    res, list_data, veclen, offset, codes);
-}
-}  // namespace codepacker
-
-/**
- * @brief Public helper API to reset the data and indices ptrs, and the list sizes. Useful for
- * externally modifying the index without going through the build stage. The data and indices of the
- * IVF lists will be lost.
- *
- * Usage example:
- * @code{.cpp}
- *   raft::resources res;
- *   using namespace cuvs::neighbors;
- *   // use default index parameters
- *   ivf_flat::index_params index_params;
- *   // initialize an empty index
- *   ivf_flat::index<int64_t> index(res, index_params, D);
- *   // reset the index's state and list sizes
- *   ivf_flat::helpers::reset_index(res, &index);
- * @endcode
- *
- * @tparam IdxT
- *
- * @param[in] res raft resource
- * @param[inout] index pointer to IVF-PQ index
- */
-template <typename T, typename IdxT>
-void reset_index(const raft::resources& res, index<T, IdxT>* index)
-{
-  auto stream = resource::get_cuda_stream(res);
-
-  utils::memzero(index->list_sizes().data_handle(), index->list_sizes().size(), stream);
-  utils::memzero(index->data_ptrs().data_handle(), index->data_ptrs().size(), stream);
-  utils::memzero(index->inds_ptrs().data_handle(), index->inds_ptrs().size(), stream);
-}
-/** @} */
-}  // namespace cuvs::neighbors::ivf_flat::helpers
diff --git a/cpp/include/cuvs/neighbors/ivf_flat_serialize.cuh b/cpp/include/cuvs/neighbors/ivf_flat_serialize.cuh
deleted file mode 100644
index 37062ea68..000000000
--- a/cpp/include/cuvs/neighbors/ivf_flat_serialize.cuh
+++ /dev/null
@@ -1,154 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "detail/ivf_flat_serialize.cuh"
-
-namespace cuvs::neighbors::ivf_flat {
-
-/**
- * \defgroup ivf_flat_serialize IVF-Flat Serialize
- * @{
- */
-
-/**
- * Write the index to an output stream
- *
- * Experimental, both the API and the serialization format are subject to change.
- *
- * @code{.cpp}
- * #include <raft/core/resources.hpp>
- *
- * raft::resources handle;
- *
- * // create an output stream
- * std::ostream os(std::cout.rdbuf());
- * // create an index with `auto index = ivf_flat::build(...);`
- * raft::serialize(handle, os, index);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- *
- * @param[in] handle the raft handle
- * @param[in] os output stream
- * @param[in] index IVF-Flat index
- *
- */
-template <typename T, typename IdxT>
-void serialize(raft::resources const& handle, std::ostream& os, const index<T, IdxT>& index)
-{
-  detail::serialize(handle, os, index);
-}
-
-/**
- * Save the index to file.
- *
- * Experimental, both the API and the serialization format are subject to change.
- *
- * @code{.cpp}
- * #include <raft/core/resources.hpp>
- *
- * raft::resources handle;
- *
- * // create a string with a filepath
- * std::string filename("/path/to/index");
- * // create an index with `auto index = ivf_flat::build(...);`
- * raft::serialize(handle, filename, index);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- *
- * @param[in] handle the raft handle
- * @param[in] filename the file name for saving the index
- * @param[in] index IVF-Flat index
- *
- */
-template <typename T, typename IdxT>
-void serialize(raft::resources const& handle,
-               const std::string& filename,
-               const index<T, IdxT>& index)
-{
-  detail::serialize(handle, filename, index);
-}
-
-/**
- * Load index from input stream
- *
- * Experimental, both the API and the serialization format are subject to change.
- *
- * @code{.cpp}
- * #include <raft/core/resources.hpp>
- *
- * raft::resources handle;
- *
- * // create an input stream
- * std::istream is(std::cin.rdbuf());
- * using T    = float; // data element type
- * using IdxT = int; // type of the index
- * auto index = raft::deserialize<T, IdxT>(handle, is);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- *
- * @param[in] handle the raft handle
- * @param[in] is input stream
- *
- * @return cuvs::neighbors::ivf_flat::index<T, IdxT>
- */
-template <typename T, typename IdxT>
-index<T, IdxT> deserialize(raft::resources const& handle, std::istream& is)
-{
-  return detail::deserialize<T, IdxT>(handle, is);
-}
-
-/**
- * Load index from file.
- *
- * Experimental, both the API and the serialization format are subject to change.
- *
- * @code{.cpp}
- * #include <raft/core/resources.hpp>
- *
- * raft::resources handle;
- *
- * // create a string with a filepath
- * std::string filename("/path/to/index");
- * using T    = float; // data element type
- * using IdxT = int; // type of the index
- * auto index = raft::deserialize<T, IdxT>(handle, filename);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- *
- * @param[in] handle the raft handle
- * @param[in] filename the name of the file that stores the index
- *
- * @return cuvs::neighbors::ivf_flat::index<T, IdxT>
- */
-template <typename T, typename IdxT>
-index<T, IdxT> deserialize(raft::resources const& handle, const std::string& filename)
-{
-  return detail::deserialize<T, IdxT>(handle, filename);
-}
-
-/**@}*/
-
-}  // namespace cuvs::neighbors::ivf_flat
diff --git a/cpp/include/cuvs/neighbors/ivf_flat_types.hpp b/cpp/include/cuvs/neighbors/ivf_flat_types.hpp
deleted file mode 100644
index 28023f474..000000000
--- a/cpp/include/cuvs/neighbors/ivf_flat_types.hpp
+++ /dev/null
@@ -1,406 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "ann_types.hpp"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/ivf_list_types.hpp>
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/error.hpp>
-#include <raft/core/host_mdarray.hpp>
-#include <raft/core/mdspan_types.hpp>
-#include <raft/core/operators.hpp>
-#include <raft/core/resource/thrust_policy.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/util/integer_utils.hpp>
-
-#include <thrust/reduce.h>
-
-#include <algorithm>  // std::max
-#include <memory>
-#include <optional>
-#include <type_traits>
-
-namespace cuvs::neighbors::ivf_flat {
-/**
- * @addtogroup ivf_flat
- * @{
- */
-
-/** Size of the interleaved group (see `index::data` description). */
-constexpr static uint32_t kIndexGroupSize = 32;
-
-struct index_params : ann::index_params {
-  /** The number of inverted lists (clusters) */
-  uint32_t n_lists = 1024;
-  /** The number of iterations searching for kmeans centers (index building). */
-  uint32_t kmeans_n_iters = 20;
-  /** The fraction of data to use during iterative kmeans building. */
-  double kmeans_trainset_fraction = 0.5;
-  /**
-   * By default (adaptive_centers = false), the cluster centers are trained in `ivf_flat::build`,
-   * and never modified in `ivf_flat::extend`. As a result, you may need to retrain the index
-   * from scratch after invoking (`ivf_flat::extend`) a few times with new data, the distribution of
-   * which is no longer representative of the original training set.
-   *
-   * The alternative behavior (adaptive_centers = true) is to update the cluster centers for new
-   * data when it is added. In this case, `index.centers()` are always exactly the centroids of the
-   * data in the corresponding clusters. The drawback of this behavior is that the centroids depend
-   * on the order of adding new data (through the classification of the added data); that is,
-   * `index.centers()` "drift" together with the changing distribution of the newly added data.
-   */
-  bool adaptive_centers = false;
-  /**
-   * By default, the algorithm allocates more space than necessary for individual clusters
-   * (`list_data`). This allows to amortize the cost of memory allocation and reduce the number of
-   * data copies during repeated calls to `extend` (extending the database).
-   *
-   * The alternative is the conservative allocation behavior; when enabled, the algorithm always
-   * allocates the minimum amount of memory required to store the given number of records. Set this
-   * flag to `true` if you prefer to use as little GPU memory for the database as possible.
-   */
-  bool conservative_memory_allocation = false;
-};
-
-struct search_params : ann::search_params {
-  /** The number of clusters to search. */
-  uint32_t n_probes = 20;
-};
-
-static_assert(std::is_aggregate_v<index_params>);
-static_assert(std::is_aggregate_v<search_params>);
-
-template <typename SizeT, typename ValueT, typename IdxT>
-struct list_spec {
-  using value_type   = ValueT;
-  using list_extents = raft::matrix_extent<SizeT>;
-  using index_type   = IdxT;
-
-  SizeT align_max;
-  SizeT align_min;
-  uint32_t dim;
-
-  constexpr list_spec(uint32_t dim, bool conservative_memory_allocation)
-    : dim(dim),
-      align_min(kIndexGroupSize),
-      align_max(conservative_memory_allocation ? kIndexGroupSize : 1024)
-  {
-  }
-
-  // Allow casting between different size-types (for safer size and offset calculations)
-  template <typename OtherSizeT>
-  constexpr explicit list_spec(const list_spec<OtherSizeT, ValueT, IdxT>& other_spec)
-    : dim{other_spec.dim}, align_min{other_spec.align_min}, align_max{other_spec.align_max}
-  {
-  }
-
-  /** Determine the extents of an array enough to hold a given amount of data. */
-  constexpr auto make_list_extents(SizeT n_rows) const -> list_extents
-  {
-    return make_extents<SizeT>(n_rows, dim);
-  }
-};
-
-template <typename ValueT, typename IdxT, typename SizeT = uint32_t>
-using list_data = ivf::list<list_spec, SizeT, ValueT, IdxT>;
-
-/**
- * @brief IVF-flat index.
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- */
-template <typename T, typename IdxT>
-struct index : ann::index {
-  static_assert(!raft::is_narrowing_v<uint32_t, IdxT>,
-                "IdxT must be able to represent all values of uint32_t");
-
- public:
-  /**
-   * Vectorized load/store size in elements, determines the size of interleaved data chunks.
-   *
-   * TODO: in theory, we can lift this to the template parameter and keep it at hardware maximum
-   * possible value by padding the `dim` of the data https://github.com/rapidsai/raft/issues/711
-   */
-  [[nodiscard]] constexpr inline auto veclen() const noexcept -> uint32_t { return veclen_; }
-  /** Distance metric used for clustering. */
-  [[nodiscard]] constexpr inline auto metric() const noexcept -> cuvs::distance::DistanceType
-  {
-    return metric_;
-  }
-  /** Whether `centers()` change upon extending the index (ivf_pq::extend). */
-  [[nodiscard]] constexpr inline auto adaptive_centers() const noexcept -> bool
-  {
-    return adaptive_centers_;
-  }
-  /**
-   * Inverted list data [size, dim].
-   *
-   * The data consists of the dataset rows, grouped by their labels (into clusters/lists).
-   * Within each list (cluster), the data is grouped into blocks of `kIndexGroupSize` interleaved
-   * vectors. Note, the total index length is slightly larger than the source dataset length,
-   * because each cluster is padded by `kIndexGroupSize` elements.
-   *
-   * Interleaving pattern:
-   * within groups of `kIndexGroupSize` rows, the data is interleaved with the block size equal to
-   * `veclen * sizeof(T)`. That is, a chunk of `veclen` consecutive components of one row is
-   * followed by a chunk of the same size of the next row, and so on.
-   *
-   * __Example__: veclen = 2, dim = 6, kIndexGroupSize = 32, list_size = 31
-   *
-   *     x[ 0, 0], x[ 0, 1], x[ 1, 0], x[ 1, 1], ... x[14, 0], x[14, 1], x[15, 0], x[15, 1],
-   *     x[16, 0], x[16, 1], x[17, 0], x[17, 1], ... x[30, 0], x[30, 1],    -    ,    -    ,
-   *     x[ 0, 2], x[ 0, 3], x[ 1, 2], x[ 1, 3], ... x[14, 2], x[14, 3], x[15, 2], x[15, 3],
-   *     x[16, 2], x[16, 3], x[17, 2], x[17, 3], ... x[30, 2], x[30, 3],    -    ,    -    ,
-   *     x[ 0, 4], x[ 0, 5], x[ 1, 4], x[ 1, 5], ... x[14, 4], x[14, 5], x[15, 4], x[15, 5],
-   *     x[16, 4], x[16, 5], x[17, 4], x[17, 5], ... x[30, 4], x[30, 5],    -    ,    -    ,
-   *
-   */
-  /** Sizes of the lists (clusters) [n_lists]
-   * NB: This may differ from the actual list size if the shared lists have been extended by another
-   * index
-   */
-  inline auto list_sizes() noexcept -> raft::device_vector_view<uint32_t, uint32_t>
-  {
-    return list_sizes_.view();
-  }
-  [[nodiscard]] inline auto list_sizes() const noexcept
-    -> raft::device_vector_view<const uint32_t, uint32_t>
-  {
-    return list_sizes_.view();
-  }
-
-  /** k-means cluster centers corresponding to the lists [n_lists, dim] */
-  inline auto centers() noexcept -> raft::device_matrix_view<float, uint32_t, raft::row_major>
-  {
-    return centers_.view();
-  }
-  [[nodiscard]] inline auto centers() const noexcept
-    -> raft::device_matrix_view<const float, uint32_t, raft::row_major>
-  {
-    return centers_.view();
-  }
-
-  /**
-   * (Optional) Precomputed norms of the `centers` w.r.t. the chosen distance metric [n_lists].
-   *
-   * NB: this may be empty if the index is empty or if the metric does not require the center norms
-   * calculation.
-   */
-  inline auto center_norms() noexcept -> std::optional<raft::device_vector_view<float, uint32_t>>
-  {
-    if (center_norms_.has_value()) {
-      return std::make_optional<device_vector_view<float, uint32_t>>(center_norms_->view());
-    } else {
-      return std::nullopt;
-    }
-  }
-  [[nodiscard]] inline auto center_norms() const noexcept
-    -> std::optional<raft::device_vector_view<const float, uint32_t>>
-  {
-    if (center_norms_.has_value()) {
-      return std::make_optional<raft::device_vector_view<const float, uint32_t>>(
-        center_norms_->view());
-    } else {
-      return std::nullopt;
-    }
-  }
-
-  /** Total length of the index. */
-  [[nodiscard]] constexpr inline auto size() const noexcept -> IdxT { return total_size_; }
-  /** Dimensionality of the data. */
-  [[nodiscard]] constexpr inline auto dim() const noexcept -> uint32_t
-  {
-    return centers_.extent(1);
-  }
-  /** Number of clusters/inverted lists. */
-  [[nodiscard]] constexpr inline auto n_lists() const noexcept -> uint32_t { return lists_.size(); }
-
-  // Don't allow copying the index for performance reasons (try avoiding copying data)
-  index(const index&)                    = delete;
-  index(index&&)                         = default;
-  auto operator=(const index&) -> index& = delete;
-  auto operator=(index&&) -> index&      = default;
-  ~index()                               = default;
-
-  /** Construct an empty index. It needs to be trained and then populated. */
-  index(raft::resources const& res,
-        cuvs::distance::DistanceType metric,
-        uint32_t n_lists,
-        bool adaptive_centers,
-        bool conservative_memory_allocation,
-        uint32_t dim)
-    : ann::index(),
-      veclen_(calculate_veclen(dim)),
-      metric_(metric),
-      adaptive_centers_(adaptive_centers),
-      conservative_memory_allocation_{conservative_memory_allocation},
-      centers_(raft::make_device_matrix<float, uint32_t>(res, n_lists, dim)),
-      center_norms_(std::nullopt),
-      lists_{n_lists},
-      list_sizes_{raft::make_device_vector<uint32_t, uint32_t>(res, n_lists)},
-      data_ptrs_{raft::make_device_vector<T*, uint32_t>(res, n_lists)},
-      inds_ptrs_{raft::make_device_vector<IdxT*, uint32_t>(res, n_lists)},
-      total_size_{0}
-  {
-    check_consistency();
-  }
-
-  /** Construct an empty index. It needs to be trained and then populated. */
-  index(raft::resources const& res, const index_params& params, uint32_t dim)
-    : index(res,
-            params.metric,
-            params.n_lists,
-            params.adaptive_centers,
-            params.conservative_memory_allocation,
-            dim)
-  {
-  }
-
-  /** Pointers to the inverted lists (clusters) data  [n_lists]. */
-  inline auto data_ptrs() noexcept -> raft::device_vector_view<T*, uint32_t>
-  {
-    return data_ptrs_.view();
-  }
-  [[nodiscard]] inline auto data_ptrs() const noexcept
-    -> raft::device_vector_view<T* const, uint32_t>
-  {
-    return data_ptrs_.view();
-  }
-
-  /** Pointers to the inverted lists (clusters) indices  [n_lists]. */
-  inline auto inds_ptrs() noexcept -> raft::device_vector_view<IdxT*, uint32_t>
-  {
-    return inds_ptrs_.view();
-  }
-  [[nodiscard]] inline auto inds_ptrs() const noexcept
-    -> raft::device_vector_view<IdxT* const, uint32_t>
-  {
-    return inds_ptrs_.view();
-  }
-  /**
-   * Whether to use convervative memory allocation when extending the list (cluster) data
-   * (see index_params.conservative_memory_allocation).
-   */
-  [[nodiscard]] constexpr inline auto conservative_memory_allocation() const noexcept -> bool
-  {
-    return conservative_memory_allocation_;
-  }
-
-  /**
-   * Update the state of the dependent index members.
-   */
-  void recompute_internal_state(raft::resources const& res)
-  {
-    auto stream = resource::get_cuda_stream(res);
-
-    // Actualize the list pointers
-    auto this_lists     = lists();
-    auto this_data_ptrs = data_ptrs();
-    auto this_inds_ptrs = inds_ptrs();
-    for (uint32_t label = 0; label < this_lists.size(); label++) {
-      auto& list          = this_lists[label];
-      const auto data_ptr = list ? list->data.data_handle() : nullptr;
-      const auto inds_ptr = list ? list->indices.data_handle() : nullptr;
-      copy(&this_data_ptrs(label), &data_ptr, 1, stream);
-      copy(&this_inds_ptrs(label), &inds_ptr, 1, stream);
-    }
-    auto this_list_sizes = list_sizes().data_handle();
-    total_size_          = thrust::reduce(raft::resource::get_thrust_policy(res),
-                                 this_list_sizes,
-                                 this_list_sizes + this_lists.size(),
-                                 0,
-                                 raft::add_op{});
-    check_consistency();
-  }
-
-  void allocate_center_norms(raft::resources const& res)
-  {
-    switch (metric_) {
-      case cuvs::distance::DistanceType::L2Expanded:
-      case cuvs::distance::DistanceType::L2SqrtExpanded:
-      case cuvs::distance::DistanceType::L2Unexpanded:
-      case cuvs::distance::DistanceType::L2SqrtUnexpanded:
-        center_norms_ = raft::make_device_vector<float, uint32_t>(res, n_lists());
-        break;
-      default: center_norms_ = std::nullopt;
-    }
-  }
-
-  /** Lists' data and indices. */
-  inline auto lists() noexcept -> std::vector<std::shared_ptr<list_data<T, IdxT>>>&
-  {
-    return lists_;
-  }
-  [[nodiscard]] inline auto lists() const noexcept
-    -> const std::vector<std::shared_ptr<list_data<T, IdxT>>>&
-  {
-    return lists_;
-  }
-
- private:
-  /**
-   * TODO: in theory, we can lift this to the template parameter and keep it at hardware maximum
-   * possible value by padding the `dim` of the data https://github.com/rapidsai/raft/issues/711
-   */
-  uint32_t veclen_;
-  cuvs::distance::DistanceType metric_;
-  bool adaptive_centers_;
-  bool conservative_memory_allocation_;
-  std::vector<std::shared_ptr<list_data<T, IdxT>>> lists_;
-  raft::device_vector<uint32_t, uint32_t> list_sizes_;
-  raft::device_matrix<float, uint32_t, raft::row_major> centers_;
-  std::optional<raft::device_vector<float, uint32_t>> center_norms_;
-
-  // Computed members
-  raft::device_vector<T*, uint32_t> data_ptrs_;
-  raft::device_vector<IdxT*, uint32_t> inds_ptrs_;
-  IdxT total_size_;
-
-  /** Throw an error if the index content is inconsistent. */
-  void check_consistency()
-  {
-    auto n_lists = lists_.size();
-    RAFT_EXPECTS(dim() % veclen_ == 0, "dimensionality is not a multiple of the veclen");
-    RAFT_EXPECTS(list_sizes_.extent(0) == n_lists, "inconsistent list size");
-    RAFT_EXPECTS(data_ptrs_.extent(0) == n_lists, "inconsistent list size");
-    RAFT_EXPECTS(inds_ptrs_.extent(0) == n_lists, "inconsistent list size");
-    RAFT_EXPECTS(                                       //
-      (centers_.extent(0) == list_sizes_.extent(0)) &&  //
-        (!center_norms_.has_value() || centers_.extent(0) == center_norms_->extent(0)),
-      "inconsistent number of lists (clusters)");
-  }
-
-  static auto calculate_veclen(uint32_t dim) -> uint32_t
-  {
-    // TODO: consider padding the dimensions and fixing veclen to its maximum possible value as a
-    // template parameter (https://github.com/rapidsai/raft/issues/711)
-
-    // NOTE: keep this consistent with the select_interleaved_scan_kernel logic
-    // in detail/ivf_flat_interleaved_scan-inl.cuh.
-    uint32_t veclen = std::max<uint32_t>(1, 16 / sizeof(T));
-    if (dim % veclen != 0) { veclen = 1; }
-    return veclen;
-  }
-};
-
-/** @} */
-
-}  // namespace cuvs::neighbors::ivf_flat
diff --git a/cpp/include/cuvs/neighbors/ivf_list.hpp b/cpp/include/cuvs/neighbors/ivf_list.hpp
deleted file mode 100644
index c395980de..000000000
--- a/cpp/include/cuvs/neighbors/ivf_list.hpp
+++ /dev/null
@@ -1,196 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/neighbors/ivf_list_types.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resource/thrust_policy.hpp>
-
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/error.hpp>
-#include <raft/core/host_mdarray.hpp>
-#include <raft/core/mdspan.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/core/serialize.hpp>
-#include <raft/util/integer_utils.hpp>
-
-#include <thrust/fill.h>
-
-#include <fstream>
-#include <memory>
-#include <type_traits>
-
-namespace cuvs::neighbors::ivf {
-
-/** The data for a single IVF list. */
-template <template <typename, typename...> typename SpecT,
-          typename SizeT,
-          typename... SpecExtraArgs>
-list<SpecT, SizeT, SpecExtraArgs...>::list(raft::resources const& res,
-                                           const spec_type& spec,
-                                           size_type n_rows)
-  : size{n_rows}, data{res}, indices{res}
-{
-  auto capacity = raft::round_up_safe<SizeT>(n_rows, spec.align_max);
-  if (n_rows < spec.align_max) {
-    capacity = bound_by_power_of_two<SizeT>(std::max<SizeT>(n_rows, spec.align_min));
-    capacity = std::min<SizeT>(capacity, spec.align_max);
-  }
-  try {
-    data    = raft::make_device_mdarray<value_type>(res, spec.make_list_extents(capacity));
-    indices = raft::make_device_vector<index_type, SizeT>(res, capacity);
-  } catch (std::bad_alloc& e) {
-    RAFT_FAIL(
-      "ivf::list: failed to allocate a big enough list to hold all data "
-      "(requested size: %zu records, selected capacity: %zu records). "
-      "Allocator exception: %s",
-      size_t(size),
-      size_t(capacity),
-      e.what());
-  }
-  // Fill the index buffer with a pre-defined marker for easier debugging
-  thrust::fill_n(raft::resource::get_thrust_policy(res),
-                 indices.data_handle(),
-                 indices.size(),
-                 ivf::kInvalidRecord<index_type>);
-}
-
-/**
- * Resize a list by the given id, so that it can contain the given number of records;
- * copy the data if necessary.
- */
-template <typename ListT>
-void resize_list(raft::resources const& res,
-                 std::shared_ptr<ListT>& orig_list,  // NOLINT
-                 const typename ListT::spec_type& spec,
-                 typename ListT::size_type new_used_size,
-                 typename ListT::size_type old_used_size)
-{
-  bool skip_resize = false;
-  if (orig_list) {
-    if (new_used_size <= orig_list->indices.extent(0)) {
-      auto shared_list_size = old_used_size;
-      if (new_used_size <= old_used_size ||
-          orig_list->size.compare_exchange_strong(shared_list_size, new_used_size)) {
-        // We don't need to resize the list if:
-        //  1. The list exists
-        //  2. The new size fits in the list
-        //  3. The list doesn't grow or no-one else has grown it yet
-        skip_resize = true;
-      }
-    }
-  } else {
-    old_used_size = 0;
-  }
-  if (skip_resize) { return; }
-  auto new_list = std::make_shared<ListT>(res, spec, new_used_size);
-  if (old_used_size > 0) {
-    auto copied_data_extents = spec.make_list_extents(old_used_size);
-    auto copied_view         = make_mdspan<typename ListT::value_type,
-                                   typename ListT::size_type,
-                                   raft::row_major,
-                                   false,
-                                   true>(new_list->data.data_handle(), copied_data_extents);
-    copy(copied_view.data_handle(),
-         orig_list->data.data_handle(),
-         copied_view.size(),
-         resource::get_cuda_stream(res));
-    copy(new_list->indices.data_handle(),
-         orig_list->indices.data_handle(),
-         old_used_size,
-         resource::get_cuda_stream(res));
-  }
-  // swap the shared pointer content with the new list
-  new_list.swap(orig_list);
-}
-
-template <typename ListT>
-auto serialize_list(const raft::resources& handle,
-                    std::ostream& os,
-                    const ListT& ld,
-                    const typename ListT::spec_type& store_spec,
-                    std::optional<typename ListT::size_type> size_override = std::nullopt)
-  -> enable_if_valid_list_t<ListT>
-{
-  using size_type = typename ListT::size_type;
-  auto size       = size_override.value_or(ld.size.load());
-  serialize_scalar(handle, os, size);
-  if (size == 0) { return; }
-
-  auto data_extents = store_spec.make_list_extents(size);
-  auto data_array =
-    raft::make_host_mdarray<typename ListT::value_type, size_type, raft::row_major>(data_extents);
-  auto inds_array = raft::make_host_mdarray<typename ListT::index_type, size_type, raft::row_major>(
-    make_extents<size_type>(size));
-  copy(data_array.data_handle(),
-       ld.data.data_handle(),
-       data_array.size(),
-       resource::get_cuda_stream(handle));
-  copy(inds_array.data_handle(),
-       ld.indices.data_handle(),
-       inds_array.size(),
-       resource::get_cuda_stream(handle));
-  resource::sync_stream(handle);
-  serialize_mdspan(handle, os, data_array.view());
-  serialize_mdspan(handle, os, inds_array.view());
-}
-
-template <typename ListT>
-auto serialize_list(const raft::resources& handle,
-                    std::ostream& os,
-                    const std::shared_ptr<ListT>& ld,
-                    const typename ListT::spec_type& store_spec,
-                    std::optional<typename ListT::size_type> size_override = std::nullopt)
-  -> enable_if_valid_list_t<ListT>
-{
-  if (ld) {
-    return serialize_list<ListT>(handle, os, *ld, store_spec, size_override);
-  } else {
-    return serialize_scalar(handle, os, typename ListT::size_type{0});
-  }
-}
-
-template <typename ListT>
-auto deserialize_list(const raft::resources& handle,
-                      std::istream& is,
-                      std::shared_ptr<ListT>& ld,
-                      const typename ListT::spec_type& store_spec,
-                      const typename ListT::spec_type& device_spec) -> enable_if_valid_list_t<ListT>
-{
-  using size_type = typename ListT::size_type;
-  auto size       = deserialize_scalar<size_type>(handle, is);
-  if (size == 0) { return ld.reset(); }
-  std::make_shared<ListT>(handle, device_spec, size).swap(ld);
-  auto data_extents = store_spec.make_list_extents(size);
-  auto data_array =
-    raft::make_host_mdarray<typename ListT::value_type, size_type, raft::row_major>(data_extents);
-  auto inds_array = raft::make_host_mdarray<typename ListT::index_type, size_type, raft::row_major>(
-    make_extents<size_type>(size));
-  deserialize_mdspan(handle, is, data_array.view());
-  deserialize_mdspan(handle, is, inds_array.view());
-  copy(ld->data.data_handle(),
-       data_array.data_handle(),
-       data_array.size(),
-       resource::get_cuda_stream(handle));
-  // NB: copying exactly 'size' indices to leave the rest 'kInvalidRecord' intact.
-  copy(
-    ld->indices.data_handle(), inds_array.data_handle(), size, resource::get_cuda_stream(handle));
-  // Make sure the data is copied from host to device before the host arrays get out of the scope.
-  resource::sync_stream(handle);
-}
-
-}  // namespace cuvs::neighbors::ivf
diff --git a/cpp/include/cuvs/neighbors/ivf_list_types.hpp b/cpp/include/cuvs/neighbors/ivf_list_types.hpp
deleted file mode 100644
index 8d57971a2..000000000
--- a/cpp/include/cuvs/neighbors/ivf_list_types.hpp
+++ /dev/null
@@ -1,79 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/resources.hpp>
-
-#include <atomic>
-#include <limits>
-#include <type_traits>
-
-namespace cuvs::neighbors::ivf {
-
-/**
- * Default value filled in the `indices` array.
- * One may encounter it trying to access a record within a list that is outside of the
- * `size` bound or whenever the list is allocated but not filled-in yet.
- */
-template <typename IdxT>
-constexpr static IdxT kInvalidRecord =
-  (std::is_signed_v<IdxT> ? IdxT{0} : std::numeric_limits<IdxT>::max()) - 1;
-
-/** The data for a single IVF list. */
-template <template <typename, typename...> typename SpecT,
-          typename SizeT,
-          typename... SpecExtraArgs>
-struct list {
-  using size_type    = SizeT;
-  using spec_type    = SpecT<size_type, SpecExtraArgs...>;
-  using value_type   = typename spec_type::value_type;
-  using index_type   = typename spec_type::index_type;
-  using list_extents = typename spec_type::list_extents;
-
-  /** Possibly encoded data; it's layout is defined by `SpecT`. */
-  raft::device_mdarray<value_type, list_extents, raft::row_major> data;
-  /** Source indices. */
-  raft::device_mdarray<index_type, raft::extent_1d<size_type>, raft::row_major> indices;
-  /** The actual size of the content. */
-  std::atomic<size_type> size;
-
-  /** Allocate a new list capable of holding at least `n_rows` data records and indices. */
-  list(raft::resources const& res, const spec_type& spec, size_type n_rows);
-};
-
-template <typename ListT, class T = void>
-struct enable_if_valid_list {};
-
-template <class T,
-          template <typename, typename...>
-          typename SpecT,
-          typename SizeT,
-          typename... SpecExtraArgs>
-struct enable_if_valid_list<list<SpecT, SizeT, SpecExtraArgs...>, T> {
-  using type = T;
-};
-
-/**
- * Designed after `std::enable_if_t`, this trait is helpful in the instance resolution;
- * plug this in the return type of a function that has an instance of `ivf::list` as
- * a template parameter.
- */
-template <typename ListT, class T = void>
-using enable_if_valid_list_t = typename enable_if_valid_list<ListT, T>::type;
-
-}  // namespace cuvs::neighbors::ivf
diff --git a/cpp/include/cuvs/neighbors/ivf_pq-ext.cuh b/cpp/include/cuvs/neighbors/ivf_pq-ext.cuh
deleted file mode 100644
index 8c12175e5..000000000
--- a/cpp/include/cuvs/neighbors/ivf_pq-ext.cuh
+++ /dev/null
@@ -1,227 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cstdint>  // int64_t
-
-#include <cuvs/neighbors/ivf_pq_types.hpp>        // cuvs::neighbors::ivf_pq::index
-#include <raft/core/device_mdspan.hpp>            // raft::device_matrix_view
-#include <raft/core/resources.hpp>                // raft::resources
-#include <raft/util/raft_explicit.hpp>            // RAFT_EXPLICIT
-#include <rmm/mr/device/per_device_resource.hpp>  // rmm::mr::device_memory_resource
-
-#ifdef RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-namespace cuvs::neighbors::ivf_pq {
-
-template <typename T, typename IdxT = uint32_t>
-index<IdxT> build(raft::resources const& handle,
-                  const index_params& params,
-                  raft::device_matrix_view<const T, IdxT, raft::row_major> dataset) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-index<IdxT> extend(
-  raft::resources const& handle,
-  raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,
-  std::optional<raft::device_vector_view<const IdxT, IdxT, raft::row_major>> new_indices,
-  const index<IdxT>& idx) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-void extend(raft::resources const& handle,
-            raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,
-            std::optional<raft::device_vector_view<const IdxT, IdxT, raft::row_major>> new_indices,
-            index<IdxT>* idx) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT, typename IvfSampleFilterT>
-void search_with_filtering(raft::resources const& handle,
-                           const search_params& params,
-                           const index<IdxT>& idx,
-                           raft::device_matrix_view<const T, uint32_t, raft::row_major> queries,
-                           raft::device_matrix_view<IdxT, uint32_t, raft::row_major> neighbors,
-                           raft::device_matrix_view<float, uint32_t, raft::row_major> distances,
-                           IvfSampleFilterT sample_filter) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-void search(raft::resources const& handle,
-            const search_params& params,
-            const index<IdxT>& idx,
-            raft::device_matrix_view<const T, uint32_t, raft::row_major> queries,
-            raft::device_matrix_view<IdxT, uint32_t, raft::row_major> neighbors,
-            raft::device_matrix_view<float, uint32_t, raft::row_major> distances) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT = uint32_t>
-auto build(raft::resources const& handle,
-           const index_params& params,
-           const T* dataset,
-           IdxT n_rows,
-           uint32_t dim) -> index<IdxT> RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-auto extend(raft::resources const& handle,
-            const index<IdxT>& idx,
-            const T* new_vectors,
-            const IdxT* new_indices,
-            IdxT n_rows) -> index<IdxT> RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-void extend(raft::resources const& handle,
-            index<IdxT>* idx,
-            const T* new_vectors,
-            const IdxT* new_indices,
-            IdxT n_rows) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT, typename IvfSampleFilterT>
-void search_with_filtering(raft::resources const& handle,
-                           const cuvs::neighbors::ivf_pq::search_params& params,
-                           const index<IdxT>& idx,
-                           const T* queries,
-                           uint32_t n_queries,
-                           uint32_t k,
-                           IdxT* neighbors,
-                           float* distances,
-                           IvfSampleFilterT sample_filter = IvfSampleFilterT{}) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-void search(raft::resources const& handle,
-            const cuvs::neighbors::ivf_pq::search_params& params,
-            const index<IdxT>& idx,
-            const T* queries,
-            uint32_t n_queries,
-            uint32_t k,
-            IdxT* neighbors,
-            float* distances) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT, typename IvfSampleFilterT>
-[[deprecated(
-  "Drop the `mr` argument and use `raft::resource::set_workspace_resource` instead")]] void
-search_with_filtering(raft::resources const& handle,
-                      const cuvs::neighbors::ivf_pq::search_params& params,
-                      const index<IdxT>& idx,
-                      const T* queries,
-                      uint32_t n_queries,
-                      uint32_t k,
-                      IdxT* neighbors,
-                      float* distances,
-                      rmm::mr::device_memory_resource* mr,
-                      IvfSampleFilterT sample_filter = IvfSampleFilterT{}) RAFT_EXPLICIT;
-
-template <typename T, typename IdxT>
-[[deprecated(
-  "Drop the `mr` argument and use `raft::resource::set_workspace_resource` instead")]] void
-search(raft::resources const& handle,
-       const cuvs::neighbors::ivf_pq::search_params& params,
-       const index<IdxT>& idx,
-       const T* queries,
-       uint32_t n_queries,
-       uint32_t k,
-       IdxT* neighbors,
-       float* distances,
-       rmm::mr::device_memory_resource* mr) RAFT_EXPLICIT;
-
-}  // namespace cuvs::neighbors::ivf_pq
-
-#endif  // RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-#define instantiate_raft_neighbors_ivf_pq_build(T, IdxT)                                        \
-  extern template cuvs::neighbors::ivf_pq::index<IdxT> cuvs::neighbors::ivf_pq::build<T, IdxT>( \
-    raft::resources const& handle,                                                              \
-    const cuvs::neighbors::ivf_pq::index_params& params,                                        \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset);                          \
-                                                                                                \
-  extern template auto cuvs::neighbors::ivf_pq::build(                                          \
-    raft::resources const& handle,                                                              \
-    const cuvs::neighbors::ivf_pq::index_params& params,                                        \
-    const T* dataset,                                                                           \
-    IdxT n_rows,                                                                                \
-    uint32_t dim)                                                                               \
-    ->cuvs::neighbors::ivf_pq::index<IdxT>;
-
-instantiate_raft_neighbors_ivf_pq_build(float, int64_t);
-instantiate_raft_neighbors_ivf_pq_build(int8_t, int64_t);
-instantiate_raft_neighbors_ivf_pq_build(uint8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_build
-
-#define instantiate_raft_neighbors_ivf_pq_extend(T, IdxT)                                        \
-  extern template cuvs::neighbors::ivf_pq::index<IdxT> cuvs::neighbors::ivf_pq::extend<T, IdxT>( \
-    raft::resources const& handle,                                                               \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,                        \
-    std::optional<raft::device_vector_view<const IdxT, IdxT, raft::row_major>> new_indices,      \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx);                                            \
-                                                                                                 \
-  extern template void cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                 \
-    raft::resources const& handle,                                                               \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,                        \
-    std::optional<raft::device_vector_view<const IdxT, IdxT, raft::row_major>> new_indices,      \
-    cuvs::neighbors::ivf_pq::index<IdxT>* idx);                                                  \
-                                                                                                 \
-  extern template auto cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                 \
-    raft::resources const& handle,                                                               \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                                             \
-    const T* new_vectors,                                                                        \
-    const IdxT* new_indices,                                                                     \
-    IdxT n_rows)                                                                                 \
-    ->cuvs::neighbors::ivf_pq::index<IdxT>;                                                      \
-                                                                                                 \
-  extern template void cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                 \
-    raft::resources const& handle,                                                               \
-    cuvs::neighbors::ivf_pq::index<IdxT>* idx,                                                   \
-    const T* new_vectors,                                                                        \
-    const IdxT* new_indices,                                                                     \
-    IdxT n_rows);
-
-instantiate_raft_neighbors_ivf_pq_extend(float, int64_t);
-instantiate_raft_neighbors_ivf_pq_extend(int8_t, int64_t);
-instantiate_raft_neighbors_ivf_pq_extend(uint8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_extend
-
-#define instantiate_raft_neighbors_ivf_pq_search(T, IdxT)                  \
-  extern template void cuvs::neighbors::ivf_pq::search<T, IdxT>(           \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_pq::search_params& params,                  \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                       \
-    raft::device_matrix_view<const T, uint32_t, raft::row_major> queries,  \
-    raft::device_matrix_view<IdxT, uint32_t, raft::row_major> neighbors,   \
-    raft::device_matrix_view<float, uint32_t, raft::row_major> distances); \
-                                                                           \
-  extern template void cuvs::neighbors::ivf_pq::search<T, IdxT>(           \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_pq::search_params& params,                  \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                       \
-    const T* queries,                                                      \
-    uint32_t n_queries,                                                    \
-    uint32_t k,                                                            \
-    IdxT* neighbors,                                                       \
-    float* distances,                                                      \
-    rmm::mr::device_memory_resource* mr);                                  \
-                                                                           \
-  extern template void cuvs::neighbors::ivf_pq::search<T, IdxT>(           \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_pq::search_params& params,                  \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                       \
-    const T* queries,                                                      \
-    uint32_t n_queries,                                                    \
-    uint32_t k,                                                            \
-    IdxT* neighbors,                                                       \
-    float* distances)
-
-instantiate_raft_neighbors_ivf_pq_search(float, int64_t);
-instantiate_raft_neighbors_ivf_pq_search(int8_t, int64_t);
-instantiate_raft_neighbors_ivf_pq_search(uint8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_search
diff --git a/cpp/include/cuvs/neighbors/ivf_pq-inl.cuh b/cpp/include/cuvs/neighbors/ivf_pq-inl.cuh
deleted file mode 100644
index b71a738be..000000000
--- a/cpp/include/cuvs/neighbors/ivf_pq-inl.cuh
+++ /dev/null
@@ -1,529 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/neighbors/detail/ivf_pq_build.cuh>
-#include <cuvs/neighbors/detail/ivf_pq_search.cuh>
-#include <cuvs/neighbors/ivf_pq_serialize.cuh>
-#include <cuvs/neighbors/ivf_pq_types.hpp>
-
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resource/device_memory_resource.hpp>
-#include <raft/core/resources.hpp>
-
-#include <rmm/mr/device/device_memory_resource.hpp>
-
-#include <memory>  // shared_ptr
-
-namespace cuvs::neighbors::ivf_pq {
-
-/**
- * @defgroup ivf_pq IVF PQ Algorithm
- * @{
- */
-
-/**
- * @brief Build the index from the dataset for efficient search.
- *
- * NB: Currently, the following distance metrics are supported:
- * - L2Expanded
- * - L2Unexpanded
- * - InnerProduct
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[in] params configure the index building
- * @param[in] dataset a device matrix view to a row-major matrix [n_rows, dim]
- *
- * @return the constructed ivf-pq index
- */
-template <typename T, typename IdxT = uint32_t>
-index<IdxT> build(raft::resources const& handle,
-                  const index_params& params,
-                  raft::device_matrix_view<const T, IdxT, raft::row_major> dataset)
-{
-  IdxT n_rows = dataset.extent(0);
-  IdxT dim    = dataset.extent(1);
-  return detail::build(handle, params, dataset.data_handle(), n_rows, dim);
-}
-
-/**
- * @brief Extend the index with the new data.
- * *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[in] new_vectors a device matrix view to a row-major matrix [n_rows, idx.dim()]
- * @param[in] new_indices a device vector view to a vector of indices [n_rows].
- *    If the original index is empty (`idx.size() == 0`), you can pass `std::nullopt`
- *    here to imply a continuous range `[0...n_rows)`.
- * @param[inout] idx
- */
-template <typename T, typename IdxT>
-index<IdxT> extend(raft::resources const& handle,
-                   raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,
-                   std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices,
-                   const index<IdxT>& idx)
-{
-  ASSERT(new_vectors.extent(1) == idx.dim(),
-         "new_vectors should have the same dimension as the index");
-
-  IdxT n_rows = new_vectors.extent(0);
-  if (new_indices.has_value()) {
-    ASSERT(n_rows == new_indices.value().extent(0),
-           "new_vectors and new_indices have different number of rows");
-  }
-
-  return detail::extend(handle,
-                        idx,
-                        new_vectors.data_handle(),
-                        new_indices.has_value() ? new_indices.value().data_handle() : nullptr,
-                        n_rows);
-}
-
-/**
- * @brief Extend the index with the new data.
- * *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[in] new_vectors a device matrix view to a row-major matrix [n_rows, idx.dim()]
- * @param[in] new_indices a device vector view to a vector of indices [n_rows].
- *    If the original index is empty (`idx.size() == 0`), you can pass `std::nullopt`
- *    here to imply a continuous range `[0...n_rows)`.
- * @param[inout] idx
- */
-template <typename T, typename IdxT>
-void extend(raft::resources const& handle,
-            raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,
-            std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices,
-            index<IdxT>* idx)
-{
-  ASSERT(new_vectors.extent(1) == idx->dim(),
-         "new_vectors should have the same dimension as the index");
-
-  IdxT n_rows = new_vectors.extent(0);
-  if (new_indices.has_value()) {
-    ASSERT(n_rows == new_indices.value().extent(0),
-           "new_vectors and new_indices have different number of rows");
-  }
-
-  *idx = detail::extend(handle,
-                        *idx,
-                        new_vectors.data_handle(),
-                        new_indices.has_value() ? new_indices.value().data_handle() : nullptr,
-                        n_rows);
-}
-
-/**
- * @brief Search ANN using the constructed index with the given filter.
- *
- * See the [ivf_pq::build](#ivf_pq::build) documentation for a usage example.
- *
- * Note, this function requires a temporary buffer to store intermediate results between cuda kernel
- * calls, which may lead to undesirable allocations and slowdown. To alleviate the problem, you can
- * pass a pool memory resource or a large enough pre-allocated memory resource to reduce or
- * eliminate entirely allocations happening within `search`.
- * The exact size of the temporary buffer depends on multiple factors and is an implementation
- * detail. However, you can safely specify a small initial size for the memory pool, so that only a
- * few allocations happen to grow it during the first invocations of the `search`.
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- * @tparam IvfSampleFilterT Device filter function, with the signature
- *         `(uint32_t query_ix, uint32 cluster_ix, uint32_t sample_ix) -> bool` or
- *         `(uint32_t query_ix, uint32 sample_ix) -> bool`
- *
- * @param[in] handle
- * @param[in] params configure the search
- * @param[in] idx ivf-pq constructed index
- * @param[in] queries a device matrix view to a row-major matrix [n_queries, index->dim()]
- * @param[out] neighbors a device matrix view to the indices of the neighbors in the source dataset
- * [n_queries, k]
- * @param[out] distances a device matrix view to the distances to the selected neighbors [n_queries,
- * k]
- * @param[in] sample_filter a device filter function that greenlights samples for a given query.
- */
-template <typename T, typename IdxT, typename IvfSampleFilterT>
-void search_with_filtering(raft::resources const& handle,
-                           const search_params& params,
-                           const index<IdxT>& idx,
-                           raft::device_matrix_view<const T, uint32_t, raft::row_major> queries,
-                           raft::device_matrix_view<IdxT, uint32_t, raft::row_major> neighbors,
-                           raft::device_matrix_view<float, uint32_t, raft::row_major> distances,
-                           IvfSampleFilterT sample_filter = IvfSampleFilterT{})
-{
-  RAFT_EXPECTS(
-    queries.extent(0) == neighbors.extent(0) && queries.extent(0) == distances.extent(0),
-    "Number of rows in output neighbors and distances matrices must equal the number of queries.");
-
-  RAFT_EXPECTS(neighbors.extent(1) == distances.extent(1),
-               "Number of columns in output neighbors and distances matrices must equal k");
-
-  RAFT_EXPECTS(queries.extent(1) == idx.dim(),
-               "Number of query dimensions should equal number of dimensions in the index.");
-
-  std::uint32_t k = neighbors.extent(1);
-  detail::search(handle,
-                 params,
-                 idx,
-                 queries.data_handle(),
-                 queries.extent(0),
-                 k,
-                 neighbors.data_handle(),
-                 distances.data_handle(),
-                 sample_filter);
-}
-
-/**
- * @brief Search ANN using the constructed index.
- *
- * See the [ivf_pq::build](#ivf_pq::build) documentation for a usage example.
- *
- * Note, this function requires a temporary buffer to store intermediate results between cuda kernel
- * calls, which may lead to undesirable allocations and slowdown. To alleviate the problem, you can
- * pass a pool memory resource or a large enough pre-allocated memory resource to reduce or
- * eliminate entirely allocations happening within `search`.
- * The exact size of the temporary buffer depends on multiple factors and is an implementation
- * detail. However, you can safely specify a small initial size for the memory pool, so that only a
- * few allocations happen to grow it during the first invocations of the `search`.
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- *
- * @param[in] handle
- * @param[in] params configure the search
- * @param[in] idx ivf-pq constructed index
- * @param[in] queries a device matrix view to a row-major matrix [n_queries, index->dim()]
- * @param[out] neighbors a device matrix view to the indices of the neighbors in the source dataset
- * [n_queries, k]
- * @param[out] distances a device matrix view to the distances to the selected neighbors [n_queries,
- * k]
- */
-template <typename T, typename IdxT>
-void search(raft::resources const& handle,
-            const search_params& params,
-            const index<IdxT>& idx,
-            raft::device_matrix_view<const T, uint32_t, raft::row_major> queries,
-            raft::device_matrix_view<IdxT, uint32_t, raft::row_major> neighbors,
-            raft::device_matrix_view<float, uint32_t, raft::row_major> distances)
-{
-  search_with_filtering(handle,
-                        params,
-                        idx,
-                        queries,
-                        neighbors,
-                        distances,
-                        cuvs::neighbors::filtering::none_ivf_sample_filter{});
-}
-
-/** @} */  // end group ivf_pq
-
-/**
- * @brief Build the index from the dataset for efficient search.
- *
- * NB: Currently, the following distance metrics are supported:
- * - L2Expanded
- * - L2Unexpanded
- * - InnerProduct
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   // use default index parameters
- *   ivf_pq::index_params index_params;
- *   // create and fill the index from a [N, D] dataset
- *   auto index = ivf_pq::build(handle, index_params, dataset, N, D);
- *   // use default search parameters
- *   ivf_pq::search_params search_params;
- *   // search K nearest neighbours for each of the N queries
- *   ivf_pq::search(handle, search_params, index, queries, N, K, out_inds, out_dists);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[in] params configure the index building
- * @param[in] dataset a device/host pointer to a row-major matrix [n_rows, dim]
- * @param[in] n_rows the number of samples
- * @param[in] dim the dimensionality of the data
- *
- * @return the constructed ivf-pq index
- */
-template <typename T, typename IdxT = uint32_t>
-auto build(raft::resources const& handle,
-           const index_params& params,
-           const T* dataset,
-           IdxT n_rows,
-           uint32_t dim) -> index<IdxT>
-{
-  return detail::build(handle, params, dataset, n_rows, dim);
-}
-
-/**
- * @brief Build a new index containing the data of the original plus new extra vectors.
- *
- * Implementation note:
- *    The new data is clustered according to existing kmeans clusters, the cluster
- *    centers are unchanged.
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   ivf_pq::index_params index_params;
- *   index_params.add_data_on_build = false;      // don't populate index on build
- *   index_params.kmeans_trainset_fraction = 1.0; // use whole dataset for kmeans training
- *   // train the index from a [N, D] dataset
- *   auto index_empty = ivf_pq::build(handle, index_params, dataset, N, D);
- *   // fill the index with the data
- *   auto index = ivf_pq::extend(handle, index_empty, dataset, nullptr, N);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[inout] idx original index
- * @param[in] new_vectors a device/host pointer to a row-major matrix [n_rows, idx.dim()]
- * @param[in] new_indices a device/host pointer to a vector of indices [n_rows].
- *    If the original index is empty (`idx.size() == 0`), you can pass `nullptr`
- *    here to imply a continuous range `[0...n_rows)`.
- * @param[in] n_rows the number of samples
- *
- * @return the constructed extended ivf-pq index
- */
-template <typename T, typename IdxT>
-auto extend(raft::resources const& handle,
-            const index<IdxT>& idx,
-            const T* new_vectors,
-            const IdxT* new_indices,
-            IdxT n_rows) -> index<IdxT>
-{
-  return detail::extend(handle, idx, new_vectors, new_indices, n_rows);
-}
-
-/**
- * @brief Extend the index with the new data.
- * *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] handle
- * @param[inout] idx
- * @param[in] new_vectors a device/host pointer to a row-major matrix [n_rows, idx.dim()]
- * @param[in] new_indices a device/host pointer to a vector of indices [n_rows].
- *    If the original index is empty (`idx.size() == 0`), you can pass `nullptr`
- *    here to imply a continuous range `[0...n_rows)`.
- * @param[in] n_rows the number of samples
- */
-template <typename T, typename IdxT>
-void extend(raft::resources const& handle,
-            index<IdxT>* idx,
-            const T* new_vectors,
-            const IdxT* new_indices,
-            IdxT n_rows)
-{
-  detail::extend(handle, idx, new_vectors, new_indices, n_rows);
-}
-
-/**
- * @brief Search ANN using the constructed index with the given filter.
- *
- * See the [ivf_pq::build](#ivf_pq::build) documentation for a usage example.
- *
- * Note, this function requires a temporary buffer to store intermediate results between cuda kernel
- * calls, which may lead to undesirable allocations and slowdown. To alleviate the problem, you can
- * pass a pool memory resource or a large enough pre-allocated memory resource to reduce or
- * eliminate entirely allocations happening within `search`:
- * @code{.cpp}
- *   ...
- *   // use default search parameters
- *   ivf_pq::search_params search_params;
- *   filtering::none_ivf_sample_filter filter;
- *   // Use the same allocator across multiple searches to reduce the number of
- *   // cuda memory allocations
- *   ivf_pq::search_with_filtering(
- *     handle, search_params, index, queries1, N1, K, out_inds1, out_dists1, filter);
- *   ivf_pq::search_with_filtering(
- *     handle, search_params, index, queries2, N2, K, out_inds2, out_dists2, filter);
- *   ivf_pq::search_with_filtering(
- *     handle, search_params, index, queries3, N3, K, out_inds3, out_dists3, nfilter);
- *   ...
- * @endcode
- * The exact size of the temporary buffer depends on multiple factors and is an implementation
- * detail. However, you can safely specify a small initial size for the memory pool, so that only a
- * few allocations happen to grow it during the first invocations of the `search`.
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- * @tparam IvfSampleFilterT Device filter function, with the signature
- *         `(uint32_t query_ix, uint32 cluster_ix, uint32_t sample_ix) -> bool` or
- *         `(uint32_t query_ix, uint32 sample_ix) -> bool`
- *
- * @param[in] handle
- * @param[in] params configure the search
- * @param[in] idx ivf-pq constructed index
- * @param[in] queries a device pointer to a row-major matrix [n_queries, index->dim()]
- * @param[in] n_queries the batch size
- * @param[in] k the number of neighbors to find for each query.
- * @param[out] neighbors a device pointer to the indices of the neighbors in the source dataset
- * [n_queries, k]
- * @param[out] distances a device pointer to the distances to the selected neighbors [n_queries, k]
- * @param[in] sample_filter a device filter function that greenlights samples for a given query
- */
-template <typename T, typename IdxT, typename IvfSampleFilterT>
-void search_with_filtering(raft::resources const& handle,
-                           const search_params& params,
-                           const index<IdxT>& idx,
-                           const T* queries,
-                           uint32_t n_queries,
-                           uint32_t k,
-                           IdxT* neighbors,
-                           float* distances,
-                           IvfSampleFilterT sample_filter = IvfSampleFilterT{})
-{
-  detail::search(handle, params, idx, queries, n_queries, k, neighbors, distances, sample_filter);
-}
-
-/**
- * This function is deprecated and will be removed in a future.
- * Please drop the `mr` argument and use `raft::resource::set_workspace_resource` instead.
- */
-template <typename T, typename IdxT, typename IvfSampleFilterT>
-[[deprecated(
-  "Drop the `mr` argument and use `raft::resource::set_workspace_resource` instead")]] void
-search_with_filtering(raft::resources const& handle,
-                      const search_params& params,
-                      const index<IdxT>& idx,
-                      const T* queries,
-                      uint32_t n_queries,
-                      uint32_t k,
-                      IdxT* neighbors,
-                      float* distances,
-                      rmm::mr::device_memory_resource* mr,
-                      IvfSampleFilterT sample_filter = IvfSampleFilterT{})
-{
-  if (mr != nullptr) {
-    // Shallow copy of the resource with the automatic lifespan:
-    //                               change the workspace resource temporarily
-    raft::resources res_local(handle);
-    resource::set_workspace_resource(
-      res_local, std::shared_ptr<rmm::mr::device_memory_resource>{mr, void_op{}});
-    return search_with_filtering(
-      res_local, params, idx, queries, n_queries, k, neighbors, distances, sample_filter);
-  } else {
-    return search_with_filtering(
-      handle, params, idx, queries, n_queries, k, neighbors, distances, sample_filter);
-  }
-}
-
-/**
- * @brief Search ANN using the constructed index.
- *
- * See the [ivf_pq::build](#ivf_pq::build) documentation for a usage example.
- *
- * Note, this function requires a temporary buffer to store intermediate results between cuda kernel
- * calls, which may lead to undesirable allocations and slowdown. To alleviate the problem, you can
- * pass a pool memory resource or a large enough pre-allocated memory resource to reduce or
- * eliminate entirely allocations happening within `search`:
- * @code{.cpp}
- *   ...
- *   // Create a pooling memory resource with a pre-defined initial size.
- *   rmm::mr::pool_memory_resource<rmm::mr::device_memory_resource> mr(
- *     rmm::mr::get_current_device_resource(), 1024 * 1024);
- *   // use default search parameters
- *   ivf_pq::search_params search_params;
- *   // Use the same allocator across multiple searches to reduce the number of
- *   // cuda memory allocations
- *   ivf_pq::search(handle, search_params, index, queries1, N1, K, out_inds1, out_dists1, &mr);
- *   ivf_pq::search(handle, search_params, index, queries2, N2, K, out_inds2, out_dists2, &mr);
- *   ivf_pq::search(handle, search_params, index, queries3, N3, K, out_inds3, out_dists3, &mr);
- *   ...
- * @endcode
- * The exact size of the temporary buffer depends on multiple factors and is an implementation
- * detail. However, you can safely specify a small initial size for the memory pool, so that only a
- * few allocations happen to grow it during the first invocations of the `search`.
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- *
- * @param[in] handle
- * @param[in] params configure the search
- * @param[in] idx ivf-pq constructed index
- * @param[in] queries a device pointer to a row-major matrix [n_queries, index->dim()]
- * @param[in] n_queries the batch size
- * @param[in] k the number of neighbors to find for each query.
- * @param[out] neighbors a device pointer to the indices of the neighbors in the source dataset
- * [n_queries, k]
- * @param[out] distances a device pointer to the distances to the selected neighbors [n_queries, k]
- */
-template <typename T, typename IdxT>
-void search(raft::resources const& handle,
-            const search_params& params,
-            const index<IdxT>& idx,
-            const T* queries,
-            uint32_t n_queries,
-            uint32_t k,
-            IdxT* neighbors,
-            float* distances)
-{
-  return search_with_filtering(handle,
-                               params,
-                               idx,
-                               queries,
-                               n_queries,
-                               k,
-                               neighbors,
-                               distances,
-                               cuvs::neighbors::filtering::none_ivf_sample_filter{});
-}
-
-/**
- * This function is deprecated and will be removed in a future.
- * Please drop the `mr` argument and use `raft::resource::set_workspace_resource` instead.
- */
-template <typename T, typename IdxT>
-[[deprecated(
-  "Drop the `mr` argument and use `raft::resource::set_workspace_resource` instead")]] void
-search(raft::resources const& handle,
-       const search_params& params,
-       const index<IdxT>& idx,
-       const T* queries,
-       uint32_t n_queries,
-       uint32_t k,
-       IdxT* neighbors,
-       float* distances,
-       rmm::mr::device_memory_resource* mr)
-{
-  return search_with_filtering(handle,
-                               params,
-                               idx,
-                               queries,
-                               n_queries,
-                               k,
-                               neighbors,
-                               distances,
-                               mr,
-                               cuvs::neighbors::filtering::none_ivf_sample_filter{});
-}
-
-}  // namespace cuvs::neighbors::ivf_pq
diff --git a/cpp/include/cuvs/neighbors/ivf_pq.cuh b/cpp/include/cuvs/neighbors/ivf_pq.cuh
deleted file mode 100644
index 2d20638f0..000000000
--- a/cpp/include/cuvs/neighbors/ivf_pq.cuh
+++ /dev/null
@@ -1,24 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#ifndef RAFT_EXPLICIT_INSTANTIATE_ONLY
-#include "ivf_pq-inl.cuh"
-#endif
-
-#ifdef RAFT_COMPILED
-#include "ivf_pq-ext.cuh"
-#endif
diff --git a/cpp/include/cuvs/neighbors/ivf_pq_helpers.cuh b/cpp/include/cuvs/neighbors/ivf_pq_helpers.cuh
deleted file mode 100644
index f02148101..000000000
--- a/cpp/include/cuvs/neighbors/ivf_pq_helpers.cuh
+++ /dev/null
@@ -1,798 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cstdio>
-#include <cuvs/neighbors/detail/ivf_pq_build.cuh>
-#include <cuvs/neighbors/ivf_pq_types.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resources.hpp>
-
-#include <cuvs/spatial/knn/detail/ann_utils.cuh>
-
-namespace cuvs::neighbors::ivf_pq::helpers {
-using namespace cuvs::spatial::knn::detail;  // NOLINT
-/**
- * @defgroup ivf_pq_helpers Helper functions for manipulationg IVF PQ Index
- * @{
- */
-
-namespace codepacker {
-/**
- * @brief Unpack `n_take` consecutive records of a single list (cluster) in the compressed index
- * starting at given `offset`.
- *
- * Bit compression is removed, which means output will have pq_dim dimensional vectors (one code per
- * byte, instead of raft::ceildiv(pq_dim * pq_bits, 8) bytes of pq codes).
- *
- * Usage example:
- * @code{.cpp}
- *   auto list_data = index.lists()[label]->data.view();
- *   // allocate the buffer for the output
- *   uint32_t n_take = 4;
- *   auto codes = raft::make_device_matrix<uint8_t>(res, n_take, index.pq_dim());
- *   uint32_t offset = 0;
- *   // unpack n_take elements from the list
- *   ivf_pq::helpers::codepacker::unpack(res, list_data, index.pq_bits(), offset, codes.view());
- * @endcode
- *
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] res raft resource
- * @param[in] list_data block to read from
- * @param[in] pq_bits bit length of encoded vector elements
- * @param[in] offset
- *   How many records in the list to skip.
- * @param[out] codes
- *   the destination buffer [n_take, index.pq_dim()].
- *   The length `n_take` defines how many records to unpack,
- *   it must be smaller than the list size.
- */
-inline void unpack(
-  raft::resources const& res,
-  raft::device_mdspan<const uint8_t, list_spec<uint32_t, uint32_t>::list_extents, raft::row_major>
-    list_data,
-  uint32_t pq_bits,
-  uint32_t offset,
-  raft::device_matrix_view<uint8_t, uint32_t, raft::row_major> codes)
-{
-  ivf_pq::detail::unpack_list_data(
-    codes, list_data, offset, pq_bits, resource::get_cuda_stream(res));
-}
-
-/**
- * @brief Unpack `n_rows` consecutive records of a single list (cluster) in the compressed index
- * starting at given `offset`. The output codes of a single vector are contiguous, not expanded to
- * one code per byte, which means the output has raft::ceildiv(pq_dim * pq_bits, 8) bytes per PQ
- * encoded vector.
- *
- * Usage example:
- * @code{.cpp}
- *   raft::resources res;
- *   auto list_data = index.lists()[label]->data.view();
- *   // allocate the buffer for the output
- *   uint32_t n_rows = 4;
- *   auto codes = raft::make_device_matrix<uint8_t>(
- *     res, n_rows, raft::ceildiv(index.pq_dim() * index.pq_bits(), 8));
- *   uint32_t offset = 0;
- *   // unpack n_rows elements from the list
- *   ivf_pq::helpers::codepacker::unpack_contiguous(
- *     res, list_data, index.pq_bits(), offset, n_rows, index.pq_dim(), codes.data_handle());
- * @endcode
- *
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] res raft resource
- * @param[in] list_data block to read from
- * @param[in] pq_bits bit length of encoded vector elements
- * @param[in] offset
- *   How many records in the list to skip.
- * @param[in] n_rows How many records to unpack
- * @param[in] pq_dim The dimensionality of the PQ compressed records
- * @param[out] codes
- *   the destination buffer [n_rows, raft::ceildiv(pq_dim * pq_bits, 8)].
- *   The length `n_rows` defines how many records to unpack,
- *   it must be smaller than the list size.
- */
-inline void unpack_contiguous(
-  raft::resources const& res,
-  raft::device_mdspan<const uint8_t, list_spec<uint32_t, uint32_t>::list_extents, raft::row_major>
-    list_data,
-  uint32_t pq_bits,
-  uint32_t offset,
-  uint32_t n_rows,
-  uint32_t pq_dim,
-  uint8_t* codes)
-{
-  ivf_pq::detail::unpack_contiguous_list_data(
-    codes, list_data, n_rows, pq_dim, offset, pq_bits, resource::get_cuda_stream(res));
-}
-
-/**
- * Write flat PQ codes into an existing list by the given offset.
- *
- * NB: no memory allocation happens here; the list must fit the data (offset + n_vec).
- *
- * Usage example:
- * @code{.cpp}
- *   auto list_data  = index.lists()[label]->data.view();
- *   // allocate the buffer for the input codes
- *   auto codes = raft::make_device_matrix<uint8_t>(res, n_vec, index.pq_dim());
- *   ... prepare n_vecs to pack into the list in codes ...
- *   // write codes into the list starting from the 42nd position
- *   ivf_pq::helpers::codepacker::pack(
- *       res, make_const_mdspan(codes.view()), index.pq_bits(), 42, list_data);
- * @endcode
- *
- * @param[in] res raft resource
- * @param[in] codes flat PQ codes, one code per byte [n_vec, pq_dim]
- * @param[in] pq_bits bit length of encoded vector elements
- * @param[in] offset how many records to skip before writing the data into the list
- * @param[in] list_data block to write into
- */
-inline void pack(
-  raft::resources const& res,
-  raft::device_matrix_view<const uint8_t, uint32_t, raft::row_major> codes,
-  uint32_t pq_bits,
-  uint32_t offset,
-  raft::device_mdspan<uint8_t, list_spec<uint32_t, uint32_t>::list_extents, raft::row_major>
-    list_data)
-{
-  ivf_pq::detail::pack_list_data(list_data, codes, offset, pq_bits, resource::get_cuda_stream(res));
-}
-
-/**
- * Write flat PQ codes into an existing list by the given offset. The input codes of a single vector
- * are contiguous (not expanded to one code per byte).
- *
- * NB: no memory allocation happens here; the list must fit the data (offset + n_rows records).
- *
- * Usage example:
- * @code{.cpp}
- *   raft::resources res;
- *   auto list_data  = index.lists()[label]->data.view();
- *   // allocate the buffer for the input codes
- *   auto codes = raft::make_device_matrix<uint8_t>(
- *     res, n_rows, raft::ceildiv(index.pq_dim() * index.pq_bits(), 8));
- *   ... prepare compressed vectors to pack into the list in codes ...
- *   // write codes into the list starting from the 42nd position. If the current size of the list
- *   // is greater than 42, this will overwrite the codes starting at this offset.
- *   ivf_pq::helpers::codepacker::pack_contiguous(
- *     res, codes.data_handle(), n_rows, index.pq_dim(), index.pq_bits(), 42, list_data);
- * @endcode
- *
- * @param[in] res raft resource
- * @param[in] codes flat PQ codes, [n_vec, raft::ceildiv(pq_dim * pq_bits, 8)]
- * @param[in] n_rows number of records
- * @param[in] pq_dim
- * @param[in] pq_bits bit length of encoded vector elements
- * @param[in] offset how many records to skip before writing the data into the list
- * @param[in] list_data block to write into
- */
-inline void pack_contiguous(
-  raft::resources const& res,
-  const uint8_t* codes,
-  uint32_t n_rows,
-  uint32_t pq_dim,
-  uint32_t pq_bits,
-  uint32_t offset,
-  raft::device_mdspan<uint8_t, list_spec<uint32_t, uint32_t>::list_extents, raft::row_major>
-    list_data)
-{
-  ivf_pq::detail::pack_contiguous_list_data(
-    list_data, codes, n_rows, pq_dim, offset, pq_bits, resource::get_cuda_stream(res));
-}
-}  // namespace codepacker
-
-/**
- * Write flat PQ codes into an existing list by the given offset.
- *
- * The list is identified by its label.
- *
- * NB: no memory allocation happens here; the list must fit the data (offset + n_vec).
- *
- * Usage example:
- * @code{.cpp}
- *   // We will write into the 137th cluster
- *   uint32_t label = 137;
- *   // allocate the buffer for the input codes
- *   auto codes = raft::make_device_matrix<const uint8_t>(res, n_vec, index.pq_dim());
- *   ... prepare n_vecs to pack into the list in codes ...
- *   // write codes into the list starting from the 42nd position
- *   ivf_pq::helpers::pack_list_data(res, &index, codes_to_pack, label, 42);
- * @endcode
- *
- * @param[in] res raft resource
- * @param[inout] index IVF-PQ index.
- * @param[in] codes flat PQ codes, one code per byte [n_rows, pq_dim]
- * @param[in] label The id of the list (cluster) into which we write.
- * @param[in] offset how many records to skip before writing the data into the list
- */
-template <typename IdxT>
-void pack_list_data(raft::resources const& res,
-                    index<IdxT>* index,
-                    raft::device_matrix_view<const uint8_t, uint32_t, raft::row_major> codes,
-                    uint32_t label,
-                    uint32_t offset)
-{
-  ivf_pq::detail::pack_list_data(res, index, codes, label, offset);
-}
-
-/**
- * Write flat PQ codes into an existing list by the given offset. Use this when the input
- * vectors are PQ encoded and not expanded to one code per byte.
- *
- * The list is identified by its label.
- *
- * NB: no memory allocation happens here; the list into which the vectors are packed must fit offset
- * + n_rows rows.
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   raft::resources res;
- *   // use default index parameters
- *   ivf_pq::index_params index_params;
- *   // create and fill the index from a [N, D] dataset
- *   auto index = ivf_pq::build(res, index_params, dataset, N, D);
- *   // allocate the buffer for n_rows input codes. Each vector occupies
- *   // raft::ceildiv(index.pq_dim() * index.pq_bits(), 8) bytes because
- *   // codes are compressed and without gaps.
- *   auto codes = raft::make_device_matrix<const uint8_t>(
- *     res, n_rows, raft::ceildiv(index.pq_dim() * index.pq_bits(), 8));
- *   ... prepare the compressed vectors to pack into the list in codes ...
- *   // the first n_rows codes in the fourth IVF list are to be overwritten.
- *   uint32_t label = 3;
- *   // write codes into the list starting from the 0th position
- *   ivf_pq::helpers::pack_contiguous_list_data(
- *     res, &index, codes.data_handle(), n_rows, label, 0);
- * @endcode
- *
- * @tparam IdxT
- *
- * @param[in] res raft resource
- * @param[inout] index pointer to IVF-PQ index
- * @param[in] codes flat contiguous PQ codes [n_rows, raft::ceildiv(pq_dim * pq_bits, 8)]
- * @param[in] n_rows how many records to pack
- * @param[in] label The id of the list (cluster) into which we write.
- * @param[in] offset how many records to skip before writing the data into the list
- */
-template <typename IdxT>
-void pack_contiguous_list_data(raft::resources const& res,
-                               index<IdxT>* index,
-                               uint8_t* codes,
-                               uint32_t n_rows,
-                               uint32_t label,
-                               uint32_t offset)
-{
-  ivf_pq::detail::pack_contiguous_list_data(res, index, codes, n_rows, label, offset);
-}
-
-/**
- * @brief Unpack `n_take` consecutive records of a single list (cluster) in the compressed index
- * starting at given `offset`, one code per byte (independently of pq_bits).
- *
- * Usage example:
- * @code{.cpp}
- *   // We will unpack the fourth cluster
- *   uint32_t label = 3;
- *   // Get the list size
- *   uint32_t list_size = 0;
- *   raft::copy(&list_size, index.list_sizes().data_handle() + label, 1,
- * resource::get_cuda_stream(res)); resource::sync_stream(res);
- *   // allocate the buffer for the output
- *   auto codes = raft::make_device_matrix<float>(res, list_size, index.pq_dim());
- *   // unpack the whole list
- *   ivf_pq::helpers::unpack_list_data(res, index, codes.view(), label, 0);
- * @endcode
- *
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] res
- * @param[in] index
- * @param[out] out_codes
- *   the destination buffer [n_take, index.pq_dim()].
- *   The length `n_take` defines how many records to unpack,
- *   it must be smaller than the list size.
- * @param[in] label
- *   The id of the list (cluster) to decode.
- * @param[in] offset
- *   How many records in the list to skip.
- */
-template <typename IdxT>
-void unpack_list_data(raft::resources const& res,
-                      const index<IdxT>& index,
-                      raft::device_matrix_view<uint8_t, uint32_t, raft::row_major> out_codes,
-                      uint32_t label,
-                      uint32_t offset)
-{
-  return ivf_pq::detail::unpack_list_data<IdxT>(res, index, out_codes, label, offset);
-}
-
-/**
- * @brief Unpack a series of records of a single list (cluster) in the compressed index
- * by their in-list offsets, one code per byte (independently of pq_bits).
- *
- * Usage example:
- * @code{.cpp}
- *   // We will unpack the fourth cluster
- *   uint32_t label = 3;
- *   // Create the selection vector
- *   auto selected_indices = raft::make_device_vector<uint32_t>(res, 4);
- *   ... fill the indices ...
- *   resource::sync_stream(res);
- *   // allocate the buffer for the output
- *   auto codes = raft::make_device_matrix<float>(res, selected_indices.size(), index.pq_dim());
- *   // decode the whole list
- *   ivf_pq::helpers::unpack_list_data(
- *       res, index, selected_indices.view(), codes.view(), label);
- * @endcode
- *
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] res raft resource
- * @param[in] index IVF-PQ index (passed by reference)
- * @param[in] in_cluster_indices
- *   The offsets of the selected indices within the cluster.
- * @param[out] out_codes
- *   the destination buffer [n_take, index.pq_dim()].
- *   The length `n_take` defines how many records to unpack,
- *   it must be smaller than the list size.
- * @param[in] label
- *   The id of the list (cluster) to decode.
- */
-template <typename IdxT>
-void unpack_list_data(raft::resources const& res,
-                      const index<IdxT>& index,
-                      raft::device_vector_view<const uint32_t> in_cluster_indices,
-                      raft::device_matrix_view<uint8_t, uint32_t, raft::row_major> out_codes,
-                      uint32_t label)
-{
-  return ivf_pq::detail::unpack_list_data<IdxT>(res, index, out_codes, label, in_cluster_indices);
-}
-
-/**
- * @brief Unpack `n_rows` consecutive PQ encoded vectors of a single list (cluster) in the
- * compressed index starting at given `offset`, not expanded to one code per byte. Each code in the
- * output buffer occupies raft::ceildiv(index.pq_dim() * index.pq_bits(), 8) bytes.
- *
- * Usage example:
- * @code{.cpp}
- *   raft::resources res;
- *   // We will unpack the whole fourth cluster
- *   uint32_t label = 3;
- *   // Get the list size
- *   uint32_t list_size = 0;
- *   raft::update_host(&list_size, index.list_sizes().data_handle() + label, 1,
- * raft::resource::get_cuda_stream(res)); raft::resource::sync_stream(res);
- *   // allocate the buffer for the output
- *   auto codes = raft::make_device_matrix<float>(res, list_size, raft::ceildiv(index.pq_dim() *
- * index.pq_bits(), 8));
- *   // unpack the whole list
- *   ivf_pq::helpers::unpack_list_data(res, index, codes.data_handle(), list_size, label, 0);
- * @endcode
- *
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] res raft resource
- * @param[in] index IVF-PQ index (passed by reference)
- * @param[out] out_codes
- *   the destination buffer [n_rows, raft::ceildiv(index.pq_dim() * index.pq_bits(), 8)].
- *   The length `n_rows` defines how many records to unpack,
- *   offset + n_rows must be smaller than or equal to the list size.
- * @param[in] n_rows how many codes to unpack
- * @param[in] label
- *   The id of the list (cluster) to decode.
- * @param[in] offset
- *   How many records in the list to skip.
- */
-template <typename IdxT>
-void unpack_contiguous_list_data(raft::resources const& res,
-                                 const index<IdxT>& index,
-                                 uint8_t* out_codes,
-                                 uint32_t n_rows,
-                                 uint32_t label,
-                                 uint32_t offset)
-{
-  return ivf_pq::detail::unpack_contiguous_list_data<IdxT>(
-    res, index, out_codes, n_rows, label, offset);
-}
-
-/**
- * @brief Decode `n_take` consecutive records of a single list (cluster) in the compressed index
- * starting at given `offset`.
- *
- * Usage example:
- * @code{.cpp}
- *   // We will reconstruct the fourth cluster
- *   uint32_t label = 3;
- *   // Get the list size
- *   uint32_t list_size = 0;
- *   raft::copy(&list_size, index.list_sizes().data_handle() + label, 1,
- *   resource::get_cuda_stream(res)); resource::sync_stream(res);
- *   // allocate the buffer for the output
- *   auto decoded_vectors = raft::make_device_matrix<float>(res, list_size, index.dim());
- *   // decode the whole list
- *   ivf_pq::helpers::reconstruct_list_data(res, index, decoded_vectors.view(), label, 0);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] res
- * @param[in] index
- * @param[out] out_vectors
- *   the destination buffer [n_take, index.dim()].
- *   The length `n_take` defines how many records to reconstruct,
- *   it must be smaller than the list size.
- * @param[in] label
- *   The id of the list (cluster) to decode.
- * @param[in] offset
- *   How many records in the list to skip.
- */
-template <typename T, typename IdxT>
-void reconstruct_list_data(raft::resources const& res,
-                           const index<IdxT>& index,
-                           raft::device_matrix_view<T, uint32_t, raft::row_major> out_vectors,
-                           uint32_t label,
-                           uint32_t offset)
-{
-  return ivf_pq::detail::reconstruct_list_data(res, index, out_vectors, label, offset);
-}
-
-/**
- * @brief Decode a series of records of a single list (cluster) in the compressed index
- * by their in-list offsets.
- *
- * Usage example:
- * @code{.cpp}
- *   // We will reconstruct the fourth cluster
- *   uint32_t label = 3;
- *   // Create the selection vector
- *   auto selected_indices = raft::make_device_vector<uint32_t>(res, 4);
- *   ... fill the indices ...
- *   resource::sync_stream(res);
- *   // allocate the buffer for the output
- *   auto decoded_vectors = raft::make_device_matrix<float>(
- *                             res, selected_indices.size(), index.dim());
- *   // decode the whole list
- *   ivf_pq::helpers::reconstruct_list_data(
- *       res, index, selected_indices.view(), decoded_vectors.view(), label);
- * @endcode
- *
- * @tparam T data element type
- * @tparam IdxT type of the indices in the source dataset
- *
- * @param[in] res
- * @param[in] index
- * @param[in] in_cluster_indices
- *   The offsets of the selected indices within the cluster.
- * @param[out] out_vectors
- *   the destination buffer [n_take, index.dim()].
- *   The length `n_take` defines how many records to reconstruct,
- *   it must be smaller than the list size.
- * @param[in] label
- *   The id of the list (cluster) to decode.
- */
-template <typename T, typename IdxT>
-void reconstruct_list_data(raft::resources const& res,
-                           const index<IdxT>& index,
-                           raft::device_vector_view<const uint32_t> in_cluster_indices,
-                           raft::device_matrix_view<T, uint32_t, raft::row_major> out_vectors,
-                           uint32_t label)
-{
-  return ivf_pq::detail::reconstruct_list_data(res, index, out_vectors, label, in_cluster_indices);
-}
-
-/**
- * @brief Extend one list of the index in-place, by the list label, skipping the classification and
- * encoding steps.
- *
- * Usage example:
- * @code{.cpp}
- *   // We will extend the fourth cluster
- *   uint32_t label = 3;
- *   // We will fill 4 new vectors
- *   uint32_t n_vec = 4;
- *   // Indices of the new vectors
- *   auto indices = raft::make_device_vector<uint32_t>(res, n_vec);
- *   ... fill the indices ...
- *   auto new_codes = raft::make_device_matrix<uint8_t, uint32_t, raft::row_major> new_codes(
- *       res, n_vec, index.pq_dim());
- *   ... fill codes ...
- *   // extend list with new codes
- *   ivf_pq::helpers::extend_list_with_codes(
- *       res, &index, codes.view(), indices.view(), label);
- * @endcode
- *
- * @tparam IdxT
- *
- * @param[in] res
- * @param[inout] index
- * @param[in] new_codes flat PQ codes, one code per byte [n_rows, index.pq_dim()]
- * @param[in] new_indices source indices [n_rows]
- * @param[in] label the id of the target list (cluster).
- */
-template <typename IdxT>
-void extend_list_with_codes(
-  raft::resources const& res,
-  index<IdxT>* index,
-  raft::device_matrix_view<const uint8_t, uint32_t, raft::row_major> new_codes,
-  raft::device_vector_view<const IdxT, uint32_t, raft::row_major> new_indices,
-  uint32_t label)
-{
-  ivf_pq::detail::extend_list_with_codes(res, index, new_codes, new_indices, label);
-}
-
-/**
- * @brief Extend one list of the index in-place, by the list label, skipping the classification
- * step.
- *
- *  Usage example:
- * @code{.cpp}
- *   // We will extend the fourth cluster
- *   uint32_t label = 3;
- *   // We will extend with 4 new vectors
- *   uint32_t n_vec = 4;
- *   // Indices of the new vectors
- *   auto indices = raft::make_device_vector<uint32_t>(res, n_vec);
- *   ... fill the indices ...
- *   auto new_vectors = raft::make_device_matrix<float, uint32_t, raft::row_major> new_codes(
- *       res, n_vec, index.dim());
- *   ... fill vectors ...
- *   // extend list with new vectors
- *   ivf_pq::helpers::extend_list(
- *       res, &index, new_vectors.view(), indices.view(), label);
- * @endcode
- *
- * @tparam T
- * @tparam IdxT
- *
- * @param[in] res
- * @param[inout] index
- * @param[in] new_vectors data to encode [n_rows, index.dim()]
- * @param[in] new_indices source indices [n_rows]
- * @param[in] label the id of the target list (cluster).
- *
- */
-template <typename T, typename IdxT>
-void extend_list(raft::resources const& res,
-                 index<IdxT>* index,
-                 raft::device_matrix_view<const T, uint32_t, raft::row_major> new_vectors,
-                 raft::device_vector_view<const IdxT, uint32_t, raft::row_major> new_indices,
-                 uint32_t label)
-{
-  ivf_pq::detail::extend_list(res, index, new_vectors, new_indices, label);
-}
-
-/**
- * @brief Remove all data from a single list (cluster) in the index.
- *
- * Usage example:
- * @code{.cpp}
- *   // We will erase the fourth cluster (label = 3)
- *   ivf_pq::helpers::erase_list(res, &index, 3);
- * @endcode
- *
- * @tparam IdxT
- *
- * @param[in] res
- * @param[inout] index
- * @param[in] label the id of the target list (cluster).
- */
-template <typename IdxT>
-void erase_list(raft::resources const& res, index<IdxT>* index, uint32_t label)
-{
-  ivf_pq::detail::erase_list(res, index, label);
-}
-
-/**
- * @brief Public helper API to reset the data and indices ptrs, and the list sizes. Useful for
- * externally modifying the index without going through the build stage. The data and indices of the
- * IVF lists will be lost.
- *
- * Usage example:
- * @code{.cpp}
- *   raft::resources res;
- *   using namespace cuvs::neighbors;
- *   // use default index parameters
- *   ivf_pq::index_params index_params;
- *   // initialize an empty index
- *   ivf_pq::index<int64_t> index(res, index_params, D);
- *   // reset the index's state and list sizes
- *   ivf_pq::helpers::reset_index(res, &index);
- * @endcode
- *
- * @tparam IdxT
- *
- * @param[in] res raft resource
- * @param[inout] index pointer to IVF-PQ index
- */
-template <typename IdxT>
-void reset_index(const raft::resources& res, index<IdxT>* index)
-{
-  auto stream = resource::get_cuda_stream(res);
-
-  utils::memzero(
-    index->accum_sorted_sizes().data_handle(), index->accum_sorted_sizes().size(), stream);
-  utils::memzero(index->list_sizes().data_handle(), index->list_sizes().size(), stream);
-  utils::memzero(index->data_ptrs().data_handle(), index->data_ptrs().size(), stream);
-  utils::memzero(index->inds_ptrs().data_handle(), index->inds_ptrs().size(), stream);
-}
-
-/**
- * @brief Public helper API exposing the computation of the index's rotation matrix.
- * NB: This is to be used only when the rotation matrix is not already computed through
- * cuvs::neighbors::ivf_pq::build.
- *
- * Usage example:
- * @code{.cpp}
- *   raft::resources res;
- *   // use default index parameters
- *   ivf_pq::index_params index_params;
- *   // force random rotation
- *   index_params.force_random_rotation = true;
- *   // initialize an empty index
- *   cuvs::neighbors::ivf_pq::index<int64_t> index(res, index_params, D);
- *   // reset the index
- *   reset_index(res, &index);
- *   // compute the rotation matrix with random_rotation
- *   cuvs::neighbors::ivf_pq::helpers::make_rotation_matrix(
- *     res, &index, index_params.force_random_rotation);
- * @endcode
- *
- * @tparam IdxT
- *
- * @param[in] res raft resource
- * @param[inout] index pointer to IVF-PQ index
- * @param[in] force_random_rotation whether to apply a random rotation matrix on the input data. See
- * cuvs::neighbors::ivf_pq::index_params for more details.
- */
-template <typename IdxT>
-void make_rotation_matrix(raft::resources const& res,
-                          index<IdxT>* index,
-                          bool force_random_rotation)
-{
-  cuvs::neighbors::ivf_pq::detail::make_rotation_matrix(res,
-                                                        force_random_rotation,
-                                                        index->rot_dim(),
-                                                        index->dim(),
-                                                        index->rotation_matrix().data_handle());
-}
-
-/**
- * @brief Public helper API for externally modifying the index's IVF centroids.
- * NB: The index must be reset before this. Use cuvs::neighbors::ivf_pq::extend to construct IVF
- lists according to new centroids.
- *
- * Usage example:
- * @code{.cpp}
- *   raft::resources res;
- *   // allocate the buffer for the input centers
- *   auto cluster_centers = raft::make_device_matrix<float, uint32_t>(res, index.n_lists(),
- index.dim());
- *   ... prepare ivf centroids in cluster_centers ...
- *   // reset the index
- *   reset_index(res, &index);
- *   // recompute the state of the index
- *   cuvs::neighbors::ivf_pq::helpers::recompute_internal_state(res, index);
- *   // Write the IVF centroids
- *   cuvs::neighbors::ivf_pq::helpers::set_centers(
-                    res,
-                    &index,
-                    cluster_centers);
- * @endcode
- *
- * @tparam IdxT
- *
- * @param[in] res raft resource
- * @param[inout] index pointer to IVF-PQ index
- * @param[in] cluster_centers new cluster centers [index.n_lists(), index.dim()]
- */
-template <typename IdxT>
-void set_centers(raft::resources const& res,
-                 index<IdxT>* index,
-                 raft::device_matrix_view<const float, uint32_t> cluster_centers)
-{
-  RAFT_EXPECTS(cluster_centers.extent(0) == index->n_lists(),
-               "Number of rows in the new centers must be equal to the number of IVF lists");
-  RAFT_EXPECTS(cluster_centers.extent(1) == index->dim(),
-               "Number of columns in the new cluster centers and index dim are different");
-  RAFT_EXPECTS(index->size() == 0, "Index must be empty");
-  ivf_pq::detail::set_centers(res, index, cluster_centers.data_handle());
-}
-
-/**
- * @brief Helper exposing the re-computation of list sizes and related arrays if IVF lists have been
- * modified.
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   raft::resources res;
- *   // use default index parameters
- *   ivf_pq::index_params index_params;
- *   // initialize an empty index
- *   ivf_pq::index<int64_t> index(res, index_params, D);
- *   ivf_pq::helpers::reset_index(res, &index);
- *   // resize the first IVF list to hold 5 records
- *   auto spec = list_spec<uint32_t, int64_t>{
- *     index->pq_bits(), index->pq_dim(), index->conservative_memory_allocation()};
- *   uint32_t new_size = 5;
- *   ivf::resize_list(res, list, spec, new_size, 0);
- *   raft::update_device(index.list_sizes(), &new_size, 1, stream);
- *   // recompute the internal state of the index
- *   ivf_pq::recompute_internal_state(res, &index);
- * @endcode
- *
- * @tparam IdxT
- *
- * @param[in] res raft resource
- * @param[inout] index pointer to IVF-PQ index
- */
-template <typename IdxT>
-void recompute_internal_state(const raft::resources& res, index<IdxT>* index)
-{
-  auto& list = index->lists()[0];
-  ivf_pq::detail::recompute_internal_state(res, *index);
-}
-
-/**
- * @brief Public helper API for fetching a trained index's IVF centroids into a buffer that may be
- * allocated on either host or device.
- *
- * Usage example:
- * @code{.cpp}
- *   raft::resources res;
- *   // allocate the buffer for the output centers
- *   auto cluster_centers = raft::make_device_matrix<float, uint32_t>(
- *     res, index.n_lists(), index.dim());
- *   // Extract the IVF centroids into the buffer
- *   cuvs::neighbors::ivf_pq::helpers::extract_centers(res, index, cluster_centers.data_handle());
- * @endcode
- *
- * @tparam IdxT
- *
- * @param[in] res raft resource
- * @param[in] index IVF-PQ index (passed by reference)
- * @param[out] cluster_centers IVF cluster centers [index.n_lists(), index.dim]
- */
-template <typename IdxT>
-void extract_centers(raft::resources const& res,
-                     const index<IdxT>& index,
-                     raft::device_matrix_view<float> cluster_centers)
-{
-  RAFT_EXPECTS(cluster_centers.extent(0) == index.n_lists(),
-               "Number of rows in the output buffer for cluster centers must be equal to the "
-               "number of IVF lists");
-  RAFT_EXPECTS(
-    cluster_centers.extent(1) == index.dim(),
-    "Number of columns in the output buffer for cluster centers and index dim are different");
-  auto stream = resource::get_cuda_stream(res);
-  RAFT_CUDA_TRY(cudaMemcpy2DAsync(cluster_centers.data_handle(),
-                                  sizeof(float) * index.dim(),
-                                  index.centers().data_handle(),
-                                  sizeof(float) * index.dim_ext(),
-                                  sizeof(float) * index.dim(),
-                                  index.n_lists(),
-                                  cudaMemcpyDefault,
-                                  stream));
-}
-/** @} */
-}  // namespace cuvs::neighbors::ivf_pq::helpers
diff --git a/cpp/include/cuvs/neighbors/ivf_pq_serialize.cuh b/cpp/include/cuvs/neighbors/ivf_pq_serialize.cuh
deleted file mode 100644
index 4f8a81afd..000000000
--- a/cpp/include/cuvs/neighbors/ivf_pq_serialize.cuh
+++ /dev/null
@@ -1,146 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "detail/ivf_pq_serialize.cuh"
-
-namespace cuvs::neighbors::ivf_pq {
-
-/**
- * \defgroup ivf_pq_serialize IVF-PQ Serialize
- * @{
- */
-
-/**
- * Write the index to an output stream
- *
- * Experimental, both the API and the serialization format are subject to change.
- *
- * @code{.cpp}
- * #include <raft/core/resources.hpp>
- *
- * raft::resources handle;
- *
- * // create an output stream
- * std::ostream os(std::cout.rdbuf());
- * // create an index with `auto index = ivf_pq::build(...);`
- * raft::serialize(handle, os, index);
- * @endcode
- *
- * @tparam IdxT type of the index
- *
- * @param[in] handle the raft handle
- * @param[in] os output stream
- * @param[in] index IVF-PQ index
- *
- */
-template <typename IdxT>
-void serialize(raft::resources const& handle, std::ostream& os, const index<IdxT>& index)
-{
-  detail::serialize(handle, os, index);
-}
-
-/**
- * Save the index to file.
- *
- * Experimental, both the API and the serialization format are subject to change.
- *
- * @code{.cpp}
- * #include <raft/core/resources.hpp>
- *
- * raft::resources handle;
- *
- * // create a string with a filepath
- * std::string filename("/path/to/index");
- * // create an index with `auto index = ivf_pq::build(...);`
- * raft::serialize(handle, filename, index);
- * @endcode
- *
- * @tparam IdxT type of the index
- *
- * @param[in] handle the raft handle
- * @param[in] filename the file name for saving the index
- * @param[in] index IVF-PQ index
- *
- */
-template <typename IdxT>
-void serialize(raft::resources const& handle, const std::string& filename, const index<IdxT>& index)
-{
-  detail::serialize(handle, filename, index);
-}
-
-/**
- * Load index from input stream
- *
- * Experimental, both the API and the serialization format are subject to change.
- *
- * @code{.cpp}
- * #include <raft/core/resources.hpp>
- *
- * raft::resources handle;
- *
- * // create an input stream
- * std::istream is(std::cin.rdbuf());
- * using IdxT = int; // type of the index
- * auto index = raft::deserialize<IdxT>(handle, is);
- * @endcode
- *
- * @tparam IdxT type of the index
- *
- * @param[in] handle the raft handle
- * @param[in] is input stream
- *
- * @return cuvs::neighbors::ivf_pq::index<IdxT>
- */
-template <typename IdxT>
-index<IdxT> deserialize(raft::resources const& handle, std::istream& is)
-{
-  return detail::deserialize<IdxT>(handle, is);
-}
-
-/**
- * Load index from file.
- *
- * Experimental, both the API and the serialization format are subject to change.
- *
- * @code{.cpp}
- * #include <raft/core/resources.hpp>
- *
- * raft::resources handle;
- *
- * // create a string with a filepath
- * std::string filename("/path/to/index");
- * using IdxT = int; // type of the index
- * auto index = raft::deserialize<IdxT>(handle, filename);
- * @endcode
- *
- * @tparam IdxT type of the index
- *
- * @param[in] handle the raft handle
- * @param[in] filename the name of the file that stores the index
- *
- * @return cuvs::neighbors::ivf_pq::index<IdxT>
- */
-template <typename IdxT>
-index<IdxT> deserialize(raft::resources const& handle, const std::string& filename)
-{
-  return detail::deserialize<IdxT>(handle, filename);
-}
-
-/**@}*/
-
-}  // namespace cuvs::neighbors::ivf_pq
diff --git a/cpp/include/cuvs/neighbors/ivf_pq_types.hpp b/cpp/include/cuvs/neighbors/ivf_pq_types.hpp
deleted file mode 100644
index 16a904fcc..000000000
--- a/cpp/include/cuvs/neighbors/ivf_pq_types.hpp
+++ /dev/null
@@ -1,580 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/neighbors/ann_types.hpp>
-#include <cuvs/neighbors/ivf_list_types.hpp>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/error.hpp>
-#include <raft/core/host_mdarray.hpp>
-#include <raft/core/mdspan_types.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/util/integer_utils.hpp>
-
-#include <thrust/fill.h>
-
-#include <memory>
-#include <type_traits>
-
-namespace cuvs::neighbors::ivf_pq {
-
-/**
- * @addtogroup ivf_pq
- * @{
- */
-
-/** A type for specifying how PQ codebooks are created. */
-enum class codebook_gen {  // NOLINT
-  PER_SUBSPACE = 0,        // NOLINT
-  PER_CLUSTER  = 1,        // NOLINT
-};
-
-struct index_params : ann::index_params {
-  /**
-   * The number of inverted lists (clusters)
-   *
-   * Hint: the number of vectors per cluster (`n_rows/n_lists`) should be approximately 1,000 to
-   * 10,000.
-   */
-  uint32_t n_lists = 1024;
-  /** The number of iterations searching for kmeans centers (index building). */
-  uint32_t kmeans_n_iters = 20;
-  /** The fraction of data to use during iterative kmeans building. */
-  double kmeans_trainset_fraction = 0.5;
-  /**
-   * The bit length of the vector element after compression by PQ.
-   *
-   * Possible values: [4, 5, 6, 7, 8].
-   *
-   * Hint: the smaller the 'pq_bits', the smaller the index size and the better the search
-   * performance, but the lower the recall.
-   */
-  uint32_t pq_bits = 8;
-  /**
-   * The dimensionality of the vector after compression by PQ. When zero, an optimal value is
-   * selected using a heuristic.
-   *
-   * NB: `pq_dim * pq_bits` must be a multiple of 8.
-   *
-   * Hint: a smaller 'pq_dim' results in a smaller index size and better search performance, but
-   * lower recall. If 'pq_bits' is 8, 'pq_dim' can be set to any number, but multiple of 8 are
-   * desirable for good performance. If 'pq_bits' is not 8, 'pq_dim' should be a multiple of 8.
-   * For good performance, it is desirable that 'pq_dim' is a multiple of 32. Ideally, 'pq_dim'
-   * should be also a divisor of the dataset dim.
-   */
-  uint32_t pq_dim = 0;
-  /** How PQ codebooks are created. */
-  codebook_gen codebook_kind = codebook_gen::PER_SUBSPACE;
-  /**
-   * Apply a random rotation matrix on the input data and queries even if `dim % pq_dim == 0`.
-   *
-   * Note: if `dim` is not multiple of `pq_dim`, a random rotation is always applied to the input
-   * data and queries to transform the working space from `dim` to `rot_dim`, which may be slightly
-   * larger than the original space and and is a multiple of `pq_dim` (`rot_dim % pq_dim == 0`).
-   * However, this transform is not necessary when `dim` is multiple of `pq_dim`
-   *   (`dim == rot_dim`, hence no need in adding "extra" data columns / features).
-   *
-   * By default, if `dim == rot_dim`, the rotation transform is initialized with the identity
-   * matrix. When `force_random_rotation == true`, a random orthogonal transform matrix is generated
-   * regardless of the values of `dim` and `pq_dim`.
-   */
-  bool force_random_rotation = false;
-  /**
-   * By default, the algorithm allocates more space than necessary for individual clusters
-   * (`list_data`). This allows to amortize the cost of memory allocation and reduce the number of
-   * data copies during repeated calls to `extend` (extending the database).
-   *
-   * The alternative is the conservative allocation behavior; when enabled, the algorithm always
-   * allocates the minimum amount of memory required to store the given number of records. Set this
-   * flag to `true` if you prefer to use as little GPU memory for the database as possible.
-   */
-  bool conservative_memory_allocation = false;
-};
-
-struct search_params : ann::search_params {
-  /** The number of clusters to search. */
-  uint32_t n_probes = 20;
-  /**
-   * Data type of look up table to be created dynamically at search time.
-   *
-   * Possible values: [CUDA_R_32F, CUDA_R_16F, CUDA_R_8U]
-   *
-   * The use of low-precision types reduces the amount of shared memory required at search time, so
-   * fast shared memory kernels can be used even for datasets with large dimansionality. Note that
-   * the recall is slightly degraded when low-precision type is selected.
-   */
-  cudaDataType_t lut_dtype = CUDA_R_32F;
-  /**
-   * Storage data type for distance/similarity computed at search time.
-   *
-   * Possible values: [CUDA_R_16F, CUDA_R_32F]
-   *
-   * If the performance limiter at search time is device memory access, selecting FP16 will improve
-   * performance slightly.
-   */
-  cudaDataType_t internal_distance_dtype = CUDA_R_32F;
-  /**
-   * Preferred fraction of SM's unified memory / L1 cache to be used as shared memory.
-   *
-   * Possible values: [0.0 - 1.0] as a fraction of the `sharedMemPerMultiprocessor`.
-   *
-   * One wants to increase the carveout to make sure a good GPU occupancy for the main search
-   * kernel, but not to keep it too high to leave some memory to be used as L1 cache. Note, this
-   * value is interpreted only as a hint. Moreover, a GPU usually allows only a fixed set of cache
-   * configurations, so the provided value is rounded up to the nearest configuration. Refer to the
-   * NVIDIA tuning guide for the target GPU architecture.
-   *
-   * Note, this is a low-level tuning parameter that can have drastic negative effects on the search
-   * performance if tweaked incorrectly.
-   */
-  double preferred_shmem_carveout = 1.0;
-};
-
-static_assert(std::is_aggregate_v<index_params>);
-static_assert(std::is_aggregate_v<search_params>);
-
-/** Size of the interleaved group. */
-constexpr static uint32_t kIndexGroupSize = 32;
-/** Stride of the interleaved group for vectorized loads. */
-constexpr static uint32_t kIndexGroupVecLen = 16;
-
-/**
- * Default value returned by `search` when the `n_probes` is too small and top-k is too large.
- * One may encounter it if the combined size of probed clusters is smaller than the requested
- * number of results per query.
- */
-template <typename IdxT>
-constexpr static IdxT kOutOfBoundsRecord = std::numeric_limits<IdxT>::max();
-
-template <typename SizeT, typename IdxT>
-struct list_spec {
-  using value_type = uint8_t;
-  using index_type = IdxT;
-  /** PQ-encoded data stored in the interleaved format:
-   *
-   *    [ raft::ceildiv(list_size, kIndexGroupSize)
-   *    , raft::ceildiv(pq_dim, (kIndexGroupVecLen * 8u) / pq_bits)
-   *    , kIndexGroupSize
-   *    , kIndexGroupVecLen
-   *    ].
-   */
-  using list_extents =
-    extents<SizeT, raft::dynamic_extent, raft::dynamic_extent, kIndexGroupSize, kIndexGroupVecLen>;
-
-  SizeT align_max;
-  SizeT align_min;
-  uint32_t pq_bits;
-  uint32_t pq_dim;
-
-  constexpr list_spec(uint32_t pq_bits, uint32_t pq_dim, bool conservative_memory_allocation)
-    : pq_bits(pq_bits),
-      pq_dim(pq_dim),
-      align_min(kIndexGroupSize),
-      align_max(conservative_memory_allocation ? kIndexGroupSize : 1024)
-  {
-  }
-
-  // Allow casting between different size-types (for safer size and offset calculations)
-  template <typename OtherSizeT>
-  constexpr explicit list_spec(const list_spec<OtherSizeT, IdxT>& other_spec)
-    : pq_bits{other_spec.pq_bits},
-      pq_dim{other_spec.pq_dim},
-      align_min{other_spec.align_min},
-      align_max{other_spec.align_max}
-  {
-  }
-
-  /** Determine the extents of an array enough to hold a given amount of data. */
-  constexpr auto make_list_extents(SizeT n_rows) const -> list_extents
-  {
-    // how many elems of pq_dim fit into one kIndexGroupVecLen-byte chunk
-    auto pq_chunk = (kIndexGroupVecLen * 8u) / pq_bits;
-    return make_extents<SizeT>(div_rounding_up_safe<SizeT>(n_rows, kIndexGroupSize),
-                               div_rounding_up_safe<SizeT>(pq_dim, pq_chunk),
-                               kIndexGroupSize,
-                               kIndexGroupVecLen);
-  }
-};
-
-template <typename IdxT, typename SizeT = uint32_t>
-using list_data = ivf::list<list_spec, SizeT, IdxT>;
-
-/**
- * @brief IVF-PQ index.
- *
- * In the IVF-PQ index, a database vector y is approximated with two level quantization:
- *
- * y = Q_1(y) + Q_2(y - Q_1(y))
- *
- * The first level quantizer (Q_1), maps the vector y to the nearest cluster center. The number of
- * clusters is n_lists.
- *
- * The second quantizer encodes the residual, and it is defined as a product quantizer [1].
- *
- * A product quantizer encodes a `dim` dimensional vector with a `pq_dim` dimensional vector.
- * First we split the input vector into `pq_dim` subvectors (denoted by u), where each u vector
- * contains `pq_len` distinct components of y
- *
- * y_1, y_2, ... y_{pq_len}, y_{pq_len+1}, ... y_{2*pq_len}, ... y_{dim-pq_len+1} ... y_{dim}
- *  \___________________/     \____________________________/      \______________________/
- *         u_1                         u_2                          u_{pq_dim}
- *
- * Then each subvector encoded with a separate quantizer q_i, end the results are concatenated
- *
- * Q_2(y) = q_1(u_1),q_2(u_2),...,q_{pq_dim}(u_pq_dim})
- *
- * Each quantizer q_i outputs a code with pq_bit bits. The second level quantizers are also defined
- * by k-means clustering in the corresponding sub-space: the reproduction values are the centroids,
- * and the set of reproduction values is the codebook.
- *
- * When the data dimensionality `dim` is not multiple of `pq_dim`, the feature space is transformed
- * using a random orthogonal matrix to have `rot_dim = pq_dim * pq_len` dimensions
- * (`rot_dim >= dim`).
- *
- * The second-level quantizers are trained either for each subspace or for each cluster:
- *   (a) codebook_gen::PER_SUBSPACE:
- *         creates `pq_dim` second-level quantizers - one for each slice of the data along features;
- *   (b) codebook_gen::PER_CLUSTER:
- *         creates `n_lists` second-level quantizers - one for each first-level cluster.
- * In either case, the centroids are again found using k-means clustering interpreting the data as
- * having pq_len dimensions.
- *
- * [1] Product quantization for nearest neighbor search Herve Jegou, Matthijs Douze, Cordelia Schmid
- *
- * @tparam IdxT type of the indices in the source dataset
- *
- */
-template <typename IdxT>
-struct index : ann::index {
-  static_assert(!raft::is_narrowing_v<uint32_t, IdxT>,
-                "IdxT must be able to represent all values of uint32_t");
-
- public:
-  /** Total length of the index. */
-  [[nodiscard]] constexpr inline auto size() const noexcept -> IdxT
-  {
-    return accum_sorted_sizes_(n_lists());
-  }
-  /** Dimensionality of the input data. */
-  [[nodiscard]] constexpr inline auto dim() const noexcept -> uint32_t { return dim_; }
-  /**
-   * Dimensionality of the cluster centers:
-   * input data dim extended with vector norms and padded to 8 elems.
-   */
-  [[nodiscard]] constexpr inline auto dim_ext() const noexcept -> uint32_t
-  {
-    return raft::round_up_safe(dim() + 1, 8u);
-  }
-  /**
-   * Dimensionality of the data after transforming it for PQ processing
-   * (rotated and augmented to be muplitple of `pq_dim`).
-   */
-  [[nodiscard]] constexpr inline auto rot_dim() const noexcept -> uint32_t
-  {
-    return pq_len() * pq_dim();
-  }
-  /** The bit length of an encoded vector element after compression by PQ. */
-  [[nodiscard]] constexpr inline auto pq_bits() const noexcept -> uint32_t { return pq_bits_; }
-  /** The dimensionality of an encoded vector after compression by PQ. */
-  [[nodiscard]] constexpr inline auto pq_dim() const noexcept -> uint32_t { return pq_dim_; }
-  /** Dimensionality of a subspaces, i.e. the number of vector components mapped to a subspace */
-  [[nodiscard]] constexpr inline auto pq_len() const noexcept -> uint32_t
-  {
-    return raft::div_rounding_up_unsafe(dim(), pq_dim());
-  }
-  /** The number of vectors in a PQ codebook (`1 << pq_bits`). */
-  [[nodiscard]] constexpr inline auto pq_book_size() const noexcept -> uint32_t
-  {
-    return 1 << pq_bits();
-  }
-  /** Distance metric used for clustering. */
-  [[nodiscard]] constexpr inline auto metric() const noexcept -> cuvs::distance::DistanceType
-  {
-    return metric_;
-  }
-  /** How PQ codebooks are created. */
-  [[nodiscard]] constexpr inline auto codebook_kind() const noexcept -> codebook_gen
-  {
-    return codebook_kind_;
-  }
-  /** Number of clusters/inverted lists (first level quantization). */
-  [[nodiscard]] constexpr inline auto n_lists() const noexcept -> uint32_t { return lists_.size(); }
-  /**
-   * Whether to use convervative memory allocation when extending the list (cluster) data
-   * (see index_params.conservative_memory_allocation).
-   */
-  [[nodiscard]] constexpr inline auto conservative_memory_allocation() const noexcept -> bool
-  {
-    return conservative_memory_allocation_;
-  }
-
-  // Don't allow copying the index for performance reasons (try avoiding copying data)
-  index(const index&)                    = delete;
-  index(index&&)                         = default;
-  auto operator=(const index&) -> index& = delete;
-  auto operator=(index&&) -> index&      = default;
-  ~index()                               = default;
-
-  /** Construct an empty index. It needs to be trained and then populated. */
-  index(raft::resources const& handle,
-        cuvs::distance::DistanceType metric,
-        codebook_gen codebook_kind,
-        uint32_t n_lists,
-        uint32_t dim,
-        uint32_t pq_bits                    = 8,
-        uint32_t pq_dim                     = 0,
-        bool conservative_memory_allocation = false)
-    : ann::index(),
-      metric_(metric),
-      codebook_kind_(codebook_kind),
-      dim_(dim),
-      pq_bits_(pq_bits),
-      pq_dim_(pq_dim == 0 ? calculate_pq_dim(dim) : pq_dim),
-      conservative_memory_allocation_(conservative_memory_allocation),
-      pq_centers_{make_device_mdarray<float>(handle, make_pq_centers_extents())},
-      lists_{n_lists},
-      rotation_matrix_{make_device_matrix<float, uint32_t>(handle, this->rot_dim(), this->dim())},
-      list_sizes_{make_device_vector<uint32_t, uint32_t>(handle, n_lists)},
-      centers_{make_device_matrix<float, uint32_t>(handle, n_lists, this->dim_ext())},
-      centers_rot_{make_device_matrix<float, uint32_t>(handle, n_lists, this->rot_dim())},
-      data_ptrs_{make_device_vector<uint8_t*, uint32_t>(handle, n_lists)},
-      inds_ptrs_{make_device_vector<IdxT*, uint32_t>(handle, n_lists)},
-      accum_sorted_sizes_{make_host_vector<IdxT, uint32_t>(n_lists + 1)}
-  {
-    check_consistency();
-    accum_sorted_sizes_(n_lists) = 0;
-  }
-
-  /** Construct an empty index. It needs to be trained and then populated. */
-  index(raft::resources const& handle, const index_params& params, uint32_t dim)
-    : index(handle,
-            params.metric,
-            params.codebook_kind,
-            params.n_lists,
-            dim,
-            params.pq_bits,
-            params.pq_dim,
-            params.conservative_memory_allocation)
-  {
-  }
-
-  using pq_centers_extents = std::experimental::
-    extents<uint32_t, raft::dynamic_extent, raft::dynamic_extent, raft::dynamic_extent>;
-  /**
-   * PQ cluster centers
-   *
-   *   - codebook_gen::PER_SUBSPACE: [pq_dim , pq_len, pq_book_size]
-   *   - codebook_gen::PER_CLUSTER:  [n_lists, pq_len, pq_book_size]
-   */
-  inline auto pq_centers() noexcept
-    -> raft::device_mdspan<float, pq_centers_extents, raft::row_major>
-  {
-    return pq_centers_.view();
-  }
-  [[nodiscard]] inline auto pq_centers() const noexcept
-    -> raft::device_mdspan<const float, pq_centers_extents, raft::row_major>
-  {
-    return pq_centers_.view();
-  }
-
-  /** Lists' data and indices. */
-  inline auto lists() noexcept -> std::vector<std::shared_ptr<list_data<IdxT>>>& { return lists_; }
-  [[nodiscard]] inline auto lists() const noexcept
-    -> const std::vector<std::shared_ptr<list_data<IdxT>>>&
-  {
-    return lists_;
-  }
-
-  /** Pointers to the inverted lists (clusters) data  [n_lists]. */
-  inline auto data_ptrs() noexcept -> raft::device_vector_view<uint8_t*, uint32_t, raft::row_major>
-  {
-    return data_ptrs_.view();
-  }
-  [[nodiscard]] inline auto data_ptrs() const noexcept
-    -> raft::device_vector_view<const uint8_t* const, uint32_t, raft::row_major>
-  {
-    return make_mdspan<const uint8_t* const, uint32_t, raft::row_major, false, true>(
-      data_ptrs_.data_handle(), data_ptrs_.extents());
-  }
-
-  /** Pointers to the inverted lists (clusters) indices  [n_lists]. */
-  inline auto inds_ptrs() noexcept -> raft::device_vector_view<IdxT*, uint32_t, raft::row_major>
-  {
-    return inds_ptrs_.view();
-  }
-  [[nodiscard]] inline auto inds_ptrs() const noexcept
-    -> raft::device_vector_view<const IdxT* const, uint32_t, raft::row_major>
-  {
-    return make_mdspan<const IdxT* const, uint32_t, raft::row_major, false, true>(
-      inds_ptrs_.data_handle(), inds_ptrs_.extents());
-  }
-
-  /** The transform matrix (original space -> rotated padded space) [rot_dim, dim] */
-  inline auto rotation_matrix() noexcept
-    -> raft::device_matrix_view<float, uint32_t, raft::row_major>
-  {
-    return rotation_matrix_.view();
-  }
-  [[nodiscard]] inline auto rotation_matrix() const noexcept
-    -> raft::device_matrix_view<const float, uint32_t, raft::row_major>
-  {
-    return rotation_matrix_.view();
-  }
-
-  /**
-   * Accumulated list sizes, sorted in descending order [n_lists + 1].
-   * The last value contains the total length of the index.
-   * The value at index zero is always zero.
-   *
-   * That is, the content of this span is as if the `list_sizes` was sorted and then accumulated.
-   *
-   * This span is used during search to estimate the maximum size of the workspace.
-   */
-  inline auto accum_sorted_sizes() noexcept
-    -> raft::host_vector_view<IdxT, uint32_t, raft::row_major>
-  {
-    return accum_sorted_sizes_.view();
-  }
-  [[nodiscard]] inline auto accum_sorted_sizes() const noexcept
-    -> raft::host_vector_view<const IdxT, uint32_t, raft::row_major>
-  {
-    return accum_sorted_sizes_.view();
-  }
-
-  /** Sizes of the lists [n_lists]. */
-  inline auto list_sizes() noexcept -> raft::device_vector_view<uint32_t, uint32_t, raft::row_major>
-  {
-    return list_sizes_.view();
-  }
-  [[nodiscard]] inline auto list_sizes() const noexcept
-    -> raft::device_vector_view<const uint32_t, uint32_t, raft::row_major>
-  {
-    return list_sizes_.view();
-  }
-
-  /** Cluster centers corresponding to the lists in the original space [n_lists, dim_ext] */
-  inline auto centers() noexcept -> raft::device_matrix_view<float, uint32_t, raft::row_major>
-  {
-    return centers_.view();
-  }
-  [[nodiscard]] inline auto centers() const noexcept
-    -> raft::device_matrix_view<const float, uint32_t, raft::row_major>
-  {
-    return centers_.view();
-  }
-
-  /** Cluster centers corresponding to the lists in the rotated space [n_lists, rot_dim] */
-  inline auto centers_rot() noexcept -> raft::device_matrix_view<float, uint32_t, raft::row_major>
-  {
-    return centers_rot_.view();
-  }
-  [[nodiscard]] inline auto centers_rot() const noexcept
-    -> raft::device_matrix_view<const float, uint32_t, raft::row_major>
-  {
-    return centers_rot_.view();
-  }
-
-  /** fetch size of a particular IVF list in bytes using the list extents.
-   * Usage example:
-   * @code{.cpp}
-   *   raft::resources res;
-   *   // use default index params
-   *   ivf_pq::index_params index_params;
-   *   // extend the IVF lists while building the index
-   *   index_params.add_data_on_build = true;
-   *   // create and fill the index from a [N, D] dataset
-   *   auto index = cuvs::neighbors::ivf_pq::build<int64_t>(res, index_params, dataset, N, D);
-   *   // Fetch the size of the fourth list
-   *   uint32_t size = index.get_list_size_in_bytes(3);
-   * @endcode
-   *
-   * @param[in] label list ID
-   */
-  inline auto get_list_size_in_bytes(uint32_t label) -> uint32_t
-  {
-    RAFT_EXPECTS(label < this->n_lists(),
-                 "Expected label to be less than number of lists in the index");
-    auto list_data = this->lists()[label]->data;
-    return list_data.size();
-  }
-
- private:
-  cuvs::distance::DistanceType metric_;
-  codebook_gen codebook_kind_;
-  uint32_t dim_;
-  uint32_t pq_bits_;
-  uint32_t pq_dim_;
-  bool conservative_memory_allocation_;
-
-  // Primary data members
-  std::vector<std::shared_ptr<list_data<IdxT>>> lists_;
-  raft::device_vector<uint32_t, uint32_t, raft::row_major> list_sizes_;
-  raft::device_mdarray<float, pq_centers_extents, raft::row_major> pq_centers_;
-  raft::device_matrix<float, uint32_t, raft::row_major> centers_;
-  raft::device_matrix<float, uint32_t, raft::row_major> centers_rot_;
-  raft::device_matrix<float, uint32_t, raft::row_major> rotation_matrix_;
-
-  // Computed members for accelerating search.
-  raft::device_vector<uint8_t*, uint32_t, raft::row_major> data_ptrs_;
-  raft::device_vector<IdxT*, uint32_t, raft::row_major> inds_ptrs_;
-  raft::host_vector<IdxT, uint32_t, raft::row_major> accum_sorted_sizes_;
-
-  /** Throw an error if the index content is inconsistent. */
-  void check_consistency()
-  {
-    RAFT_EXPECTS(pq_bits() >= 4 && pq_bits() <= 8,
-                 "`pq_bits` must be within closed range [4,8], but got %u.",
-                 pq_bits());
-    RAFT_EXPECTS((pq_bits() * pq_dim()) % 8 == 0,
-                 "`pq_bits * pq_dim` must be a multiple of 8, but got %u * %u = %u.",
-                 pq_bits(),
-                 pq_dim(),
-                 pq_bits() * pq_dim());
-  }
-
-  auto make_pq_centers_extents() -> pq_centers_extents
-  {
-    switch (codebook_kind()) {
-      case codebook_gen::PER_SUBSPACE:
-        return make_extents<uint32_t>(pq_dim(), pq_len(), pq_book_size());
-      case codebook_gen::PER_CLUSTER:
-        return make_extents<uint32_t>(n_lists(), pq_len(), pq_book_size());
-      default: RAFT_FAIL("Unreachable code");
-    }
-  }
-
-  static inline auto calculate_pq_dim(uint32_t dim) -> uint32_t
-  {
-    // If the dimensionality is large enough, we can reduce it to improve performance
-    if (dim >= 128) { dim /= 2; }
-    // Round it down to 32 to improve performance.
-    auto r = raft::round_down_safe<uint32_t>(dim, 32);
-    if (r > 0) return r;
-    // If the dimensionality is really low, round it to the closest power-of-two
-    r = 1;
-    while ((r << 1) <= dim) {
-      r = r << 1;
-    }
-    return r;
-  }
-};
-
-/** @} */
-
-}  // namespace cuvs::neighbors::ivf_pq
diff --git a/cpp/include/cuvs/neighbors/neighbors_types.hpp b/cpp/include/cuvs/neighbors/neighbors_types.hpp
deleted file mode 100644
index c6c30655c..000000000
--- a/cpp/include/cuvs/neighbors/neighbors_types.hpp
+++ /dev/null
@@ -1,63 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/mdspan_types.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::neighbors {
-
-/** A single batch of nearest neighbors in device memory */
-template <typename T, typename IdxT>
-class batch {
- public:
-  /** Create a new empty batch of data */
-  batch(raft::resources const& res, int64_t rows, int64_t cols)
-    : indices_(raft::make_device_matrix<IdxT, int64_t>(res, rows, cols)),
-      distances_(raft::make_device_matrix<T, int64_t>(res, rows, cols))
-  {
-  }
-
-  void resize(raft::resources const& res, int64_t rows, int64_t cols)
-  {
-    indices_   = raft::make_device_matrix<IdxT, int64_t>(res, rows, cols);
-    distances_ = raft::make_device_matrix<T, int64_t>(res, rows, cols);
-  }
-
-  /** Returns the indices for the batch */
-  raft::device_matrix_view<const IdxT, int64_t> indices() const
-  {
-    return raft::make_const_mdspan(indices_.view());
-  }
-  raft::device_matrix_view<IdxT, int64_t> indices() { return indices_.view(); }
-
-  /** Returns the distances for the batch */
-  raft::device_matrix_view<const T, int64_t> distances() const
-  {
-    return raft::make_const_mdspan(distances_.view());
-  }
-  raft::device_matrix_view<T, int64_t> distances() { return distances_.view(); }
-
-  /** Returns the size of the batch */
-  int64_t batch_size() const { return indices().extent(1); }
-
- protected:
-  raft::device_matrix<IdxT, int64_t> indices_;
-  raft::device_matrix<T, int64_t> distances_;
-};
-}  // namespace cuvs::neighbors
diff --git a/cpp/include/cuvs/neighbors/nn_descent.cuh b/cpp/include/cuvs/neighbors/nn_descent.cuh
deleted file mode 100644
index 0ed5cfd4a..000000000
--- a/cpp/include/cuvs/neighbors/nn_descent.cuh
+++ /dev/null
@@ -1,181 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "detail/nn_descent.cuh"
-
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/host_mdspan.hpp>
-
-namespace cuvs::neighbors::experimental::nn_descent {
-
-/**
- * @defgroup nn-descent CUDA gradient descent nearest neighbor
- * @{
- */
-
-/**
- * @brief Build nn-descent Index with dataset in device memory
- *
- * The following distance metrics are supported:
- * - L2
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors::experimental;
- *   // use default index parameters
- *   nn_descent::index_params index_params;
- *   // create and fill the index from a [N, D] raft::device_matrix_view dataset
- *   auto index = cagra::build(res, index_params, dataset);
- *   // index.graph() provides a raft::host_matrix_view of an
- *   // all-neighbors knn graph of dimensions [N, k] of the input
- *   // dataset
- * @endcode
- *
- * @tparam T data-type of the input dataset
- * @tparam IdxT data-type for the output index
- * @param[in] res raft::resources is an object mangaging resources
- * @param[in] params an instance of nn_descent::index_params that are parameters
- *               to run the nn-descent algorithm
- * @param[in] dataset raft::device_matrix_view input dataset expected to be located
- *                in device memory
- * @return index<IdxT> index containing all-neighbors knn graph in host memory
- */
-template <typename T, typename IdxT = uint32_t>
-index<IdxT> build(raft::resources const& res,
-                  index_params const& params,
-                  raft::device_matrix_view<const T, int64_t, raft::row_major> dataset)
-{
-  return detail::build<T, IdxT>(res, params, dataset);
-}
-
-/**
- * @brief Build nn-descent Index with dataset in device memory
- *
- * The following distance metrics are supported:
- * - L2
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors::experimental;
- *   // use default index parameters
- *   nn_descent::index_params index_params;
- *   // create and fill the index from a [N, D] raft::device_matrix_view dataset
- *   auto knn_graph = raft::make_host_matrix<uint32_t, int64_t>(N, D);
- *   auto index = nn_descent::index{res, knn_graph.view()};
- *   cagra::build(res, index_params, dataset, index);
- *   // index.graph() provides a raft::host_matrix_view of an
- *   // all-neighbors knn graph of dimensions [N, k] of the input
- *   // dataset
- * @endcode
- *
- * @tparam T data-type of the input dataset
- * @tparam IdxT data-type for the output index
- * @param res raft::resources is an object mangaging resources
- * @param[in] params an instance of nn_descent::index_params that are parameters
- *               to run the nn-descent algorithm
- * @param[in] dataset raft::device_matrix_view input dataset expected to be located
- *                in device memory
- * @param[out] idx cuvs::neighbors::experimental::nn_descentindex containing all-neighbors knn graph
- * in host memory
- */
-template <typename T, typename IdxT = uint32_t>
-void build(raft::resources const& res,
-           index_params const& params,
-           raft::device_matrix_view<const T, int64_t, raft::row_major> dataset,
-           index<IdxT>& idx)
-{
-  detail::build<T, IdxT>(res, params, dataset, idx);
-}
-
-/**
- * @brief Build nn-descent Index with dataset in host memory
- *
- * The following distance metrics are supported:
- * - L2
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors::experimental;
- *   // use default index parameters
- *   nn_descent::index_params index_params;
- *   // create and fill the index from a [N, D] raft::host_matrix_view dataset
- *   auto index = cagra::build(res, index_params, dataset);
- *   // index.graph() provides a raft::host_matrix_view of an
- *   // all-neighbors knn graph of dimensions [N, k] of the input
- *   // dataset
- * @endcode
- *
- * @tparam T data-type of the input dataset
- * @tparam IdxT data-type for the output index
- * @param res raft::resources is an object mangaging resources
- * @param[in] params an instance of nn_descent::index_params that are parameters
- *               to run the nn-descent algorithm
- * @param[in] dataset raft::host_matrix_view input dataset expected to be located
- *                in host memory
- * @return index<IdxT> index containing all-neighbors knn graph in host memory
- */
-template <typename T, typename IdxT = uint32_t>
-index<IdxT> build(raft::resources const& res,
-                  index_params const& params,
-                  raft::host_matrix_view<const T, int64_t, raft::row_major> dataset)
-{
-  return detail::build<T, IdxT>(res, params, dataset);
-}
-
-/**
- * @brief Build nn-descent Index with dataset in host memory
- *
- * The following distance metrics are supported:
- * - L2
- *
- * Usage example:
- * @code{.cpp}
- *   using namespace cuvs::neighbors::experimental;
- *   // use default index parameters
- *   nn_descent::index_params index_params;
- *   // create and fill the index from a [N, D] raft::host_matrix_view dataset
- *   auto knn_graph = raft::make_host_matrix<uint32_t, int64_t>(N, D);
- *   auto index = nn_descent::index{res, knn_graph.view()};
- *   cagra::build(res, index_params, dataset, index);
- *   // index.graph() provides a raft::host_matrix_view of an
- *   // all-neighbors knn graph of dimensions [N, k] of the input
- *   // dataset
- * @endcode
- *
- * @tparam T data-type of the input dataset
- * @tparam IdxT data-type for the output index
- * @param[in] res raft::resources is an object mangaging resources
- * @param[in] params an instance of nn_descent::index_params that are parameters
- *               to run the nn-descent algorithm
- * @param[in] dataset raft::host_matrix_view input dataset expected to be located
- *                in host memory
- * @param[out] idx cuvs::neighbors::experimental::nn_descentindex containing all-neighbors knn graph
- * in host memory
- */
-template <typename T, typename IdxT = uint32_t>
-void build(raft::resources const& res,
-           index_params const& params,
-           raft::host_matrix_view<const T, int64_t, raft::row_major> dataset,
-           index<IdxT>& idx)
-{
-  detail::build<T, IdxT>(res, params, dataset, idx);
-}
-
-/** @} */  // end group nn-descent
-
-}  // namespace cuvs::neighbors::experimental::nn_descent
diff --git a/cpp/include/cuvs/neighbors/nn_descent_types.hpp b/cpp/include/cuvs/neighbors/nn_descent_types.hpp
deleted file mode 100644
index 66991755c..000000000
--- a/cpp/include/cuvs/neighbors/nn_descent_types.hpp
+++ /dev/null
@@ -1,147 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "ann_types.hpp"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/host_mdarray.hpp>
-#include <raft/core/host_mdspan.hpp>
-#include <raft/core/mdspan_types.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::neighbors::experimental::nn_descent {
-/**
- * @ingroup nn_descent
- * @{
- */
-
-/**
- * @brief Parameters used to build an nn-descent index
- *
- * `graph_degree`: For an input dataset of dimensions (N, D),
- * determines the final dimensions of the all-neighbors knn graph
- * which turns out to be of dimensions (N, graph_degree)
- * `intermediate_graph_degree`: Internally, nn-descent builds an
- * all-neighbors knn graph of dimensions (N, intermediate_graph_degree)
- * before selecting the final `graph_degree` neighbors. It's recommended
- * that `intermediate_graph_degree` >= 1.5 * graph_degree
- * `max_iterations`: The number of iterations that nn-descent will refine
- * the graph for. More iterations produce a better quality graph at cost of performance
- * `termination_threshold`: The delta at which nn-descent will terminate its iterations
- *
- */
-struct index_params : ann::index_params {
-  size_t graph_degree              = 64;      // Degree of output graph.
-  size_t intermediate_graph_degree = 128;     // Degree of input graph for pruning.
-  size_t max_iterations            = 20;      // Number of nn-descent iterations.
-  float termination_threshold      = 0.0001;  // Termination threshold of nn-descent.
-};
-
-/**
- * @brief nn-descent Build an nn-descent index
- * The index contains an all-neighbors graph of the input dataset
- * stored in host memory of dimensions (n_rows, n_cols)
- *
- * @tparam IdxT dtype to be used for constructing knn-graph
- */
-template <typename IdxT>
-struct index : ann::index {
- public:
-  /**
-   * @brief Construct a new index object
-   *
-   * This constructor creates an nn-descent index which is a knn-graph in host memory.
-   * The type of the knn-graph is a dense raft::host_matrix and dimensions are
-   * (n_rows, n_cols).
-   *
-   * @param res raft::resources is an object mangaging resources
-   * @param n_rows number of rows in knn-graph
-   * @param n_cols number of cols in knn-graph
-   */
-  index(raft::resources const& res, int64_t n_rows, int64_t n_cols)
-    : ann::index(),
-      res_{res},
-      metric_{cuvs::distance::DistanceType::L2Expanded},
-      graph_{raft::make_host_matrix<IdxT, int64_t, raft::row_major>(n_rows, n_cols)},
-      graph_view_{graph_.view()}
-  {
-  }
-
-  /**
-   * @brief Construct a new index object
-   *
-   * This constructor creates an nn-descent index using a user allocated host memory knn-graph.
-   * The type of the knn-graph is a dense raft::host_matrix and dimensions are
-   * (n_rows, n_cols).
-   *
-   * @param res raft::resources is an object mangaging resources
-   * @param graph_view raft::host_matrix_view<IdxT, int64_t, raft::row_major> for storing knn-graph
-   */
-  index(raft::resources const& res,
-        raft::host_matrix_view<IdxT, int64_t, raft::row_major> graph_view)
-    : ann::index(),
-      res_{res},
-      metric_{cuvs::distance::DistanceType::L2Expanded},
-      graph_{raft::make_host_matrix<IdxT, int64_t, raft::row_major>(0, 0)},
-      graph_view_{graph_view}
-  {
-  }
-
-  /** Distance metric used for clustering. */
-  [[nodiscard]] constexpr inline auto metric() const noexcept -> cuvs::distance::DistanceType
-  {
-    return metric_;
-  }
-
-  // /** Total length of the index (number of vectors). */
-  [[nodiscard]] constexpr inline auto size() const noexcept -> IdxT
-  {
-    return graph_view_.extent(0);
-  }
-
-  /** Graph degree */
-  [[nodiscard]] constexpr inline auto graph_degree() const noexcept -> uint32_t
-  {
-    return graph_view_.extent(1);
-  }
-
-  /** neighborhood graph [size, graph-degree] */
-  [[nodiscard]] inline auto graph() noexcept -> host_matrix_view<IdxT, int64_t, raft::row_major>
-  {
-    return graph_view_;
-  }
-
-  // Don't allow copying the index for performance reasons (try avoiding copying data)
-  index(const index&)                    = delete;
-  index(index&&)                         = default;
-  auto operator=(const index&) -> index& = delete;
-  auto operator=(index&&) -> index&      = default;
-  ~index()                               = default;
-
- private:
-  raft::resources const& res_;
-  cuvs::distance::DistanceType metric_;
-  raft::host_matrix<IdxT, int64_t, raft::row_major> graph_;  // graph to return for non-int IdxT
-  raft::host_matrix_view<IdxT, int64_t, raft::row_major>
-    graph_view_;  // view of graph for user provided matrix
-};
-
-/** @} */
-
-}  // namespace cuvs::neighbors::experimental::nn_descent
diff --git a/cpp/include/cuvs/neighbors/refine-ext.cuh b/cpp/include/cuvs/neighbors/refine-ext.cuh
deleted file mode 100644
index 49bfd7301..000000000
--- a/cpp/include/cuvs/neighbors/refine-ext.cuh
+++ /dev/null
@@ -1,78 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cstdint>  // int64_t
-
-#include <cuvs/distance/distance_types.hpp>  // cuvs::distance::DistanceType
-#include <raft/core/device_mdspan.hpp>       // raft::device_matrix_view
-#include <raft/core/host_mdspan.hpp>         // // raft::host_matrix_view
-#include <raft/core/resources.hpp>           // raft::resources
-#include <raft/util/raft_explicit.hpp>       // RAFT_EXPLICIT
-
-#ifdef RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-namespace cuvs::neighbors {
-
-template <typename idx_t, typename data_t, typename distance_t, typename matrix_idx>
-void refine(raft::resources const& handle,
-            raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,
-            raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> queries,
-            raft::device_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates,
-            raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,
-            raft::device_matrix_view<distance_t, matrix_idx, raft::row_major> distances,
-            cuvs::distance::DistanceType metric = distance::DistanceType::L2Unexpanded)
-  RAFT_EXPLICIT;
-
-template <typename idx_t, typename data_t, typename distance_t, typename matrix_idx>
-void refine(raft::resources const& handle,
-            raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,
-            raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> queries,
-            raft::host_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates,
-            raft::host_matrix_view<idx_t, matrix_idx, raft::row_major> indices,
-            raft::host_matrix_view<distance_t, matrix_idx, raft::row_major> distances,
-            cuvs::distance::DistanceType metric = distance::DistanceType::L2Unexpanded)
-  RAFT_EXPLICIT;
-
-}  // namespace cuvs::neighbors
-
-#endif  // RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-#define instantiate_raft_neighbors_refine(idx_t, data_t, distance_t, matrix_idx)            \
-  extern template void cuvs::neighbors::refine<idx_t, data_t, distance_t, matrix_idx>(      \
-    raft::resources const& handle,                                                          \
-    raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,            \
-    raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> queries,            \
-    raft::device_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates, \
-    raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                   \
-    raft::device_matrix_view<distance_t, matrix_idx, raft::row_major> distances,            \
-    cuvs::distance::DistanceType metric);                                                   \
-                                                                                            \
-  extern template void cuvs::neighbors::refine<idx_t, data_t, distance_t, matrix_idx>(      \
-    raft::resources const& handle,                                                          \
-    raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,              \
-    raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> queries,              \
-    raft::host_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates,   \
-    raft::host_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                     \
-    raft::host_matrix_view<distance_t, matrix_idx, raft::row_major> distances,              \
-    cuvs::distance::DistanceType metric);
-
-instantiate_raft_neighbors_refine(int64_t, float, float, int64_t);
-instantiate_raft_neighbors_refine(int64_t, int8_t, float, int64_t);
-instantiate_raft_neighbors_refine(int64_t, uint8_t, float, int64_t);
-
-#undef instantiate_raft_neighbors_refine
diff --git a/cpp/include/cuvs/neighbors/refine-inl.cuh b/cpp/include/cuvs/neighbors/refine-inl.cuh
deleted file mode 100644
index 1a9b9fe34..000000000
--- a/cpp/include/cuvs/neighbors/refine-inl.cuh
+++ /dev/null
@@ -1,104 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/neighbors/detail/refine.cuh>
-#include <cuvs/spatial/knn/detail/ann_utils.cuh>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/host_mdspan.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::neighbors {
-
-/**
- * @defgroup ann_refine Approximate Nearest Neighbors Refinement
- * @{
- */
-
-/**
- * @brief Refine nearest neighbor search.
- *
- * Refinement is an operation that follows an approximate NN search. The approximate search has
- * already selected n_candidates neighbor candidates for each query. We narrow it down to k
- * neighbors. For each query, we calculate the exact distance between the query and its
- * n_candidates neighbor candidate, and select the k nearest ones.
- *
- * The k nearest neighbors and distances are returned.
- *
- * Example usage
- * @code{.cpp}
- *   using namespace cuvs::neighbors;
- *   // use default index parameters
- *   ivf_pq::index_params index_params;
- *   // create and fill the index from a [N, D] dataset
- *   auto index = ivf_pq::build(handle, index_params, dataset, N, D);
- *   // use default search parameters
- *   ivf_pq::search_params search_params;
- *   // search m = 4 * k nearest neighbours for each of the N queries
- *   ivf_pq::search(handle, search_params, index, queries, N, 4 * k, neighbor_candidates,
- *                  out_dists_tmp);
- *   // refine it to the k nearest one
- *   refine(handle, dataset, queries, neighbor_candidates, out_indices, out_dists,
- *           index.metric());
- * @endcode
- *
- *
- * @param[in] handle the raft handle
- * @param[in] dataset device matrix that stores the dataset [n_rows, dims]
- * @param[in] queries device matrix of the queries [n_queris, dims]
- * @param[in] neighbor_candidates indices of candidate vectors [n_queries, n_candidates], where
- *   n_candidates >= k
- * @param[out] indices device matrix that stores the refined indices [n_queries, k]
- * @param[out] distances device matrix that stores the refined distances [n_queries, k]
- * @param[in] metric distance metric to use. Euclidean (L2) is used by default
- */
-template <typename idx_t, typename data_t, typename distance_t, typename matrix_idx>
-void refine(raft::resources const& handle,
-            raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,
-            raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> queries,
-            raft::device_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates,
-            raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,
-            raft::device_matrix_view<distance_t, matrix_idx, raft::row_major> distances,
-            distance::DistanceType metric = distance::DistanceType::L2Unexpanded)
-{
-  detail::refine_device(handle, dataset, queries, neighbor_candidates, indices, distances, metric);
-}
-
-/** Same as above, but all input and out data is in host memory.
- * @param[in] handle the raft handle
- * @param[in] dataset host matrix that stores the dataset [n_rows, dims]
- * @param[in] queries host matrix of the queries [n_queris, dims]
- * @param[in] neighbor_candidates host matrix with indices of candidate vectors [n_queries,
- *   n_candidates], where n_candidates >= k
- * @param[out] indices host matrix that stores the refined indices [n_queries, k]
- * @param[out] distances host matrix that stores the refined distances [n_queries, k]
- * @param[in] metric distance metric to use. Euclidean (L2) is used by default
- */
-template <typename idx_t, typename data_t, typename distance_t, typename matrix_idx>
-void refine(raft::resources const& handle,
-            raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,
-            raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> queries,
-            raft::host_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates,
-            raft::host_matrix_view<idx_t, matrix_idx, raft::row_major> indices,
-            raft::host_matrix_view<distance_t, matrix_idx, raft::row_major> distances,
-            distance::DistanceType metric = distance::DistanceType::L2Unexpanded)
-{
-  detail::refine_host(dataset, queries, neighbor_candidates, indices, distances, metric);
-}
-
-/** @} */  // end group ann_refine
-}  // namespace cuvs::neighbors
diff --git a/cpp/include/cuvs/neighbors/refine.cuh b/cpp/include/cuvs/neighbors/refine.cuh
deleted file mode 100644
index 15f2b0292..000000000
--- a/cpp/include/cuvs/neighbors/refine.cuh
+++ /dev/null
@@ -1,24 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#ifndef RAFT_EXPLICIT_INSTANTIATE_ONLY
-#include "refine-inl.cuh"
-#endif
-
-#ifdef RAFT_COMPILED
-#include "refine-ext.cuh"
-#endif
diff --git a/cpp/include/cuvs/neighbors/sample_filter.cuh b/cpp/include/cuvs/neighbors/sample_filter.cuh
deleted file mode 100644
index 1d3fc618f..000000000
--- a/cpp/include/cuvs/neighbors/sample_filter.cuh
+++ /dev/null
@@ -1,49 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cstddef>
-#include <cstdint>
-
-#include <raft/core/bitset.cuh>
-
-namespace cuvs::neighbors::filtering {
-/**
- * @brief Filter an index with a bitset
- *
- * @tparam index_t Indexing type
- */
-template <typename bitset_t, typename index_t>
-struct bitset_filter {
-  // View of the bitset to use as a filter
-  const raft::core::bitset_view<bitset_t, index_t> bitset_view_;
-
-  bitset_filter(const raft::core::bitset_view<bitset_t, index_t> bitset_for_filtering)
-    : bitset_view_{bitset_for_filtering}
-  {
-  }
-  inline _RAFT_HOST_DEVICE bool operator()(
-    // query index
-    const uint32_t query_ix,
-    // the index of the current sample
-    const uint32_t sample_ix) const
-  {
-    return bitset_view_.test(sample_ix);
-  }
-};
-
-}  // namespace cuvs::neighbors::filtering
diff --git a/cpp/include/cuvs/neighbors/sample_filter_types.hpp b/cpp/include/cuvs/neighbors/sample_filter_types.hpp
deleted file mode 100644
index 517c447f0..000000000
--- a/cpp/include/cuvs/neighbors/sample_filter_types.hpp
+++ /dev/null
@@ -1,175 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cstddef>
-#include <cstdint>
-
-#include <raft/core/detail/macros.hpp>
-
-namespace cuvs::neighbors::filtering {
-
-/* A filter that filters nothing. This is the default behavior. */
-struct none_ivf_sample_filter {
-  inline _RAFT_HOST_DEVICE bool operator()(
-    // query index
-    const uint32_t query_ix,
-    // the current inverted list index
-    const uint32_t cluster_ix,
-    // the index of the current sample inside the current inverted list
-    const uint32_t sample_ix) const
-  {
-    return true;
-  }
-};
-
-/* A filter that filters nothing. This is the default behavior. */
-struct none_cagra_sample_filter {
-  inline _RAFT_HOST_DEVICE bool operator()(
-    // query index
-    const uint32_t query_ix,
-    // the index of the current sample
-    const uint32_t sample_ix) const
-  {
-    return true;
-  }
-};
-
-template <typename filter_t, typename = void>
-struct takes_three_args : std::false_type {};
-template <typename filter_t>
-struct takes_three_args<
-  filter_t,
-  std::void_t<decltype(std::declval<filter_t>()(uint32_t{}, uint32_t{}, uint32_t{}))>>
-  : std::true_type {};
-
-/**
- * @brief Filter used to convert the cluster index and sample index
- * of an IVF search into a sample index. This can be used as an
- * intermediate filter.
- *
- * @tparam index_t Indexing type
- * @tparam filter_t
- */
-template <typename index_t, typename filter_t>
-struct ivf_to_sample_filter {
-  const index_t* const* inds_ptrs_;
-  const filter_t next_filter_;
-
-  ivf_to_sample_filter(const index_t* const* inds_ptrs, const filter_t next_filter)
-    : inds_ptrs_{inds_ptrs}, next_filter_{next_filter}
-  {
-  }
-
-  /** If the original filter takes three arguments, then don't modify the arguments.
-   * If the original filter takes two arguments, then we are using `inds_ptr_` to obtain the sample
-   * index.
-   */
-  inline _RAFT_HOST_DEVICE bool operator()(
-    // query index
-    const uint32_t query_ix,
-    // the current inverted list index
-    const uint32_t cluster_ix,
-    // the index of the current sample inside the current inverted list
-    const uint32_t sample_ix) const
-  {
-    if constexpr (takes_three_args<filter_t>::value) {
-      return next_filter_(query_ix, cluster_ix, sample_ix);
-    } else {
-      return next_filter_(query_ix, inds_ptrs_[cluster_ix][sample_ix]);
-    }
-  }
-};
-/**
- * If the filtering depends on the index of a sample, then the following
- * filter template can be used:
- *
- * template <typename IdxT>
- * struct index_ivf_sample_filter {
- *   using index_type = IdxT;
- *
- *   const index_type* const* inds_ptr = nullptr;
- *
- *   index_ivf_sample_filter() {}
- *   index_ivf_sample_filter(const index_type* const* _inds_ptr)
- *       : inds_ptr{_inds_ptr} {}
- *   index_ivf_sample_filter(const index_ivf_sample_filter&) = default;
- *   index_ivf_sample_filter(index_ivf_sample_filter&&) = default;
- *   index_ivf_sample_filter& operator=(const index_ivf_sample_filter&) = default;
- *   index_ivf_sample_filter& operator=(index_ivf_sample_filter&&) = default;
- *
- *   inline _RAFT_HOST_DEVICE bool operator()(
- *       const uint32_t query_ix,
- *       const uint32_t cluster_ix,
- *       const uint32_t sample_ix) const {
- *     index_type database_idx = inds_ptr[cluster_ix][sample_ix];
- *
- *     // return true or false, depending on the database_idx
- *     return true;
- *   }
- * };
- *
- * Initialize it as:
- *   using filter_type = index_ivf_sample_filter<idx_t>;
- *   filter_type filter(raft_ivfpq_index.inds_ptrs().data_handle());
- *
- * Use it as:
- *   cuvs::neighbors::ivf_pq::search_with_filtering<data_t, idx_t, filter_type>(
- *     ...regular parameters here...,
- *     filter
- *   );
- *
- * Another example would be the following filter that greenlights samples according
- * to a contiguous bit mask vector.
- *
- * template <typename IdxT>
- * struct bitmask_ivf_sample_filter {
- *   using index_type = IdxT;
- *
- *   const index_type* const* inds_ptr = nullptr;
- *   const uint64_t* const bit_mask_ptr = nullptr;
- *   const int64_t bit_mask_stride_64 = 0;
- *
- *   bitmask_ivf_sample_filter() {}
- *   bitmask_ivf_sample_filter(
- *       const index_type* const* _inds_ptr,
- *       const uint64_t* const _bit_mask_ptr,
- *       const int64_t _bit_mask_stride_64)
- *       : inds_ptr{_inds_ptr},
- *         bit_mask_ptr{_bit_mask_ptr},
- *         bit_mask_stride_64{_bit_mask_stride_64} {}
- *   bitmask_ivf_sample_filter(const bitmask_ivf_sample_filter&) = default;
- *   bitmask_ivf_sample_filter(bitmask_ivf_sample_filter&&) = default;
- *   bitmask_ivf_sample_filter& operator=(const bitmask_ivf_sample_filter&) = default;
- *   bitmask_ivf_sample_filter& operator=(bitmask_ivf_sample_filter&&) = default;
- *
- *   inline _RAFT_HOST_DEVICE bool operator()(
- *       const uint32_t query_ix,
- *       const uint32_t cluster_ix,
- *       const uint32_t sample_ix) const {
- *     const index_type database_idx = inds_ptr[cluster_ix][sample_ix];
- *     const uint64_t bit_mask_element =
- *         bit_mask_ptr[query_ix * bit_mask_stride_64 + database_idx / 64];
- *     const uint64_t masked_bool =
- *         bit_mask_element & (1ULL << (uint64_t)(database_idx % 64));
- *     const bool is_bit_set = (masked_bool != 0);
- *
- *     return is_bit_set;
- *   }
- * };
- */
-}  // namespace cuvs::neighbors::filtering
diff --git a/cpp/include/cuvs/neighbors/specializations.cuh b/cpp/include/cuvs/neighbors/specializations.cuh
deleted file mode 100644
index ed0b6848a..000000000
--- a/cpp/include/cuvs/neighbors/specializations.cuh
+++ /dev/null
@@ -1,22 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#pragma message(                                            \
-    __FILE__                                                \
-    " is deprecated and will be removed."                   \
-    " Including specializations is not necessary any more." \
-    " For more information, see: https://docs.rapids.ai/api/raft/nightly/using_libraft.html")
diff --git a/cpp/include/cuvs/spatial/knn/ann.cuh b/cpp/include/cuvs/spatial/knn/ann.cuh
deleted file mode 100644
index 99f5f12eb..000000000
--- a/cpp/include/cuvs/spatial/knn/ann.cuh
+++ /dev/null
@@ -1,83 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "ann_common.h"
-#include "detail/ann_quantized.cuh"
-
-#include <raft/core/nvtx.hpp>
-
-namespace cuvs::spatial::knn {
-
-/**
- * @brief Flat C++ API function to build an approximate nearest neighbors index
- * from an index array and a set of parameters.
- *
- * @param[in] handle RAFT handle
- * @param[out] index index to be built
- * @param[in] params parametrization of the index to be built
- * @param[in] metric distance metric to use. Euclidean (L2) is used by default
- * @param[in] metricArg metric argument
- * @param[in] index_array the index array to build the index with
- * @param[in] n number of rows in the index array
- * @param[in] D the dimensionality of the index array
- */
-template <typename T = float, typename value_idx = int>
-[[deprecated("Consider using new-style cuvs::spatial::knn::*::build functions")]] inline void
-approx_knn_build_index(raft::resources& handle,
-                       cuvs::spatial::knn::knnIndex* index,
-                       knnIndexParam* params,
-                       cuvs::distance::DistanceType metric,
-                       float metricArg,
-                       T* index_array,
-                       value_idx n,
-                       value_idx D)
-{
-  raft::common::nvtx::range<raft::common::nvtx::domain::raft> fun_scope(
-    "legacy approx_knn_build_index(n_rows = %u, dim = %u)", n, D);
-  detail::approx_knn_build_index(handle, index, params, metric, metricArg, index_array, n, D);
-}
-
-/**
- * @brief Flat C++ API function to perform an approximate nearest neighbors
- * search from previously built index and a query array
- *
- * @param[in] handle RAFT handle
- * @param[out] distances distances of the nearest neighbors toward
- *                       their query point
- * @param[out] indices indices of the nearest neighbors
- * @param[in] index index to perform a search with
- * @param[in] k the number of nearest neighbors to search for
- * @param[in] query_array the query to perform a search with
- * @param[in] n number of rows in the query array
- */
-template <typename T = float, typename value_idx = int>
-[[deprecated("Consider using new-style cuvs::spatial::knn::*::search functions")]] inline void
-approx_knn_search(raft::resources& handle,
-                  float* distances,
-                  int64_t* indices,
-                  cuvs::spatial::knn::knnIndex* index,
-                  value_idx k,
-                  T* query_array,
-                  value_idx n)
-{
-  raft::common::nvtx::range<raft::common::nvtx::domain::raft> fun_scope(
-    "legacy approx_knn_search(k = %u, n_queries = %u)", k, n);
-  detail::approx_knn_search(handle, distances, indices, index, k, query_array, n);
-}
-
-}  // namespace cuvs::spatial::knn
diff --git a/cpp/include/cuvs/spatial/knn/ann_common.h b/cpp/include/cuvs/spatial/knn/ann_common.h
deleted file mode 100644
index b3954a2d3..000000000
--- a/cpp/include/cuvs/spatial/knn/ann_common.h
+++ /dev/null
@@ -1,103 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma message(                                              \
-    __FILE__                                                  \
-    " is deprecated and will be removed in a future release." \
-    " Please use the other approximate KNN implementations defined in spatial/knn/*.")
-
-#pragma once
-
-#include "detail/processing.hpp"
-#include "ivf_flat_types.hpp"
-#include <cuvs/neighbors/ivf_pq_types.hpp>
-
-#include <cuvs/distance/distance_types.hpp>
-
-namespace cuvs {
-namespace spatial {
-namespace knn {
-
-struct knnIndex {
-  cuvs::distance::DistanceType metric;
-  float metricArg;
-  int nprobe;
-  std::unique_ptr<MetricProcessor<float>> metric_processor;
-
-  std::unique_ptr<const ivf_flat::index<float, int64_t>> ivf_flat_float_;
-  std::unique_ptr<const ivf_flat::index<uint8_t, int64_t>> ivf_flat_uint8_t_;
-  std::unique_ptr<const ivf_flat::index<int8_t, int64_t>> ivf_flat_int8_t_;
-
-  std::unique_ptr<const cuvs::neighbors::ivf_pq::index<int64_t>> ivf_pq;
-
-  int device;
-
-  template <typename T, typename IdxT>
-  auto ivf_flat() -> std::unique_ptr<const ivf_flat::index<T, IdxT>>&;
-};
-
-template <>
-inline auto knnIndex::ivf_flat<float, int64_t>()
-  -> std::unique_ptr<const ivf_flat::index<float, int64_t>>&
-{
-  return ivf_flat_float_;
-}
-
-template <>
-inline auto knnIndex::ivf_flat<uint8_t, int64_t>()
-  -> std::unique_ptr<const ivf_flat::index<uint8_t, int64_t>>&
-{
-  return ivf_flat_uint8_t_;
-}
-
-template <>
-inline auto knnIndex::ivf_flat<int8_t, int64_t>()
-  -> std::unique_ptr<const ivf_flat::index<int8_t, int64_t>>&
-{
-  return ivf_flat_int8_t_;
-}
-
-struct knnIndexParam {
-  virtual ~knnIndexParam() {}
-};
-
-struct IVFParam : knnIndexParam {
-  int nlist;
-  int nprobe;
-};
-
-struct IVFFlatParam : IVFParam {};
-
-struct IVFPQParam : IVFParam {
-  int M;
-  int n_bits;
-  bool usePrecomputedTables;
-};
-
-inline auto from_legacy_index_params(const IVFFlatParam& legacy,
-                                     cuvs::distance::DistanceType metric,
-                                     float metric_arg)
-{
-  ivf_flat::index_params params;
-  params.metric     = metric;
-  params.metric_arg = metric_arg;
-  params.n_lists    = legacy.nlist;
-  return params;
-}
-
-};  // namespace knn
-};  // namespace spatial
-};  // namespace cuvs
diff --git a/cpp/include/cuvs/spatial/knn/ann_types.hpp b/cpp/include/cuvs/spatial/knn/ann_types.hpp
deleted file mode 100644
index 3c2ae75ad..000000000
--- a/cpp/include/cuvs/spatial/knn/ann_types.hpp
+++ /dev/null
@@ -1,45 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/distance/distance_types.hpp>
-
-namespace cuvs::spatial::knn {
-
-/** The base for approximate KNN index structures. */
-struct index {};
-
-/** The base for KNN index parameters. */
-struct index_params {
-  /** Distance type. */
-  cuvs::distance::DistanceType metric = distance::DistanceType::L2Expanded;
-  /** The argument used by some distance metrics. */
-  float metric_arg = 2.0f;
-  /**
-   * Whether to add the dataset content to the index, i.e.:
-   *
-   *  - `true` means the index is filled with the dataset vectors and ready to search after calling
-   * `build`.
-   *  - `false` means `build` only trains the underlying model (e.g. quantizer or clustering), but
-   * the index is left empty; you'd need to call `extend` on the index afterwards to populate it.
-   */
-  bool add_data_on_build = true;
-};
-
-struct search_params {};
-
-};  // namespace cuvs::spatial::knn
diff --git a/cpp/include/cuvs/spatial/knn/ball_cover.cuh b/cpp/include/cuvs/spatial/knn/ball_cover.cuh
deleted file mode 100644
index 605387443..000000000
--- a/cpp/include/cuvs/spatial/knn/ball_cover.cuh
+++ /dev/null
@@ -1,70 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/**
- * This file is deprecated and will be removed in release 22.06.
- * Please use the cuh version instead.
- */
-
-/**
- * DISCLAIMER: this file is deprecated: use epsilon_neighborhood.cuh instead
- */
-
-#pragma once
-
-#pragma message(__FILE__                                                    \
-                  " is deprecated and will be removed in a future release." \
-                  " Please use the cuvs::neighbors version instead.")
-
-#include <cuvs/neighbors/ball_cover.cuh>
-#include <cuvs/spatial/knn/ball_cover_types.hpp>
-
-namespace cuvs::spatial::knn {
-
-template <typename idx_t, typename value_t, typename int_t, typename matrix_idx_t>
-void rbc_build_index(raft::resources const& handle,
-                     BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index)
-{
-  cuvs::neighbors::ball_cover::build_index(handle, index);
-}
-
-template <typename idx_t, typename value_t, typename int_t, typename matrix_idx_t>
-void rbc_all_knn_query(raft::resources const& handle,
-                       BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index,
-                       int_t k,
-                       idx_t* inds,
-                       value_t* dists,
-                       bool perform_post_filtering = true,
-                       float weight                = 1.0)
-{
-  cuvs::neighbors::ball_cover::all_knn_query(
-    handle, index, k, inds, dists, perform_post_filtering, weight);
-}
-
-template <typename idx_t, typename value_t, typename int_t>
-void rbc_knn_query(raft::resources const& handle,
-                   const BallCoverIndex<idx_t, value_t, int_t>& index,
-                   int_t k,
-                   const value_t* query,
-                   int_t n_query_pts,
-                   idx_t* inds,
-                   value_t* dists,
-                   bool perform_post_filtering = true,
-                   float weight                = 1.0)
-{
-  cuvs::neighbors::ball_cover::knn_query(
-    handle, index, k, query, n_query_pts, inds, dists, perform_post_filtering, weight);
-}
-}  // namespace cuvs::spatial::knn
diff --git a/cpp/include/cuvs/spatial/knn/ball_cover_types.hpp b/cpp/include/cuvs/spatial/knn/ball_cover_types.hpp
deleted file mode 100644
index 4b36c6cb2..000000000
--- a/cpp/include/cuvs/spatial/knn/ball_cover_types.hpp
+++ /dev/null
@@ -1,37 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/**
- * This file is deprecated and will be removed in release 22.06.
- * Please use the cuh version instead.
- */
-
-/**
- * DISCLAIMER: this file is deprecated: use epsilon_neighborhood.cuh instead
- */
-
-#pragma once
-
-#pragma message(__FILE__                                                    \
-                  " is deprecated and will be removed in a future release." \
-                  " Please use the cuvs::neighbors version instead.")
-
-#include <cuvs/neighbors/ball_cover_types.hpp>
-
-namespace cuvs::spatial::knn {
-
-using cuvs::neighbors::ball_cover::BallCoverIndex;
-
-}  // namespace cuvs::spatial::knn
diff --git a/cpp/include/cuvs/spatial/knn/common.hpp b/cpp/include/cuvs/spatial/knn/common.hpp
deleted file mode 100644
index 1b8780c42..000000000
--- a/cpp/include/cuvs/spatial/knn/common.hpp
+++ /dev/null
@@ -1,23 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/**
- * This file is deprecated and will be removed in a future release.
- * Please use the ann_types.hpp version instead.
- */
-
-#pragma once
-
-#include <cuvs/spatial/knn/ann_types.hpp>
diff --git a/cpp/include/cuvs/spatial/knn/detail/ann_quantized.cuh b/cpp/include/cuvs/spatial/knn/detail/ann_quantized.cuh
deleted file mode 100644
index fef621f85..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/ann_quantized.cuh
+++ /dev/null
@@ -1,147 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "../ann_common.h"
-#include "../ivf_flat.cuh"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include "processing.cuh"
-#include <raft/core/operators.hpp>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-
-#include <cuvs/distance/distance.cuh>
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/ivf_pq.cuh>
-#include <raft/label/classlabels.cuh>
-
-#include <raft/core/device_mdspan.hpp>
-
-#include <rmm/cuda_stream_view.hpp>
-
-#include <thrust/iterator/transform_iterator.h>
-
-namespace cuvs::spatial::knn::detail {
-
-template <typename T = float, typename IntType = int>
-void approx_knn_build_index(raft::resources const& handle,
-                            knnIndex* index,
-                            knnIndexParam* params,
-                            cuvs::distance::DistanceType metric,
-                            float metricArg,
-                            T* index_array,
-                            IntType n,
-                            IntType D)
-{
-  auto stream      = resource::get_cuda_stream(handle);
-  index->metric    = metric;
-  index->metricArg = metricArg;
-  if (dynamic_cast<const IVFParam*>(params)) {
-    index->nprobe = dynamic_cast<const IVFParam*>(params)->nprobe;
-  }
-  auto ivf_ft_pams = dynamic_cast<IVFFlatParam*>(params);
-  auto ivf_pq_pams = dynamic_cast<IVFPQParam*>(params);
-
-  if constexpr (std::is_same_v<T, float>) {
-    index->metric_processor = create_processor<float>(metric, n, D, 0, false, stream);
-    // For cosine/correlation distance, the metric processor translates distance
-    // to inner product via pre/post processing - pass the translated metric to
-    // ANN index
-    if (metric == cuvs::distance::DistanceType::CosineExpanded ||
-        metric == cuvs::distance::DistanceType::CorrelationExpanded) {
-      metric = index->metric = cuvs::distance::DistanceType::InnerProduct;
-    }
-  }
-  if constexpr (std::is_same_v<T, float>) { index->metric_processor->preprocess(index_array); }
-
-  if (ivf_ft_pams) {
-    auto new_params               = from_legacy_index_params(*ivf_ft_pams, metric, metricArg);
-    index->ivf_flat<T, int64_t>() = std::make_unique<const ivf_flat::index<T, int64_t>>(
-      ivf_flat::build(handle, new_params, index_array, int64_t(n), D));
-  } else if (ivf_pq_pams) {
-    neighbors::ivf_pq::index_params params;
-    params.metric     = metric;
-    params.metric_arg = metricArg;
-    params.n_lists    = ivf_pq_pams->nlist;
-    params.pq_bits    = ivf_pq_pams->n_bits;
-    params.pq_dim     = ivf_pq_pams->M;
-    // TODO: handle ivf_pq_pams.usePrecomputedTables ?
-
-    auto index_view = raft::make_device_matrix_view<const T, int64_t>(index_array, n, D);
-    index->ivf_pq   = std::make_unique<const neighbors::ivf_pq::index<int64_t>>(
-      neighbors::ivf_pq::build(handle, params, index_view));
-  } else {
-    RAFT_FAIL("Unrecognized index type.");
-  }
-
-  if constexpr (std::is_same_v<T, float>) { index->metric_processor->revert(index_array); }
-}
-
-template <typename T = float, typename IntType = int>
-void approx_knn_search(raft::resources const& handle,
-                       float* distances,
-                       int64_t* indices,
-                       knnIndex* index,
-                       IntType k,
-                       T* query_array,
-                       IntType n)
-{
-  if constexpr (std::is_same_v<T, float>) {
-    index->metric_processor->preprocess(query_array);
-    index->metric_processor->set_num_queries(k);
-  }
-
-  // search
-  if (index->ivf_flat<T, int64_t>()) {
-    ivf_flat::search_params params;
-    params.n_probes = index->nprobe;
-    ivf_flat::search(
-      handle, params, *(index->ivf_flat<T, int64_t>()), query_array, n, k, indices, distances);
-  } else if (index->ivf_pq) {
-    neighbors::ivf_pq::search_params params;
-    params.n_probes = index->nprobe;
-
-    auto query_view =
-      raft::make_device_matrix_view<const T, uint32_t>(query_array, n, index->ivf_pq->dim());
-    auto indices_view   = raft::make_device_matrix_view<int64_t, uint32_t>(indices, n, k);
-    auto distances_view = raft::make_device_matrix_view<float, uint32_t>(distances, n, k);
-    neighbors::ivf_pq::search(
-      handle, params, *index->ivf_pq, query_view, indices_view, distances_view);
-  } else {
-    RAFT_FAIL("The model is not trained");
-  }
-
-  // revert changes to the query
-  if constexpr (std::is_same_v<T, float>) { index->metric_processor->revert(query_array); }
-
-  // perform post-processing to show the real distances
-  if (index->metric == cuvs::distance::DistanceType::L2SqrtExpanded ||
-      index->metric == cuvs::distance::DistanceType::L2SqrtUnexpanded ||
-      index->metric == cuvs::distance::DistanceType::LpUnexpanded) {
-    /**
-     * post-processing
-     */
-    float p = 0.5;  // standard l2
-    if (index->metric == cuvs::distance::DistanceType::LpUnexpanded) p = 1.0 / index->metricArg;
-    raft::linalg::unaryOp<float>(
-      distances, distances, n * k, raft::pow_const_op<float>(p), resource::get_cuda_stream(handle));
-  }
-  if constexpr (std::is_same_v<T, float>) { index->metric_processor->postprocess(distances); }
-}
-
-}  // namespace cuvs::spatial::knn::detail
diff --git a/cpp/include/cuvs/spatial/knn/detail/ann_utils.cuh b/cpp/include/cuvs/spatial/knn/detail/ann_utils.cuh
deleted file mode 100644
index 461479e11..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/ann_utils.cuh
+++ /dev/null
@@ -1,576 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/logger.hpp>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <raft/util/integer_utils.hpp>
-
-#include <rmm/cuda_stream_view.hpp>
-#include <rmm/device_scalar.hpp>
-#include <rmm/device_uvector.hpp>
-
-#include <memory>
-#include <optional>
-
-#include <cuda_fp16.hpp>
-
-namespace cuvs::spatial::knn::detail::utils {
-
-/** Whether pointers are accessible on the device or on the host. */
-enum class pointer_residency {
-  /** Some of the pointers are on the device, some on the host. */
-  mixed,
-  /** All pointers accessible from both the device and the host. */
-  host_and_device,
-  /** All pointers are host accessible. */
-  host_only,
-  /** All poitners are device accessible. */
-  device_only
-};
-
-template <typename... Types>
-struct pointer_residency_count {};
-
-template <>
-struct pointer_residency_count<> {
-  static inline auto run() -> std::tuple<int, int> { return std::make_tuple(0, 0); }
-};
-
-template <typename Type, typename... Types>
-struct pointer_residency_count<Type, Types...> {
-  static inline auto run(const Type* ptr, const Types*... ptrs) -> std::tuple<int, int>
-  {
-    auto [on_device, on_host] = pointer_residency_count<Types...>::run(ptrs...);
-    cudaPointerAttributes attr;
-    RAFT_CUDA_TRY(cudaPointerGetAttributes(&attr, ptr));
-    switch (attr.type) {
-      case cudaMemoryTypeUnregistered: return std::make_tuple(on_device, on_host + 1);
-      case cudaMemoryTypeHost:
-        return std::make_tuple(on_device + int(attr.devicePointer == ptr), on_host + 1);
-      case cudaMemoryTypeDevice: return std::make_tuple(on_device + 1, on_host);
-      case cudaMemoryTypeManaged: return std::make_tuple(on_device + 1, on_host + 1);
-      default: return std::make_tuple(on_device, on_host);
-    }
-  }
-};
-
-/** Check if all argument pointers reside on the host or on the device. */
-template <typename... Types>
-auto check_pointer_residency(const Types*... ptrs) -> pointer_residency
-{
-  auto [on_device, on_host] = pointer_residency_count<Types...>::run(ptrs...);
-  int n_args                = sizeof...(Types);
-  if (on_device == n_args && on_host == n_args) { return pointer_residency::host_and_device; }
-  if (on_device == n_args) { return pointer_residency::device_only; }
-  if (on_host == n_args) { return pointer_residency::host_only; }
-  return pointer_residency::mixed;
-}
-
-/** RAII helper to access the host data from gpu when necessary. */
-template <typename PtrT, typename Action>
-struct with_mapped_memory_t {
-  with_mapped_memory_t(PtrT ptr, size_t size, Action action) : action_(action)
-  {
-    if (ptr == nullptr) { return; }
-    switch (utils::check_pointer_residency(ptr)) {
-      case utils::pointer_residency::device_only:
-      case utils::pointer_residency::host_and_device: {
-        dev_ptr_ = (void*)ptr;  // NOLINT
-      } break;
-      default: {
-        host_ptr_ = (void*)ptr;  // NOLINT
-        RAFT_CUDA_TRY(cudaHostRegister(host_ptr_, size, choose_flags(ptr)));
-        RAFT_CUDA_TRY(cudaHostGetDevicePointer(&dev_ptr_, host_ptr_, 0));
-      } break;
-    }
-  }
-
-  ~with_mapped_memory_t()
-  {
-    if (host_ptr_ != nullptr) { cudaHostUnregister(host_ptr_); }
-  }
-
-  auto operator()() { return action_((PtrT)dev_ptr_); }  // NOLINT
-
- private:
-  Action action_;
-  void* host_ptr_ = nullptr;
-  void* dev_ptr_  = nullptr;
-
-  template <typename T>
-  static auto choose_flags(const T*) -> unsigned int
-  {
-    int dev_id, readonly_supported;
-    RAFT_CUDA_TRY(cudaGetDevice(&dev_id));
-    RAFT_CUDA_TRY(cudaDeviceGetAttribute(
-      &readonly_supported, cudaDevAttrHostRegisterReadOnlySupported, dev_id));
-    if (readonly_supported) {
-      return cudaHostRegisterMapped | cudaHostRegisterReadOnly;
-    } else {
-      return cudaHostRegisterMapped;
-    }
-  }
-
-  template <typename T>
-  static auto choose_flags(T*) -> unsigned int
-  {
-    return cudaHostRegisterMapped;
-  }
-};
-
-template <typename T>
-struct config {};
-
-template <>
-struct config<double> {
-  using value_t                    = double;
-  static constexpr double kDivisor = 1.0;
-};
-template <>
-struct config<float> {
-  using value_t                    = float;
-  static constexpr double kDivisor = 1.0;
-};
-template <>
-struct config<half> {
-  using value_t                    = half;
-  static constexpr double kDivisor = 1.0;
-};
-template <>
-struct config<uint8_t> {
-  using value_t                    = uint32_t;
-  static constexpr double kDivisor = 256.0;
-};
-template <>
-struct config<int8_t> {
-  using value_t                    = int32_t;
-  static constexpr double kDivisor = 128.0;
-};
-
-/**
- * @brief Converting values between the types taking into account scaling factors
- * for the integral types.
- *
- * @tparam T target type of the mapping.
- */
-template <typename T>
-struct mapping {
-  /**
-   * @defgroup
-   * @brief Cast and possibly scale a value of the source type `S` to the target type `T`.
-   *
-   * @tparam S source type
-   * @param x source value
-   * @{
-   */
-  template <typename S>
-  HDI constexpr auto operator()(const S& x) const -> std::enable_if_t<std::is_same_v<S, T>, T>
-  {
-    return x;
-  };
-
-  template <typename S>
-  HDI constexpr auto operator()(const S& x) const -> std::enable_if_t<!std::is_same_v<S, T>, T>
-  {
-    constexpr double kMult = config<T>::kDivisor / config<S>::kDivisor;
-    if constexpr (std::is_floating_point_v<S>) { return static_cast<T>(x * static_cast<S>(kMult)); }
-    if constexpr (std::is_floating_point_v<T>) { return static_cast<T>(x) * static_cast<T>(kMult); }
-    return static_cast<T>(static_cast<float>(x) * static_cast<float>(kMult));
-  };
-  /** @} */
-};
-
-/**
- * @brief Sets the first num bytes of the block of memory pointed by ptr to the specified value.
- *
- * @param[out] ptr host or device pointer
- * @param[in] value
- * @param[in] n_bytes
- */
-template <typename T, typename IdxT>
-inline void memzero(T* ptr, IdxT n_elems, rmm::cuda_stream_view stream)
-{
-  switch (check_pointer_residency(ptr)) {
-    case pointer_residency::host_and_device:
-    case pointer_residency::device_only: {
-      RAFT_CUDA_TRY(cudaMemsetAsync(ptr, 0, n_elems * sizeof(T), stream));
-    } break;
-    case pointer_residency::host_only: {
-      stream.synchronize();
-      ::memset(ptr, 0, n_elems * sizeof(T));
-    } break;
-    default: RAFT_FAIL("memset: unreachable code");
-  }
-}
-
-template <typename T, typename IdxT>
-RAFT_KERNEL outer_add_kernel(const T* a, IdxT len_a, const T* b, IdxT len_b, T* c)
-{
-  IdxT gid = threadIdx.x + blockDim.x * static_cast<IdxT>(blockIdx.x);
-  IdxT i   = gid / len_b;
-  IdxT j   = gid % len_b;
-  if (i >= len_a) return;
-  c[gid] = (a == nullptr ? T(0) : a[i]) + (b == nullptr ? T(0) : b[j]);
-}
-
-template <typename T, typename IdxT>
-RAFT_KERNEL block_copy_kernel(const IdxT* in_offsets,
-                              const IdxT* out_offsets,
-                              IdxT n_blocks,
-                              const T* in_data,
-                              T* out_data,
-                              IdxT n_mult)
-{
-  IdxT i = static_cast<IdxT>(blockDim.x) * static_cast<IdxT>(blockIdx.x) + threadIdx.x;
-  // find the source offset using the binary search.
-  uint32_t l     = 0;
-  uint32_t r     = n_blocks;
-  IdxT in_offset = 0;
-  if (in_offsets[r] * n_mult <= i) return;
-  while (l + 1 < r) {
-    uint32_t c = (l + r) >> 1;
-    IdxT o     = in_offsets[c] * n_mult;
-    if (o <= i) {
-      l         = c;
-      in_offset = o;
-    } else {
-      r = c;
-    }
-  }
-  // copy the data
-  out_data[out_offsets[l] * n_mult - in_offset + i] = in_data[i];
-}
-
-/**
- * Copy chunks of data from one array to another at given offsets.
- *
- * @tparam T element type
- * @tparam IdxT index type
- *
- * @param[in] in_offsets
- * @param[in] out_offsets
- * @param n_blocks size of the offset arrays minus one.
- * @param[in] in_data
- * @param[out] out_data
- * @param n_mult constant multiplier for offset values (such as e.g. `dim`)
- * @param stream
- */
-template <typename T, typename IdxT>
-void block_copy(const IdxT* in_offsets,
-                const IdxT* out_offsets,
-                IdxT n_blocks,
-                const T* in_data,
-                T* out_data,
-                IdxT n_mult,
-                rmm::cuda_stream_view stream)
-{
-  IdxT in_size;
-  update_host(&in_size, in_offsets + n_blocks, 1, stream);
-  stream.synchronize();
-  dim3 threads(128, 1, 1);
-  dim3 blocks(raft::ceildiv<IdxT>(in_size * n_mult, threads.x), 1, 1);
-  block_copy_kernel<<<blocks, threads, 0, stream>>>(
-    in_offsets, out_offsets, n_blocks, in_data, out_data, n_mult);
-}
-
-/**
- * @brief Fill matrix `c` with all combinations of sums of vectors `a` and `b`.
- *
- * NB: device-only function
- *
- * @tparam T    element type
- * @tparam IdxT index type
- *
- * @param[in] a device pointer to a vector [len_a]
- * @param len_a number of elements in `a`
- * @param[in] b device pointer to a vector [len_b]
- * @param len_b number of elements in `b`
- * @param[out] c row-major matrix [len_a, len_b]
- * @param stream
- */
-template <typename T, typename IdxT>
-void outer_add(const T* a, IdxT len_a, const T* b, IdxT len_b, T* c, rmm::cuda_stream_view stream)
-{
-  dim3 threads(128, 1, 1);
-  dim3 blocks(raft::ceildiv<IdxT>(len_a * len_b, threads.x), 1, 1);
-  outer_add_kernel<<<blocks, threads, 0, stream>>>(a, len_a, b, len_b, c);
-}
-
-template <typename T, typename S, typename IdxT, typename LabelT>
-RAFT_KERNEL copy_selected_kernel(
-  IdxT n_rows, IdxT n_cols, const S* src, const LabelT* row_ids, IdxT ld_src, T* dst, IdxT ld_dst)
-{
-  IdxT gid   = threadIdx.x + blockDim.x * static_cast<IdxT>(blockIdx.x);
-  IdxT j     = gid % n_cols;
-  IdxT i_dst = gid / n_cols;
-  if (i_dst >= n_rows) return;
-  auto i_src              = static_cast<IdxT>(row_ids[i_dst]);
-  dst[ld_dst * i_dst + j] = mapping<T>{}(src[ld_src * i_src + j]);
-}
-
-/**
- * @brief Copy selected rows of a matrix while mapping the data from the source to the target
- * type.
- *
- * @tparam T      target type
- * @tparam S      source type
- * @tparam IdxT   index type
- * @tparam LabelT label type
- *
- * @param n_rows
- * @param n_cols
- * @param[in] src input matrix [..., ld_src]
- * @param[in] row_ids selection of rows to be copied [n_rows]
- * @param ld_src number of cols in the input (ld_src >= n_cols)
- * @param[out] dst output matrix [n_rows, ld_dst]
- * @param ld_dst number of cols in the output (ld_dst >= n_cols)
- * @param stream
- */
-template <typename T, typename S, typename IdxT, typename LabelT>
-void copy_selected(IdxT n_rows,
-                   IdxT n_cols,
-                   const S* src,
-                   const LabelT* row_ids,
-                   IdxT ld_src,
-                   T* dst,
-                   IdxT ld_dst,
-                   rmm::cuda_stream_view stream)
-{
-  switch (check_pointer_residency(src, dst, row_ids)) {
-    case pointer_residency::host_and_device:
-    case pointer_residency::device_only: {
-      IdxT block_dim = 128;
-      IdxT grid_dim  = raft::ceildiv(n_rows * n_cols, block_dim);
-      copy_selected_kernel<T, S>
-        <<<grid_dim, block_dim, 0, stream>>>(n_rows, n_cols, src, row_ids, ld_src, dst, ld_dst);
-    } break;
-    case pointer_residency::host_only: {
-      stream.synchronize();
-      for (IdxT i_dst = 0; i_dst < n_rows; i_dst++) {
-        auto i_src = static_cast<IdxT>(row_ids[i_dst]);
-        for (IdxT j = 0; j < n_cols; j++) {
-          dst[ld_dst * i_dst + j] = mapping<T>{}(src[ld_src * i_src + j]);
-        }
-      }
-      stream.synchronize();
-    } break;
-    default: RAFT_FAIL("All pointers must reside on the same side, host or device.");
-  }
-}
-
-/**
- * A batch input iterator over the data source.
- * Given an input pointer, it decides whether the current device has the access to the data and
- * gives it back to the user in batches. Three scenarios are possible:
- *
- *  1. if `source == nullptr`: then `batch.data() == nullptr`
- *  2. if `source` is accessible from the device, `batch.data()` points directly at the source at
- *     the proper offsets on each iteration.
- *  3. if `source` is not accessible from the device, `batch.data()` points to an intermediate
- *     buffer; the corresponding data is copied in the given `stream` on every iterator dereference
- *     (i.e. batches can be skipped). Dereferencing the same batch two times in a row does not force
- *     the copy.
- *
- * In all three scenarios, the number of iterations, batch offsets and sizes are the same.
- *
- * The iterator can be reused. If the number of iterations is one, at most one copy will ever be
- * invoked (i.e. small datasets are not reloaded multiple times).
- */
-template <typename T>
-struct batch_load_iterator {
-  using size_type = size_t;
-
-  /** A single batch of data residing in device memory. */
-  struct batch {
-    /** Logical width of a single row in a batch, in elements of type `T`. */
-    [[nodiscard]] auto row_width() const -> size_type { return row_width_; }
-    /** Logical offset of the batch, in rows (`row_width()`) */
-    [[nodiscard]] auto offset() const -> size_type { return pos_.value_or(0) * batch_size_; }
-    /** Logical size of the batch, in rows (`row_width()`) */
-    [[nodiscard]] auto size() const -> size_type { return batch_len_; }
-    /** Logical size of the batch, in rows (`row_width()`) */
-    [[nodiscard]] auto data() const -> const T* { return const_cast<const T*>(dev_ptr_); }
-    /** Whether this batch copies the data (i.e. the source is inaccessible from the device). */
-    [[nodiscard]] auto does_copy() const -> bool { return needs_copy_; }
-
-   private:
-    batch(const T* source,
-          size_type n_rows,
-          size_type row_width,
-          size_type batch_size,
-          rmm::cuda_stream_view stream,
-          rmm::mr::device_memory_resource* mr)
-      : stream_(stream),
-        buf_(0, stream, mr),
-        source_(source),
-        dev_ptr_(nullptr),
-        n_rows_(n_rows),
-        row_width_(row_width),
-        batch_size_(std::min(batch_size, n_rows)),
-        pos_(std::nullopt),
-        n_iters_(raft::div_rounding_up_safe(n_rows, batch_size)),
-        needs_copy_(false)
-    {
-      if (source_ == nullptr) { return; }
-      cudaPointerAttributes attr;
-      RAFT_CUDA_TRY(cudaPointerGetAttributes(&attr, source_));
-      dev_ptr_ = reinterpret_cast<T*>(attr.devicePointer);
-      if (dev_ptr_ == nullptr) {
-        buf_.resize(row_width_ * batch_size_, stream);
-        dev_ptr_    = buf_.data();
-        needs_copy_ = true;
-      }
-    }
-    rmm::cuda_stream_view stream_;
-    rmm::device_uvector<T> buf_;
-    const T* source_;
-    size_type n_rows_;
-    size_type row_width_;
-    size_type batch_size_;
-    size_type n_iters_;
-    bool needs_copy_;
-
-    std::optional<size_type> pos_;
-    size_type batch_len_;
-    T* dev_ptr_;
-
-    friend class batch_load_iterator<T>;
-
-    /**
-     * Changes the state of the batch to point at the `pos` index.
-     * If necessary, copies the data from the source in the registered stream.
-     */
-    void load(const size_type& pos)
-    {
-      // No-op if the data is already loaded, or it's the end of the input.
-      if (pos == pos_ || pos >= n_iters_) { return; }
-      pos_.emplace(pos);
-      batch_len_ = std::min(batch_size_, n_rows_ - std::min(offset(), n_rows_));
-      if (source_ == nullptr) { return; }
-      if (needs_copy_) {
-        if (size() > 0) {
-          RAFT_LOG_TRACE("batch_load_iterator::copy(offset = %zu, size = %zu, row_width = %zu)",
-                         size_t(offset()),
-                         size_t(size()),
-                         size_t(row_width()));
-          copy(dev_ptr_, source_ + offset() * row_width(), size() * row_width(), stream_);
-        }
-      } else {
-        dev_ptr_ = const_cast<T*>(source_) + offset() * row_width();
-      }
-    }
-  };
-
-  using value_type = batch;
-  using reference  = const value_type&;
-  using pointer    = const value_type*;
-
-  /**
-   * Create a batch iterator over the data `source`.
-   *
-   * For convenience, the data `source` is read in logical units of size `row_width`; batch sizes
-   * and offsets are calculated in logical rows. Hence, can interpret the data as a contiguous
-   * row-major matrix of size [n_rows, row_width], and the batches are the sub-matrices of size
-   * [x<=batch_size, n_rows].
-   *
-   * @param source the input data -- host, device, or nullptr.
-   * @param n_rows the size of the input in logical rows.
-   * @param row_width the size of the logical row in the elements of type `T`.
-   * @param batch_size the desired size of the batch.
-   * @param stream the ordering for the host->device copies, if applicable.
-   * @param mr a custom memory resource for the intermediate buffer, if applicable.
-   */
-  batch_load_iterator(const T* source,
-                      size_type n_rows,
-                      size_type row_width,
-                      size_type batch_size,
-                      rmm::cuda_stream_view stream,
-                      rmm::mr::device_memory_resource* mr = rmm::mr::get_current_device_resource())
-    : cur_batch_(new batch(source, n_rows, row_width, batch_size, stream, mr)), cur_pos_(0)
-  {
-  }
-  /**
-   * Whether this iterator copies the data on every iteration
-   * (i.e. the source is inaccessible from the device).
-   */
-  [[nodiscard]] auto does_copy() const -> bool { return cur_batch_->does_copy(); }
-  /** Reset the iterator position to `begin()` */
-  void reset() { cur_pos_ = 0; }
-  /** Reset the iterator position to `end()` */
-  void reset_to_end() { cur_pos_ = cur_batch_->n_iters_; }
-  [[nodiscard]] auto begin() const -> const batch_load_iterator<T>
-  {
-    batch_load_iterator<T> x(*this);
-    x.reset();
-    return x;
-  }
-  [[nodiscard]] auto end() const -> const batch_load_iterator<T>
-  {
-    batch_load_iterator<T> x(*this);
-    x.reset_to_end();
-    return x;
-  }
-  [[nodiscard]] auto operator*() const -> reference
-  {
-    cur_batch_->load(cur_pos_);
-    return *cur_batch_;
-  }
-  [[nodiscard]] auto operator->() const -> pointer
-  {
-    cur_batch_->load(cur_pos_);
-    return cur_batch_.get();
-  }
-  friend auto operator==(const batch_load_iterator<T>& x, const batch_load_iterator<T>& y) -> bool
-  {
-    return x.cur_batch_ == y.cur_batch_ && x.cur_pos_ == y.cur_pos_;
-  };
-  friend auto operator!=(const batch_load_iterator<T>& x, const batch_load_iterator<T>& y) -> bool
-  {
-    return x.cur_batch_ != y.cur_batch_ || x.cur_pos_ != y.cur_pos_;
-  };
-  auto operator++() -> batch_load_iterator<T>&
-  {
-    ++cur_pos_;
-    return *this;
-  }
-  auto operator++(int) -> batch_load_iterator<T>
-  {
-    batch_load_iterator<T> x(*this);
-    ++cur_pos_;
-    return x;
-  }
-  auto operator--() -> batch_load_iterator<T>&
-  {
-    --cur_pos_;
-    return *this;
-  }
-  auto operator--(int) -> batch_load_iterator<T>
-  {
-    batch_load_iterator<T> x(*this);
-    --cur_pos_;
-    return x;
-  }
-
- private:
-  std::shared_ptr<value_type> cur_batch_;
-  size_type cur_pos_;
-};
-
-}  // namespace cuvs::spatial::knn::detail::utils
diff --git a/cpp/include/cuvs/spatial/knn/detail/ball_cover.cuh b/cpp/include/cuvs/spatial/knn/detail/ball_cover.cuh
deleted file mode 100644
index f467600dd..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/ball_cover.cuh
+++ /dev/null
@@ -1,549 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resource/thrust_policy.hpp>
-#include <raft/core/resources.hpp>
-
-#include "../ball_cover_types.hpp"
-#include "ball_cover/common.cuh"
-#include "ball_cover/registers.cuh"
-#include "haversine_distance.cuh"
-
-#include <cstdint>
-#include <limits.h>
-
-#include <raft/util/cuda_utils.cuh>
-
-#include <cuvs/neighbors/detail/faiss_select/key_value_block_select.cuh>
-
-#include <cuvs/neighbors/brute_force.cuh>
-#include <raft/matrix/copy.cuh>
-#include <raft/random/rng.cuh>
-#include <raft/sparse/convert/csr.cuh>
-
-#include <rmm/device_uvector.hpp>
-#include <rmm/exec_policy.hpp>
-
-#include <thrust/fill.h>
-#include <thrust/for_each.h>
-#include <thrust/functional.h>
-#include <thrust/iterator/counting_iterator.h>
-#include <thrust/iterator/zip_iterator.h>
-#include <thrust/reduce.h>
-#include <thrust/sequence.h>
-#include <thrust/sort.h>
-#include <thrust/tuple.h>
-
-namespace cuvs {
-namespace spatial {
-namespace knn {
-namespace detail {
-
-/**
- * Given a set of points in row-major order which are to be
- * used as a set of index points, uniformly samples a subset
- * of points to be used as landmarks.
- * @tparam value_idx
- * @tparam value_t
- * @param handle
- * @param index
- */
-template <typename value_idx, typename value_t, typename value_int = std::uint32_t>
-void sample_landmarks(raft::resources const& handle,
-                      BallCoverIndex<value_idx, value_t, value_int>& index)
-{
-  rmm::device_uvector<value_idx> R_1nn_cols2(index.n_landmarks, resource::get_cuda_stream(handle));
-  rmm::device_uvector<value_t> R_1nn_ones(index.m, resource::get_cuda_stream(handle));
-  rmm::device_uvector<value_idx> R_indices(index.n_landmarks, resource::get_cuda_stream(handle));
-
-  thrust::sequence(raft::resource::get_thrust_policy(handle),
-                   index.get_R_1nn_cols().data_handle(),
-                   index.get_R_1nn_cols().data_handle() + index.m,
-                   (value_idx)0);
-
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               R_1nn_ones.data(),
-               R_1nn_ones.data() + R_1nn_ones.size(),
-               1.0);
-
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               R_indices.data(),
-               R_indices.data() + R_indices.size(),
-               0.0);
-
-  /**
-   * 1. Randomly sample sqrt(n) points from X
-   */
-  raft::random::RngState rng_state(12345);
-  raft::random::sampleWithoutReplacement(handle,
-                                         rng_state,
-                                         R_indices.data(),
-                                         R_1nn_cols2.data(),
-                                         index.get_R_1nn_cols().data_handle(),
-                                         R_1nn_ones.data(),
-                                         (value_idx)index.n_landmarks,
-                                         (value_idx)index.m);
-
-  // index.get_X() returns the wrong indextype (uint32_t where we need value_idx), so need to
-  // create new device_matrix_view here
-  auto x = index.get_X();
-  auto r = index.get_R();
-
-  raft::matrix::copy_rows<value_t, value_idx>(
-    handle,
-    make_device_matrix_view<const value_t, value_idx>(x.data_handle(), x.extent(0), x.extent(1)),
-    make_device_matrix_view<value_t, value_idx>(r.data_handle(), r.extent(0), r.extent(1)),
-    make_device_vector_view(R_1nn_cols2.data(), index.n_landmarks));
-}
-
-/**
- * Constructs a 1-nn index mapping each landmark to their closest points.
- * @tparam value_idx
- * @tparam value_t
- * @param handle
- * @param R_knn_inds_ptr
- * @param R_knn_dists_ptr
- * @param k
- * @param index
- */
-template <typename value_idx, typename value_t, typename value_int = std::uint32_t>
-void construct_landmark_1nn(raft::resources const& handle,
-                            const value_idx* R_knn_inds_ptr,
-                            const value_t* R_knn_dists_ptr,
-                            value_int k,
-                            BallCoverIndex<value_idx, value_t, value_int>& index)
-{
-  rmm::device_uvector<value_idx> R_1nn_inds(index.m, resource::get_cuda_stream(handle));
-
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               R_1nn_inds.data(),
-               R_1nn_inds.data() + index.m,
-               std::numeric_limits<value_idx>::max());
-
-  value_idx* R_1nn_inds_ptr = R_1nn_inds.data();
-  value_t* R_1nn_dists_ptr  = index.get_R_1nn_dists().data_handle();
-
-  auto idxs = thrust::make_counting_iterator<value_idx>(0);
-  thrust::for_each(
-    resource::get_thrust_policy(handle), idxs, idxs + index.m, [=] __device__(value_idx i) {
-      R_1nn_inds_ptr[i]  = R_knn_inds_ptr[i * k];
-      R_1nn_dists_ptr[i] = R_knn_dists_ptr[i * k];
-    });
-
-  auto keys = thrust::make_zip_iterator(
-    thrust::make_tuple(R_1nn_inds.data(), index.get_R_1nn_dists().data_handle()));
-
-  // group neighborhoods for each reference landmark and sort each group by distance
-  thrust::sort_by_key(raft::resource::get_thrust_policy(handle),
-                      keys,
-                      keys + index.m,
-                      index.get_R_1nn_cols().data_handle(),
-                      NNComp());
-
-  // convert to CSR for fast lookup
-  raft::sparse::convert::sorted_coo_to_csr(R_1nn_inds.data(),
-                                           index.m,
-                                           index.get_R_indptr().data_handle(),
-                                           index.n_landmarks + 1,
-                                           resource::get_cuda_stream(handle));
-}
-
-/**
- * Computes the k closest landmarks to a set of query points.
- * @tparam value_idx
- * @tparam value_t
- * @tparam value_int
- * @param handle
- * @param index
- * @param query_pts
- * @param n_query_pts
- * @param k
- * @param R_knn_inds
- * @param R_knn_dists
- */
-template <typename value_idx, typename value_t, typename value_int = std::uint32_t>
-void k_closest_landmarks(raft::resources const& handle,
-                         const BallCoverIndex<value_idx, value_t, value_int>& index,
-                         const value_t* query_pts,
-                         value_int n_query_pts,
-                         value_int k,
-                         value_idx* R_knn_inds,
-                         value_t* R_knn_dists)
-{
-  std::vector<raft::device_matrix_view<const value_t, value_int>> inputs = {index.get_R()};
-
-  cuvs::neighbors::brute_force::knn<value_idx, value_t, value_int>(
-    handle,
-    inputs,
-    make_device_matrix_view(query_pts, n_query_pts, inputs[0].extent(1)),
-    make_device_matrix_view(R_knn_inds, n_query_pts, k),
-    make_device_matrix_view(R_knn_dists, n_query_pts, k),
-    index.get_metric());
-}
-
-/**
- * Uses the sorted data points in the 1-nn landmark index to compute
- * an array of radii for each landmark.
- * @tparam value_idx
- * @tparam value_t
- * @param handle
- * @param index
- */
-template <typename value_idx, typename value_t, typename value_int = std::uint32_t>
-void compute_landmark_radii(raft::resources const& handle,
-                            BallCoverIndex<value_idx, value_t, value_int>& index)
-{
-  auto entries = thrust::make_counting_iterator<value_idx>(0);
-
-  const value_idx* R_indptr_ptr  = index.get_R_indptr().data_handle();
-  const value_t* R_1nn_dists_ptr = index.get_R_1nn_dists().data_handle();
-  value_t* R_radius_ptr          = index.get_R_radius().data_handle();
-  thrust::for_each(raft::resource::get_thrust_policy(handle),
-                   entries,
-                   entries + index.n_landmarks,
-                   [=] __device__(value_idx input) {
-                     value_idx last_row_idx = R_indptr_ptr[input + 1] - 1;
-                     R_radius_ptr[input]    = R_1nn_dists_ptr[last_row_idx];
-                   });
-}
-
-/**
- * 4. Perform k-select over original KNN, using L_r to filter distances
- *
- * a. Map 1 row to each warp/block
- * b. Add closest k R points to heap
- * c. Iterate through batches of R, having each thread in the warp load a set
- * of distances y from R (only if d(q, r) < 3 * distance to closest r) and
- * marking the distance to be computed between x, y only
- * if knn[k].distance >= d(x_i, R_k) + d(R_k, y)
- */
-template <typename value_idx,
-          typename value_t,
-          typename value_int = std::uint32_t,
-          typename dist_func>
-void perform_rbc_query(raft::resources const& handle,
-                       const BallCoverIndex<value_idx, value_t, value_int>& index,
-                       const value_t* query,
-                       value_int n_query_pts,
-                       std::uint32_t k,
-                       const value_idx* R_knn_inds,
-                       const value_t* R_knn_dists,
-                       dist_func dfunc,
-                       value_idx* inds,
-                       value_t* dists,
-                       value_int* dists_counter,
-                       value_int* post_dists_counter,
-                       float weight                = 1.0,
-                       bool perform_post_filtering = true)
-{
-  // initialize output inds and dists
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               inds,
-               inds + (k * n_query_pts),
-               std::numeric_limits<value_idx>::max());
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               dists,
-               dists + (k * n_query_pts),
-               std::numeric_limits<value_t>::max());
-
-  if (index.n == 2) {
-    // Compute nearest k for each neighborhood in each closest R
-    rbc_low_dim_pass_one<value_idx, value_t, value_int, 2>(handle,
-                                                           index,
-                                                           query,
-                                                           n_query_pts,
-                                                           k,
-                                                           R_knn_inds,
-                                                           R_knn_dists,
-                                                           dfunc,
-                                                           inds,
-                                                           dists,
-                                                           weight,
-                                                           dists_counter);
-
-    if (perform_post_filtering) {
-      rbc_low_dim_pass_two<value_idx, value_t, value_int, 2>(handle,
-                                                             index,
-                                                             query,
-                                                             n_query_pts,
-                                                             k,
-                                                             R_knn_inds,
-                                                             R_knn_dists,
-                                                             dfunc,
-                                                             inds,
-                                                             dists,
-                                                             weight,
-                                                             post_dists_counter);
-    }
-
-  } else if (index.n == 3) {
-    // Compute nearest k for each neighborhood in each closest R
-    rbc_low_dim_pass_one<value_idx, value_t, value_int, 3>(handle,
-                                                           index,
-                                                           query,
-                                                           n_query_pts,
-                                                           k,
-                                                           R_knn_inds,
-                                                           R_knn_dists,
-                                                           dfunc,
-                                                           inds,
-                                                           dists,
-                                                           weight,
-                                                           dists_counter);
-
-    if (perform_post_filtering) {
-      rbc_low_dim_pass_two<value_idx, value_t, value_int, 3>(handle,
-                                                             index,
-                                                             query,
-                                                             n_query_pts,
-                                                             k,
-                                                             R_knn_inds,
-                                                             R_knn_dists,
-                                                             dfunc,
-                                                             inds,
-                                                             dists,
-                                                             weight,
-                                                             post_dists_counter);
-    }
-  }
-}
-
-/**
- * Similar to a ball tree, the random ball cover algorithm
- * uses the triangle inequality to prune distance computations
- * in any metric space with a guarantee of sqrt(n) * c^{3/2}
- * where `c` is an expansion constant based on the distance
- * metric.
- *
- * This function variant performs an all nearest neighbors
- * query which is useful for algorithms that need to perform
- * A * A.T.
- */
-template <typename value_idx = std::int64_t,
-          typename value_t,
-          typename value_int = std::uint32_t,
-          typename distance_func>
-void rbc_build_index(raft::resources const& handle,
-                     BallCoverIndex<value_idx, value_t, value_int>& index,
-                     distance_func dfunc)
-{
-  ASSERT(index.n <= 3, "only 2d and 3d vectors are supported in current implementation");
-  ASSERT(!index.is_index_trained(), "index cannot be previously trained");
-
-  rmm::device_uvector<value_idx> R_knn_inds(index.m, resource::get_cuda_stream(handle));
-
-  // Initialize the uvectors
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               R_knn_inds.begin(),
-               R_knn_inds.end(),
-               std::numeric_limits<value_idx>::max());
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               index.get_R_closest_landmark_dists().data_handle(),
-               index.get_R_closest_landmark_dists().data_handle() + index.m,
-               std::numeric_limits<value_t>::max());
-
-  /**
-   * 1. Randomly sample sqrt(n) points from X
-   */
-  sample_landmarks<value_idx, value_t>(handle, index);
-
-  /**
-   * 2. Perform knn = bfknn(X, R, k)
-   */
-  value_int k = 1;
-  k_closest_landmarks(handle,
-                      index,
-                      index.get_X().data_handle(),
-                      index.m,
-                      k,
-                      R_knn_inds.data(),
-                      index.get_R_closest_landmark_dists().data_handle());
-
-  /**
-   * 3. Create L_r = knn[:,0].T (CSR)
-   *
-   * Slice closest neighboring R
-   * Secondary sort by (R_knn_inds, R_knn_dists)
-   */
-  construct_landmark_1nn(
-    handle, R_knn_inds.data(), index.get_R_closest_landmark_dists().data_handle(), k, index);
-
-  /**
-   * Compute radius of each R for filtering: p(q, r) <= p(q, q_r) + radius(r)
-   * (need to take the
-   */
-  compute_landmark_radii(handle, index);
-}
-
-/**
- * Performs an all neighbors knn query (e.g. index == query)
- */
-template <typename value_idx = std::int64_t,
-          typename value_t,
-          typename value_int = std::uint32_t,
-          typename distance_func>
-void rbc_all_knn_query(raft::resources const& handle,
-                       BallCoverIndex<value_idx, value_t, value_int>& index,
-                       value_int k,
-                       value_idx* inds,
-                       value_t* dists,
-                       distance_func dfunc,
-                       // approximate nn options
-                       bool perform_post_filtering = true,
-                       float weight                = 1.0)
-{
-  ASSERT(index.n <= 3, "only 2d and 3d vectors are supported in current implementation");
-  ASSERT(index.n_landmarks >= k, "number of landmark samples must be >= k");
-  ASSERT(!index.is_index_trained(), "index cannot be previously trained");
-
-  rmm::device_uvector<value_idx> R_knn_inds(k * index.m, raft::resource::get_cuda_stream(handle));
-  rmm::device_uvector<value_t> R_knn_dists(k * index.m, raft::resource::get_cuda_stream(handle));
-
-  // Initialize the uvectors
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               R_knn_inds.begin(),
-               R_knn_inds.end(),
-               std::numeric_limits<value_idx>::max());
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               R_knn_dists.begin(),
-               R_knn_dists.end(),
-               std::numeric_limits<value_t>::max());
-
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               inds,
-               inds + (k * index.m),
-               std::numeric_limits<value_idx>::max());
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               dists,
-               dists + (k * index.m),
-               std::numeric_limits<value_t>::max());
-
-  // For debugging / verification. Remove before releasing
-  rmm::device_uvector<value_int> dists_counter(index.m, raft::resource::get_cuda_stream(handle));
-  rmm::device_uvector<value_int> post_dists_counter(index.m,
-                                                    raft::resource::get_cuda_stream(handle));
-
-  sample_landmarks<value_idx, value_t>(handle, index);
-
-  k_closest_landmarks(
-    handle, index, index.get_X().data_handle(), index.m, k, R_knn_inds.data(), R_knn_dists.data());
-
-  construct_landmark_1nn(handle, R_knn_inds.data(), R_knn_dists.data(), k, index);
-
-  compute_landmark_radii(handle, index);
-
-  perform_rbc_query(handle,
-                    index,
-                    index.get_X().data_handle(),
-                    index.m,
-                    k,
-                    R_knn_inds.data(),
-                    R_knn_dists.data(),
-                    dfunc,
-                    inds,
-                    dists,
-                    dists_counter.data(),
-                    post_dists_counter.data(),
-                    weight,
-                    perform_post_filtering);
-}
-
-/**
- * Performs a knn query against an index. This assumes the index has
- * already been built.
- */
-template <typename value_idx = std::int64_t,
-          typename value_t,
-          typename value_int = std::uint32_t,
-          typename distance_func>
-void rbc_knn_query(raft::resources const& handle,
-                   const BallCoverIndex<value_idx, value_t, value_int>& index,
-                   value_int k,
-                   const value_t* query,
-                   value_int n_query_pts,
-                   value_idx* inds,
-                   value_t* dists,
-                   distance_func dfunc,
-                   // approximate nn options
-                   bool perform_post_filtering = true,
-                   float weight                = 1.0)
-{
-  ASSERT(index.n <= 3, "only 2d and 3d vectors are supported in current implementation");
-  ASSERT(index.n_landmarks >= k, "number of landmark samples must be >= k");
-  ASSERT(index.is_index_trained(), "index must be previously trained");
-
-  rmm::device_uvector<value_idx> R_knn_inds(k * n_query_pts,
-                                            raft::resource::get_cuda_stream(handle));
-  rmm::device_uvector<value_t> R_knn_dists(k * n_query_pts,
-                                           raft::resource::get_cuda_stream(handle));
-
-  // Initialize the uvectors
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               R_knn_inds.begin(),
-               R_knn_inds.end(),
-               std::numeric_limits<value_idx>::max());
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               R_knn_dists.begin(),
-               R_knn_dists.end(),
-               std::numeric_limits<value_t>::max());
-
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               inds,
-               inds + (k * n_query_pts),
-               std::numeric_limits<value_idx>::max());
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               dists,
-               dists + (k * n_query_pts),
-               std::numeric_limits<value_t>::max());
-
-  k_closest_landmarks(handle, index, query, n_query_pts, k, R_knn_inds.data(), R_knn_dists.data());
-
-  // For debugging / verification. Remove before releasing
-  rmm::device_uvector<value_int> dists_counter(index.m, raft::resource::get_cuda_stream(handle));
-  rmm::device_uvector<value_int> post_dists_counter(index.m,
-                                                    raft::resource::get_cuda_stream(handle));
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               post_dists_counter.data(),
-               post_dists_counter.data() + post_dists_counter.size(),
-               0);
-  thrust::fill(raft::resource::get_thrust_policy(handle),
-               dists_counter.data(),
-               dists_counter.data() + dists_counter.size(),
-               0);
-
-  perform_rbc_query(handle,
-                    index,
-                    query,
-                    n_query_pts,
-                    k,
-                    R_knn_inds.data(),
-                    R_knn_dists.data(),
-                    dfunc,
-                    inds,
-                    dists,
-                    dists_counter.data(),
-                    post_dists_counter.data(),
-                    weight,
-                    perform_post_filtering);
-}
-
-};  // namespace detail
-};  // namespace knn
-};  // namespace spatial
-};  // namespace cuvs
diff --git a/cpp/include/cuvs/spatial/knn/detail/ball_cover/common.cuh b/cpp/include/cuvs/spatial/knn/detail/ball_cover/common.cuh
deleted file mode 100644
index 05aa1439a..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/ball_cover/common.cuh
+++ /dev/null
@@ -1,73 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "../haversine_distance.cuh"
-#include "registers_types.cuh"
-#include <cstdint>
-#include <thrust/functional.h>
-#include <thrust/tuple.h>
-
-namespace cuvs {
-namespace spatial {
-namespace knn {
-namespace detail {
-
-struct NNComp {
-  template <typename one, typename two>
-  __host__ __device__ bool operator()(const one& t1, const two& t2)
-  {
-    // sort first by each sample's reference landmark,
-    if (thrust::get<0>(t1) < thrust::get<0>(t2)) return true;
-    if (thrust::get<0>(t1) > thrust::get<0>(t2)) return false;
-
-    // then by closest neighbor,
-    return thrust::get<1>(t1) < thrust::get<1>(t2);
-  }
-};
-
-/**
- * Zeros the bit at location h in a one-hot encoded 32-bit int array
- */
-__device__ inline void _zero_bit(std::uint32_t* arr, std::uint32_t h)
-{
-  int bit = h % 32;
-  int idx = h / 32;
-
-  std::uint32_t assumed;
-  std::uint32_t old = arr[idx];
-  do {
-    assumed = old;
-    old     = atomicCAS(arr + idx, assumed, assumed & ~(1 << bit));
-  } while (assumed != old);
-}
-
-/**
- * Returns whether or not bit at location h is nonzero in a one-hot
- * encoded 32-bit in array.
- */
-__device__ inline bool _get_val(std::uint32_t* arr, std::uint32_t h)
-{
-  int bit = h % 32;
-  int idx = h / 32;
-  return (arr[idx] & (1 << bit)) > 0;
-}
-
-};  // namespace detail
-};  // namespace knn
-};  // namespace spatial
-};  // namespace cuvs
diff --git a/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers-ext.cuh b/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers-ext.cuh
deleted file mode 100644
index a96c329a6..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers-ext.cuh
+++ /dev/null
@@ -1,129 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "../../ball_cover_types.hpp"   // BallCoverIndex
-#include "registers_types.cuh"          // DistFunc
-#include <cstdint>                      // uint32_t
-#include <raft/util/raft_explicit.hpp>  //RAFT_EXPLICIT
-
-#if defined(RAFT_EXPLICIT_INSTANTIATE_ONLY)
-
-namespace cuvs::spatial::knn::detail {
-
-template <typename value_idx,
-          typename value_t,
-          typename value_int = std::uint32_t,
-          int dims           = 2,
-          typename dist_func>
-void rbc_low_dim_pass_one(raft::resources const& handle,
-                          const BallCoverIndex<value_idx, value_t, value_int>& index,
-                          const value_t* query,
-                          const value_int n_query_rows,
-                          value_int k,
-                          const value_idx* R_knn_inds,
-                          const value_t* R_knn_dists,
-                          dist_func& dfunc,
-                          value_idx* inds,
-                          value_t* dists,
-                          float weight,
-                          value_int* dists_counter) RAFT_EXPLICIT;
-
-template <typename value_idx,
-          typename value_t,
-          typename value_int = std::uint32_t,
-          int dims           = 2,
-          typename dist_func>
-void rbc_low_dim_pass_two(raft::resources const& handle,
-                          const BallCoverIndex<value_idx, value_t, value_int>& index,
-                          const value_t* query,
-                          const value_int n_query_rows,
-                          value_int k,
-                          const value_idx* R_knn_inds,
-                          const value_t* R_knn_dists,
-                          dist_func& dfunc,
-                          value_idx* inds,
-                          value_t* dists,
-                          float weight,
-                          value_int* post_dists_counter) RAFT_EXPLICIT;
-
-};  // namespace cuvs::spatial::knn::detail
-
-#endif  // RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  extern template void                                                                       \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_one<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  extern template void                                                                       \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_two<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::HaversineFunc);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::HaversineFunc);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::EuclideanFunc);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::EuclideanFunc);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::DistFunc);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::DistFunc);
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::HaversineFunc);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::HaversineFunc);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::EuclideanFunc);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::EuclideanFunc);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::DistFunc);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::DistFunc);
-
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one
diff --git a/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh b/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh
deleted file mode 100644
index f1cb45e97..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh
+++ /dev/null
@@ -1,794 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "common.cuh"
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resource/thrust_policy.hpp>
-
-#include "../../ball_cover_types.hpp"
-#include "../haversine_distance.cuh"
-#include "registers_types.cuh"  // DistFunc
-
-#include <cstdint>
-#include <limits.h>
-
-#include <cuvs/neighbors/detail/faiss_select/key_value_block_select.cuh>
-#include <raft/util/cuda_utils.cuh>
-
-#include <thrust/fill.h>
-
-namespace cuvs {
-namespace spatial {
-namespace knn {
-namespace detail {
-
-/**
- * To find exact neighbors, we perform a post-processing stage
- * that filters out those points which might have neighbors outside
- * of their k closest landmarks. This is usually a very small portion
- * of the total points.
- * @tparam value_idx
- * @tparam value_t
- * @tparam value_int
- * @tparam tpb
- * @param X
- * @param n_cols
- * @param R_knn_inds
- * @param R_knn_dists
- * @param R_radius
- * @param landmarks
- * @param n_landmarks
- * @param bitset_size
- * @param k
- * @param output
- * @param weight
- */
-template <typename value_idx,
-          typename value_t,
-          typename value_int = std::uint32_t,
-          int col_q          = 2,
-          int tpb            = 32,
-          typename distance_func>
-RAFT_KERNEL perform_post_filter_registers(const value_t* X,
-                                          value_int n_cols,
-                                          const value_idx* R_knn_inds,
-                                          const value_t* R_knn_dists,
-                                          const value_t* R_radius,
-                                          const value_t* landmarks,
-                                          int n_landmarks,
-                                          value_int bitset_size,
-                                          value_int k,
-                                          distance_func dfunc,
-                                          std::uint32_t* output,
-                                          float weight = 1.0)
-{
-  // allocate array of size n_landmarks / 32 ints
-  extern __shared__ std::uint32_t shared_mem[];
-
-  // Start with all bits on
-  for (value_int i = threadIdx.x; i < bitset_size; i += tpb) {
-    shared_mem[i] = 0xffffffff;
-  }
-
-  __syncthreads();
-
-  // TODO: Would it be faster to use L1 for this?
-  value_t local_x_ptr[col_q];
-  for (value_int j = 0; j < n_cols; ++j) {
-    local_x_ptr[j] = X[n_cols * blockIdx.x + j];
-  }
-
-  value_t closest_R_dist = R_knn_dists[blockIdx.x * k + (k - 1)];
-
-  // zero out bits for closest k landmarks
-  for (value_int j = threadIdx.x; j < k; j += tpb) {
-    _zero_bit(shared_mem, (std::uint32_t)R_knn_inds[blockIdx.x * k + j]);
-  }
-
-  __syncthreads();
-
-  // Discard any landmarks where p(q, r) > p(q, r_q) + radius(r)
-  // That is, the distance between the current point and the current
-  // landmark is > the distance between the current point and
-  // its closest landmark + the radius of the current landmark.
-  for (value_int l = threadIdx.x; l < n_landmarks; l += tpb) {
-    // compute p(q, r)
-    value_t dist = dfunc(local_x_ptr, landmarks + (n_cols * l), n_cols);
-    if (dist > weight * (closest_R_dist + R_radius[l]) || dist > 3 * closest_R_dist) {
-      _zero_bit(shared_mem, l);
-    }
-  }
-
-  __syncthreads();
-
-  /**
-   * Output bitset
-   */
-  for (value_int l = threadIdx.x; l < bitset_size; l += tpb) {
-    output[blockIdx.x * bitset_size + l] = shared_mem[l];
-  }
-}
-
-/**
- * @tparam value_idx
- * @tparam value_t
- * @tparam value_int
- * @tparam bitset_type
- * @tparam warp_q number of registers to use per warp
- * @tparam thread_q number of registers to use within each thread
- * @tparam tpb number of threads per block
- * @param X
- * @param n_cols
- * @param bitset
- * @param bitset_size
- * @param R_knn_dists
- * @param R_indptr
- * @param R_1nn_inds
- * @param R_1nn_dists
- * @param knn_inds
- * @param knn_dists
- * @param n_landmarks
- * @param k
- * @param dist_counter
- */
-template <typename value_idx,
-          typename value_t,
-          typename value_int   = std::uint32_t,
-          typename bitset_type = std::uint32_t,
-          typename dist_func,
-          int warp_q   = 32,
-          int thread_q = 2,
-          int tpb      = 128,
-          int col_q    = 2>
-RAFT_KERNEL compute_final_dists_registers(const value_t* X_index,
-                                          const value_t* X,
-                                          const value_int n_cols,
-                                          bitset_type* bitset,
-                                          value_int bitset_size,
-                                          const value_t* R_closest_landmark_dists,
-                                          const value_idx* R_indptr,
-                                          const value_idx* R_1nn_inds,
-                                          const value_t* R_1nn_dists,
-                                          value_idx* knn_inds,
-                                          value_t* knn_dists,
-                                          value_int n_landmarks,
-                                          value_int k,
-                                          dist_func dfunc,
-                                          value_int* dist_counter)
-{
-  static constexpr int kNumWarps = tpb / raft::WarpSize;
-
-  __shared__ value_t shared_memK[kNumWarps * warp_q];
-  __shared__ KeyValuePair<value_t, value_idx> shared_memV[kNumWarps * warp_q];
-
-  const value_t* x_ptr = X + (n_cols * blockIdx.x);
-  value_t local_x_ptr[col_q];
-  for (value_int j = 0; j < n_cols; ++j) {
-    local_x_ptr[j] = x_ptr[j];
-  }
-
-  using namespace cuvs::neighbors::detail::faiss_select;
-  KeyValueBlockSelect<value_t, value_idx, false, Comparator<value_t>, warp_q, thread_q, tpb> heap(
-    std::numeric_limits<value_t>::max(),
-    std::numeric_limits<value_t>::max(),
-    -1,
-    shared_memK,
-    shared_memV,
-    k);
-
-  const value_int n_k = raft::Pow2<WarpSize>::roundDown(k);
-  value_int i         = threadIdx.x;
-  for (; i < n_k; i += tpb) {
-    value_idx ind = knn_inds[blockIdx.x * k + i];
-    heap.add(knn_dists[blockIdx.x * k + i], R_closest_landmark_dists[ind], ind);
-  }
-
-  if (i < k) {
-    value_idx ind = knn_inds[blockIdx.x * k + i];
-    heap.addThreadQ(knn_dists[blockIdx.x * k + i], R_closest_landmark_dists[ind], ind);
-  }
-
-  heap.checkThreadQ();
-
-  for (value_int cur_R_ind = 0; cur_R_ind < n_landmarks; ++cur_R_ind) {
-    // if cur R overlaps cur point's closest R, it could be a
-    // candidate
-    if (_get_val(bitset + (blockIdx.x * bitset_size), cur_R_ind)) {
-      value_idx R_start_offset = R_indptr[cur_R_ind];
-      value_idx R_stop_offset  = R_indptr[cur_R_ind + 1];
-      value_idx R_size         = R_stop_offset - R_start_offset;
-
-      // Loop through R's neighborhood in parallel
-
-      // Round R_size to the nearest warp threads so they can
-      // all be computing in parallel.
-
-      const value_int limit = raft::Pow2<WarpSize>::roundDown(R_size);
-
-      i = threadIdx.x;
-      for (; i < limit; i += tpb) {
-        value_idx cur_candidate_ind = R_1nn_inds[R_start_offset + i];
-        value_t cur_candidate_dist  = R_1nn_dists[R_start_offset + i];
-
-        value_t z = heap.warpKTopRDist == 0.00 ? 0.0
-                                               : (abs(heap.warpKTop - heap.warpKTopRDist) *
-                                                    abs(heap.warpKTopRDist - cur_candidate_dist) -
-                                                  heap.warpKTop * cur_candidate_dist) /
-                                                   heap.warpKTopRDist;
-        z         = isnan(z) || isinf(z) ? 0.0 : z;
-
-        // If lower bound on distance could possibly be in
-        // the closest k neighbors, compute it and add to k-select
-        value_t dist = std::numeric_limits<value_t>::max();
-        if (z <= heap.warpKTop) {
-          const value_t* y_ptr = X_index + (n_cols * cur_candidate_ind);
-          value_t local_y_ptr[col_q];
-          for (value_int j = 0; j < n_cols; ++j) {
-            local_y_ptr[j] = y_ptr[j];
-          }
-
-          dist = dfunc(local_x_ptr, local_y_ptr, n_cols);
-        }
-
-        heap.add(dist, cur_candidate_dist, cur_candidate_ind);
-      }
-
-      // second round guarantees to be only a single warp.
-      if (i < R_size) {
-        value_idx cur_candidate_ind = R_1nn_inds[R_start_offset + i];
-        value_t cur_candidate_dist  = R_1nn_dists[R_start_offset + i];
-
-        value_t z = heap.warpKTopRDist == 0.00 ? 0.0
-                                               : (abs(heap.warpKTop - heap.warpKTopRDist) *
-                                                    abs(heap.warpKTopRDist - cur_candidate_dist) -
-                                                  heap.warpKTop * cur_candidate_dist) /
-                                                   heap.warpKTopRDist;
-
-        z = isnan(z) || isinf(z) ? 0.0 : z;
-
-        // If lower bound on distance could possibly be in
-        // the closest k neighbors, compute it and add to k-select
-        value_t dist = std::numeric_limits<value_t>::max();
-        if (z <= heap.warpKTop) {
-          const value_t* y_ptr = X_index + (n_cols * cur_candidate_ind);
-          value_t local_y_ptr[col_q];
-          for (value_int j = 0; j < n_cols; ++j) {
-            local_y_ptr[j] = y_ptr[j];
-          }
-          dist = dfunc(local_x_ptr, local_y_ptr, n_cols);
-        }
-        heap.addThreadQ(dist, cur_candidate_dist, cur_candidate_ind);
-      }
-      heap.checkThreadQ();
-    }
-  }
-
-  heap.reduce();
-
-  for (value_int i = threadIdx.x; i < k; i += tpb) {
-    knn_dists[blockIdx.x * k + i] = shared_memK[i];
-    knn_inds[blockIdx.x * k + i]  = shared_memV[i].value;
-  }
-}
-
-/**
- * Random ball cover kernel for n_dims == 2
- * @tparam value_idx
- * @tparam value_t
- * @tparam warp_q
- * @tparam thread_q
- * @tparam tpb
- * @tparam value_idx
- * @tparam value_t
- * @param R_knn_inds
- * @param R_knn_dists
- * @param m
- * @param k
- * @param R_indptr
- * @param R_1nn_cols
- * @param R_1nn_dists
- */
-template <typename value_idx = std::int64_t,
-          typename value_t,
-          int warp_q         = 32,
-          int thread_q       = 2,
-          int tpb            = 128,
-          int col_q          = 2,
-          typename value_int = std::uint32_t,
-          typename distance_func>
-RAFT_KERNEL block_rbc_kernel_registers(const value_t* X_index,
-                                       const value_t* X,
-                                       value_int n_cols,  // n_cols should be 2 or 3 dims
-                                       const value_idx* R_knn_inds,
-                                       const value_t* R_knn_dists,
-                                       value_int m,
-                                       value_int k,
-                                       const value_idx* R_indptr,
-                                       const value_idx* R_1nn_cols,
-                                       const value_t* R_1nn_dists,
-                                       value_idx* out_inds,
-                                       value_t* out_dists,
-                                       value_int* dist_counter,
-                                       const value_t* R_radius,
-                                       distance_func dfunc,
-                                       float weight = 1.0)
-{
-  static constexpr value_int kNumWarps = tpb / raft::WarpSize;
-
-  __shared__ value_t shared_memK[kNumWarps * warp_q];
-  __shared__ KeyValuePair<value_t, value_idx> shared_memV[kNumWarps * warp_q];
-
-  // TODO: Separate kernels for different widths:
-  // 1. Very small (between 3 and 32) just use registers for columns of "blockIdx.x"
-  // 2. Can fit comfortably in shared memory (32 to a few thousand?)
-  // 3. Load each time individually.
-  const value_t* x_ptr = X + (n_cols * blockIdx.x);
-
-  // Use registers only for 2d or 3d
-  value_t local_x_ptr[col_q];
-  for (value_int i = 0; i < n_cols; ++i) {
-    local_x_ptr[i] = x_ptr[i];
-  }
-
-  // Each warp works on 1 R
-  using namespace cuvs::neighbors::detail::faiss_select;
-  KeyValueBlockSelect<value_t, value_idx, false, Comparator<value_t>, warp_q, thread_q, tpb> heap(
-    std::numeric_limits<value_t>::max(),
-    std::numeric_limits<value_t>::max(),
-    -1,
-    shared_memK,
-    shared_memV,
-    k);
-
-  value_t min_R_dist         = R_knn_dists[blockIdx.x * k + (k - 1)];
-  value_int n_dists_computed = 0;
-
-  /**
-   * First add distances for k closest neighbors of R
-   * to the heap
-   */
-  // Start iterating through elements of each set from closest R elements,
-  // determining if the distance could even potentially be in the heap.
-  for (value_int cur_k = 0; cur_k < k; ++cur_k) {
-    // index and distance to current blockIdx.x's closest landmark
-    value_t cur_R_dist  = R_knn_dists[blockIdx.x * k + cur_k];
-    value_idx cur_R_ind = R_knn_inds[blockIdx.x * k + cur_k];
-
-    // Equation (2) in Cayton's paper- prune out R's which are > 3 * p(q, r_q)
-    if (cur_R_dist > weight * (min_R_dist + R_radius[cur_R_ind])) continue;
-    if (cur_R_dist > 3 * min_R_dist) return;
-
-    // The whole warp should iterate through the elements in the current R
-    value_idx R_start_offset = R_indptr[cur_R_ind];
-    value_idx R_stop_offset  = R_indptr[cur_R_ind + 1];
-
-    value_idx R_size = R_stop_offset - R_start_offset;
-
-    value_int limit = raft::Pow2<WarpSize>::roundDown(R_size);
-    value_int i     = threadIdx.x;
-    for (; i < limit; i += tpb) {
-      // Index and distance of current candidate's nearest landmark
-      value_idx cur_candidate_ind = R_1nn_cols[R_start_offset + i];
-      value_t cur_candidate_dist  = R_1nn_dists[R_start_offset + i];
-
-      // Take 2 landmarks l_1 and l_2 where l_1 is the furthest point in the heap
-      // and l_2 is the current landmark R. s is the current data point and
-      // t is the new candidate data point. We know that:
-      // d(s, t) cannot possibly be any smaller than | d(s, l_1) - d(l_1, l_2) | * | d(l_1, l_2) -
-      // d(l_2, t) | - d(s, l_1) * d(l_2, t)
-
-      // Therefore, if d(s, t) >= d(s, l_1) from the computation above, we know that the distance to
-      // the candidate point cannot possibly be in the nearest neighbors. However, if d(s, t) < d(s,
-      // l_1) then we should compute the distance because it's possible it could be smaller.
-      //
-      value_t z = heap.warpKTopRDist == 0.00 ? 0.0
-                                             : (abs(heap.warpKTop - heap.warpKTopRDist) *
-                                                  abs(heap.warpKTopRDist - cur_candidate_dist) -
-                                                heap.warpKTop * cur_candidate_dist) /
-                                                 heap.warpKTopRDist;
-
-      z            = isnan(z) || isinf(z) ? 0.0 : z;
-      value_t dist = std::numeric_limits<value_t>::max();
-
-      if (z <= heap.warpKTop) {
-        const value_t* y_ptr = X_index + (n_cols * cur_candidate_ind);
-        value_t local_y_ptr[col_q];
-        for (value_int j = 0; j < n_cols; ++j) {
-          local_y_ptr[j] = y_ptr[j];
-        }
-        dist = dfunc(local_x_ptr, local_y_ptr, n_cols);
-        ++n_dists_computed;
-      }
-
-      heap.add(dist, cur_candidate_dist, cur_candidate_ind);
-    }
-
-    if (i < R_size) {
-      value_idx cur_candidate_ind = R_1nn_cols[R_start_offset + i];
-      value_t cur_candidate_dist  = R_1nn_dists[R_start_offset + i];
-      value_t z                   = heap.warpKTopRDist == 0.0 ? 0.0
-                                                              : (abs(heap.warpKTop - heap.warpKTopRDist) *
-                                                 abs(heap.warpKTopRDist - cur_candidate_dist) -
-                                               heap.warpKTop * cur_candidate_dist) /
-                                                heap.warpKTopRDist;
-
-      z            = isnan(z) || isinf(z) ? 0.0 : z;
-      value_t dist = std::numeric_limits<value_t>::max();
-
-      if (z <= heap.warpKTop) {
-        const value_t* y_ptr = X_index + (n_cols * cur_candidate_ind);
-        value_t local_y_ptr[col_q];
-        for (value_int j = 0; j < n_cols; ++j) {
-          local_y_ptr[j] = y_ptr[j];
-        }
-        dist = dfunc(local_x_ptr, local_y_ptr, n_cols);
-        ++n_dists_computed;
-      }
-
-      heap.addThreadQ(dist, cur_candidate_dist, cur_candidate_ind);
-    }
-
-    heap.checkThreadQ();
-  }
-
-  heap.reduce();
-
-  for (int i = threadIdx.x; i < k; i += tpb) {
-    out_dists[blockIdx.x * k + i] = shared_memK[i];
-    out_inds[blockIdx.x * k + i]  = shared_memV[i].value;
-  }
-}
-
-template <typename value_idx,
-          typename value_t,
-          typename value_int = std::uint32_t,
-          int dims           = 2,
-          typename dist_func>
-void rbc_low_dim_pass_one(raft::resources const& handle,
-                          const BallCoverIndex<value_idx, value_t, value_int>& index,
-                          const value_t* query,
-                          const value_int n_query_rows,
-                          value_int k,
-                          const value_idx* R_knn_inds,
-                          const value_t* R_knn_dists,
-                          dist_func& dfunc,
-                          value_idx* inds,
-                          value_t* dists,
-                          float weight,
-                          value_int* dists_counter)
-{
-  if (k <= 32)
-    block_rbc_kernel_registers<value_idx, value_t, 32, 2, 128, dims, value_int>
-      <<<n_query_rows, 128, 0, resource::get_cuda_stream(handle)>>>(
-        index.get_X().data_handle(),
-        query,
-        index.n,
-        R_knn_inds,
-        R_knn_dists,
-        index.m,
-        k,
-        index.get_R_indptr().data_handle(),
-        index.get_R_1nn_cols().data_handle(),
-        index.get_R_1nn_dists().data_handle(),
-        inds,
-        dists,
-        dists_counter,
-        index.get_R_radius().data_handle(),
-        dfunc,
-        weight);
-
-  else if (k <= 64)
-    block_rbc_kernel_registers<value_idx, value_t, 64, 3, 128, 2, value_int>
-      <<<n_query_rows, 128, 0, resource::get_cuda_stream(handle)>>>(
-        index.get_X().data_handle(),
-        query,
-        index.n,
-        R_knn_inds,
-        R_knn_dists,
-        index.m,
-        k,
-        index.get_R_indptr().data_handle(),
-        index.get_R_1nn_cols().data_handle(),
-        index.get_R_1nn_dists().data_handle(),
-        inds,
-        dists,
-        dists_counter,
-        index.get_R_radius().data_handle(),
-        dfunc,
-        weight);
-  else if (k <= 128)
-    block_rbc_kernel_registers<value_idx, value_t, 128, 3, 128, dims, value_int>
-      <<<n_query_rows, 128, 0, resource::get_cuda_stream(handle)>>>(
-        index.get_X().data_handle(),
-        query,
-        index.n,
-        R_knn_inds,
-        R_knn_dists,
-        index.m,
-        k,
-        index.get_R_indptr().data_handle(),
-        index.get_R_1nn_cols().data_handle(),
-        index.get_R_1nn_dists().data_handle(),
-        inds,
-        dists,
-        dists_counter,
-        index.get_R_radius().data_handle(),
-        dfunc,
-        weight);
-
-  else if (k <= 256)
-    block_rbc_kernel_registers<value_idx, value_t, 256, 4, 128, dims, value_int>
-      <<<n_query_rows, 128, 0, resource::get_cuda_stream(handle)>>>(
-        index.get_X().data_handle(),
-        query,
-        index.n,
-        R_knn_inds,
-        R_knn_dists,
-        index.m,
-        k,
-        index.get_R_indptr().data_handle(),
-        index.get_R_1nn_cols().data_handle(),
-        index.get_R_1nn_dists().data_handle(),
-        inds,
-        dists,
-        dists_counter,
-        index.get_R_radius().data_handle(),
-        dfunc,
-        weight);
-
-  else if (k <= 512)
-    block_rbc_kernel_registers<value_idx, value_t, 512, 8, 64, dims, value_int>
-      <<<n_query_rows, 64, 0, resource::get_cuda_stream(handle)>>>(
-        index.get_X().data_handle(),
-        query,
-        index.n,
-        R_knn_inds,
-        R_knn_dists,
-        index.m,
-        k,
-        index.get_R_indptr().data_handle(),
-        index.get_R_1nn_cols().data_handle(),
-        index.get_R_1nn_dists().data_handle(),
-        inds,
-        dists,
-        dists_counter,
-        index.get_R_radius().data_handle(),
-        dfunc,
-        weight);
-
-  else if (k <= 1024)
-    block_rbc_kernel_registers<value_idx, value_t, 1024, 8, 64, dims, value_int>
-      <<<n_query_rows, 64, 0, resource::get_cuda_stream(handle)>>>(
-        index.get_X().data_handle(),
-        query,
-        index.n,
-        R_knn_inds,
-        R_knn_dists,
-        index.m,
-        k,
-        index.get_R_indptr().data_handle(),
-        index.get_R_1nn_cols().data_handle(),
-        index.get_R_1nn_dists().data_handle(),
-        inds,
-        dists,
-        dists_counter,
-        index.get_R_radius().data_handle(),
-        dfunc,
-        weight);
-}
-
-template <typename value_idx,
-          typename value_t,
-          typename value_int = std::uint32_t,
-          int dims           = 2,
-          typename dist_func>
-void rbc_low_dim_pass_two(raft::resources const& handle,
-                          const BallCoverIndex<value_idx, value_t, value_int>& index,
-                          const value_t* query,
-                          const value_int n_query_rows,
-                          value_int k,
-                          const value_idx* R_knn_inds,
-                          const value_t* R_knn_dists,
-                          dist_func& dfunc,
-                          value_idx* inds,
-                          value_t* dists,
-                          float weight,
-                          value_int* post_dists_counter)
-{
-  const value_int bitset_size = ceil(index.n_landmarks / 32.0);
-
-  rmm::device_uvector<std::uint32_t> bitset(bitset_size * n_query_rows,
-                                            resource::get_cuda_stream(handle));
-  thrust::fill(
-    resource::get_thrust_policy(handle), bitset.data(), bitset.data() + bitset.size(), 0);
-
-  perform_post_filter_registers<value_idx, value_t, value_int, dims, 128>
-    <<<n_query_rows, 128, bitset_size * sizeof(std::uint32_t), resource::get_cuda_stream(handle)>>>(
-      query,
-      index.n,
-      R_knn_inds,
-      R_knn_dists,
-      index.get_R_radius().data_handle(),
-      index.get_R().data_handle(),
-      index.n_landmarks,
-      bitset_size,
-      k,
-      dfunc,
-      bitset.data(),
-      weight);
-
-  if (k <= 32)
-    compute_final_dists_registers<value_idx,
-                                  value_t,
-                                  value_int,
-                                  std::uint32_t,
-                                  dist_func,
-                                  32,
-                                  2,
-                                  128,
-                                  dims>
-      <<<n_query_rows, 128, 0, resource::get_cuda_stream(handle)>>>(
-        index.get_X().data_handle(),
-        query,
-        index.n,
-        bitset.data(),
-        bitset_size,
-        index.get_R_closest_landmark_dists().data_handle(),
-        index.get_R_indptr().data_handle(),
-        index.get_R_1nn_cols().data_handle(),
-        index.get_R_1nn_dists().data_handle(),
-        inds,
-        dists,
-        index.n_landmarks,
-        k,
-        dfunc,
-        post_dists_counter);
-  else if (k <= 64)
-    compute_final_dists_registers<value_idx,
-                                  value_t,
-                                  value_int,
-                                  std::uint32_t,
-                                  dist_func,
-                                  64,
-                                  3,
-                                  128,
-                                  dims>
-      <<<n_query_rows, 128, 0, resource::get_cuda_stream(handle)>>>(
-        index.get_X().data_handle(),
-        query,
-        index.n,
-        bitset.data(),
-        bitset_size,
-        index.get_R_closest_landmark_dists().data_handle(),
-        index.get_R_indptr().data_handle(),
-        index.get_R_1nn_cols().data_handle(),
-        index.get_R_1nn_dists().data_handle(),
-        inds,
-        dists,
-        index.n_landmarks,
-        k,
-        dfunc,
-        post_dists_counter);
-  else if (k <= 128)
-    compute_final_dists_registers<value_idx,
-                                  value_t,
-                                  value_int,
-                                  std::uint32_t,
-                                  dist_func,
-                                  128,
-                                  3,
-                                  128,
-                                  dims>
-      <<<n_query_rows, 128, 0, resource::get_cuda_stream(handle)>>>(
-        index.get_X().data_handle(),
-        query,
-        index.n,
-        bitset.data(),
-        bitset_size,
-        index.get_R_closest_landmark_dists().data_handle(),
-        index.get_R_indptr().data_handle(),
-        index.get_R_1nn_cols().data_handle(),
-        index.get_R_1nn_dists().data_handle(),
-        inds,
-        dists,
-        index.n_landmarks,
-        k,
-        dfunc,
-        post_dists_counter);
-  else if (k <= 256)
-    compute_final_dists_registers<value_idx,
-                                  value_t,
-                                  value_int,
-                                  std::uint32_t,
-                                  dist_func,
-                                  256,
-                                  4,
-                                  128,
-                                  dims>
-      <<<n_query_rows, 128, 0, resource::get_cuda_stream(handle)>>>(
-        index.get_X().data_handle(),
-        query,
-        index.n,
-        bitset.data(),
-        bitset_size,
-        index.get_R_closest_landmark_dists().data_handle(),
-        index.get_R_indptr().data_handle(),
-        index.get_R_1nn_cols().data_handle(),
-        index.get_R_1nn_dists().data_handle(),
-        inds,
-        dists,
-        index.n_landmarks,
-        k,
-        dfunc,
-        post_dists_counter);
-  else if (k <= 512)
-    compute_final_dists_registers<value_idx,
-                                  value_t,
-                                  value_int,
-                                  std::uint32_t,
-                                  dist_func,
-                                  512,
-                                  8,
-                                  64,
-                                  dims><<<n_query_rows, 64, 0, resource::get_cuda_stream(handle)>>>(
-      index.get_X().data_handle(),
-      query,
-      index.n,
-      bitset.data(),
-      bitset_size,
-      index.get_R_closest_landmark_dists().data_handle(),
-      index.get_R_indptr().data_handle(),
-      index.get_R_1nn_cols().data_handle(),
-      index.get_R_1nn_dists().data_handle(),
-      inds,
-      dists,
-      index.n_landmarks,
-      k,
-      dfunc,
-      post_dists_counter);
-  else if (k <= 1024)
-    compute_final_dists_registers<value_idx,
-                                  value_t,
-                                  value_int,
-                                  std::uint32_t,
-                                  dist_func,
-                                  1024,
-                                  8,
-                                  64,
-                                  dims><<<n_query_rows, 64, 0, resource::get_cuda_stream(handle)>>>(
-      index.get_X().data_handle(),
-      query,
-      index.n,
-      bitset.data(),
-      bitset_size,
-      index.get_R_closest_landmark_dists().data_handle(),
-      index.get_R_indptr().data_handle(),
-      index.get_R_1nn_cols().data_handle(),
-      index.get_R_1nn_dists().data_handle(),
-      inds,
-      dists,
-      index.n_landmarks,
-      k,
-      dfunc,
-      post_dists_counter);
-}
-
-};  // namespace detail
-};  // namespace knn
-};  // namespace spatial
-};  // namespace cuvs
diff --git a/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers.cuh b/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers.cuh
deleted file mode 100644
index 8bd57b47c..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers.cuh
+++ /dev/null
@@ -1,24 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#ifndef RAFT_EXPLICIT_INSTANTIATE_ONLY
-#include "registers-inl.cuh"
-#endif
-
-#ifdef RAFT_COMPILED
-#include "registers-ext.cuh"
-#endif
diff --git a/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers_types.cuh b/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers_types.cuh
deleted file mode 100644
index 792f46828..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/ball_cover/registers_types.cuh
+++ /dev/null
@@ -1,66 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include "../haversine_distance.cuh"  // compute_haversine
-#include <cstdint>                    // uint32_t
-
-namespace cuvs {
-namespace spatial {
-namespace knn {
-namespace detail {
-
-template <typename value_t, typename value_int = std::uint32_t>
-struct DistFunc {
-  virtual __device__ __host__ __forceinline__ value_t operator()(const value_t* a,
-                                                                 const value_t* b,
-                                                                 const value_int n_dims)
-  {
-    return -1;
-  };
-};
-
-template <typename value_t, typename value_int = std::uint32_t>
-struct HaversineFunc : public DistFunc<value_t, value_int> {
-  __device__ __host__ __forceinline__ value_t operator()(const value_t* a,
-                                                         const value_t* b,
-                                                         const value_int n_dims) override
-  {
-    return cuvs::spatial::knn::detail::compute_haversine(a[0], b[0], a[1], b[1]);
-  }
-};
-
-template <typename value_t, typename value_int = std::uint32_t>
-struct EuclideanFunc : public DistFunc<value_t, value_int> {
-  __device__ __host__ __forceinline__ value_t operator()(const value_t* a,
-                                                         const value_t* b,
-                                                         const value_int n_dims) override
-  {
-    value_t sum_sq = 0;
-    for (value_int i = 0; i < n_dims; ++i) {
-      value_t diff = a[i] - b[i];
-      sum_sq += diff * diff;
-    }
-
-    return raft::sqrt(sum_sq);
-  }
-};
-
-};  // namespace detail
-};  // namespace knn
-};  // namespace spatial
-};  // namespace cuvs
diff --git a/cpp/include/cuvs/spatial/knn/detail/epsilon_neighborhood.cuh b/cpp/include/cuvs/spatial/knn/detail/epsilon_neighborhood.cuh
deleted file mode 100644
index 55f838c56..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/epsilon_neighborhood.cuh
+++ /dev/null
@@ -1,241 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <raft/linalg/contractions.cuh>
-#include <raft/util/device_utils.cuh>
-
-namespace cuvs {
-namespace spatial {
-namespace knn {
-namespace detail {
-
-template <typename DataT,
-          typename IdxT,
-          typename Policy,
-          typename BaseClass = raft::linalg::Contractions_NT<DataT, IdxT, Policy>>
-struct EpsUnexpL2SqNeighborhood : public BaseClass {
- private:
-  typedef Policy P;
-
-  bool* adj;
-  DataT eps;
-  IdxT* vd;
-
-  char* smem;  // for final reductions
-
-  DataT acc[P::AccRowsPerTh][P::AccColsPerTh];
-
- public:
-  DI EpsUnexpL2SqNeighborhood(bool* _adj,
-                              IdxT* _vd,
-                              const DataT* _x,
-                              const DataT* _y,
-                              IdxT _m,
-                              IdxT _n,
-                              IdxT _k,
-                              DataT _eps,
-                              char* _smem)
-    : BaseClass(_x, _y, _m, _n, _k, _smem), adj(_adj), eps(_eps), vd(_vd), smem(_smem)
-  {
-  }
-
-  DI void run()
-  {
-    prolog();
-    loop();
-    epilog();
-  }
-
- private:
-  DI void prolog()
-  {
-    this->ldgXY(IdxT(blockIdx.x) * P::Mblk, IdxT(blockIdx.y) * P::Nblk, 0);
-#pragma unroll
-    for (int i = 0; i < P::AccRowsPerTh; ++i) {
-#pragma unroll
-      for (int j = 0; j < P::AccColsPerTh; ++j) {
-        acc[i][j] = BaseClass::Zero;
-      }
-    }
-    this->stsXY();
-    __syncthreads();
-    this->switch_write_buffer();
-  }
-
-  DI void loop()
-  {
-    for (int kidx = P::Kblk; kidx < this->k; kidx += P::Kblk) {
-      this->ldgXY(IdxT(blockIdx.x) * P::Mblk, IdxT(blockIdx.y) * P::Nblk, kidx);
-      accumulate();  // on the previous k-block
-      this->stsXY();
-      __syncthreads();
-      this->switch_write_buffer();
-      this->switch_read_buffer();
-    }
-    accumulate();  // last iteration
-  }
-
-  DI void epilog()
-  {
-    IdxT startx = blockIdx.x * P::Mblk + this->accrowid;
-    IdxT starty = blockIdx.y * P::Nblk + this->acccolid;
-    auto lid    = raft::laneId();
-    IdxT sums[P::AccColsPerTh];
-#pragma unroll
-    for (int j = 0; j < P::AccColsPerTh; ++j) {
-      sums[j] = 0;
-    }
-#pragma unroll
-    for (int i = 0; i < P::AccRowsPerTh; ++i) {
-      auto xid = startx + i * P::AccThRows;
-#pragma unroll
-      for (int j = 0; j < P::AccColsPerTh; ++j) {
-        auto yid      = starty + j * P::AccThCols;
-        auto is_neigh = acc[i][j] <= eps;
-        ///@todo: fix uncoalesced writes using shared mem
-        if (xid < this->m && yid < this->n) {
-          adj[xid * this->n + yid] = is_neigh;
-          sums[j] += is_neigh;
-        }
-      }
-    }
-    // perform reduction of adjacency values to compute vertex degrees
-    if (vd != nullptr) { updateVertexDegree(sums); }
-  }
-
-  DI void accumulate()
-  {
-#pragma unroll
-    for (int ki = 0; ki < P::Kblk; ki += P::Veclen) {
-      this->ldsXY(ki);
-#pragma unroll
-      for (int i = 0; i < P::AccRowsPerTh; ++i) {
-#pragma unroll
-        for (int j = 0; j < P::AccColsPerTh; ++j) {
-#pragma unroll
-          for (int v = 0; v < P::Veclen; ++v) {
-            auto diff = this->regx[i][v] - this->regy[j][v];
-            acc[i][j] += diff * diff;
-          }
-        }
-      }
-    }
-  }
-
-  DI void updateVertexDegree(IdxT (&sums)[P::AccColsPerTh])
-  {
-    __syncthreads();  // so that we can safely reuse smem
-    int gid       = threadIdx.x / P::AccThCols;
-    int lid       = threadIdx.x % P::AccThCols;
-    auto cidx     = IdxT(blockIdx.y) * P::Nblk + lid;
-    IdxT totalSum = 0;
-    // update the individual vertex degrees
-#pragma unroll
-    for (int i = 0; i < P::AccColsPerTh; ++i) {
-      sums[i]  = batchedBlockReduce<IdxT, P::AccThCols>(sums[i], smem);
-      auto cid = cidx + i * P::AccThCols;
-      if (gid == 0 && cid < this->n) {
-        atomicUpdate(cid, sums[i]);
-        totalSum += sums[i];
-      }
-      __syncthreads();  // for safe smem reuse
-    }
-    // update the total edge count
-    totalSum = raft::blockReduce<IdxT>(totalSum, smem);
-    if (threadIdx.x == 0) { atomicUpdate(this->n, totalSum); }
-  }
-
-  DI void atomicUpdate(IdxT addrId, IdxT val)
-  {
-    if (sizeof(IdxT) == 4) {
-      raft::myAtomicAdd<unsigned>((unsigned*)(vd + addrId), val);
-    } else if (sizeof(IdxT) == 8) {
-      raft::myAtomicAdd<unsigned long long>((unsigned long long*)(vd + addrId), val);
-    }
-  }
-};  // struct EpsUnexpL2SqNeighborhood
-
-template <typename DataT, typename IdxT, typename Policy>
-__launch_bounds__(Policy::Nthreads, 2) RAFT_KERNEL epsUnexpL2SqNeighKernel(
-  bool* adj, IdxT* vd, const DataT* x, const DataT* y, IdxT m, IdxT n, IdxT k, DataT eps)
-{
-  extern __shared__ char smem[];
-  EpsUnexpL2SqNeighborhood<DataT, IdxT, Policy> obj(adj, vd, x, y, m, n, k, eps, smem);
-  obj.run();
-}
-
-template <typename DataT, typename IdxT, int VecLen>
-void epsUnexpL2SqNeighImpl(bool* adj,
-                           IdxT* vd,
-                           const DataT* x,
-                           const DataT* y,
-                           IdxT m,
-                           IdxT n,
-                           IdxT k,
-                           DataT eps,
-                           cudaStream_t stream)
-{
-  typedef typename raft::linalg::Policy4x4<DataT, VecLen>::Policy Policy;
-  dim3 grid(raft::ceildiv<int>(m, Policy::Mblk), raft::ceildiv<int>(n, Policy::Nblk));
-  dim3 blk(Policy::Nthreads);
-  epsUnexpL2SqNeighKernel<DataT, IdxT, Policy>
-    <<<grid, blk, Policy::SmemSize, stream>>>(adj, vd, x, y, m, n, k, eps);
-  RAFT_CUDA_TRY(cudaGetLastError());
-}
-
-/**
- * @brief Computes epsilon neighborhood for the L2-Squared distance metric
- *
- * @tparam DataT   IO and math type
- * @tparam IdxT    Index type
- *
- * @param[out] adj    adjacency matrix [row-major] [on device] [dim = m x n]
- * @param[out] vd     vertex degree array [on device] [len = m + 1]
- *                    `vd + m` stores the total number of edges in the adjacency
- *                    matrix. Pass a nullptr if you don't need this info.
- * @param[in]  x      first matrix [row-major] [on device] [dim = m x k]
- * @param[in]  y      second matrix [row-major] [on device] [dim = n x k]
- * @param[in]  eps    defines epsilon neighborhood radius (should be passed as
- *                    squared as we compute L2-squared distance in this method)
- * @param[in]  fop    device lambda to do any other custom functions
- * @param[in]  stream cuda stream
- */
-template <typename DataT, typename IdxT>
-void epsUnexpL2SqNeighborhood(bool* adj,
-                              IdxT* vd,
-                              const DataT* x,
-                              const DataT* y,
-                              IdxT m,
-                              IdxT n,
-                              IdxT k,
-                              DataT eps,
-                              cudaStream_t stream)
-{
-  size_t bytes = sizeof(DataT) * k;
-  if (16 % sizeof(DataT) == 0 && bytes % 16 == 0) {
-    epsUnexpL2SqNeighImpl<DataT, IdxT, 16 / sizeof(DataT)>(adj, vd, x, y, m, n, k, eps, stream);
-  } else if (8 % sizeof(DataT) == 0 && bytes % 8 == 0) {
-    epsUnexpL2SqNeighImpl<DataT, IdxT, 8 / sizeof(DataT)>(adj, vd, x, y, m, n, k, eps, stream);
-  } else {
-    epsUnexpL2SqNeighImpl<DataT, IdxT, 1>(adj, vd, x, y, m, n, k, eps, stream);
-  }
-}
-}  // namespace detail
-}  // namespace knn
-}  // namespace spatial
-}  // namespace cuvs
\ No newline at end of file
diff --git a/cpp/include/cuvs/spatial/knn/detail/fused_l2_knn-ext.cuh b/cpp/include/cuvs/spatial/knn/detail/fused_l2_knn-ext.cuh
deleted file mode 100644
index 93ad6737d..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/fused_l2_knn-ext.cuh
+++ /dev/null
@@ -1,74 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#include <cstddef>                           // size_t
-#include <cstdint>                           // uint32_t
-#include <cuvs/distance/distance_types.hpp>  // DistanceType
-#include <raft/util/raft_explicit.hpp>       // RAFT_EXPLICIT
-
-#if defined(RAFT_EXPLICIT_INSTANTIATE_ONLY)
-
-namespace cuvs::spatial::knn::detail {
-
-template <typename value_idx, typename value_t, bool usePrevTopKs = false>
-void fusedL2Knn(size_t D,
-                value_idx* out_inds,
-                value_t* out_dists,
-                const value_t* index,
-                const value_t* query,
-                size_t n_index_rows,
-                size_t n_query_rows,
-                int k,
-                bool rowMajorIndex,
-                bool rowMajorQuery,
-                cudaStream_t stream,
-                cuvs::distance::DistanceType metric,
-                const value_t* index_norms = NULL,
-                const value_t* query_norms = NULL) RAFT_EXPLICIT;
-
-}  // namespace cuvs::spatial::knn::detail
-
-#endif  // RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-#define instantiate_raft_spatial_knn_detail_fusedL2Knn(Mvalue_idx, Mvalue_t, MusePrevTopKs) \
-  extern template void                                                                      \
-  cuvs::spatial::knn::detail::fusedL2Knn<Mvalue_idx, Mvalue_t, MusePrevTopKs>(              \
-    size_t D,                                                                               \
-    Mvalue_idx * out_inds,                                                                  \
-    Mvalue_t * out_dists,                                                                   \
-    const Mvalue_t* index,                                                                  \
-    const Mvalue_t* query,                                                                  \
-    size_t n_index_rows,                                                                    \
-    size_t n_query_rows,                                                                    \
-    int k,                                                                                  \
-    bool rowMajorIndex,                                                                     \
-    bool rowMajorQuery,                                                                     \
-    cudaStream_t stream,                                                                    \
-    cuvs::distance::DistanceType metric,                                                    \
-    const Mvalue_t* index_norms,                                                            \
-    const Mvalue_t* query_norms);
-
-instantiate_raft_spatial_knn_detail_fusedL2Knn(int32_t, float, true);
-instantiate_raft_spatial_knn_detail_fusedL2Knn(int32_t, float, false);
-instantiate_raft_spatial_knn_detail_fusedL2Knn(int64_t, float, true);
-instantiate_raft_spatial_knn_detail_fusedL2Knn(int64_t, float, false);
-
-// These are used by brute_force_knn:
-instantiate_raft_spatial_knn_detail_fusedL2Knn(uint32_t, float, true);
-instantiate_raft_spatial_knn_detail_fusedL2Knn(uint32_t, float, false);
-
-#undef instantiate_raft_spatial_knn_detail_fusedL2Knn
diff --git a/cpp/include/cuvs/spatial/knn/detail/fused_l2_knn-inl.cuh b/cpp/include/cuvs/spatial/knn/detail/fused_l2_knn-inl.cuh
deleted file mode 100644
index e4ac0fb54..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/fused_l2_knn-inl.cuh
+++ /dev/null
@@ -1,1062 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-#include <cub/cub.cuh>
-#include <cuvs/neighbors/detail/faiss_select/Select.cuh>
-#include <limits>
-#include <raft/linalg/norm.cuh>
-// TODO: Need to hide the PairwiseDistance class impl and expose to public API
-#include "processing.cuh"
-#include <cuvs/distance/detail/distance.cuh>
-#include <cuvs/distance/detail/distance_ops/l2_exp.cuh>
-#include <cuvs/distance/detail/distance_ops/l2_unexp.cuh>
-#include <cuvs/distance/detail/pairwise_distance_base.cuh>
-#include <raft/core/operators.hpp>
-#include <raft/util/cuda_utils.cuh>
-
-namespace cuvs {
-namespace spatial {
-namespace knn {
-namespace detail {
-
-template <typename Policy, typename Pair, typename myWarpSelect, typename IdxT>
-DI void loadAllWarpQShmem(myWarpSelect** heapArr,
-                          Pair* shDumpKV,
-                          const IdxT m,
-                          const unsigned int numOfNN)
-{
-  const int lid = raft::laneId();
-#pragma unroll
-  for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-    const auto rowId = (threadIdx.x / Policy::AccThCols) + i * Policy::AccThRows;
-    if (rowId < m) {
-#pragma unroll
-      for (int j = 0; j < myWarpSelect::kNumWarpQRegisters; ++j) {
-        const int idx = j * warpSize + lid;
-        if (idx < numOfNN) {
-          Pair KVPair          = shDumpKV[rowId * numOfNN + idx];
-          heapArr[i]->warpV[j] = KVPair.key;
-          heapArr[i]->warpK[j] = KVPair.value;
-        }
-      }
-    }
-  }
-}
-
-template <typename Policy, typename Pair, typename myWarpSelect>
-DI void loadWarpQShmem(myWarpSelect* heapArr,
-                       Pair* shDumpKV,
-                       const int rowId,
-                       const unsigned int numOfNN)
-{
-  const int lid = raft::laneId();
-#pragma unroll
-  for (int j = 0; j < myWarpSelect::kNumWarpQRegisters; ++j) {
-    const int idx = j * warpSize + lid;
-    if (idx < numOfNN) {
-      Pair KVPair       = shDumpKV[rowId * numOfNN + idx];
-      heapArr->warpV[j] = KVPair.key;
-      heapArr->warpK[j] = KVPair.value;
-    }
-  }
-}
-
-template <typename Policy, typename Pair, typename myWarpSelect, typename IdxT>
-DI void storeWarpQShmem(myWarpSelect* heapArr,
-                        Pair* shDumpKV,
-                        const IdxT rowId,
-                        const unsigned int numOfNN)
-{
-  const int lid = raft::laneId();
-
-#pragma unroll
-  for (int j = 0; j < myWarpSelect::kNumWarpQRegisters; ++j) {
-    const int idx = j * warpSize + lid;
-    if (idx < numOfNN) {
-      Pair otherKV                    = Pair(heapArr->warpV[j], heapArr->warpK[j]);
-      shDumpKV[rowId * numOfNN + idx] = otherKV;
-    }
-  }
-}
-
-template <typename Policy, typename Pair, typename myWarpSelect, typename IdxT, typename OutT>
-DI void storeWarpQGmem(myWarpSelect** heapArr,
-                       volatile OutT* out_dists,
-                       volatile IdxT* out_inds,
-                       const IdxT m,
-                       const unsigned int numOfNN,
-                       const IdxT starty)
-{
-  const int lid = raft::laneId();
-#pragma unroll
-  for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-    const auto gmemRowId = starty + i * Policy::AccThRows;
-    if (gmemRowId < m) {
-#pragma unroll
-      for (int j = 0; j < myWarpSelect::kNumWarpQRegisters; ++j) {
-        const auto idx = j * warpSize + lid;
-        if (idx < numOfNN) {
-          out_dists[std::size_t(gmemRowId) * numOfNN + idx] = heapArr[i]->warpK[j];
-          out_inds[std::size_t(gmemRowId) * numOfNN + idx]  = (IdxT)heapArr[i]->warpV[j];
-        }
-      }
-    }
-  }
-}
-
-template <typename Policy, typename Pair, typename myWarpSelect, typename IdxT, typename OutT>
-DI void loadPrevTopKsGmemWarpQ(myWarpSelect** heapArr,
-                               volatile OutT* out_dists,
-                               volatile IdxT* out_inds,
-                               const IdxT m,
-                               const unsigned int numOfNN,
-                               const IdxT starty)
-{
-  const int lid = raft::laneId();
-#pragma unroll
-  for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-    const auto gmemRowId = starty + i * Policy::AccThRows;
-    if (gmemRowId < m) {
-#pragma unroll
-      for (int j = 0; j < myWarpSelect::kNumWarpQRegisters; ++j) {
-        const auto idx = j * warpSize + lid;
-        if (idx < numOfNN) {
-          heapArr[i]->warpK[j] = out_dists[std::size_t(gmemRowId) * numOfNN + idx];
-          heapArr[i]->warpV[j] = (uint32_t)out_inds[std::size_t(gmemRowId) * numOfNN + idx];
-        }
-      }
-      static constexpr auto kLaneWarpKTop = myWarpSelect::kNumWarpQRegisters - 1;
-      heapArr[i]->warpKTop = raft::shfl(heapArr[i]->warpK[kLaneWarpKTop], heapArr[i]->kLane);
-    }
-  }
-}
-
-template <typename Pair, int NumWarpQRegs, typename myWarpSelect>
-DI void updateSortedWarpQ(
-  myWarpSelect& heapArr, Pair* allWarpTopKs, int rowId, int finalNumVals, int startId = 0)
-{
-  constexpr uint32_t mask = 0xffffffffu;
-  const int lid           = raft::laneId();
-  // calculate srcLane such that tid 0 -> 31, 1 -> 0,... 31 -> 30.
-  // warp around 0 to 31 required for NN > 32
-  const auto srcLane = (warpSize + (lid - 1)) & (warpSize - 1);
-
-  for (int k = startId; k < finalNumVals; k++) {
-    Pair KVPair = allWarpTopKs[rowId * (256) + k];
-#pragma unroll
-    for (int i = 0; i < NumWarpQRegs; i++) {
-      unsigned activeLanes = __ballot_sync(mask, KVPair.value < heapArr->warpK[i]);
-      if (activeLanes) {
-        Pair tempKV;
-        tempKV.value               = raft::shfl(heapArr->warpK[i], srcLane);
-        tempKV.key                 = raft::shfl(heapArr->warpV[i], srcLane);
-        const auto firstActiveLane = __ffs(activeLanes) - 1;
-        if (firstActiveLane == lid) {
-          heapArr->warpK[i] = KVPair.value;
-          heapArr->warpV[i] = KVPair.key;
-        } else if (lid > firstActiveLane) {
-          heapArr->warpK[i] = tempKV.value;
-          heapArr->warpV[i] = tempKV.key;
-        }
-        if (i == 0 && NumWarpQRegs > 1) {
-          heapArr->warpK[1] = __shfl_up_sync(mask, heapArr->warpK[1], 1);
-          heapArr->warpV[1] = __shfl_up_sync(mask, heapArr->warpV[1], 1);
-          if (lid == 0) {
-            heapArr->warpK[1] = tempKV.value;
-            heapArr->warpV[1] = tempKV.key;
-          }
-          break;
-        }
-      }
-    }
-  }
-}
-
-template <typename DataT,
-          typename OutT,
-          typename IdxT,
-          typename Policy,
-          typename OpT,
-          typename FinalLambda,
-          int NumWarpQ,
-          int NumThreadQ,
-          bool usePrevTopKs = false,
-          bool isRowMajor   = true>
-__launch_bounds__(Policy::Nthreads, 2) RAFT_KERNEL fusedL2kNN(const DataT* x,
-                                                              const DataT* y,
-                                                              const DataT* _xn,
-                                                              const DataT* _yn,
-                                                              const IdxT m,
-                                                              const IdxT n,
-                                                              const IdxT k,
-                                                              const IdxT lda,
-                                                              const IdxT ldb,
-                                                              const IdxT ldd,
-                                                              OpT distance_op,
-                                                              FinalLambda fin_op,
-                                                              unsigned int numOfNN,
-                                                              volatile int* mutexes,
-                                                              volatile OutT* out_dists,
-                                                              volatile IdxT* out_inds)
-{
-  using AccT = typename OpT::AccT;
-  extern __shared__ char smem[];
-
-  typedef cub::KeyValuePair<uint32_t, AccT> Pair;
-  constexpr auto identity = std::numeric_limits<AccT>::max();
-  constexpr auto keyMax   = std::numeric_limits<uint32_t>::max();
-  constexpr auto Dir      = false;
-  using namespace cuvs::neighbors::detail::faiss_select;
-  typedef WarpSelect<AccT, uint32_t, Dir, Comparator<AccT>, NumWarpQ, NumThreadQ, 32> myWarpSelect;
-
-  auto rowEpilog_lambda =
-    [m, n, &distance_op, numOfNN, out_dists, out_inds, mutexes] __device__(IdxT gridStrideY) {
-      if (gridDim.x == 1) { return; }
-
-      // Use ::template to disambiguate (See:
-      // https://en.cppreference.com/w/cpp/language/dependent_name)
-      int smem_offset = OpT::template shared_mem_size<Policy>();
-      Pair* shDumpKV  = (Pair*)(&smem[smem_offset]);
-
-      const int lid     = threadIdx.x % warpSize;
-      const IdxT starty = gridStrideY + (threadIdx.x / Policy::AccThCols);
-
-      //  0 -> consumer done consuming the buffer.
-      // -1 -> consumer started consuming the buffer
-      // -2 -> producer done filling the buffer
-      //  1 -> prod acquired to fill the buffer
-      if (blockIdx.x == 0) {
-        auto cta_processed = 0;
-        myWarpSelect heapArr1(identity, keyMax, numOfNN);
-        myWarpSelect heapArr2(identity, keyMax, numOfNN);
-        myWarpSelect* heapArr[] = {&heapArr1, &heapArr2};
-        __syncwarp();
-
-        loadAllWarpQShmem<Policy, Pair>(heapArr, &shDumpKV[0], m, numOfNN);
-
-        while (cta_processed < gridDim.x - 1) {
-          if (threadIdx.x == 0) {
-            while (atomicCAS((int*)&mutexes[gridStrideY / Policy::Mblk], -2, -1) != -2)
-              ;
-          }
-          __threadfence();
-          __syncthreads();
-
-#pragma unroll
-          for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-            const auto rowId = starty + i * Policy::AccThRows;
-            if (rowId < m) {
-#pragma unroll
-              for (int j = 0; j < myWarpSelect::kNumWarpQRegisters; ++j) {
-                Pair otherKV;
-                otherKV.value  = identity;
-                otherKV.key    = keyMax;
-                const auto idx = j * warpSize + lid;
-                if (idx < numOfNN) {
-                  otherKV.value         = out_dists[rowId * numOfNN + idx];
-                  otherKV.key           = (uint32_t)out_inds[rowId * numOfNN + idx];
-                  const auto shMemRowId = (threadIdx.x / Policy::AccThCols) + i * Policy::AccThRows;
-                  shDumpKV[shMemRowId * numOfNN + idx] = otherKV;
-                }
-              }
-            }
-          }
-          __threadfence();
-          __syncthreads();
-
-          if (threadIdx.x == 0) { atomicExch((int*)&mutexes[gridStrideY / Policy::Mblk], 0); }
-          __threadfence();
-
-        // Perform merging of otherKV with topk's across warp.
-#pragma unroll
-          for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-            const auto rowId = starty + i * Policy::AccThRows;
-            if (rowId < m) {
-#pragma unroll
-              for (int j = 0; j < myWarpSelect::kNumWarpQRegisters; ++j) {
-                Pair otherKV;
-                otherKV.value  = identity;
-                otherKV.key    = keyMax;
-                const auto idx = j * warpSize + lid;
-                if (idx < numOfNN) {
-                  const auto shMemRowId = (threadIdx.x / Policy::AccThCols) + i * Policy::AccThRows;
-                  otherKV               = shDumpKV[shMemRowId * numOfNN + idx];
-                }
-                heapArr[i]->add(otherKV.value, otherKV.key);
-              }
-            }
-          }
-          cta_processed++;
-        }
-#pragma unroll
-        for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-          const auto rowId = starty + i * Policy::AccThRows;
-          if (rowId < m) {
-            bool needSort = (heapArr[i]->numVals > 0);
-            needSort      = __any_sync(0xffffffff, needSort);
-            if (needSort) { heapArr[i]->reduce(); }
-          }
-        }
-        storeWarpQGmem<Policy, Pair>(heapArr, out_dists, out_inds, m, numOfNN, starty);
-      } else {
-        if (threadIdx.x == 0) {
-          while (atomicCAS((int*)&mutexes[gridStrideY / Policy::Mblk], 0, 1) != 0)
-            ;
-        }
-        __threadfence();
-        __syncthreads();
-
-#pragma unroll
-        for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-          const auto rowId = starty + i * Policy::AccThRows;
-          if (rowId < m) {
-            for (int idx = lid; idx < numOfNN; idx += warpSize) {
-              const auto shMemRowId = (threadIdx.x / Policy::AccThCols) + i * Policy::AccThRows;
-              Pair KVPair           = shDumpKV[shMemRowId * numOfNN + idx];
-              out_dists[rowId * numOfNN + idx] = KVPair.value;
-              out_inds[rowId * numOfNN + idx]  = (IdxT)KVPair.key;
-            }
-          }
-        }
-        __threadfence();
-        __syncthreads();
-
-        if (threadIdx.x == 0) { atomicExch((int*)&mutexes[gridStrideY / Policy::Mblk], -2); }
-        __threadfence();
-      }
-    };
-
-  // epilogue operation lambda for final value calculation
-  auto epilog_lambda =
-    [&distance_op, numOfNN, m, n, ldd, out_dists, out_inds, keyMax, identity] __device__(
-      AccT acc[Policy::AccRowsPerTh][Policy::AccColsPerTh],
-      DataT * regxn,
-      DataT * regyn,
-      IdxT gridStrideX,
-      IdxT gridStrideY) {
-      // Use ::template to disambiguate (See:
-      // https://en.cppreference.com/w/cpp/language/dependent_name)
-      int smem_offset = OpT::template shared_mem_size<Policy>();
-      Pair* shDumpKV  = (Pair*)(&smem[smem_offset]);
-
-      constexpr uint32_t mask = 0xffffffffu;
-      const IdxT starty       = gridStrideY + (threadIdx.x / Policy::AccThCols);
-      const IdxT startx       = gridStrideX + (threadIdx.x % Policy::AccThCols);
-      const int lid           = raft::laneId();
-
-      myWarpSelect heapArr1(identity, keyMax, numOfNN);
-      myWarpSelect heapArr2(identity, keyMax, numOfNN);
-      myWarpSelect* heapArr[] = {&heapArr1, &heapArr2};
-      if (usePrevTopKs) {
-        if (gridStrideX == blockIdx.x * Policy::Nblk) {
-          loadPrevTopKsGmemWarpQ<Policy, Pair>(heapArr, out_dists, out_inds, m, numOfNN, starty);
-        }
-      }
-
-      if (gridStrideX > blockIdx.x * Policy::Nblk) {
-#pragma unroll
-        for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-          const auto rowId     = (threadIdx.x / Policy::AccThCols) + i * Policy::AccThRows;
-          Pair tempKV          = shDumpKV[(rowId * numOfNN) + numOfNN - 1];
-          heapArr[i]->warpKTop = tempKV.value;
-        }
-
-        // total vals can atmost be 256, (32*8)
-        int numValsWarpTopK[Policy::AccRowsPerTh];
-        int anyWarpTopKs = 0;
-#pragma unroll
-        for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-          const auto rowId   = starty + i * Policy::AccThRows;
-          numValsWarpTopK[i] = 0;
-          if (rowId < m) {
-#pragma unroll
-            for (int j = 0; j < Policy::AccColsPerTh; ++j) {
-              const auto colId = startx + j * Policy::AccThCols;
-              if (colId < ldd) {
-                if (acc[i][j] < heapArr[i]->warpKTop) { numValsWarpTopK[i]++; }
-              }
-            }
-            anyWarpTopKs += numValsWarpTopK[i];
-          }
-        }
-        anyWarpTopKs = __syncthreads_or(anyWarpTopKs > 0);
-        if (anyWarpTopKs) {
-          Pair* allWarpTopKs = (Pair*)(&smem[0]);
-          uint32_t needScanSort[Policy::AccRowsPerTh];
-
-#pragma unroll
-          for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-            const auto gmemRowId = starty + i * Policy::AccThRows;
-            needScanSort[i]      = 0;
-            if (gmemRowId < m) {
-              int myVals      = numValsWarpTopK[i];
-              needScanSort[i] = __ballot_sync(mask, myVals > 0);
-              if (needScanSort[i]) {
-#pragma unroll
-                for (unsigned int k = 1; k <= 16; k *= 2) {
-                  const unsigned int n = __shfl_up_sync(mask, numValsWarpTopK[i], k);
-                  if (lid >= k) { numValsWarpTopK[i] += n; }
-                }
-              }
-              // As each thread will know its total vals to write.
-              // we only store its starting location.
-              numValsWarpTopK[i] -= myVals;
-            }
-
-            if (needScanSort[i]) {
-              const auto rowId = (threadIdx.x / Policy::AccThCols) + i * Policy::AccThRows;
-              if (gmemRowId < m) {
-                if (needScanSort[i] & ((uint32_t)1 << lid)) {
-#pragma unroll
-                  for (int j = 0; j < Policy::AccColsPerTh; ++j) {
-                    const auto colId = startx + j * Policy::AccThCols;
-                    if (colId < ldd) {
-                      if (acc[i][j] < heapArr[i]->warpKTop) {
-                        Pair otherKV                                     = {colId, acc[i][j]};
-                        allWarpTopKs[rowId * (256) + numValsWarpTopK[i]] = otherKV;
-                        numValsWarpTopK[i]++;
-                      }
-                    }
-                  }
-                }
-                __syncwarp();
-                const int finalNumVals = raft::shfl(numValsWarpTopK[i], 31);
-                loadWarpQShmem<Policy, Pair>(heapArr[i], &shDumpKV[0], rowId, numOfNN);
-                updateSortedWarpQ<Pair, myWarpSelect::kNumWarpQRegisters>(
-                  heapArr[i], &allWarpTopKs[0], rowId, finalNumVals);
-              }
-            }
-          }
-          __syncthreads();
-#pragma unroll
-          for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-            if (needScanSort[i]) {
-              const auto rowId     = (threadIdx.x / Policy::AccThCols) + i * Policy::AccThRows;
-              const auto gmemRowId = starty + i * Policy::AccThRows;
-              if (gmemRowId < m) {
-                storeWarpQShmem<Policy, Pair>(heapArr[i], shDumpKV, rowId, numOfNN);
-              }
-            }
-          }
-        }
-      } else {
-#pragma unroll
-        for (int i = 0; i < Policy::AccRowsPerTh; ++i) {
-          const auto gmemRowId  = starty + i * Policy::AccThRows;
-          const auto shMemRowId = (threadIdx.x / Policy::AccThCols) + i * Policy::AccThRows;
-          if (gmemRowId < m) {
-#pragma unroll
-            for (int j = 0; j < Policy::AccColsPerTh; ++j) {
-              const auto colId = startx + j * Policy::AccThCols;
-              Pair otherKV     = {keyMax, identity};
-              if (colId < ldd) {
-                otherKV.value = acc[i][j];
-                otherKV.key   = colId;
-              }
-              heapArr[i]->add(otherKV.value, otherKV.key);
-            }
-
-            bool needSort = (heapArr[i]->numVals > 0);
-            needSort      = __any_sync(mask, needSort);
-            if (needSort) { heapArr[i]->reduce(); }
-            storeWarpQShmem<Policy, Pair>(heapArr[i], shDumpKV, shMemRowId, numOfNN);
-          }
-        }
-      }
-
-      if (((gridStrideX + Policy::Nblk * gridDim.x) >= n) && gridDim.x == 1) {
-        // This is last iteration of grid stride X
-        loadAllWarpQShmem<Policy, Pair>(heapArr, &shDumpKV[0], m, numOfNN);
-        storeWarpQGmem<Policy, Pair>(heapArr, out_dists, out_inds, m, numOfNN, starty);
-      }
-    };
-
-  constexpr bool write_out = false;
-  cuvs::distance::detail::PairwiseDistances<DataT,
-                                            OutT,
-                                            IdxT,
-                                            Policy,
-                                            OpT,
-                                            decltype(epilog_lambda),
-                                            FinalLambda,
-                                            decltype(rowEpilog_lambda),
-                                            isRowMajor,
-                                            write_out>
-    obj(x,
-        y,
-        m,
-        n,
-        k,
-        lda,
-        ldb,
-        ldd,
-        _xn,
-        _yn,
-        nullptr,  // output ptr, can be null as write_out == false.
-        smem,
-        distance_op,
-        epilog_lambda,
-        fin_op,
-        rowEpilog_lambda);
-  obj.run();
-}
-
-template <typename DataT,
-          typename AccT,
-          typename OutT,
-          typename IdxT,
-          int VecLen,
-          bool usePrevTopKs,
-          bool isRowMajor>
-void fusedL2UnexpKnnImpl(const DataT* x,
-                         const DataT* y,
-                         IdxT m,
-                         IdxT n,
-                         IdxT k,
-                         IdxT lda,
-                         IdxT ldb,
-                         IdxT ldd,
-                         bool sqrt,
-                         OutT* out_dists,
-                         IdxT* out_inds,
-                         IdxT numOfNN,
-                         cudaStream_t stream,
-                         void* workspace,
-                         size_t& worksize)
-{
-  typedef typename raft::linalg::Policy2x8<DataT, 1>::Policy RowPolicy;
-  typedef typename raft::linalg::Policy4x4<DataT, VecLen>::ColPolicy ColPolicy;
-
-  typedef typename std::conditional<true, RowPolicy, ColPolicy>::type KPolicy;
-
-  ASSERT(isRowMajor, "Only Row major inputs are allowed");
-
-  dim3 blk(KPolicy::Nthreads);
-  // Accumulation operation lambda
-  typedef cub::KeyValuePair<uint32_t, AccT> Pair;
-
-  cuvs::distance::detail::ops::l2_unexp_distance_op<DataT, AccT, IdxT> distance_op{sqrt};
-  raft::identity_op fin_op{};
-
-  if constexpr (isRowMajor) {
-    constexpr auto fusedL2UnexpKnn32RowMajor = fusedL2kNN<DataT,
-                                                          OutT,
-                                                          IdxT,
-                                                          KPolicy,
-                                                          decltype(distance_op),
-                                                          decltype(fin_op),
-                                                          32,
-                                                          2,
-                                                          usePrevTopKs,
-                                                          isRowMajor>;
-    constexpr auto fusedL2UnexpKnn64RowMajor = fusedL2kNN<DataT,
-                                                          OutT,
-                                                          IdxT,
-                                                          KPolicy,
-                                                          decltype(distance_op),
-                                                          decltype(fin_op),
-                                                          64,
-                                                          3,
-                                                          usePrevTopKs,
-                                                          isRowMajor>;
-
-    auto fusedL2UnexpKnnRowMajor = fusedL2UnexpKnn32RowMajor;
-    if (numOfNN <= 32) {
-      fusedL2UnexpKnnRowMajor = fusedL2UnexpKnn32RowMajor;
-    } else if (numOfNN <= 64) {
-      fusedL2UnexpKnnRowMajor = fusedL2UnexpKnn64RowMajor;
-    } else {
-      ASSERT(numOfNN <= 64, "fusedL2kNN: num of nearest neighbors must be <= 64");
-    }
-
-    const auto sharedMemSize =
-      distance_op.template shared_mem_size<KPolicy>() + KPolicy::Mblk * numOfNN * sizeof(Pair);
-
-    dim3 grid = cuvs::distance::detail::launchConfigGenerator<KPolicy>(
-      m, n, sharedMemSize, fusedL2UnexpKnnRowMajor);
-
-    if (grid.x > 1) {
-      const auto numMutexes = raft::ceildiv<int>(m, KPolicy::Mblk);
-      if (workspace == nullptr || worksize < (sizeof(int32_t) * numMutexes)) {
-        worksize = sizeof(int32_t) * numMutexes;
-        return;
-      } else {
-        RAFT_CUDA_TRY(cudaMemsetAsync(workspace, 0, sizeof(int32_t) * numMutexes, stream));
-      }
-    }
-
-    fusedL2UnexpKnnRowMajor<<<grid, blk, sharedMemSize, stream>>>(x,
-                                                                  y,
-                                                                  nullptr,
-                                                                  nullptr,
-                                                                  m,
-                                                                  n,
-                                                                  k,
-                                                                  lda,
-                                                                  ldb,
-                                                                  ldd,
-                                                                  distance_op,
-                                                                  fin_op,
-                                                                  (uint32_t)numOfNN,
-                                                                  (int*)workspace,
-                                                                  out_dists,
-                                                                  out_inds);
-  } else {
-  }
-
-  RAFT_CUDA_TRY(cudaGetLastError());
-}
-
-template <typename DataT,
-          typename AccT,
-          typename OutT,
-          typename IdxT,
-          bool usePrevTopKs,
-          bool isRowMajor>
-void fusedL2UnexpKnn(IdxT m,
-                     IdxT n,
-                     IdxT k,
-                     IdxT lda,
-                     IdxT ldb,
-                     IdxT ldd,
-                     const DataT* x,
-                     const DataT* y,
-                     bool sqrt,
-                     OutT* out_dists,
-                     IdxT* out_inds,
-                     IdxT numOfNN,
-                     cudaStream_t stream,
-                     void* workspace,
-                     size_t& worksize)
-{
-  size_t bytesA = sizeof(DataT) * lda;
-  size_t bytesB = sizeof(DataT) * ldb;
-  if (16 % sizeof(DataT) == 0 && bytesA % 16 == 0 && bytesB % 16 == 0) {
-    fusedL2UnexpKnnImpl<DataT, AccT, OutT, IdxT, 16 / sizeof(DataT), usePrevTopKs, isRowMajor>(
-      x,
-      y,
-      m,
-      n,
-      k,
-      lda,
-      ldb,
-      ldd,
-      sqrt,
-      out_dists,
-      out_inds,
-      numOfNN,
-      stream,
-      workspace,
-      worksize);
-  } else if (8 % sizeof(DataT) == 0 && bytesA % 8 == 0 && bytesB % 8 == 0) {
-    fusedL2UnexpKnnImpl<DataT, AccT, OutT, IdxT, 8 / sizeof(DataT), usePrevTopKs, isRowMajor>(
-      x,
-      y,
-      m,
-      n,
-      k,
-      lda,
-      ldb,
-      ldd,
-      sqrt,
-      out_dists,
-      out_inds,
-      numOfNN,
-      stream,
-      workspace,
-      worksize);
-  } else {
-    fusedL2UnexpKnnImpl<DataT, AccT, OutT, IdxT, 1, usePrevTopKs, isRowMajor>(x,
-                                                                              y,
-                                                                              m,
-                                                                              n,
-                                                                              k,
-                                                                              lda,
-                                                                              ldb,
-                                                                              ldd,
-                                                                              sqrt,
-                                                                              out_dists,
-                                                                              out_inds,
-                                                                              numOfNN,
-                                                                              stream,
-                                                                              workspace,
-                                                                              worksize);
-  }
-}
-
-template <typename DataT,
-          typename AccT,
-          typename OutT,
-          typename IdxT,
-          int VecLen,
-          bool usePrevTopKs,
-          bool isRowMajor>
-void fusedL2ExpKnnImpl(const DataT* x,
-                       const DataT* y,
-                       const DataT* xn,
-                       const DataT* yn,
-                       IdxT m,
-                       IdxT n,
-                       IdxT k,
-                       IdxT lda,
-                       IdxT ldb,
-                       IdxT ldd,
-                       bool sqrt,
-                       OutT* out_dists,
-                       IdxT* out_inds,
-                       IdxT numOfNN,
-                       cudaStream_t stream,
-                       void* workspace,
-                       size_t& worksize)
-{
-  typedef typename raft::linalg::Policy2x8<DataT, 1>::Policy RowPolicy;
-  typedef typename raft::linalg::Policy4x4<DataT, VecLen>::ColPolicy ColPolicy;
-
-  typedef typename std::conditional<true, RowPolicy, ColPolicy>::type KPolicy;
-
-  ASSERT(isRowMajor, "Only Row major inputs are allowed");
-
-  ASSERT(!(((x != y) && (worksize < (m + n) * sizeof(AccT))) || (worksize < m * sizeof(AccT))),
-         "workspace size error");
-  ASSERT(workspace != nullptr, "workspace is null");
-
-  dim3 blk(KPolicy::Nthreads);
-
-  typedef cub::KeyValuePair<uint32_t, AccT> Pair;
-
-  cuvs::distance::detail::ops::l2_exp_distance_op<DataT, AccT, IdxT> distance_op{sqrt};
-  raft::identity_op fin_op{};
-
-  if constexpr (isRowMajor) {
-    constexpr auto fusedL2ExpKnn32RowMajor = fusedL2kNN<DataT,
-                                                        OutT,
-                                                        IdxT,
-                                                        KPolicy,
-                                                        decltype(distance_op),
-                                                        decltype(fin_op),
-                                                        32,
-                                                        2,
-                                                        usePrevTopKs,
-                                                        isRowMajor>;
-    constexpr auto fusedL2ExpKnn64RowMajor = fusedL2kNN<DataT,
-                                                        OutT,
-                                                        IdxT,
-                                                        KPolicy,
-                                                        decltype(distance_op),
-                                                        decltype(fin_op),
-                                                        64,
-                                                        3,
-                                                        usePrevTopKs,
-                                                        isRowMajor>;
-
-    auto fusedL2ExpKnnRowMajor = fusedL2ExpKnn32RowMajor;
-    if (numOfNN <= 32) {
-      fusedL2ExpKnnRowMajor = fusedL2ExpKnn32RowMajor;
-    } else if (numOfNN <= 64) {
-      fusedL2ExpKnnRowMajor = fusedL2ExpKnn64RowMajor;
-    } else {
-      ASSERT(numOfNN <= 64, "fusedL2kNN: num of nearest neighbors must be <= 64");
-    }
-
-    const auto sharedMemSize =
-      distance_op.template shared_mem_size<KPolicy>() + (KPolicy::Mblk * numOfNN * sizeof(Pair));
-    dim3 grid = cuvs::distance::detail::launchConfigGenerator<KPolicy>(
-      m, n, sharedMemSize, fusedL2ExpKnnRowMajor);
-    int32_t* mutexes = nullptr;
-    if (grid.x > 1) {
-      const auto numMutexes   = raft::ceildiv<int>(m, KPolicy::Mblk);
-      const auto normsSize    = (x != y) ? (m + n) * sizeof(DataT) : n * sizeof(DataT);
-      const auto requiredSize = sizeof(int32_t) * numMutexes + normsSize;
-      if (worksize < requiredSize) {
-        worksize = requiredSize;
-        return;
-      } else {
-        mutexes = (int32_t*)((char*)workspace + normsSize);
-        RAFT_CUDA_TRY(cudaMemsetAsync(mutexes, 0, sizeof(int32_t) * numMutexes, stream));
-      }
-    }
-
-    // calculate norms if they haven't been passed in
-    if (!xn) {
-      DataT* xn_ = (DataT*)workspace;
-      workspace  = xn_ + m;
-      raft::linalg::rowNorm(
-        xn_, x, k, m, raft::linalg::L2Norm, isRowMajor, stream, raft::identity_op{});
-      xn = xn_;
-    }
-    if (!yn) {
-      if (x == y) {
-        yn = xn;
-      } else {
-        DataT* yn_ = (DataT*)(workspace);
-        raft::linalg::rowNorm(
-          yn_, y, k, n, raft::linalg::L2Norm, isRowMajor, stream, raft::identity_op{});
-        yn = yn_;
-      }
-    }
-
-    fusedL2ExpKnnRowMajor<<<grid, blk, sharedMemSize, stream>>>(x,
-                                                                y,
-                                                                xn,
-                                                                yn,
-                                                                m,
-                                                                n,
-                                                                k,
-                                                                lda,
-                                                                ldb,
-                                                                ldd,
-                                                                distance_op,
-                                                                fin_op,
-                                                                (uint32_t)numOfNN,
-                                                                mutexes,
-                                                                out_dists,
-                                                                out_inds);
-  } else {
-  }
-
-  RAFT_CUDA_TRY(cudaGetLastError());
-}
-
-template <typename DataT,
-          typename AccT,
-          typename OutT,
-          typename IdxT,
-          bool usePrevTopKs,
-          bool isRowMajor>
-void fusedL2ExpKnn(IdxT m,
-                   IdxT n,
-                   IdxT k,
-                   IdxT lda,
-                   IdxT ldb,
-                   IdxT ldd,
-                   const DataT* x,
-                   const DataT* y,
-                   const DataT* xn,
-                   const DataT* yn,
-                   bool sqrt,
-                   OutT* out_dists,
-                   IdxT* out_inds,
-                   IdxT numOfNN,
-                   cudaStream_t stream,
-                   void* workspace,
-                   size_t& worksize)
-{
-  size_t bytesA = sizeof(DataT) * lda;
-  size_t bytesB = sizeof(DataT) * ldb;
-  if (16 % sizeof(DataT) == 0 && bytesA % 16 == 0 && bytesB % 16 == 0) {
-    fusedL2ExpKnnImpl<DataT, AccT, OutT, IdxT, 16 / sizeof(DataT), usePrevTopKs, isRowMajor>(
-      x,
-      y,
-      xn,
-      yn,
-      m,
-      n,
-      k,
-      lda,
-      ldb,
-      ldd,
-      sqrt,
-      out_dists,
-      out_inds,
-      numOfNN,
-      stream,
-      workspace,
-      worksize);
-  } else if (8 % sizeof(DataT) == 0 && bytesA % 8 == 0 && bytesB % 8 == 0) {
-    fusedL2ExpKnnImpl<DataT, AccT, OutT, IdxT, 8 / sizeof(DataT), usePrevTopKs, isRowMajor>(
-      x,
-      y,
-      xn,
-      yn,
-      m,
-      n,
-      k,
-      lda,
-      ldb,
-      ldd,
-      sqrt,
-      out_dists,
-      out_inds,
-      numOfNN,
-      stream,
-      workspace,
-      worksize);
-  } else {
-    fusedL2ExpKnnImpl<DataT, AccT, OutT, IdxT, 1, usePrevTopKs, isRowMajor>(x,
-                                                                            y,
-                                                                            xn,
-                                                                            yn,
-                                                                            m,
-                                                                            n,
-                                                                            k,
-                                                                            lda,
-                                                                            ldb,
-                                                                            ldd,
-                                                                            sqrt,
-                                                                            out_dists,
-                                                                            out_inds,
-                                                                            numOfNN,
-                                                                            stream,
-                                                                            workspace,
-                                                                            worksize);
-  }
-}
-
-/**
- * Compute the k-nearest neighbors using L2 expanded/unexpanded distance.
-
- * @tparam value_idx
- * @tparam value_t
- * @param[out] out_inds output indices array on device (size n_query_rows * k)
- * @param[out] out_dists output dists array on device (size n_query_rows * k)
- * @param[in] index input index array on device (size n_index_rows * D)
- * @param[in] query input query array on device (size n_query_rows * D)
- * @param[in] n_index_rows number of rows in index array
- * @param[in] n_query_rows number of rows in query array
- * @param[in] k number of closest neighbors to return
- * @param[in] rowMajorIndex are the index arrays in row-major layout?
- * @param[in] rowMajorQuery are the query array in row-major layout?
- * @param[in] stream stream to order kernel launch
- */
-template <typename value_idx, typename value_t, bool usePrevTopKs = false>
-void fusedL2Knn(size_t D,
-                value_idx* out_inds,
-                value_t* out_dists,
-                const value_t* index,
-                const value_t* query,
-                size_t n_index_rows,
-                size_t n_query_rows,
-                int k,
-                bool rowMajorIndex,
-                bool rowMajorQuery,
-                cudaStream_t stream,
-                cuvs::distance::DistanceType metric,
-                const value_t* index_norms = NULL,
-                const value_t* query_norms = NULL)
-{
-  // Validate the input data
-  ASSERT(k > 0, "l2Knn: k must be > 0");
-  ASSERT(D > 0, "l2Knn: D must be > 0");
-  ASSERT(n_index_rows > 0, "l2Knn: n_index_rows must be > 0");
-  ASSERT(index, "l2Knn: index must be provided (passed null)");
-  ASSERT(n_query_rows > 0, "l2Knn: n_query_rows must be > 0");
-  ASSERT(query, "l2Knn: query must be provided (passed null)");
-  ASSERT(out_dists, "l2Knn: out_dists must be provided (passed null)");
-  ASSERT(out_inds, "l2Knn: out_inds must be provided (passed null)");
-  // Currently we only support same layout for x & y inputs.
-  ASSERT(rowMajorIndex == rowMajorQuery,
-         "l2Knn: rowMajorIndex and rowMajorQuery should have same layout");
-  // TODO: Add support for column major layout
-  ASSERT(rowMajorIndex == true, "l2Knn: only rowMajor inputs are supported for now.");
-
-  // Even for L2 Sqrt distance case we use non-sqrt version as FAISS bfKNN only support
-  // non-sqrt metric & some tests in RAFT/cuML (like Linkage) fails if we use L2 sqrt.
-  constexpr bool sqrt = false;
-
-  size_t worksize = 0, tempWorksize = 0;
-  rmm::device_uvector<char> workspace(worksize, stream);
-  value_idx lda = D, ldb = D, ldd = n_index_rows;
-
-  switch (metric) {
-    case cuvs::distance::DistanceType::L2SqrtExpanded:
-    case cuvs::distance::DistanceType::L2Expanded:
-      tempWorksize = cuvs::distance::detail::
-        getWorkspaceSize<cuvs::distance::DistanceType::L2Expanded, float, float, float, value_idx>(
-          query, index, n_query_rows, n_index_rows, D);
-      worksize = tempWorksize;
-      workspace.resize(worksize, stream);
-      fusedL2ExpKnn<value_t, value_t, value_t, value_idx, usePrevTopKs, true>(n_query_rows,
-                                                                              n_index_rows,
-                                                                              D,
-                                                                              lda,
-                                                                              ldb,
-                                                                              ldd,
-                                                                              query,
-                                                                              index,
-                                                                              query_norms,
-                                                                              index_norms,
-                                                                              sqrt,
-                                                                              out_dists,
-                                                                              out_inds,
-                                                                              k,
-                                                                              stream,
-                                                                              workspace.data(),
-                                                                              worksize);
-      if (worksize > tempWorksize) {
-        workspace.resize(worksize, stream);
-        fusedL2ExpKnn<value_t, value_t, value_t, value_idx, usePrevTopKs, true>(n_query_rows,
-                                                                                n_index_rows,
-                                                                                D,
-                                                                                lda,
-                                                                                ldb,
-                                                                                ldd,
-                                                                                query,
-                                                                                index,
-                                                                                query_norms,
-                                                                                index_norms,
-                                                                                sqrt,
-                                                                                out_dists,
-                                                                                out_inds,
-                                                                                k,
-                                                                                stream,
-                                                                                workspace.data(),
-                                                                                worksize);
-      }
-      break;
-    case cuvs::distance::DistanceType::L2Unexpanded:
-    case cuvs::distance::DistanceType::L2SqrtUnexpanded:
-      fusedL2UnexpKnn<value_t, value_t, value_t, value_idx, usePrevTopKs, true>(n_query_rows,
-                                                                                n_index_rows,
-                                                                                D,
-                                                                                lda,
-                                                                                ldb,
-                                                                                ldd,
-                                                                                query,
-                                                                                index,
-                                                                                sqrt,
-                                                                                out_dists,
-                                                                                out_inds,
-                                                                                k,
-                                                                                stream,
-                                                                                workspace.data(),
-                                                                                worksize);
-      if (worksize) {
-        workspace.resize(worksize, stream);
-        fusedL2UnexpKnn<value_t, value_t, value_t, value_idx, usePrevTopKs, true>(n_query_rows,
-                                                                                  n_index_rows,
-                                                                                  D,
-                                                                                  lda,
-                                                                                  ldb,
-                                                                                  ldd,
-                                                                                  query,
-                                                                                  index,
-                                                                                  sqrt,
-                                                                                  out_dists,
-                                                                                  out_inds,
-                                                                                  k,
-                                                                                  stream,
-                                                                                  workspace.data(),
-                                                                                  worksize);
-      }
-      break;
-    default: printf("only L2 distance metric is supported\n"); break;
-  };
-}
-
-}  // namespace detail
-}  // namespace knn
-}  // namespace spatial
-}  // namespace cuvs
diff --git a/cpp/include/cuvs/spatial/knn/detail/fused_l2_knn.cuh b/cpp/include/cuvs/spatial/knn/detail/fused_l2_knn.cuh
deleted file mode 100644
index 8cc02c7c7..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/fused_l2_knn.cuh
+++ /dev/null
@@ -1,24 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#ifndef RAFT_EXPLICIT_INSTANTIATE_ONLY
-#include "fused_l2_knn-inl.cuh"
-#endif
-
-#ifdef RAFT_COMPILED
-#include "fused_l2_knn-ext.cuh"
-#endif
diff --git a/cpp/include/cuvs/spatial/knn/detail/haversine_distance.cuh b/cpp/include/cuvs/spatial/knn/detail/haversine_distance.cuh
deleted file mode 100644
index 992d27912..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/haversine_distance.cuh
+++ /dev/null
@@ -1,143 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <raft/util/pow2_utils.cuh>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/detail/faiss_select/Select.cuh>
-#include <raft/core/resources.hpp>
-
-namespace cuvs {
-namespace spatial {
-namespace knn {
-namespace detail {
-
-template <typename value_t>
-DI value_t compute_haversine(value_t x1, value_t y1, value_t x2, value_t y2)
-{
-  value_t sin_0 = raft::sin(0.5 * (x1 - y1));
-  value_t sin_1 = raft::sin(0.5 * (x2 - y2));
-  value_t rdist = sin_0 * sin_0 + raft::cos(x1) * raft::cos(y1) * sin_1 * sin_1;
-
-  return 2 * raft::asin(raft::sqrt(rdist));
-}
-
-/**
- * @tparam value_idx data type of indices
- * @tparam value_t data type of values and distances
- * @tparam warp_q
- * @tparam thread_q
- * @tparam tpb
- * @param[out] out_inds output indices
- * @param[out] out_dists output distances
- * @param[in] index index array
- * @param[in] query query array
- * @param[in] n_index_rows number of rows in index array
- * @param[in] k number of closest neighbors to return
- */
-template <typename value_idx, typename value_t, int warp_q = 1024, int thread_q = 8, int tpb = 128>
-RAFT_KERNEL haversine_knn_kernel(value_idx* out_inds,
-                                 value_t* out_dists,
-                                 const value_t* index,
-                                 const value_t* query,
-                                 size_t n_index_rows,
-                                 int k)
-{
-  constexpr int kNumWarps = tpb / raft::WarpSize;
-
-  __shared__ value_t smemK[kNumWarps * warp_q];
-  __shared__ value_idx smemV[kNumWarps * warp_q];
-
-  using namespace cuvs::neighbors::detail::faiss_select;
-  BlockSelect<value_t, value_idx, false, Comparator<value_t>, warp_q, thread_q, tpb> heap(
-    std::numeric_limits<value_t>::max(), std::numeric_limits<value_idx>::max(), smemK, smemV, k);
-
-  // Grid is exactly sized to rows available
-  int limit = raft::Pow2<raft::WarpSize>::roundDown(n_index_rows);
-
-  const value_t* query_ptr = query + (blockIdx.x * 2);
-  value_t x1               = query_ptr[0];
-  value_t x2               = query_ptr[1];
-
-  int i = threadIdx.x;
-
-  for (; i < limit; i += tpb) {
-    const value_t* idx_ptr = index + (i * 2);
-    value_t y1             = idx_ptr[0];
-    value_t y2             = idx_ptr[1];
-
-    value_t dist = compute_haversine(x1, y1, x2, y2);
-
-    heap.add(dist, i);
-  }
-
-  // Handle last remainder fraction of a warp of elements
-  if (i < n_index_rows) {
-    const value_t* idx_ptr = index + (i * 2);
-    value_t y1             = idx_ptr[0];
-    value_t y2             = idx_ptr[1];
-
-    value_t dist = compute_haversine(x1, y1, x2, y2);
-
-    heap.addThreadQ(dist, i);
-  }
-
-  heap.reduce();
-
-  for (int i = threadIdx.x; i < k; i += tpb) {
-    out_dists[blockIdx.x * k + i] = smemK[i];
-    out_inds[blockIdx.x * k + i]  = smemV[i];
-  }
-}
-
-/**
- * Conmpute the k-nearest neighbors using the Haversine
- * (great circle arc) distance. Input is assumed to have
- * 2 dimensions (latitude, longitude) in radians.
-
- * @tparam value_idx
- * @tparam value_t
- * @param[out] out_inds output indices array on device (size n_query_rows * k)
- * @param[out] out_dists output dists array on device (size n_query_rows * k)
- * @param[in] index input index array on device (size n_index_rows * 2)
- * @param[in] query input query array on device (size n_query_rows * 2)
- * @param[in] n_index_rows number of rows in index array
- * @param[in] n_query_rows number of rows in query array
- * @param[in] k number of closest neighbors to return
- * @param[in] stream stream to order kernel launch
- */
-template <typename value_idx, typename value_t>
-void haversine_knn(value_idx* out_inds,
-                   value_t* out_dists,
-                   const value_t* index,
-                   const value_t* query,
-                   size_t n_index_rows,
-                   size_t n_query_rows,
-                   int k,
-                   cudaStream_t stream)
-{
-  haversine_knn_kernel<<<n_query_rows, 128, 0, stream>>>(
-    out_inds, out_dists, index, query, n_index_rows, k);
-}
-
-}  // namespace detail
-}  // namespace knn
-}  // namespace spatial
-}  // namespace cuvs
diff --git a/cpp/include/cuvs/spatial/knn/detail/processing.cuh b/cpp/include/cuvs/spatial/knn/detail/processing.cuh
deleted file mode 100644
index 5e364cf5b..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/processing.cuh
+++ /dev/null
@@ -1,189 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#include "processing.hpp"
-
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/operators.hpp>
-#include <raft/linalg/matrix_vector_op.cuh>
-#include <raft/linalg/norm.cuh>
-#include <raft/linalg/unary_op.cuh>
-#include <raft/stats/mean.cuh>
-#include <raft/stats/mean_center.cuh>
-#include <rmm/device_uvector.hpp>
-
-namespace cuvs {
-namespace spatial {
-namespace knn {
-
-template <typename math_t>
-class CosineMetricProcessor : public MetricProcessor<math_t> {
- protected:
-  int k_;
-  bool row_major_;
-  size_t n_rows_;
-  size_t n_cols_;
-  cudaStream_t stream_;
-  rmm::device_uvector<math_t> colsums_;
-
- public:
-  CosineMetricProcessor(size_t n_rows, size_t n_cols, int k, bool row_major, cudaStream_t stream)
-    : stream_(stream),
-      colsums_(n_rows, stream),
-      n_cols_(n_cols),
-      n_rows_(n_rows),
-      row_major_(row_major),
-      k_(k)
-  {
-  }
-
-  void preprocess(math_t* data)
-  {
-    raft::linalg::rowNorm(colsums_.data(),
-                          data,
-                          n_cols_,
-                          n_rows_,
-                          raft::linalg::NormType::L2Norm,
-                          row_major_,
-                          stream_,
-                          raft::sqrt_op{});
-
-    raft::linalg::matrixVectorOp(
-      data, data, colsums_.data(), n_cols_, n_rows_, row_major_, false, raft::div_op{}, stream_);
-  }
-
-  void revert(math_t* data)
-  {
-    raft::linalg::matrixVectorOp(
-      data, data, colsums_.data(), n_cols_, n_rows_, row_major_, false, raft::mul_op{}, stream_);
-  }
-
-  void postprocess(math_t* data)
-  {
-    raft::linalg::unaryOp(
-      data, data, k_ * n_rows_, [] __device__(math_t in) { return 1 - in; }, stream_);
-  }
-
-  void set_num_queries(int k) override { k_ = k; }
-
-  ~CosineMetricProcessor() = default;
-};
-
-template <typename math_t>
-class CorrelationMetricProcessor : public CosineMetricProcessor<math_t> {
-  using cosine = CosineMetricProcessor<math_t>;
-
- public:
-  CorrelationMetricProcessor(
-    size_t n_rows, size_t n_cols, int k, bool row_major, cudaStream_t stream)
-    : CosineMetricProcessor<math_t>(n_rows, n_cols, k, row_major, stream), means_(n_rows, stream)
-  {
-  }
-
-  void preprocess(math_t* data)
-  {
-    math_t normalizer_const = 1.0 / (math_t)cosine::n_cols_;
-
-    raft::linalg::reduce(means_.data(),
-                         data,
-                         cosine::n_cols_,
-                         cosine::n_rows_,
-                         (math_t)0.0,
-                         cosine::row_major_,
-                         true,
-                         cosine::stream_);
-
-    raft::linalg::unaryOp(means_.data(),
-                          means_.data(),
-                          cosine::n_rows_,
-                          raft::mul_const_op<math_t>(normalizer_const),
-                          cosine::stream_);
-
-    raft::stats::meanCenter(data,
-                            data,
-                            means_.data(),
-                            cosine::n_cols_,
-                            cosine::n_rows_,
-                            cosine::row_major_,
-                            false,
-                            cosine::stream_);
-
-    CosineMetricProcessor<math_t>::preprocess(data);
-  }
-
-  void revert(math_t* data)
-  {
-    CosineMetricProcessor<math_t>::revert(data);
-
-    raft::stats::meanAdd(data,
-                         data,
-                         means_.data(),
-                         cosine::n_cols_,
-                         cosine::n_rows_,
-                         cosine::row_major_,
-                         false,
-                         cosine::stream_);
-  }
-
-  void postprocess(math_t* data) { CosineMetricProcessor<math_t>::postprocess(data); }
-
-  ~CorrelationMetricProcessor() = default;
-
-  rmm::device_uvector<math_t> means_;
-};
-
-template <typename math_t>
-class DefaultMetricProcessor : public MetricProcessor<math_t> {
- public:
-  void preprocess(math_t* data) {}
-
-  void revert(math_t* data) {}
-
-  void postprocess(math_t* data) {}
-
-  ~DefaultMetricProcessor() = default;
-};
-
-template <typename math_t>
-inline std::unique_ptr<MetricProcessor<math_t>> create_processor(
-  distance::DistanceType metric, int n, int D, int k, bool rowMajorQuery, cudaStream_t userStream)
-{
-  MetricProcessor<math_t>* mp = nullptr;
-
-  switch (metric) {
-    case distance::DistanceType::CosineExpanded:
-      mp = new CosineMetricProcessor<math_t>(n, D, k, rowMajorQuery, userStream);
-      break;
-
-    case distance::DistanceType::CorrelationExpanded:
-      mp = new CorrelationMetricProcessor<math_t>(n, D, k, rowMajorQuery, userStream);
-      break;
-    default: mp = new DefaultMetricProcessor<math_t>();
-  }
-
-  return std::unique_ptr<MetricProcessor<math_t>>(mp);
-}
-
-// Currently only being used by floats
-template class MetricProcessor<float>;
-template class CosineMetricProcessor<float>;
-template class CorrelationMetricProcessor<float>;
-template class DefaultMetricProcessor<float>;
-
-}  // namespace knn
-}  // namespace spatial
-}  // namespace cuvs
diff --git a/cpp/include/cuvs/spatial/knn/detail/processing.hpp b/cpp/include/cuvs/spatial/knn/detail/processing.hpp
deleted file mode 100644
index d1f7349b8..000000000
--- a/cpp/include/cuvs/spatial/knn/detail/processing.hpp
+++ /dev/null
@@ -1,45 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-namespace cuvs {
-namespace spatial {
-namespace knn {
-
-/**
- * @brief A virtual class defining pre- and post-processing
- * for metrics. This class will temporarily modify its given
- * state in `preprocess()` and undo those modifications in
- * `postprocess()`
- */
-
-template <typename math_t>
-class MetricProcessor {
- public:
-  virtual void preprocess(math_t* data) {}
-
-  virtual void revert(math_t* data) {}
-
-  virtual void postprocess(math_t* data) {}
-
-  virtual void set_num_queries(int k) {}
-
-  virtual ~MetricProcessor() = default;
-};
-
-}  // namespace knn
-}  // namespace spatial
-}  // namespace cuvs
diff --git a/cpp/include/cuvs/spatial/knn/epsilon_neighborhood.cuh b/cpp/include/cuvs/spatial/knn/epsilon_neighborhood.cuh
deleted file mode 100644
index a896a2288..000000000
--- a/cpp/include/cuvs/spatial/knn/epsilon_neighborhood.cuh
+++ /dev/null
@@ -1,38 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/**
- * This file is deprecated and will be removed in release 22.06.
- * Please use the cuh version instead.
- */
-
-/**
- * DISCLAIMER: this file is deprecated: use epsilon_neighborhood.cuh instead
- */
-
-#pragma once
-
-#pragma message(__FILE__                                                    \
-                  " is deprecated and will be removed in a future release." \
-                  " Please use the cuvs::neighbors version instead.")
-
-#include <cuvs/neighbors/epsilon_neighborhood.cuh>
-
-namespace cuvs::spatial::knn {
-
-using cuvs::neighbors::epsilon_neighborhood::eps_neighbors_l2sq;
-using cuvs::neighbors::epsilon_neighborhood::epsUnexpL2SqNeighborhood;
-
-}  // namespace cuvs::spatial::knn
diff --git a/cpp/include/cuvs/spatial/knn/ivf_flat.cuh b/cpp/include/cuvs/spatial/knn/ivf_flat.cuh
deleted file mode 100644
index 4b3b759b8..000000000
--- a/cpp/include/cuvs/spatial/knn/ivf_flat.cuh
+++ /dev/null
@@ -1,39 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/**
- * This file is deprecated and will be removed in release 22.06.
- * Please use the cuh version instead.
- */
-
-/**
- * DISCLAIMER: this file is deprecated: use epsilon_neighborhood.cuh instead
- */
-
-#pragma once
-
-#pragma message(__FILE__                                                    \
-                  " is deprecated and will be removed in a future release." \
-                  " Please use the cuvs::neighbors version instead.")
-
-#include <cuvs/neighbors/ivf_flat.cuh>
-
-namespace cuvs::spatial::knn::ivf_flat {
-
-using cuvs::neighbors::ivf_flat::build;
-using cuvs::neighbors::ivf_flat::extend;
-using cuvs::neighbors::ivf_flat::search;
-
-};  // namespace cuvs::spatial::knn::ivf_flat
diff --git a/cpp/include/cuvs/spatial/knn/ivf_flat_types.hpp b/cpp/include/cuvs/spatial/knn/ivf_flat_types.hpp
deleted file mode 100644
index 1cd832ee1..000000000
--- a/cpp/include/cuvs/spatial/knn/ivf_flat_types.hpp
+++ /dev/null
@@ -1,40 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/**
- * This file is deprecated and will be removed in release 22.06.
- * Please use the cuh version instead.
- */
-
-/**
- * DISCLAIMER: this file is deprecated: use epsilon_neighborhood.cuh instead
- */
-
-#pragma once
-
-#pragma message(__FILE__                                                    \
-                  " is deprecated and will be removed in a future release." \
-                  " Please use the cuvs::neighbors version instead.")
-
-#include <cuvs/neighbors/ivf_flat_types.hpp>
-
-namespace cuvs::spatial::knn::ivf_flat {
-
-using cuvs::neighbors::ivf_flat::index;
-using cuvs::neighbors::ivf_flat::index_params;
-using cuvs::neighbors::ivf_flat::kIndexGroupSize;
-using cuvs::neighbors::ivf_flat::search_params;
-
-};  // namespace cuvs::spatial::knn::ivf_flat
diff --git a/cpp/include/cuvs/spatial/knn/ivf_pq.cuh b/cpp/include/cuvs/spatial/knn/ivf_pq.cuh
deleted file mode 100644
index 915fced5c..000000000
--- a/cpp/include/cuvs/spatial/knn/ivf_pq.cuh
+++ /dev/null
@@ -1,39 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/**
- * This file is deprecated and will be removed in release 22.06.
- * Please use the cuh version instead.
- */
-
-/**
- * DISCLAIMER: this file is deprecated: use epsilon_neighborhood.cuh instead
- */
-
-#pragma once
-
-#pragma message(__FILE__                                                    \
-                  " is deprecated and will be removed in a future release." \
-                  " Please use the cuvs::neighbors version instead.")
-
-#include <cuvs/neighbors/ivf_pq.cuh>
-
-namespace cuvs::spatial::knn::ivf_pq {
-
-using cuvs::neighbors::ivf_pq::build;
-using cuvs::neighbors::ivf_pq::extend;
-using cuvs::neighbors::ivf_pq::search;
-
-}  // namespace cuvs::spatial::knn::ivf_pq
diff --git a/cpp/include/cuvs/spatial/knn/ivf_pq_types.hpp b/cpp/include/cuvs/spatial/knn/ivf_pq_types.hpp
deleted file mode 100644
index 1202f8968..000000000
--- a/cpp/include/cuvs/spatial/knn/ivf_pq_types.hpp
+++ /dev/null
@@ -1,40 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-/**
- * This file is deprecated and will be removed in release 22.06.
- * Please use the cuh version instead.
- */
-
-/**
- * DISCLAIMER: this file is deprecated: use epsilon_neighborhood.cuh instead
- */
-
-#pragma once
-
-#pragma message(__FILE__                                                    \
-                  " is deprecated and will be removed in a future release." \
-                  " Please use the cuvs::neighbors version instead.")
-
-#include <cuvs/neighbors/ivf_pq_types.hpp>
-
-namespace cuvs::spatial::knn::ivf_pq {
-
-using cuvs::neighbors::ivf_pq::codebook_gen;
-using cuvs::neighbors::ivf_pq::index;
-using cuvs::neighbors::ivf_pq::index_params;
-using cuvs::neighbors::ivf_pq::search_params;
-
-}  // namespace cuvs::spatial::knn::ivf_pq
diff --git a/cpp/include/cuvs/spatial/knn/knn.cuh b/cpp/include/cuvs/spatial/knn/knn.cuh
deleted file mode 100644
index f6267feb5..000000000
--- a/cpp/include/cuvs/spatial/knn/knn.cuh
+++ /dev/null
@@ -1,231 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/neighbors/detail/knn_brute_force.cuh>
-#include <cuvs/neighbors/detail/selection_faiss.cuh>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/nvtx.hpp>
-#include <raft/matrix/detail/select_radix.cuh>
-#include <raft/matrix/detail/select_warpsort.cuh>
-
-namespace cuvs::spatial::knn {
-
-/**
- * Performs a k-select across row partitioned index/distance
- * matrices formatted like the following:
- * row1: k0, k1, k2
- * row2: k0, k1, k2
- * row3: k0, k1, k2
- * row1: k0, k1, k2
- * row2: k0, k1, k2
- * row3: k0, k1, k2
- *
- * etc...
- *
- * @tparam idx_t
- * @tparam value_t
- * @param in_keys
- * @param in_values
- * @param out_keys
- * @param out_values
- * @param n_samples
- * @param n_parts
- * @param k
- * @param stream
- * @param translations
- */
-template <typename idx_t = int64_t, typename value_t = float>
-inline void knn_merge_parts(const value_t* in_keys,
-                            const idx_t* in_values,
-                            value_t* out_keys,
-                            idx_t* out_values,
-                            size_t n_samples,
-                            int n_parts,
-                            int k,
-                            cudaStream_t stream,
-                            idx_t* translations)
-{
-  cuvs::neighbors::detail::knn_merge_parts(
-    in_keys, in_values, out_keys, out_values, n_samples, n_parts, k, stream, translations);
-}
-
-/** Choose an implementation for the select-top-k, */
-enum class SelectKAlgo {
-  /** Adapted from the faiss project. Result: sorted (not stable). */
-  FAISS,
-  /** Incomplete series of radix sort passes, comparing 8 bits per pass. Result: unsorted. */
-  RADIX_8_BITS,
-  /** Incomplete series of radix sort passes, comparing 11 bits per pass. Result: unsorted. */
-  RADIX_11_BITS,
-  /** Filtering with a bitonic-sort-based priority queue. Result: sorted (not stable). */
-  WARP_SORT
-};
-
-/**
- * Select k smallest or largest key/values from each row in the input data.
- *
- * If you think of the input data `in_keys` as a row-major matrix with input_len columns and
- * n_inputs rows, then this function selects k smallest/largest values in each row and fills
- * in the row-major matrix `out_keys` of size (n_inputs, k).
- *
- * Note, depending on the selected algorithm, the values within rows of `out_keys` are not
- * necessarily sorted. See the `SelectKAlgo` enumeration for more details.
- *
- * Note: This call is deprecated, please use `raft/matrix/select_k.cuh`
- *
- * @tparam idx_t
- *   the payload type (what is being selected together with the keys).
- * @tparam value_t
- *   the type of the keys (what is being compared).
- *
- * @param[in] in_keys
- *   contiguous device array of inputs of size (input_len * n_inputs);
- *   these are compared and selected.
- * @param[in] in_values
- *   contiguous device array of inputs of size (input_len * n_inputs);
- *   typically, these are indices of the corresponding in_keys.
- *   You can pass `NULL` as an argument here; this would imply `in_values` is a homogeneous array
- *   of indices from `0` to `input_len - 1` for every input and reduce the usage of memory
- *   bandwidth.
- * @param[in] n_inputs
- *   number of input rows, i.e. the batch size.
- * @param[in] input_len
- *   length of a single input array (row); also sometimes referred as n_cols.
- *   Invariant: input_len >= k.
- * @param[out] out_keys
- *   contiguous device array of outputs of size (k * n_inputs);
- *   the k smallest/largest values from each row of the `in_keys`.
- * @param[out] out_values
- *   contiguous device array of outputs of size (k * n_inputs);
- *   the payload selected together with `out_keys`.
- * @param[in] select_min
- *   whether to select k smallest (true) or largest (false) keys.
- * @param[in] k
- *   the number of outputs to select in each input row.
- * @param[in] stream
- * @param[in] algo
- *   the implementation of the algorithm
- */
-template <typename idx_t = int, typename value_t = float>
-[[deprecated("Use function `select_k` from `raft/matrix/select_k.cuh`")]] inline void select_k(
-  const value_t* in_keys,
-  const idx_t* in_values,
-  size_t n_inputs,
-  size_t input_len,
-  value_t* out_keys,
-  idx_t* out_values,
-  bool select_min,
-  int k,
-  cudaStream_t stream,
-  SelectKAlgo algo = SelectKAlgo::FAISS)
-{
-  raft::common::nvtx::range<raft::common::nvtx::domain::raft> fun_scope(
-    "select-%s-%d (%zu, %zu) algo-%d",
-    select_min ? "min" : "max",
-    k,
-    n_inputs,
-    input_len,
-    int(algo));
-  ASSERT(size_t(input_len) >= size_t(k),
-         "Size of the input (input_len = %zu) must be not smaller than the selection (k = %zu).",
-         size_t(input_len),
-         size_t(k));
-
-  switch (algo) {
-    case SelectKAlgo::FAISS:
-      neighbors::detail::select_k(
-        in_keys, in_values, n_inputs, input_len, out_keys, out_values, select_min, k, stream);
-      break;
-
-    case SelectKAlgo::RADIX_8_BITS:
-      raft::matrix::detail::select::radix::select_k<value_t, idx_t, 8, 512>(
-        in_keys, in_values, n_inputs, input_len, k, out_keys, out_values, select_min, true, stream);
-      break;
-
-    case SelectKAlgo::RADIX_11_BITS:
-      raft::matrix::detail::select::radix::select_k<value_t, idx_t, 11, 512>(
-        in_keys, in_values, n_inputs, input_len, k, out_keys, out_values, select_min, true, stream);
-      break;
-
-    case SelectKAlgo::WARP_SORT:
-      raft::matrix::detail::select::warpsort::select_k<value_t, idx_t>(
-        in_keys, in_values, n_inputs, input_len, k, out_keys, out_values, select_min, stream);
-      break;
-
-    default: ASSERT(false, "Unknown algorithm (id = %d)", int(algo));
-  }
-}
-
-/**
- * @brief Flat C++ API function to perform a brute force knn on
- * a series of input arrays and combine the results into a single
- * output array for indexes and distances.
- *
- * @param[in] handle the cuml handle to use
- * @param[in] input vector of pointers to the input arrays
- * @param[in] sizes vector of sizes of input arrays
- * @param[in] D the dimensionality of the arrays
- * @param[in] search_items array of items to search of dimensionality D
- * @param[in] n number of rows in search_items
- * @param[out] res_I the resulting index array of size n * k
- * @param[out] res_D the resulting distance array of size n * k
- * @param[in] k the number of nearest neighbors to return
- * @param[in] rowMajorIndex are the index arrays in row-major order?
- * @param[in] rowMajorQuery are the query arrays in row-major order?
- * @param[in] metric distance metric to use. Euclidean (L2) is used by
- * 			   default
- * @param[in] metric_arg the value of `p` for Minkowski (l-p) distances. This
- * 					 is ignored if the metric_type is not Minkowski.
- * @param[in] translations starting offsets for partitions. should be the same size
- *            as input vector.
- */
-template <typename idx_t = std::int64_t, typename value_t = float, typename value_int = int>
-void brute_force_knn(raft::resources const& handle,
-                     std::vector<value_t*>& input,
-                     std::vector<value_int>& sizes,
-                     value_int D,
-                     value_t* search_items,
-                     value_int n,
-                     idx_t* res_I,
-                     value_t* res_D,
-                     value_int k,
-                     bool rowMajorIndex               = true,
-                     bool rowMajorQuery               = true,
-                     std::vector<idx_t>* translations = nullptr,
-                     distance::DistanceType metric    = distance::DistanceType::L2Unexpanded,
-                     float metric_arg                 = 2.0f)
-{
-  ASSERT(input.size() == sizes.size(), "input and sizes vectors must be the same size");
-
-  cuvs::neighbors::detail::brute_force_knn_impl(handle,
-                                                input,
-                                                sizes,
-                                                D,
-                                                search_items,
-                                                n,
-                                                res_I,
-                                                res_D,
-                                                k,
-                                                rowMajorIndex,
-                                                rowMajorQuery,
-                                                translations,
-                                                metric,
-                                                metric_arg);
-}
-
-}  // namespace cuvs::spatial::knn
diff --git a/cpp/include/cuvs/spatial/knn/specializations.cuh b/cpp/include/cuvs/spatial/knn/specializations.cuh
deleted file mode 100644
index ed0b6848a..000000000
--- a/cpp/include/cuvs/spatial/knn/specializations.cuh
+++ /dev/null
@@ -1,22 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#pragma message(                                            \
-    __FILE__                                                \
-    " is deprecated and will be removed."                   \
-    " Including specializations is not necessary any more." \
-    " For more information, see: https://docs.rapids.ai/api/raft/nightly/using_libraft.html")
diff --git a/cpp/include/cuvs/spatial/knn/specializations/knn.cuh b/cpp/include/cuvs/spatial/knn/specializations/knn.cuh
deleted file mode 100644
index ed0b6848a..000000000
--- a/cpp/include/cuvs/spatial/knn/specializations/knn.cuh
+++ /dev/null
@@ -1,22 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#pragma message(                                            \
-    __FILE__                                                \
-    " is deprecated and will be removed."                   \
-    " Including specializations is not necessary any more." \
-    " For more information, see: https://docs.rapids.ai/api/raft/nightly/using_libraft.html")
diff --git a/cpp/include/cuvs_runtime/cluster/kmeans.hpp b/cpp/include/cuvs_runtime/cluster/kmeans.hpp
deleted file mode 100644
index d2fb700f2..000000000
--- a/cpp/include/cuvs_runtime/cluster/kmeans.hpp
+++ /dev/null
@@ -1,97 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/host_mdspan.hpp>
-#include <raft/core/resources.hpp>
-
-#include <cuvs/cluster/kmeans_types.hpp>
-
-namespace cuvs::runtime::cluster::kmeans {
-
-/**
- * @defgroup kmeans_runtime Kmeans Runtime API
- * @{
- */
-
-void update_centroids(raft::resources const& handle,
-                      const float* X,
-                      int n_samples,
-                      int n_features,
-                      int n_clusters,
-                      const float* sample_weights,
-                      const float* centroids,
-                      const int* labels,
-                      float* new_centroids,
-                      float* weight_per_cluster);
-
-void update_centroids(raft::resources const& handle,
-                      const double* X,
-                      int n_samples,
-                      int n_features,
-                      int n_clusters,
-                      const double* sample_weights,
-                      const double* centroids,
-                      const int* labels,
-                      double* new_centroids,
-                      double* weight_per_cluster);
-
-void fit(raft::resources const& handle,
-         const cuvs::cluster::kmeans::KMeansParams& params,
-         raft::device_matrix_view<const float, int, row_major> X,
-         std::optional<raft::device_vector_view<const float, int>> sample_weight,
-         raft::device_matrix_view<float, int, row_major> centroids,
-         raft::host_scalar_view<float, int> inertia,
-         raft::host_scalar_view<int, int> n_iter);
-
-void fit(raft::resources const& handle,
-         const cuvs::cluster::kmeans::KMeansParams& params,
-         raft::device_matrix_view<const double, int, row_major> X,
-         std::optional<raft::device_vector_view<const double, int>> sample_weight,
-         raft::device_matrix_view<double, int, row_major> centroids,
-         raft::host_scalar_view<double, int> inertia,
-         raft::host_scalar_view<int, int> n_iter);
-
-void init_plus_plus(raft::resources const& handle,
-                    const cuvs::cluster::kmeans::KMeansParams& params,
-                    raft::device_matrix_view<const float, int, row_major> X,
-                    raft::device_matrix_view<float, int, row_major> centroids);
-
-void init_plus_plus(raft::resources const& handle,
-                    const cuvs::cluster::kmeans::KMeansParams& params,
-                    raft::device_matrix_view<const double, int, row_major> X,
-                    raft::device_matrix_view<double, int, row_major> centroids);
-
-void cluster_cost(raft::resources const& handle,
-                  const float* X,
-                  int n_samples,
-                  int n_features,
-                  int n_clusters,
-                  const float* centroids,
-                  float* cost);
-
-void cluster_cost(raft::resources const& handle,
-                  const double* X,
-                  int n_samples,
-                  int n_features,
-                  int n_clusters,
-                  const double* centroids,
-                  double* cost);
-
-/** @} */  // end group kmeans_runtime
-
-}  // namespace cuvs::runtime::cluster::kmeans
diff --git a/cpp/include/cuvs_runtime/distance/fused_l2_nn.hpp b/cpp/include/cuvs_runtime/distance/fused_l2_nn.hpp
deleted file mode 100644
index 797ea8e27..000000000
--- a/cpp/include/cuvs_runtime/distance/fused_l2_nn.hpp
+++ /dev/null
@@ -1,65 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/resources.hpp>
-
-namespace raft::runtime::distance {
-
-/**
- * @defgroup fused_l2_nn_min_arg_runtime Fused L2 1NN Runtime API
- * @{
- */
-
-/**
- * @brief Wrapper around fusedL2NN with minimum reduction operators.
- *
- * fusedL2NN cannot be compiled in the distance library due to the lambda
- * operators, so this wrapper covers the most common case (minimum).
- *
- * @param[in] handle         raft handle
- * @param[out] min           will contain the reduced output (Length = `m`)
- *                           (on device)
- * @param[in]  x             first matrix. Row major. Dim = `m x k`.
- *                           (on device).
- * @param[in]  y             second matrix. Row major. Dim = `n x k`.
- *                           (on device).
- * @param[in]  m             gemm m
- * @param[in]  n             gemm n
- * @param[in]  k             gemm k
- * @param[in]  sqrt          Whether the output `minDist` should contain L2-sqrt
- */
-void fused_l2_nn_min_arg(raft::resources const& handle,
-                         int* min,
-                         const float* x,
-                         const float* y,
-                         int m,
-                         int n,
-                         int k,
-                         bool sqrt);
-
-void fused_l2_nn_min_arg(raft::resources const& handle,
-                         int* min,
-                         const double* x,
-                         const double* y,
-                         int m,
-                         int n,
-                         int k,
-                         bool sqrt);
-
-/** @} */  // end group fused_l2_nn_min_arg_runtime
-
-}  // end namespace raft::runtime::distance
diff --git a/cpp/include/cuvs_runtime/distance/pairwise_distance.hpp b/cpp/include/cuvs_runtime/distance/pairwise_distance.hpp
deleted file mode 100644
index 64343a18d..000000000
--- a/cpp/include/cuvs_runtime/distance/pairwise_distance.hpp
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/distance/distance_types.hpp>
-
-namespace raft::runtime::distance {
-
-/**
- * @defgroup pairwise_distance_runtime Pairwise Distances Runtime API
- * @{
- */
-
-void pairwise_distance(raft::resources const& handle,
-                       float* x,
-                       float* y,
-                       float* dists,
-                       int m,
-                       int n,
-                       int k,
-                       cuvs::distance::DistanceType metric,
-                       bool isRowMajor,
-                       float metric_arg);
-
-void pairwise_distance(raft::resources const& handle,
-                       double* x,
-                       double* y,
-                       double* dists,
-                       int m,
-                       int n,
-                       int k,
-                       cuvs::distance::DistanceType metric,
-                       bool isRowMajor,
-                       float metric_arg);
-
-/** @} */  // end group pairwise_distance_runtime
-
-}  // namespace raft::runtime::distance
diff --git a/cpp/include/cuvs_runtime/matrix/select_k.hpp b/cpp/include/cuvs_runtime/matrix/select_k.hpp
deleted file mode 100644
index dcd40aac3..000000000
--- a/cpp/include/cuvs_runtime/matrix/select_k.hpp
+++ /dev/null
@@ -1,32 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resources.hpp>
-
-#include <optional>
-
-namespace raft::runtime::matrix {
-void select_k(const resources& handle,
-              raft::device_matrix_view<const float, int64_t, row_major> in_val,
-              std::optional<raft::device_matrix_view<const int64_t, int64_t, row_major>> in_idx,
-              raft::device_matrix_view<float, int64_t, row_major> out_val,
-              raft::device_matrix_view<int64_t, int64_t, row_major> out_idx,
-              bool select_min);
-
-}  // namespace raft::runtime::matrix
diff --git a/cpp/include/cuvs_runtime/neighbors/brute_force.hpp b/cpp/include/cuvs_runtime/neighbors/brute_force.hpp
deleted file mode 100644
index 3bc19b2a2..000000000
--- a/cpp/include/cuvs_runtime/neighbors/brute_force.hpp
+++ /dev/null
@@ -1,38 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resources.hpp>
-
-namespace raft::runtime::neighbors::brute_force {
-
-#define RAFT_INST_BFKNN(IDX_T, DATA_T, MATRIX_IDX_T, INDEX_LAYOUT, SEARCH_LAYOUT)        \
-  void knn(raft::resources const& handle,                                                \
-           raft::device_matrix_view<const DATA_T, MATRIX_IDX_T, INDEX_LAYOUT> index,     \
-           raft::device_matrix_view<const DATA_T, MATRIX_IDX_T, SEARCH_LAYOUT> search,   \
-           raft::device_matrix_view<IDX_T, MATRIX_IDX_T, row_major> indices,             \
-           raft::device_matrix_view<DATA_T, MATRIX_IDX_T, row_major> distances,          \
-           distance::DistanceType metric         = distance::DistanceType::L2Unexpanded, \
-           std::optional<float> metric_arg       = std::make_optional<float>(2.0f),      \
-           std::optional<IDX_T> global_id_offset = std::nullopt);
-
-RAFT_INST_BFKNN(int64_t, float, int64_t, raft::row_major, raft::row_major);
-
-#undef RAFT_INST_BFKNN
-
-}  // namespace raft::runtime::neighbors::brute_force
diff --git a/cpp/include/cuvs_runtime/neighbors/ivf_flat.hpp b/cpp/include/cuvs_runtime/neighbors/ivf_flat.hpp
deleted file mode 100644
index bc3fab58c..000000000
--- a/cpp/include/cuvs_runtime/neighbors/ivf_flat.hpp
+++ /dev/null
@@ -1,83 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/neighbors/ivf_flat_types.hpp>
-#include <string>
-
-namespace raft::runtime::neighbors::ivf_flat {
-
-// We define overloads for build and extend with void return type. This is used in the Cython
-// wrappers, where exception handling is not compatible with return type that has nontrivial
-// constructor.
-#define RAFT_INST_BUILD_EXTEND(T, IdxT)                                              \
-  auto build(raft::resources const& handle,                                          \
-             const cuvs::neighbors::ivf_flat::index_params& params,                  \
-             raft::device_matrix_view<const T, IdxT, row_major> dataset)             \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                                     \
-                                                                                     \
-  auto extend(raft::resources const& handle,                                         \
-              raft::device_matrix_view<const T, IdxT, row_major> new_vectors,        \
-              std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \
-              const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index)           \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                                     \
-                                                                                     \
-  void build(raft::resources const& handle,                                          \
-             const cuvs::neighbors::ivf_flat::index_params& params,                  \
-             raft::device_matrix_view<const T, IdxT, row_major> dataset,             \
-             cuvs::neighbors::ivf_flat::index<T, IdxT>& idx);                        \
-                                                                                     \
-  void extend(raft::resources const& handle,                                         \
-              raft::device_matrix_view<const T, IdxT, row_major> new_vectors,        \
-              std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \
-              cuvs::neighbors::ivf_flat::index<T, IdxT>* idx);                       \
-                                                                                     \
-  void serialize_file(raft::resources const& handle,                                 \
-                      const std::string& filename,                                   \
-                      const cuvs::neighbors::ivf_flat::index<T, IdxT>& index);       \
-                                                                                     \
-  void deserialize_file(raft::resources const& handle,                               \
-                        const std::string& filename,                                 \
-                        cuvs::neighbors::ivf_flat::index<T, IdxT>* index);           \
-  void serialize(raft::resources const& handle,                                      \
-                 std::string& str,                                                   \
-                 const cuvs::neighbors::ivf_flat::index<T, IdxT>& index);            \
-  void deserialize(raft::resources const& handle,                                    \
-                   const std::string& str,                                           \
-                   cuvs::neighbors::ivf_flat::index<T, IdxT>*);
-
-RAFT_INST_BUILD_EXTEND(float, int64_t)
-RAFT_INST_BUILD_EXTEND(int8_t, int64_t)
-RAFT_INST_BUILD_EXTEND(uint8_t, int64_t)
-
-#undef RAFT_INST_BUILD_EXTEND
-
-#define RAFT_INST_SEARCH(T, IdxT)                                 \
-  void search(raft::resources const&,                             \
-              cuvs::neighbors::ivf_flat::search_params const&,    \
-              cuvs::neighbors::ivf_flat::index<T, IdxT> const&,   \
-              raft::device_matrix_view<const T, IdxT, row_major>, \
-              raft::device_matrix_view<IdxT, IdxT, row_major>,    \
-              raft::device_matrix_view<float, IdxT, row_major>);
-
-RAFT_INST_SEARCH(float, int64_t);
-RAFT_INST_SEARCH(int8_t, int64_t);
-RAFT_INST_SEARCH(uint8_t, int64_t);
-
-#undef RAFT_INST_SEARCH
-
-}  // namespace raft::runtime::neighbors::ivf_flat
diff --git a/cpp/include/cuvs_runtime/neighbors/ivf_pq.hpp b/cpp/include/cuvs_runtime/neighbors/ivf_pq.hpp
deleted file mode 100644
index be218fb71..000000000
--- a/cpp/include/cuvs_runtime/neighbors/ivf_pq.hpp
+++ /dev/null
@@ -1,96 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <cuvs/neighbors/ivf_pq_types.hpp>
-
-namespace raft::runtime::neighbors::ivf_pq {
-
-// We define overloads for build and extend with void return type. This is used in the Cython
-// wrappers, where exception handling is not compatible with return type that has nontrivial
-// constructor.
-#define RAFT_DECL_BUILD_EXTEND(T, IdxT)                                              \
-  [[nodiscard]] cuvs::neighbors::ivf_pq::index<IdxT> build(                          \
-    raft::resources const& handle,                                                   \
-    const cuvs::neighbors::ivf_pq::index_params& params,                             \
-    raft::device_matrix_view<const T, IdxT, row_major> dataset);                     \
-                                                                                     \
-  void build(raft::resources const& handle,                                          \
-             const cuvs::neighbors::ivf_pq::index_params& params,                    \
-             raft::device_matrix_view<const T, IdxT, row_major> dataset,             \
-             cuvs::neighbors::ivf_pq::index<IdxT>* idx);                             \
-                                                                                     \
-  [[nodiscard]] cuvs::neighbors::ivf_pq::index<IdxT> extend(                         \
-    raft::resources const& handle,                                                   \
-    raft::device_matrix_view<const T, IdxT, row_major> new_vectors,                  \
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices,           \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx);                                \
-                                                                                     \
-  void extend(raft::resources const& handle,                                         \
-              raft::device_matrix_view<const T, IdxT, row_major> new_vectors,        \
-              std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \
-              cuvs::neighbors::ivf_pq::index<IdxT>* idx);
-
-RAFT_DECL_BUILD_EXTEND(float, int64_t);
-RAFT_DECL_BUILD_EXTEND(int8_t, int64_t);
-RAFT_DECL_BUILD_EXTEND(uint8_t, int64_t);
-
-#undef RAFT_DECL_BUILD_EXTEND
-
-#define RAFT_DECL_SEARCH(T, IdxT)                                         \
-  void search(raft::resources const& handle,                              \
-              const cuvs::neighbors::ivf_pq::search_params& params,       \
-              const cuvs::neighbors::ivf_pq::index<IdxT>& idx,            \
-              raft::device_matrix_view<const T, IdxT, row_major> queries, \
-              raft::device_matrix_view<IdxT, IdxT, row_major> neighbors,  \
-              raft::device_matrix_view<float, IdxT, row_major> distances);
-
-RAFT_DECL_SEARCH(float, int64_t);
-RAFT_DECL_SEARCH(int8_t, int64_t);
-RAFT_DECL_SEARCH(uint8_t, int64_t);
-
-#undef RAFT_DECL_SEARCH
-
-/**
- * Save the index to file.
- *
- * Experimental, both the API and the serialization format are subject to change.
- *
- * @param[in] handle the raft handle
- * @param[in] filename the filename for saving the index
- * @param[in] index IVF-PQ index
- *
- */
-void serialize(raft::resources const& handle,
-               const std::string& filename,
-               const cuvs::neighbors::ivf_pq::index<int64_t>& index);
-
-/**
- * Load index from file.
- *
- * Experimental, both the API and the serialization format are subject to change.
- *
- * @param[in] handle the raft handle
- * @param[in] filename the name of the file that stores the index
- * @param[in] index IVF-PQ index
- *
- */
-void deserialize(raft::resources const& handle,
-                 const std::string& filename,
-                 cuvs::neighbors::ivf_pq::index<int64_t>* index);
-
-}  // namespace raft::runtime::neighbors::ivf_pq
diff --git a/cpp/include/cuvs_runtime/neighbors/refine.hpp b/cpp/include/cuvs_runtime/neighbors/refine.hpp
deleted file mode 100644
index fba7d0fc0..000000000
--- a/cpp/include/cuvs_runtime/neighbors/refine.hpp
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resources.hpp>
-// #include <raft/core/host_mdspan.hpp>
-
-namespace raft::runtime::neighbors {
-
-#define RAFT_INST_REFINE(IDX_T, DATA_T)                                                      \
-  void refine(raft::resources const& handle,                                                 \
-              raft::device_matrix_view<const DATA_T, int64_t, row_major> dataset,            \
-              raft::device_matrix_view<const DATA_T, int64_t, row_major> queries,            \
-              raft::device_matrix_view<const IDX_T, int64_t, row_major> neighbor_candidates, \
-              raft::device_matrix_view<IDX_T, int64_t, row_major> indices,                   \
-              raft::device_matrix_view<float, int64_t, row_major> distances,                 \
-              distance::DistanceType metric);                                                \
-                                                                                             \
-  void refine(raft::resources const& handle,                                                 \
-              raft::host_matrix_view<const DATA_T, int64_t, row_major> dataset,              \
-              raft::host_matrix_view<const DATA_T, int64_t, row_major> queries,              \
-              raft::host_matrix_view<const IDX_T, int64_t, row_major> neighbor_candidates,   \
-              raft::host_matrix_view<IDX_T, int64_t, row_major> indices,                     \
-              raft::host_matrix_view<float, int64_t, row_major> distances,                   \
-              distance::DistanceType metric);
-
-RAFT_INST_REFINE(int64_t, float);
-RAFT_INST_REFINE(int64_t, uint8_t);
-RAFT_INST_REFINE(int64_t, int8_t);
-
-#undef RAFT_INST_REFINE
-
-}  // namespace raft::runtime::neighbors
diff --git a/cpp/src/cuvs_runtime/cluster/cluster_cost.cuh b/cpp/src/cuvs_runtime/cluster/cluster_cost.cuh
deleted file mode 100644
index d00e712f8..000000000
--- a/cpp/src/cuvs_runtime/cluster/cluster_cost.cuh
+++ /dev/null
@@ -1,87 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/cluster/kmeans.cuh>
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/distance/fused_l2_nn.cuh>
-#include <raft/core/operators.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resource/thrust_policy.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/util/cuda_utils.cuh>
-
-namespace cuvs::runtime::cluster::kmeans {
-template <typename ElementType, typename IndexType>
-void cluster_cost(raft::resources const& handle,
-                  const ElementType* X,
-                  IndexType n_samples,
-                  IndexType n_features,
-                  IndexType n_clusters,
-                  const ElementType* centroids,
-                  ElementType* cost)
-{
-  rmm::device_uvector<char> workspace(n_samples * sizeof(IndexType),
-                                      resource::get_cuda_stream(handle));
-
-  rmm::device_uvector<ElementType> x_norms(n_samples, resource::get_cuda_stream(handle));
-  rmm::device_uvector<ElementType> centroid_norms(n_clusters, resource::get_cuda_stream(handle));
-  raft::linalg::rowNorm(x_norms.data(),
-                        X,
-                        n_features,
-                        n_samples,
-                        raft::linalg::L2Norm,
-                        true,
-                        resource::get_cuda_stream(handle));
-  raft::linalg::rowNorm(centroid_norms.data(),
-                        centroids,
-                        n_features,
-                        n_clusters,
-                        raft::linalg::L2Norm,
-                        true,
-                        resource::get_cuda_stream(handle));
-
-  auto min_cluster_distance =
-    raft::make_device_vector<raft::KeyValuePair<IndexType, ElementType>>(handle, n_samples);
-  cuvs::distance::fusedL2NNMinReduce(min_cluster_distance.data_handle(),
-                                     X,
-                                     centroids,
-                                     x_norms.data(),
-                                     centroid_norms.data(),
-                                     n_samples,
-                                     n_clusters,
-                                     n_features,
-                                     (void*)workspace.data(),
-                                     false,
-                                     true,
-                                     resource::get_cuda_stream(handle));
-
-  auto distances = raft::make_device_vector<ElementType, IndexType>(handle, n_samples);
-  thrust::transform(resource::get_thrust_policy(handle),
-                    min_cluster_distance.data_handle(),
-                    min_cluster_distance.data_handle() + n_samples,
-                    distances.data_handle(),
-                    raft::value_op{});
-
-  rmm::device_scalar<ElementType> device_cost(0, resource::get_cuda_stream(handle));
-  cuvs::cluster::kmeans::cluster_cost(handle,
-                                      distances.view(),
-                                      workspace,
-                                      make_device_scalar_view<ElementType>(device_cost.data()),
-                                      raft::add_op{});
-
-  raft::update_host(cost, device_cost.data(), 1, resource::get_cuda_stream(handle));
-}
-}  // namespace cuvs::runtime::cluster::kmeans
diff --git a/cpp/src/cuvs_runtime/cluster/cluster_cost_double.cu b/cpp/src/cuvs_runtime/cluster/cluster_cost_double.cu
deleted file mode 100644
index c516d0804..000000000
--- a/cpp/src/cuvs_runtime/cluster/cluster_cost_double.cu
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "cluster_cost.cuh"
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::runtime::cluster::kmeans {
-
-void cluster_cost(raft::resources const& handle,
-                  const double* X,
-                  int n_samples,
-                  int n_features,
-                  int n_clusters,
-                  const double* centroids,
-                  double* cost)
-{
-  cluster_cost<double, int>(handle, X, n_samples, n_features, n_clusters, centroids, cost);
-}
-}  // namespace cuvs::runtime::cluster::kmeans
diff --git a/cpp/src/cuvs_runtime/cluster/cluster_cost_float.cu b/cpp/src/cuvs_runtime/cluster/cluster_cost_float.cu
deleted file mode 100644
index 135fc0c17..000000000
--- a/cpp/src/cuvs_runtime/cluster/cluster_cost_float.cu
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "cluster_cost.cuh"
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::runtime::cluster::kmeans {
-
-void cluster_cost(raft::resources const& handle,
-                  const float* X,
-                  int n_samples,
-                  int n_features,
-                  int n_clusters,
-                  const float* centroids,
-                  float* cost)
-{
-  cluster_cost<float, int>(handle, X, n_samples, n_features, n_clusters, centroids, cost);
-}
-}  // namespace cuvs::runtime::cluster::kmeans
diff --git a/cpp/src/cuvs_runtime/cluster/kmeans_fit_double.cu b/cpp/src/cuvs_runtime/cluster/kmeans_fit_double.cu
deleted file mode 100644
index b2f518760..000000000
--- a/cpp/src/cuvs_runtime/cluster/kmeans_fit_double.cu
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/cluster/kmeans.cuh>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::runtime::cluster::kmeans {
-
-void fit(raft::resources const& handle,
-         const cuvs::cluster::kmeans::KMeansParams& params,
-         raft::device_matrix_view<const double, int> X,
-         std::optional<raft::device_vector_view<const double, int>> sample_weight,
-         raft::device_matrix_view<double, int> centroids,
-         raft::host_scalar_view<double, int> inertia,
-         raft::host_scalar_view<int, int> n_iter)
-{
-  cuvs::cluster::kmeans::fit<double, int>(
-    handle, params, X, sample_weight, centroids, inertia, n_iter);
-}
-}  // namespace cuvs::runtime::cluster::kmeans
diff --git a/cpp/src/cuvs_runtime/cluster/kmeans_fit_float.cu b/cpp/src/cuvs_runtime/cluster/kmeans_fit_float.cu
deleted file mode 100644
index d7b8b9492..000000000
--- a/cpp/src/cuvs_runtime/cluster/kmeans_fit_float.cu
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/cluster/kmeans.cuh>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::runtime::cluster::kmeans {
-
-void fit(raft::resources const& handle,
-         const cuvs::cluster::kmeans::KMeansParams& params,
-         raft::device_matrix_view<const float, int> X,
-         std::optional<raft::device_vector_view<const float, int>> sample_weight,
-         raft::device_matrix_view<float, int> centroids,
-         raft::host_scalar_view<float, int> inertia,
-         raft::host_scalar_view<int, int> n_iter)
-{
-  cuvs::cluster::kmeans::fit<float, int>(
-    handle, params, X, sample_weight, centroids, inertia, n_iter);
-}
-}  // namespace cuvs::runtime::cluster::kmeans
diff --git a/cpp/src/cuvs_runtime/cluster/kmeans_init_plus_plus_double.cu b/cpp/src/cuvs_runtime/cluster/kmeans_init_plus_plus_double.cu
deleted file mode 100644
index 486e87a31..000000000
--- a/cpp/src/cuvs_runtime/cluster/kmeans_init_plus_plus_double.cu
+++ /dev/null
@@ -1,31 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/cluster/kmeans.cuh>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::runtime::cluster::kmeans {
-
-void init_plus_plus(raft::resources const& handle,
-                    const cuvs::cluster::kmeans::KMeansParams& params,
-                    raft::device_matrix_view<const double, int> X,
-                    raft::device_matrix_view<double, int> centroids)
-{
-  rmm::device_uvector<char> workspace(0, resource::get_cuda_stream(handle));
-  cuvs::cluster::kmeans::init_plus_plus<double, int>(handle, params, X, centroids, workspace);
-}
-}  // namespace cuvs::runtime::cluster::kmeans
diff --git a/cpp/src/cuvs_runtime/cluster/kmeans_init_plus_plus_float.cu b/cpp/src/cuvs_runtime/cluster/kmeans_init_plus_plus_float.cu
deleted file mode 100644
index fbf255a95..000000000
--- a/cpp/src/cuvs_runtime/cluster/kmeans_init_plus_plus_float.cu
+++ /dev/null
@@ -1,31 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/cluster/kmeans.cuh>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::runtime::cluster::kmeans {
-
-void init_plus_plus(raft::resources const& handle,
-                    const cuvs::cluster::kmeans::KMeansParams& params,
-                    raft::device_matrix_view<const float, int> X,
-                    raft::device_matrix_view<float, int> centroids)
-{
-  rmm::device_uvector<char> workspace(0, resource::get_cuda_stream(handle));
-  cuvs::cluster::kmeans::init_plus_plus<float, int>(handle, params, X, centroids, workspace);
-}
-}  // namespace cuvs::runtime::cluster::kmeans
diff --git a/cpp/src/cuvs_runtime/cluster/update_centroids.cuh b/cpp/src/cuvs_runtime/cluster/update_centroids.cuh
deleted file mode 100644
index d1c6c00f1..000000000
--- a/cpp/src/cuvs_runtime/cluster/update_centroids.cuh
+++ /dev/null
@@ -1,72 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/cluster/kmeans.cuh>
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resource/thrust_policy.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/linalg/norm.cuh>
-
-namespace cuvs::runtime::cluster::kmeans {
-
-template <typename DataT, typename IndexT>
-void update_centroids(raft::resources const& handle,
-                      const DataT* X,
-                      int n_samples,
-                      int n_features,
-                      int n_clusters,
-                      const DataT* sample_weights,
-                      const DataT* centroids,
-                      const IndexT* labels,
-                      DataT* new_centroids,
-                      DataT* weight_per_cluster)
-{
-  auto X_view = raft::make_device_matrix_view<const DataT, IndexT>(X, n_samples, n_features);
-  auto centroids_view =
-    raft::make_device_matrix_view<const DataT, IndexT>(centroids, n_clusters, n_features);
-
-  rmm::device_uvector<DataT> sample_weights_uvec(0, resource::get_cuda_stream(handle));
-  if (sample_weights == nullptr) {
-    sample_weights_uvec.resize(n_samples, resource::get_cuda_stream(handle));
-    DataT weight = 1.0 / n_samples;
-    thrust::fill(resource::get_thrust_policy(handle),
-                 sample_weights_uvec.data(),
-                 sample_weights_uvec.data() + n_samples,
-                 weight);
-  }
-  auto sample_weights_view = raft::make_device_vector_view<const DataT, IndexT>(
-    sample_weights == nullptr ? sample_weights_uvec.data() : sample_weights, n_samples);
-
-  auto new_centroids_view =
-    raft::make_device_matrix_view<DataT, IndexT>(new_centroids, n_clusters, n_features);
-  rmm::device_uvector<DataT> weight_per_cluster_uvec(0, resource::get_cuda_stream(handle));
-  if (weight_per_cluster == nullptr) {
-    weight_per_cluster_uvec.resize(n_clusters, resource::get_cuda_stream(handle));
-  }
-  auto weight_per_cluster_view = raft::make_device_vector_view<DataT, IndexT>(
-    weight_per_cluster == nullptr ? weight_per_cluster_uvec.data() : weight_per_cluster,
-    n_clusters);
-
-  cuvs::cluster::kmeans::update_centroids<DataT, IndexT>(handle,
-                                                         X_view,
-                                                         sample_weights_view,
-                                                         centroids_view,
-                                                         labels,
-                                                         weight_per_cluster_view,
-                                                         new_centroids_view);
-}
-}  // namespace cuvs::runtime::cluster::kmeans
\ No newline at end of file
diff --git a/cpp/src/cuvs_runtime/cluster/update_centroids_double.cu b/cpp/src/cuvs_runtime/cluster/update_centroids_double.cu
deleted file mode 100644
index 6cd853b54..000000000
--- a/cpp/src/cuvs_runtime/cluster/update_centroids_double.cu
+++ /dev/null
@@ -1,46 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "update_centroids.cuh"
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::runtime::cluster::kmeans {
-
-void update_centroids(raft::resources const& handle,
-                      const double* X,
-                      int n_samples,
-                      int n_features,
-                      int n_clusters,
-                      const double* sample_weights,
-                      const double* centroids,
-                      const int* labels,
-                      double* new_centroids,
-                      double* weight_per_cluster)
-{
-  update_centroids<double, int>(handle,
-                                X,
-                                n_samples,
-                                n_features,
-                                n_clusters,
-                                sample_weights,
-                                centroids,
-                                labels,
-                                new_centroids,
-                                weight_per_cluster);
-}
-
-}  // namespace cuvs::runtime::cluster::kmeans
\ No newline at end of file
diff --git a/cpp/src/cuvs_runtime/cluster/update_centroids_float.cu b/cpp/src/cuvs_runtime/cluster/update_centroids_float.cu
deleted file mode 100644
index f2294120f..000000000
--- a/cpp/src/cuvs_runtime/cluster/update_centroids_float.cu
+++ /dev/null
@@ -1,46 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "update_centroids.cuh"
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::runtime::cluster::kmeans {
-
-void update_centroids(raft::resources const& handle,
-                      const float* X,
-                      int n_samples,
-                      int n_features,
-                      int n_clusters,
-                      const float* sample_weights,
-                      const float* centroids,
-                      const int* labels,
-                      float* new_centroids,
-                      float* weight_per_cluster)
-{
-  update_centroids<float, int>(handle,
-                               X,
-                               n_samples,
-                               n_features,
-                               n_clusters,
-                               sample_weights,
-                               centroids,
-                               labels,
-                               new_centroids,
-                               weight_per_cluster);
-}
-
-}  // namespace cuvs::runtime::cluster::kmeans
\ No newline at end of file
diff --git a/cpp/src/cuvs_runtime/distance/fused_l2_min_arg.cu b/cpp/src/cuvs_runtime/distance/fused_l2_min_arg.cu
deleted file mode 100644
index 19022c8f7..000000000
--- a/cpp/src/cuvs_runtime/distance/fused_l2_min_arg.cu
+++ /dev/null
@@ -1,105 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/distance/fused_l2_nn.cuh>
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/kvp.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resource/thrust_policy.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/linalg/norm.cuh>
-#include <thrust/for_each.h>
-#include <thrust/tuple.h>
-
-namespace cuvs::runtime::distance {
-
-template <typename IndexT, typename DataT>
-struct KeyValueIndexOp {
-  __host__ __device__ __forceinline__ IndexT
-  operator()(const raft::KeyValuePair<IndexT, DataT>& a) const
-  {
-    return a.key;
-  }
-};
-
-template <typename value_t, typename idx_t>
-void compute_fused_l2_nn_min_arg(raft::resources const& handle,
-                                 idx_t* min,
-                                 const value_t* x,
-                                 const value_t* y,
-                                 idx_t m,
-                                 idx_t n,
-                                 idx_t k,
-                                 bool sqrt)
-{
-  rmm::device_uvector<int> workspace(m, resource::get_cuda_stream(handle));
-  auto kvp = raft::make_device_vector<raft::KeyValuePair<idx_t, value_t>>(handle, m);
-
-  rmm::device_uvector<value_t> x_norms(m, resource::get_cuda_stream(handle));
-  rmm::device_uvector<value_t> y_norms(n, resource::get_cuda_stream(handle));
-  raft::linalg::rowNorm(
-    x_norms.data(), x, k, m, raft::linalg::L2Norm, true, resource::get_cuda_stream(handle));
-  raft::linalg::rowNorm(
-    y_norms.data(), y, k, n, raft::linalg::L2Norm, true, resource::get_cuda_stream(handle));
-
-  cuvs::distance::fusedL2NNMinReduce(kvp.data_handle(),
-                                     x,
-                                     y,
-                                     x_norms.data(),
-                                     y_norms.data(),
-                                     m,
-                                     n,
-                                     k,
-                                     (void*)workspace.data(),
-                                     sqrt,
-                                     true,
-                                     resource::get_cuda_stream(handle));
-
-  KeyValueIndexOp<idx_t, value_t> conversion_op;
-  thrust::transform(resource::get_thrust_policy(handle),
-                    kvp.data_handle(),
-                    kvp.data_handle() + m,
-                    min,
-                    conversion_op);
-  resource::sync_stream(handle);
-}
-
-void fused_l2_nn_min_arg(raft::resources const& handle,
-                         int* min,
-                         const float* x,
-                         const float* y,
-                         int m,
-                         int n,
-                         int k,
-                         bool sqrt)
-{
-  compute_fused_l2_nn_min_arg<float, int>(handle, min, x, y, m, n, k, sqrt);
-}
-
-void fused_l2_nn_min_arg(raft::resources const& handle,
-                         int* min,
-                         const double* x,
-                         const double* y,
-                         int m,
-                         int n,
-                         int k,
-                         bool sqrt)
-{
-  compute_fused_l2_nn_min_arg<double, int>(handle, min, x, y, m, n, k, sqrt);
-}
-
-}  // end namespace cuvs::runtime::distance
diff --git a/cpp/src/cuvs_runtime/distance/pairwise_distance.cu b/cpp/src/cuvs_runtime/distance/pairwise_distance.cu
deleted file mode 100644
index 735863b98..000000000
--- a/cpp/src/cuvs_runtime/distance/pairwise_distance.cu
+++ /dev/null
@@ -1,52 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/distance/distance.cuh>
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/resources.hpp>
-
-namespace cuvs::runtime::distance {
-
-void pairwise_distance(raft::resources const& handle,
-                       float* x,
-                       float* y,
-                       float* dists,
-                       int m,
-                       int n,
-                       int k,
-                       cuvs::distance::DistanceType metric,
-                       bool isRowMajor,
-                       float metric_arg)
-{
-  cuvs::distance::pairwise_distance<float, int>(
-    handle, x, y, dists, m, n, k, metric, isRowMajor, metric_arg);
-}
-
-void pairwise_distance(raft::resources const& handle,
-                       double* x,
-                       double* y,
-                       double* dists,
-                       int m,
-                       int n,
-                       int k,
-                       cuvs::distance::DistanceType metric,
-                       bool isRowMajor,
-                       float metric_arg)
-{
-  cuvs::distance::pairwise_distance<double, int>(
-    handle, x, y, dists, m, n, k, metric, isRowMajor, metric_arg);
-}
-}  // namespace cuvs::runtime::distance
\ No newline at end of file
diff --git a/cpp/src/cuvs_runtime/matrix/select_k_float_int64_t.cu b/cpp/src/cuvs_runtime/matrix/select_k_float_int64_t.cu
deleted file mode 100644
index 204ef66dc..000000000
--- a/cpp/src/cuvs_runtime/matrix/select_k_float_int64_t.cu
+++ /dev/null
@@ -1,36 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/matrix/select_k.cuh>
-
-#include <raft_runtime/matrix/select_k.hpp>
-
-#include <vector>
-
-namespace cuvs::runtime::matrix {
-
-void select_k(const resources& handle,
-              raft::device_matrix_view<const float, int64_t, row_major> in_val,
-              std::optional<raft::device_matrix_view<const int64_t, int64_t, row_major>> in_idx,
-              raft::device_matrix_view<float, int64_t, row_major> out_val,
-              raft::device_matrix_view<int64_t, int64_t, row_major> out_idx,
-              bool select_min)
-{
-  raft::matrix::select_k(handle, in_val, in_idx, out_val, out_idx, select_min);
-}
-}  // namespace cuvs::runtime::matrix
diff --git a/cpp/src/cuvs_runtime/neighbors/brute_force_knn_int64_t_float.cu b/cpp/src/cuvs_runtime/neighbors/brute_force_knn_int64_t_float.cu
deleted file mode 100644
index f53510b9c..000000000
--- a/cpp/src/cuvs_runtime/neighbors/brute_force_knn_int64_t_float.cu
+++ /dev/null
@@ -1,47 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/brute_force.cuh>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resources.hpp>
-
-#include <raft_runtime/neighbors/brute_force.hpp>
-
-#include <vector>
-
-namespace cuvs::runtime::neighbors::brute_force {
-
-#define RAFT_INST_BFKNN(IDX_T, DATA_T, MATRIX_IDX_T, INDEX_LAYOUT, SEARCH_LAYOUT)        \
-  void knn(raft::resources const& handle,                                                \
-           raft::device_matrix_view<const DATA_T, MATRIX_IDX_T, INDEX_LAYOUT> index,     \
-           raft::device_matrix_view<const DATA_T, MATRIX_IDX_T, SEARCH_LAYOUT> search,   \
-           raft::device_matrix_view<IDX_T, MATRIX_IDX_T, row_major> indices,             \
-           raft::device_matrix_view<DATA_T, MATRIX_IDX_T, row_major> distances,          \
-           distance::DistanceType metric,                                                \
-           std::optional<float> metric_arg,                                              \
-           std::optional<IDX_T> global_id_offset)                                        \
-  {                                                                                      \
-    std::vector<raft::device_matrix_view<const DATA_T, MATRIX_IDX_T, INDEX_LAYOUT>> vec; \
-    vec.push_back(index);                                                                \
-    cuvs::neighbors::brute_force::knn(                                                   \
-      handle, vec, search, indices, distances, metric, metric_arg, global_id_offset);    \
-  }
-
-RAFT_INST_BFKNN(int64_t, float, int64_t, raft::row_major, raft::row_major);
-
-#undef RAFT_INST_BFKNN
-
-}  // namespace cuvs::runtime::neighbors::brute_force
diff --git a/cpp/src/cuvs_runtime/neighbors/ivf_flat_build.cu b/cpp/src/cuvs_runtime/neighbors/ivf_flat_build.cu
deleted file mode 100644
index c2f444a5f..000000000
--- a/cpp/src/cuvs_runtime/neighbors/ivf_flat_build.cu
+++ /dev/null
@@ -1,62 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_flat.cuh>
-#include <raft_runtime/neighbors/ivf_flat.hpp>
-
-namespace cuvs::runtime::neighbors::ivf_flat {
-
-#define RAFT_INST_BUILD_EXTEND(T, IdxT)                                                \
-  auto build(raft::resources const& handle,                                            \
-             const cuvs::neighbors::ivf_flat::index_params& params,                    \
-             raft::device_matrix_view<const T, IdxT, row_major> dataset)               \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>                                        \
-  {                                                                                    \
-    return cuvs::neighbors::ivf_flat::build<T, IdxT>(handle, params, dataset);         \
-  }                                                                                    \
-  auto extend(raft::resources const& handle,                                           \
-              raft::device_matrix_view<const T, IdxT, row_major> new_vectors,          \
-              std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices,   \
-              const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index)             \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>                                        \
-  {                                                                                    \
-    return cuvs::neighbors::ivf_flat::extend<T, IdxT>(                                 \
-      handle, new_vectors, new_indices, orig_index);                                   \
-  }                                                                                    \
-                                                                                       \
-  void build(raft::resources const& handle,                                            \
-             const cuvs::neighbors::ivf_flat::index_params& params,                    \
-             raft::device_matrix_view<const T, IdxT, row_major> dataset,               \
-             cuvs::neighbors::ivf_flat::index<T, IdxT>& idx)                           \
-  {                                                                                    \
-    idx = build(handle, params, dataset);                                              \
-  }                                                                                    \
-                                                                                       \
-  void extend(raft::resources const& handle,                                           \
-              raft::device_matrix_view<const T, IdxT, row_major> new_vectors,          \
-              std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices,   \
-              cuvs::neighbors::ivf_flat::index<T, IdxT>* idx)                          \
-  {                                                                                    \
-    cuvs::neighbors::ivf_flat::extend<T, IdxT>(handle, new_vectors, new_indices, idx); \
-  }
-
-RAFT_INST_BUILD_EXTEND(float, int64_t);
-RAFT_INST_BUILD_EXTEND(int8_t, int64_t);
-RAFT_INST_BUILD_EXTEND(uint8_t, int64_t);
-
-#undef RAFT_INST_BUILD_EXTEND
-
-}  // namespace cuvs::runtime::neighbors::ivf_flat
diff --git a/cpp/src/cuvs_runtime/neighbors/ivf_flat_search.cu b/cpp/src/cuvs_runtime/neighbors/ivf_flat_search.cu
deleted file mode 100644
index 341a214f6..000000000
--- a/cpp/src/cuvs_runtime/neighbors/ivf_flat_search.cu
+++ /dev/null
@@ -1,40 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_flat.cuh>
-#include <raft_runtime/neighbors/ivf_flat.hpp>
-
-namespace cuvs::runtime::neighbors::ivf_flat {
-
-#define RAFT_INST_SEARCH(T, IdxT)                                         \
-  void search(raft::resources const& handle,                              \
-              cuvs::neighbors::ivf_flat::search_params const& params,     \
-              const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,     \
-              raft::device_matrix_view<const T, IdxT, row_major> queries, \
-              raft::device_matrix_view<IdxT, IdxT, row_major> neighbors,  \
-              raft::device_matrix_view<float, IdxT, row_major> distances) \
-  {                                                                       \
-    cuvs::neighbors::ivf_flat::search<T, IdxT>(                           \
-      handle, params, index, queries, neighbors, distances);              \
-  }
-
-RAFT_INST_SEARCH(float, int64_t);
-RAFT_INST_SEARCH(int8_t, int64_t);
-RAFT_INST_SEARCH(uint8_t, int64_t);
-
-#undef RAFT_INST_SEARCH
-
-}  // namespace cuvs::runtime::neighbors::ivf_flat
diff --git a/cpp/src/cuvs_runtime/neighbors/ivf_flat_serialize.cu b/cpp/src/cuvs_runtime/neighbors/ivf_flat_serialize.cu
deleted file mode 100644
index 5d8e47094..000000000
--- a/cpp/src/cuvs_runtime/neighbors/ivf_flat_serialize.cu
+++ /dev/null
@@ -1,65 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <sstream>
-#include <string>
-
-#include <cuvs/neighbors/ivf_flat_serialize.cuh>
-#include <cuvs/neighbors/ivf_flat_types.hpp>
-#include <raft/core/device_resources.hpp>
-#include <raft_runtime/neighbors/ivf_flat.hpp>
-
-namespace cuvs::runtime::neighbors::ivf_flat {
-
-#define RAFT_IVF_FLAT_SERIALIZE_INST(DTYPE)                                            \
-  void serialize_file(raft::resources const& handle,                                   \
-                      const std::string& filename,                                     \
-                      const cuvs::neighbors::ivf_flat::index<DTYPE, int64_t>& index)   \
-  {                                                                                    \
-    cuvs::neighbors::ivf_flat::serialize(handle, filename, index);                     \
-  };                                                                                   \
-                                                                                       \
-  void deserialize_file(raft::resources const& handle,                                 \
-                        const std::string& filename,                                   \
-                        cuvs::neighbors::ivf_flat::index<DTYPE, int64_t>* index)       \
-  {                                                                                    \
-    if (!index) { RAFT_FAIL("Invalid index pointer"); }                                \
-    *index = cuvs::neighbors::ivf_flat::deserialize<DTYPE, int64_t>(handle, filename); \
-  };                                                                                   \
-  void serialize(raft::resources const& handle,                                        \
-                 std::string& str,                                                     \
-                 const cuvs::neighbors::ivf_flat::index<DTYPE, int64_t>& index)        \
-  {                                                                                    \
-    std::stringstream os;                                                              \
-    cuvs::neighbors::ivf_flat::serialize(handle, os, index);                           \
-    str = os.str();                                                                    \
-  }                                                                                    \
-                                                                                       \
-  void deserialize(raft::resources const& handle,                                      \
-                   const std::string& str,                                             \
-                   cuvs::neighbors::ivf_flat::index<DTYPE, int64_t>* index)            \
-  {                                                                                    \
-    std::istringstream is(str);                                                        \
-    if (!index) { RAFT_FAIL("Invalid index pointer"); }                                \
-    *index = cuvs::neighbors::ivf_flat::deserialize<DTYPE, int64_t>(handle, is);       \
-  }
-
-RAFT_IVF_FLAT_SERIALIZE_INST(float);
-RAFT_IVF_FLAT_SERIALIZE_INST(int8_t);
-RAFT_IVF_FLAT_SERIALIZE_INST(uint8_t);
-
-#undef RAFT_IVF_FLAT_SERIALIZE_INST
-}  // namespace cuvs::runtime::neighbors::ivf_flat
diff --git a/cpp/src/cuvs_runtime/neighbors/ivfpq_build.cu b/cpp/src/cuvs_runtime/neighbors/ivfpq_build.cu
deleted file mode 100644
index db5a26a31..000000000
--- a/cpp/src/cuvs_runtime/neighbors/ivfpq_build.cu
+++ /dev/null
@@ -1,59 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq.cuh>
-#include <raft_runtime/neighbors/ivf_pq.hpp>
-
-namespace cuvs::runtime::neighbors::ivf_pq {
-
-#define RAFT_INST_BUILD_EXTEND(T, IdxT)                                                     \
-  cuvs::neighbors::ivf_pq::index<IdxT> build(                                               \
-    raft::resources const& handle,                                                          \
-    const cuvs::neighbors::ivf_pq::index_params& params,                                    \
-    raft::device_matrix_view<const T, IdxT, row_major> dataset)                             \
-  {                                                                                         \
-    return cuvs::neighbors::ivf_pq::build<T, IdxT>(handle, params, dataset);                \
-  }                                                                                         \
-  void build(raft::resources const& handle,                                                 \
-             const cuvs::neighbors::ivf_pq::index_params& params,                           \
-             raft::device_matrix_view<const T, IdxT, row_major> dataset,                    \
-             cuvs::neighbors::ivf_pq::index<IdxT>* idx)                                     \
-  {                                                                                         \
-    *idx = cuvs::neighbors::ivf_pq::build<T, IdxT>(handle, params, dataset);                \
-  }                                                                                         \
-  cuvs::neighbors::ivf_pq::index<IdxT> extend(                                              \
-    raft::resources const& handle,                                                          \
-    raft::device_matrix_view<const T, IdxT, row_major> new_vectors,                         \
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices,                  \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx)                                        \
-  {                                                                                         \
-    return cuvs::neighbors::ivf_pq::extend<T, IdxT>(handle, new_vectors, new_indices, idx); \
-  }                                                                                         \
-  void extend(raft::resources const& handle,                                                \
-              raft::device_matrix_view<const T, IdxT, row_major> new_vectors,               \
-              std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices,        \
-              cuvs::neighbors::ivf_pq::index<IdxT>* idx)                                    \
-  {                                                                                         \
-    cuvs::neighbors::ivf_pq::extend<T, IdxT>(handle, new_vectors, new_indices, idx);        \
-  }
-
-RAFT_INST_BUILD_EXTEND(float, int64_t);
-RAFT_INST_BUILD_EXTEND(int8_t, int64_t);
-RAFT_INST_BUILD_EXTEND(uint8_t, int64_t);
-
-#undef RAFT_INST_BUILD_EXTEND
-
-}  // namespace cuvs::runtime::neighbors::ivf_pq
diff --git a/cpp/src/cuvs_runtime/neighbors/ivfpq_deserialize.cu b/cpp/src/cuvs_runtime/neighbors/ivfpq_deserialize.cu
deleted file mode 100644
index f65c4fdcb..000000000
--- a/cpp/src/cuvs_runtime/neighbors/ivfpq_deserialize.cu
+++ /dev/null
@@ -1,31 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq.cuh>
-#include <cuvs/neighbors/ivf_pq_serialize.cuh>
-
-#include <raft_runtime/neighbors/ivf_pq.hpp>
-
-namespace cuvs::runtime::neighbors::ivf_pq {
-
-void deserialize(raft::resources const& handle,
-                 const std::string& filename,
-                 cuvs::neighbors::ivf_pq::index<int64_t>* index)
-{
-  if (!index) { RAFT_FAIL("Invalid index pointer"); }
-  *index = cuvs::neighbors::ivf_pq::deserialize<int64_t>(handle, filename);
-};
-}  // namespace cuvs::runtime::neighbors::ivf_pq
diff --git a/cpp/src/cuvs_runtime/neighbors/ivfpq_search_float_int64_t.cu b/cpp/src/cuvs_runtime/neighbors/ivfpq_search_float_int64_t.cu
deleted file mode 100644
index 1cec55f74..000000000
--- a/cpp/src/cuvs_runtime/neighbors/ivfpq_search_float_int64_t.cu
+++ /dev/null
@@ -1,38 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq.cuh>
-
-#include <raft_runtime/neighbors/ivf_pq.hpp>
-
-namespace cuvs::runtime::neighbors::ivf_pq {
-
-#define RAFT_SEARCH_INST(T, IdxT)                                                                 \
-  void search(raft::resources const& handle,                                                      \
-              const cuvs::neighbors::ivf_pq::search_params& params,                               \
-              const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                                    \
-              raft::device_matrix_view<const T, IdxT, row_major> queries,                         \
-              raft::device_matrix_view<IdxT, IdxT, row_major> neighbors,                          \
-              raft::device_matrix_view<float, IdxT, row_major> distances)                         \
-  {                                                                                               \
-    cuvs::neighbors::ivf_pq::search<T, IdxT>(handle, params, idx, queries, neighbors, distances); \
-  }
-
-RAFT_SEARCH_INST(float, int64_t);
-
-#undef RAFT_INST_SEARCH
-
-}  // namespace cuvs::runtime::neighbors::ivf_pq
diff --git a/cpp/src/cuvs_runtime/neighbors/ivfpq_search_int8_t_int64_t.cu b/cpp/src/cuvs_runtime/neighbors/ivfpq_search_int8_t_int64_t.cu
deleted file mode 100644
index 1cc3aa7c5..000000000
--- a/cpp/src/cuvs_runtime/neighbors/ivfpq_search_int8_t_int64_t.cu
+++ /dev/null
@@ -1,38 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq.cuh>
-
-#include <raft_runtime/neighbors/ivf_pq.hpp>
-
-namespace cuvs::runtime::neighbors::ivf_pq {
-
-#define RAFT_SEARCH_INST(T, IdxT)                                                                 \
-  void search(raft::resources const& handle,                                                      \
-              const cuvs::neighbors::ivf_pq::search_params& params,                               \
-              const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                                    \
-              raft::device_matrix_view<const T, IdxT, row_major> queries,                         \
-              raft::device_matrix_view<IdxT, IdxT, row_major> neighbors,                          \
-              raft::device_matrix_view<float, IdxT, row_major> distances)                         \
-  {                                                                                               \
-    cuvs::neighbors::ivf_pq::search<T, IdxT>(handle, params, idx, queries, neighbors, distances); \
-  }
-
-RAFT_SEARCH_INST(int8_t, int64_t);
-
-#undef RAFT_INST_SEARCH
-
-}  // namespace cuvs::runtime::neighbors::ivf_pq
diff --git a/cpp/src/cuvs_runtime/neighbors/ivfpq_search_uint8_t_int64_t.cu b/cpp/src/cuvs_runtime/neighbors/ivfpq_search_uint8_t_int64_t.cu
deleted file mode 100644
index dce617ea5..000000000
--- a/cpp/src/cuvs_runtime/neighbors/ivfpq_search_uint8_t_int64_t.cu
+++ /dev/null
@@ -1,38 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq.cuh>
-
-#include <raft_runtime/neighbors/ivf_pq.hpp>
-
-namespace cuvs::runtime::neighbors::ivf_pq {
-
-#define RAFT_SEARCH_INST(T, IdxT)                                                                 \
-  void search(raft::resources const& handle,                                                      \
-              const cuvs::neighbors::ivf_pq::search_params& params,                               \
-              const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                                    \
-              raft::device_matrix_view<const T, IdxT, row_major> queries,                         \
-              raft::device_matrix_view<IdxT, IdxT, row_major> neighbors,                          \
-              raft::device_matrix_view<float, IdxT, row_major> distances)                         \
-  {                                                                                               \
-    cuvs::neighbors::ivf_pq::search<T, IdxT>(handle, params, idx, queries, neighbors, distances); \
-  }
-
-RAFT_SEARCH_INST(uint8_t, int64_t);
-
-#undef RAFT_INST_SEARCH
-
-}  // namespace cuvs::runtime::neighbors::ivf_pq
diff --git a/cpp/src/cuvs_runtime/neighbors/ivfpq_serialize.cu b/cpp/src/cuvs_runtime/neighbors/ivfpq_serialize.cu
deleted file mode 100644
index 941e339ae..000000000
--- a/cpp/src/cuvs_runtime/neighbors/ivfpq_serialize.cu
+++ /dev/null
@@ -1,31 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq.cuh>
-#include <cuvs/neighbors/ivf_pq_serialize.cuh>
-
-#include <raft_runtime/neighbors/ivf_pq.hpp>
-
-namespace cuvs::runtime::neighbors::ivf_pq {
-
-void serialize(raft::resources const& handle,
-               const std::string& filename,
-               const cuvs::neighbors::ivf_pq::index<int64_t>& index)
-{
-  cuvs::neighbors::ivf_pq::serialize(handle, filename, index);
-};
-
-}  // namespace cuvs::runtime::neighbors::ivf_pq
diff --git a/cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_float.cu b/cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_float.cu
deleted file mode 100644
index 916273c4f..000000000
--- a/cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_float.cu
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/refine.cuh>
-
-namespace cuvs::runtime::neighbors {
-
-void refine(raft::resources const& handle,
-            raft::device_matrix_view<const float, int64_t, row_major> dataset,
-            raft::device_matrix_view<const float, int64_t, row_major> queries,
-            raft::device_matrix_view<const int64_t, int64_t, row_major> neighbor_candidates,
-            raft::device_matrix_view<int64_t, int64_t, row_major> indices,
-            raft::device_matrix_view<float, int64_t, row_major> distances,
-            distance::DistanceType metric)
-{
-  cuvs::neighbors::refine<int64_t, float, float, int64_t>(
-    handle, dataset, queries, neighbor_candidates, indices, distances, metric);
-}
-
-}  // namespace cuvs::runtime::neighbors
diff --git a/cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_int8_t.cu b/cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_int8_t.cu
deleted file mode 100644
index 2d867fb62..000000000
--- a/cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_int8_t.cu
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/refine.cuh>
-
-namespace cuvs::runtime::neighbors {
-
-void refine(raft::resources const& handle,
-            raft::device_matrix_view<const int8_t, int64_t, row_major> dataset,
-            raft::device_matrix_view<const int8_t, int64_t, row_major> queries,
-            raft::device_matrix_view<const int64_t, int64_t, row_major> neighbor_candidates,
-            raft::device_matrix_view<int64_t, int64_t, row_major> indices,
-            raft::device_matrix_view<float, int64_t, row_major> distances,
-            distance::DistanceType metric)
-{
-  cuvs::neighbors::refine<int64_t, int8_t, float, int64_t>(
-    handle, dataset, queries, neighbor_candidates, indices, distances, metric);
-}
-
-}  // namespace cuvs::runtime::neighbors
diff --git a/cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_uint8_t.cu b/cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_uint8_t.cu
deleted file mode 100644
index 8eea64748..000000000
--- a/cpp/src/cuvs_runtime/neighbors/refine_d_int64_t_uint8_t.cu
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/refine.cuh>
-
-namespace cuvs::runtime::neighbors {
-
-void refine(raft::resources const& handle,
-            raft::device_matrix_view<const uint8_t, int64_t, row_major> dataset,
-            raft::device_matrix_view<const uint8_t, int64_t, row_major> queries,
-            raft::device_matrix_view<const int64_t, int64_t, row_major> neighbor_candidates,
-            raft::device_matrix_view<int64_t, int64_t, row_major> indices,
-            raft::device_matrix_view<float, int64_t, row_major> distances,
-            distance::DistanceType metric)
-{
-  cuvs::neighbors::refine<int64_t, uint8_t, float, int64_t>(
-    handle, dataset, queries, neighbor_candidates, indices, distances, metric);
-}
-
-}  // namespace cuvs::runtime::neighbors
diff --git a/cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_float.cu b/cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_float.cu
deleted file mode 100644
index 65cdcbfe8..000000000
--- a/cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_float.cu
+++ /dev/null
@@ -1,34 +0,0 @@
-
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/refine.cuh>
-
-namespace cuvs::runtime::neighbors {
-
-void refine(raft::resources const& handle,
-            raft::host_matrix_view<const float, int64_t, row_major> dataset,
-            raft::host_matrix_view<const float, int64_t, row_major> queries,
-            raft::host_matrix_view<const int64_t, int64_t, row_major> neighbor_candidates,
-            raft::host_matrix_view<int64_t, int64_t, row_major> indices,
-            raft::host_matrix_view<float, int64_t, row_major> distances,
-            distance::DistanceType metric)
-{
-  cuvs::neighbors::refine<int64_t, float, float, int64_t>(
-    handle, dataset, queries, neighbor_candidates, indices, distances, metric);
-}
-
-}  // namespace cuvs::runtime::neighbors
diff --git a/cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_int8_t.cu b/cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_int8_t.cu
deleted file mode 100644
index fc7021074..000000000
--- a/cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_int8_t.cu
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/refine.cuh>
-
-namespace cuvs::runtime::neighbors {
-
-void refine(raft::resources const& handle,
-            raft::host_matrix_view<const int8_t, int64_t, row_major> dataset,
-            raft::host_matrix_view<const int8_t, int64_t, row_major> queries,
-            raft::host_matrix_view<const int64_t, int64_t, row_major> neighbor_candidates,
-            raft::host_matrix_view<int64_t, int64_t, row_major> indices,
-            raft::host_matrix_view<float, int64_t, row_major> distances,
-            distance::DistanceType metric)
-{
-  cuvs::neighbors::refine<int64_t, int8_t, float, int64_t>(
-    handle, dataset, queries, neighbor_candidates, indices, distances, metric);
-}
-
-}  // namespace cuvs::runtime::neighbors
diff --git a/cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_uint8_t.cu b/cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_uint8_t.cu
deleted file mode 100644
index 31241e55f..000000000
--- a/cpp/src/cuvs_runtime/neighbors/refine_h_int64_t_uint8_t.cu
+++ /dev/null
@@ -1,33 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/refine.cuh>
-
-namespace cuvs::runtime::neighbors {
-
-void refine(raft::resources const& handle,
-            raft::host_matrix_view<const uint8_t, int64_t, row_major> dataset,
-            raft::host_matrix_view<const uint8_t, int64_t, row_major> queries,
-            raft::host_matrix_view<const int64_t, int64_t, row_major> neighbor_candidates,
-            raft::host_matrix_view<int64_t, int64_t, row_major> indices,
-            raft::host_matrix_view<float, int64_t, row_major> distances,
-            distance::DistanceType metric)
-{
-  cuvs::neighbors::refine<int64_t, uint8_t, float, int64_t>(
-    handle, dataset, queries, neighbor_candidates, indices, distances, metric);
-}
-
-}  // namespace cuvs::runtime::neighbors
diff --git a/cpp/src/cuvs_runtime/random/common.cuh b/cpp/src/cuvs_runtime/random/common.cuh
deleted file mode 100644
index f5f8a7c0a..000000000
--- a/cpp/src/cuvs_runtime/random/common.cuh
+++ /dev/null
@@ -1,41 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/random/rmat_rectangular_generator.cuh>
-#include <raft_runtime/random/rmat_rectangular_generator.hpp>
-
-#define FUNC_DEF(IdxT, ProbT)                                                          \
-  void rmat_rectangular_gen(raft::resources const& handle,                             \
-                            IdxT* out,                                                 \
-                            IdxT* out_src,                                             \
-                            IdxT* out_dst,                                             \
-                            const ProbT* theta,                                        \
-                            IdxT r_scale,                                              \
-                            IdxT c_scale,                                              \
-                            IdxT n_edges,                                              \
-                            raft::random::RngState& r)                                 \
-  {                                                                                    \
-    raft::random::rmat_rectangular_gen<IdxT, ProbT>(out,                               \
-                                                    out_src,                           \
-                                                    out_dst,                           \
-                                                    theta,                             \
-                                                    r_scale,                           \
-                                                    c_scale,                           \
-                                                    n_edges,                           \
-                                                    resource::get_cuda_stream(handle), \
-                                                    r);                                \
-  }
diff --git a/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int64_double.cu b/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int64_double.cu
deleted file mode 100644
index 392cbaf09..000000000
--- a/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int64_double.cu
+++ /dev/null
@@ -1,23 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "common.cuh"
-
-namespace cuvs::runtime::random {
-
-FUNC_DEF(int64_t, double);
-
-}  // namespace cuvs::runtime::random
diff --git a/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int64_float.cu b/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int64_float.cu
deleted file mode 100644
index 19ab29e81..000000000
--- a/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int64_float.cu
+++ /dev/null
@@ -1,23 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "common.cuh"
-
-namespace cuvs::runtime::random {
-
-FUNC_DEF(int64_t, float);
-
-}  // namespace cuvs::runtime::random
diff --git a/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int_double.cu b/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int_double.cu
deleted file mode 100644
index 4075d9b6a..000000000
--- a/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int_double.cu
+++ /dev/null
@@ -1,23 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "common.cuh"
-
-namespace cuvs::runtime::random {
-
-FUNC_DEF(int, double);
-
-}  // namespace cuvs::runtime::random
diff --git a/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int_float.cu b/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int_float.cu
deleted file mode 100644
index b7f131e40..000000000
--- a/cpp/src/cuvs_runtime/random/rmat_rectangular_generator_int_float.cu
+++ /dev/null
@@ -1,23 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "common.cuh"
-
-namespace cuvs::runtime::random {
-
-FUNC_DEF(int, float);
-
-}  // namespace cuvs::runtime::random
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_00_generate.py b/cpp/src/distance/detail/pairwise_matrix/dispatch_00_generate.py
deleted file mode 100644
index 7022783c5..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_00_generate.py
+++ /dev/null
@@ -1,194 +0,0 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-# NOTE: this template is not perfectly formatted. Use pre-commit to get
-# everything in shape again.
-header = """/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <raft/core/operators.hpp> // raft::identity_op
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>  // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh> // dispatch
-"""
-
-
-macro = """
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \\
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \\
-  template void cuvs::distance::detail::                                               \\
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \\
-      OpT<DataT, AccT, IdxT> distance_op,                                              \\
-      IdxT m,                                                                          \\
-      IdxT n,                                                                          \\
-      IdxT k,                                                                          \\
-      const DataT* x,                                                                  \\
-      const DataT* y,                                                                  \\
-      const DataT* x_norm,                                                             \\
-      const DataT* y_norm,                                                             \\
-      OutT* out,                                                                       \\
-      FinOpT fin_op,                                                                   \\
-      cudaStream_t stream,                                                             \\
-      bool is_row_major)
-"""
-
-data_type_instances = [
-    dict(
-        DataT="float",
-        AccT="float",
-        OutT="float",
-        IdxT="int",
-    ),
-    dict(
-        DataT="double",
-        AccT="double",
-        OutT="double",
-        IdxT="int",
-    ),
-]
-
-op_instances = [
-    dict(
-        path_prefix="canberra",
-        OpT="cuvs::distance::detail::ops::canberra_distance_op",
-        archs = [60],
-    ),
-    dict(
-        path_prefix="correlation",
-        OpT="cuvs::distance::detail::ops::correlation_distance_op",
-        archs = [60],
-    ),
-    dict(
-        path_prefix="cosine",
-        OpT="cuvs::distance::detail::ops::cosine_distance_op",
-        archs = [60, 80],
-    ),
-    dict(
-        path_prefix="hamming_unexpanded",
-        OpT="cuvs::distance::detail::ops::hamming_distance_op",
-        archs = [60],
-    ),
-    dict(
-        path_prefix="hellinger_expanded",
-        OpT="cuvs::distance::detail::ops::hellinger_distance_op",
-        archs = [60],
-    ),
-    # inner product is handled by cublas.
-    dict(
-        path_prefix="jensen_shannon",
-        OpT="cuvs::distance::detail::ops::jensen_shannon_distance_op",
-        archs = [60],
-    ),
-    dict(
-        path_prefix="kl_divergence",
-        OpT="cuvs::distance::detail::ops::kl_divergence_op",
-        archs = [60],
-    ),
-    dict(
-        path_prefix="l1",
-        OpT="cuvs::distance::detail::ops::l1_distance_op",
-        archs = [60],
-    ),
-    dict(
-        path_prefix="l2_expanded",
-        OpT="cuvs::distance::detail::ops::l2_exp_distance_op",
-        archs = [60, 80],
-    ),
-    dict(
-        path_prefix="l2_unexpanded",
-        OpT="cuvs::distance::detail::ops::l2_unexp_distance_op",
-        archs = [60],
-    ),
-    dict(
-        path_prefix="l_inf",
-        OpT="cuvs::distance::detail::ops::l_inf_distance_op",
-        archs = [60],
-    ),
-    dict(
-        path_prefix="lp_unexpanded",
-        OpT="cuvs::distance::detail::ops::lp_unexp_distance_op",
-        archs = [60],
-    ),
-    dict(
-        path_prefix="russel_rao",
-        OpT="cuvs::distance::detail::ops::russel_rao_distance_op",
-        archs = [60],
-     ),
-]
-
-def arch_headers(archs):
-    include_headers ="\n".join([
-        f"#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm{arch}.cuh>"
-        for arch in archs
-    ])
-    return include_headers
-
-
-
-for op in op_instances:
-    for dt in data_type_instances:
-        DataT, AccT, OutT, IdxT = (dt[k] for k in ["DataT", "AccT", "OutT", "IdxT"]);
-        path = f"dispatch_{op['path_prefix']}_{DataT}_{AccT}_{OutT}_{IdxT}.cu"
-        with open(path, "w") as f:
-            f.write(header)
-            f.write(arch_headers(op["archs"]))
-            f.write(macro)
-
-            OpT = op['OpT']
-            FinOpT = "raft::identity_op"
-            f.write(f"\ninstantiate_raft_distance_detail_pairwise_matrix_dispatch({OpT}, {DataT}, {AccT}, {OutT}, {FinOpT}, {IdxT});\n")
-            f.write("\n#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch\n")
-        print(f"src/distance/detail/pairwise_matrix/{path}")
-
-# Dispatch kernels for with the RBF fin op.
-with open("dispatch_rbf.cu", "w") as f:
-        OpT="cuvs::distance::detail::ops::l2_unexp_distance_op"
-        archs = [60]
-
-        f.write(header)
-        f.write("#include <cuvs/distance/detail/kernels/rbf_fin_op.cuh> // rbf_fin_op\n")
-        f.write(arch_headers(archs))
-        f.write(macro)
-
-        for dt in data_type_instances:
-            DataT, AccT, OutT, IdxT = (dt[k] for k in ["DataT", "AccT", "OutT", "IdxT"]);
-            IdxT = "int64_t"    # overwrite IdxT
-
-            FinOpT = f"cuvs::distance::kernels::detail::rbf_fin_op<{DataT}>"
-            f.write(f"\ninstantiate_raft_distance_detail_pairwise_matrix_dispatch({OpT}, {DataT}, {AccT}, {OutT}, {FinOpT}, {IdxT});\n")
-
-        f.write("\n#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch\n")
-
-print("src/distance/detail/pairwise_matrix/dispatch_rbf.cu")
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_canberra_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_canberra_double_double_double_int.cu
deleted file mode 100644
index d94ad8097..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_canberra_double_double_double_int.cu
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::canberra_distance_op,
-  double,
-  double,
-  double,
-  raft::identity_op,
-  int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_canberra_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_canberra_float_float_float_int.cu
deleted file mode 100644
index 6d6fb1062..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_canberra_float_float_float_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::canberra_distance_op, float, float, float, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_correlation_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_correlation_double_double_double_int.cu
deleted file mode 100644
index 4783dc4f8..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_correlation_double_double_double_int.cu
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::correlation_distance_op,
-  double,
-  double,
-  double,
-  raft::identity_op,
-  int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_correlation_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_correlation_float_float_float_int.cu
deleted file mode 100644
index dc989a24d..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_correlation_float_float_float_int.cu
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::correlation_distance_op,
-  float,
-  float,
-  float,
-  raft::identity_op,
-  int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_cosine_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_cosine_double_double_double_int.cu
deleted file mode 100644
index f46c9b5be..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_cosine_double_double_double_int.cu
+++ /dev/null
@@ -1,51 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm80.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::cosine_distance_op, double, double, double, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_cosine_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_cosine_float_float_float_int.cu
deleted file mode 100644
index 47b9d18d7..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_cosine_float_float_float_int.cu
+++ /dev/null
@@ -1,51 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm80.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::cosine_distance_op, float, float, float, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_double_double_double_int.cu
deleted file mode 100644
index ae9cb608d..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_double_double_double_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::hamming_distance_op, double, double, double, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_float_float_float_int.cu
deleted file mode 100644
index 3a8654bac..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_hamming_unexpanded_float_float_float_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::hamming_distance_op, float, float, float, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_hellinger_expanded_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_hellinger_expanded_double_double_double_int.cu
deleted file mode 100644
index 07b350a6f..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_hellinger_expanded_double_double_double_int.cu
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::hellinger_distance_op,
-  double,
-  double,
-  double,
-  raft::identity_op,
-  int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_hellinger_expanded_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_hellinger_expanded_float_float_float_int.cu
deleted file mode 100644
index 2e3d16c61..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_hellinger_expanded_float_float_float_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::hellinger_distance_op, float, float, float, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_jensen_shannon_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_jensen_shannon_double_double_double_int.cu
deleted file mode 100644
index b736fa024..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_jensen_shannon_double_double_double_int.cu
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::jensen_shannon_distance_op,
-  double,
-  double,
-  double,
-  raft::identity_op,
-  int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_jensen_shannon_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_jensen_shannon_float_float_float_int.cu
deleted file mode 100644
index 5be1dcf8c..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_jensen_shannon_float_float_float_int.cu
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::jensen_shannon_distance_op,
-  float,
-  float,
-  float,
-  raft::identity_op,
-  int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_kl_divergence_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_kl_divergence_double_double_double_int.cu
deleted file mode 100644
index de783ffa2..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_kl_divergence_double_double_double_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::kl_divergence_op, double, double, double, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_kl_divergence_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_kl_divergence_float_float_float_int.cu
deleted file mode 100644
index b175ec2a9..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_kl_divergence_float_float_float_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::kl_divergence_op, float, float, float, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_l1_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_l1_double_double_double_int.cu
deleted file mode 100644
index 71fb2c361..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_l1_double_double_double_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::l1_distance_op, double, double, double, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_l1_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_l1_float_float_float_int.cu
deleted file mode 100644
index cef6d6317..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_l1_float_float_float_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::l1_distance_op, float, float, float, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_expanded_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_expanded_double_double_double_int.cu
deleted file mode 100644
index bb03eb04e..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_expanded_double_double_double_int.cu
+++ /dev/null
@@ -1,51 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm80.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::l2_exp_distance_op, double, double, double, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_expanded_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_expanded_float_float_float_int.cu
deleted file mode 100644
index 55735f403..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_expanded_float_float_float_int.cu
+++ /dev/null
@@ -1,51 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm80.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::l2_exp_distance_op, float, float, float, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_unexpanded_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_unexpanded_double_double_double_int.cu
deleted file mode 100644
index 75aed7c74..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_unexpanded_double_double_double_int.cu
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::l2_unexp_distance_op,
-  double,
-  double,
-  double,
-  raft::identity_op,
-  int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_unexpanded_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_unexpanded_float_float_float_int.cu
deleted file mode 100644
index 79696a1a9..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_l2_unexpanded_float_float_float_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::l2_unexp_distance_op, float, float, float, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_l_inf_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_l_inf_double_double_double_int.cu
deleted file mode 100644
index 8fc41e5ff..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_l_inf_double_double_double_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::l_inf_distance_op, double, double, double, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_l_inf_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_l_inf_float_float_float_int.cu
deleted file mode 100644
index c996e90a6..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_l_inf_float_float_float_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::l_inf_distance_op, float, float, float, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_lp_unexpanded_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_lp_unexpanded_double_double_double_int.cu
deleted file mode 100644
index 16a9c87c3..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_lp_unexpanded_double_double_double_int.cu
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::lp_unexp_distance_op,
-  double,
-  double,
-  double,
-  raft::identity_op,
-  int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_lp_unexpanded_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_lp_unexpanded_float_float_float_int.cu
deleted file mode 100644
index f9fe4e6c7..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_lp_unexpanded_float_float_float_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::lp_unexp_distance_op, float, float, float, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_rbf.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_rbf.cu
deleted file mode 100644
index 634346d61..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_rbf.cu
+++ /dev/null
@@ -1,64 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/kernels/rbf_fin_op.cuh>            // rbf_fin_op
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::l2_unexp_distance_op,
-  float,
-  float,
-  float,
-  cuvs::distance::kernels::detail::rbf_fin_op<float>,
-  int64_t);
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::l2_unexp_distance_op,
-  double,
-  double,
-  double,
-  cuvs::distance::kernels::detail::rbf_fin_op<double>,
-  int64_t);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_russel_rao_double_double_double_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_russel_rao_double_double_double_int.cu
deleted file mode 100644
index f79f8a67c..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_russel_rao_double_double_double_int.cu
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::russel_rao_distance_op,
-  double,
-  double,
-  double,
-  raft::identity_op,
-  int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/detail/pairwise_matrix/dispatch_russel_rao_float_float_float_int.cu b/cpp/src/distance/detail/pairwise_matrix/dispatch_russel_rao_float_float_float_int.cu
deleted file mode 100644
index c3878b945..000000000
--- a/cpp/src/distance/detail/pairwise_matrix/dispatch_russel_rao_float_float_float_int.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by dispatch_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python dispatch_00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/distance_ops/all_ops.cuh>          // ops::*
-#include <cuvs/distance/detail/pairwise_matrix/dispatch-inl.cuh>  // dispatch
-#include <cuvs/distance/detail/pairwise_matrix/dispatch_sm60.cuh>
-#include <raft/core/operators.hpp>  // raft::identity_op
-#define instantiate_raft_distance_detail_pairwise_matrix_dispatch(                     \
-  OpT, DataT, AccT, OutT, FinOpT, IdxT)                                                \
-  template void cuvs::distance::detail::                                               \
-    pairwise_matrix_dispatch<OpT<DataT, AccT, IdxT>, DataT, AccT, OutT, FinOpT, IdxT>( \
-      OpT<DataT, AccT, IdxT> distance_op,                                              \
-      IdxT m,                                                                          \
-      IdxT n,                                                                          \
-      IdxT k,                                                                          \
-      const DataT* x,                                                                  \
-      const DataT* y,                                                                  \
-      const DataT* x_norm,                                                             \
-      const DataT* y_norm,                                                             \
-      OutT* out,                                                                       \
-      FinOpT fin_op,                                                                   \
-      cudaStream_t stream,                                                             \
-      bool is_row_major)
-
-instantiate_raft_distance_detail_pairwise_matrix_dispatch(
-  cuvs::distance::detail::ops::russel_rao_distance_op, float, float, float, raft::identity_op, int);
-
-#undef instantiate_raft_distance_detail_pairwise_matrix_dispatch
diff --git a/cpp/src/distance/distance.cu b/cpp/src/distance/distance.cu
deleted file mode 100644
index c3af91fe1..000000000
--- a/cpp/src/distance/distance.cu
+++ /dev/null
@@ -1,934 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/distance/detail/kernels/rbf_fin_op.cuh>  // rbf_fin_op
-#include <cuvs/distance/distance-inl.cuh>
-
-/*
- * Hierarchy of instantiations:
- *
- * This file defines the template instantiations for the public API of
- * cuvs::distance. To improve compile times, the compilation of the distance
- * kernels is handled in distance/detail/pairwise_matrix/dispatch_*.cu.
- *
- */
-
-#define instantiate_raft_distance_distance(DT, DataT, AccT, OutT, FinalLambda, IdxT) \
-  template void cuvs::distance::distance<DT, DataT, AccT, OutT, FinalLambda, IdxT>(  \
-    raft::resources const& handle,                                                   \
-    const DataT* x,                                                                  \
-    const DataT* y,                                                                  \
-    OutT* dist,                                                                      \
-    IdxT m,                                                                          \
-    IdxT n,                                                                          \
-    IdxT k,                                                                          \
-    void* workspace,                                                                 \
-    size_t worksize,                                                                 \
-    FinalLambda fin_op,                                                              \
-    bool isRowMajor,                                                                 \
-    DataT metric_arg)
-
-// The following two instances are used in test/distance/gram.cu. Note the use
-// of int64_t for the index type.
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2Unexpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   cuvs::distance::kernels::detail::rbf_fin_op<float>,
-                                   int64_t);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2Unexpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   cuvs::distance::kernels::detail::rbf_fin_op<double>,
-                                   int64_t);
-
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Canberra, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Canberra, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::CorrelationExpanded, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::CorrelationExpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::identity_op,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::CosineExpanded, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::CosineExpanded, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HammingUnexpanded, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HammingUnexpanded, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HellingerExpanded, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HellingerExpanded, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::InnerProduct, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::InnerProduct, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::JensenShannon, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::JensenShannon, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::KLDivergence, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::KLDivergence, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L1, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L1, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Expanded, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Expanded, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtExpanded, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtExpanded, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtUnexpanded, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtUnexpanded, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Unexpanded, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Unexpanded, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Linf, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Linf, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::LpUnexpanded, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::LpUnexpanded, double, double, double, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::RusselRaoExpanded, float, float, float, raft::identity_op, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::RusselRaoExpanded, double, double, double, raft::identity_op, int);
-
-#undef instantiate_raft_distance_distance
-
-// Same, but without raft::identity_op
-#define instantiate_raft_distance_distance(DT, DataT, AccT, OutT, IdxT) \
-  template void cuvs::distance::distance<DT, DataT, AccT, OutT, IdxT>(  \
-    raft::resources const& handle,                                      \
-    const DataT* x,                                                     \
-    const DataT* y,                                                     \
-    OutT* dist,                                                         \
-    IdxT m,                                                             \
-    IdxT n,                                                             \
-    IdxT k,                                                             \
-    void* workspace,                                                    \
-    size_t worksize,                                                    \
-    bool isRowMajor,                                                    \
-    DataT metric_arg)
-
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Canberra, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Canberra, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::CorrelationExpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::CorrelationExpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::CosineExpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::CosineExpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HammingUnexpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HammingUnexpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HellingerExpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HellingerExpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::InnerProduct, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::InnerProduct, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::JensenShannon, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::JensenShannon, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::KLDivergence, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::KLDivergence, double, double, double, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L1, float, float, float, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L1, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Expanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Expanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtExpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtExpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtUnexpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtUnexpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Unexpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Unexpanded, double, double, double, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::Linf, float, float, float, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::Linf, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::LpUnexpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::LpUnexpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::RusselRaoExpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::RusselRaoExpanded, double, double, double, int);
-
-#undef instantiate_raft_distance_distance
-
-// Same, but without workspace
-#define instantiate_raft_distance_distance(DT, DataT, AccT, OutT, IdxT) \
-  template void cuvs::distance::distance<DT, DataT, AccT, OutT, IdxT>(  \
-    raft::resources const& handle,                                      \
-    const DataT* x,                                                     \
-    const DataT* y,                                                     \
-    OutT* dist,                                                         \
-    IdxT m,                                                             \
-    IdxT n,                                                             \
-    IdxT k,                                                             \
-    bool isRowMajor,                                                    \
-    DataT metric_arg)
-
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Canberra, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Canberra, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::CorrelationExpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::CorrelationExpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::CosineExpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::CosineExpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HammingUnexpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HammingUnexpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HellingerExpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::HellingerExpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::InnerProduct, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::InnerProduct, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::JensenShannon, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::JensenShannon, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::KLDivergence, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::KLDivergence, double, double, double, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L1, float, float, float, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L1, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Expanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Expanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtExpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtExpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtUnexpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2SqrtUnexpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Unexpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Unexpanded, double, double, double, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::Linf, float, float, float, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::Linf, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::LpUnexpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::LpUnexpanded, double, double, double, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::RusselRaoExpanded, float, float, float, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::RusselRaoExpanded, double, double, double, int);
-
-#undef instantiate_raft_distance_distance
-
-#define instantiate_raft_distance_getWorkspaceSize(DistT, DataT, AccT, OutT, IdxT)  \
-  template size_t cuvs::distance::getWorkspaceSize<DistT, DataT, AccT, OutT, IdxT>( \
-    const DataT* x, const DataT* y, IdxT m, IdxT n, IdxT k)
-
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::Canberra, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::Canberra, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::CorrelationExpanded, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::CorrelationExpanded, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::CosineExpanded, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::CosineExpanded, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::HammingUnexpanded, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::HammingUnexpanded, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::HellingerExpanded, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::HellingerExpanded, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::InnerProduct, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::InnerProduct, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::JensenShannon, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::JensenShannon, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::KLDivergence, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::KLDivergence, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L1, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L1, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2SqrtExpanded, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2SqrtExpanded, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2SqrtUnexpanded, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2SqrtUnexpanded, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Unexpanded, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Unexpanded, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::Linf, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::Linf, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::LpUnexpanded, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::LpUnexpanded, double, double, double, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::RusselRaoExpanded, float, float, float, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::RusselRaoExpanded, double, double, double, int);
-
-#undef instantiate_raft_distance_getWorkspaceSize
-
-#define instantiate_raft_distance_getWorkspaceSize(DistT, DataT, AccT, OutT, IdxT, layout)  \
-  template size_t cuvs::distance::getWorkspaceSize<DistT, DataT, AccT, OutT, IdxT, layout>( \
-    raft::device_matrix_view<DataT, IdxT, layout> const& x,                                 \
-    raft::device_matrix_view<DataT, IdxT, layout> const& y)
-
-// We could consider not taking template parameters for this function. The
-// number of instantiations seems a bit excessive..
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::Canberra, float, float, float, int, raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::Canberra, double, double, double, int, raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::Canberra, float, float, float, int, raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::Canberra, double, double, double, int, raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::CorrelationExpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::CorrelationExpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::CorrelationExpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::CorrelationExpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::CosineExpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::CosineExpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::CosineExpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::CosineExpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::HammingUnexpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::HammingUnexpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::HammingUnexpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::HammingUnexpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::HellingerExpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::HellingerExpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::HellingerExpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::HellingerExpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::InnerProduct, float, float, float, int, raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::InnerProduct,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::InnerProduct, float, float, float, int, raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::InnerProduct,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::JensenShannon, float, float, float, int, raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::JensenShannon,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::JensenShannon, float, float, float, int, raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::JensenShannon,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::KLDivergence, float, float, float, int, raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::KLDivergence,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::KLDivergence, float, float, float, int, raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::KLDivergence,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L1, float, float, float, int, raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L1, double, double, double, int, raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L1, float, float, float, int, raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L1, double, double, double, int, raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, float, float, float, int, raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, double, double, double, int, raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, float, float, float, int, raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, double, double, double, int, raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::L2SqrtExpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::L2SqrtExpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::L2SqrtExpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::L2SqrtExpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::L2SqrtUnexpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::L2SqrtUnexpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::L2SqrtUnexpanded,
-                                           float,
-                                           float,
-                                           float,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::L2SqrtUnexpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_f_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Unexpanded, float, float, float, int, raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(cuvs::distance::DistanceType::L2Unexpanded,
-                                           double,
-                                           double,
-                                           double,
-                                           int,
-                                           raft::layout_c_contiguous);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Unexpanded, float, float, float, int, raft::layout_f_contiguous);
-
-#undef instantiate_raft_distance_getWorkspaceSize
-
-#define instantiate_raft_distance_pairwise_distance(DataT, IdxT)                        \
-  template void cuvs::distance::pairwise_distance(raft::resources const& handle,        \
-                                                  const DataT* x,                       \
-                                                  const DataT* y,                       \
-                                                  DataT* dist,                          \
-                                                  IdxT m,                               \
-                                                  IdxT n,                               \
-                                                  IdxT k,                               \
-                                                  rmm::device_uvector<char>& workspace, \
-                                                  cuvs::distance::DistanceType metric,  \
-                                                  bool isRowMajor,                      \
-                                                  DataT metric_arg)
-
-instantiate_raft_distance_pairwise_distance(float, int);
-instantiate_raft_distance_pairwise_distance(double, int);
-
-#undef instantiate_raft_distance_pairwise_distance
-
-// Same, but without workspace
-#define instantiate_raft_distance_pairwise_distance(DataT, IdxT)                       \
-  template void cuvs::distance::pairwise_distance(raft::resources const& handle,       \
-                                                  const DataT* x,                      \
-                                                  const DataT* y,                      \
-                                                  DataT* dist,                         \
-                                                  IdxT m,                              \
-                                                  IdxT n,                              \
-                                                  IdxT k,                              \
-                                                  cuvs::distance::DistanceType metric, \
-                                                  bool isRowMajor,                     \
-                                                  DataT metric_arg)
-
-instantiate_raft_distance_pairwise_distance(float, int);
-instantiate_raft_distance_pairwise_distance(double, int);
-
-#undef instantiate_raft_distance_pairwise_distance
-
-// Version with mdspan
-#define instantiate_raft_distance_distance(DistT, DataT, AccT, OutT, layout, IdxT) \
-  template void cuvs::distance::distance<DistT, DataT, AccT, OutT, layout, IdxT>(  \
-    raft::resources const& handle,                                                 \
-    raft::device_matrix_view<DataT, IdxT, layout> const x,                         \
-    raft::device_matrix_view<DataT, IdxT, layout> const y,                         \
-    raft::device_matrix_view<OutT, IdxT, layout> dist,                             \
-    DataT metric_arg)
-
-// Again, we might want to consider reigning in the number of instantiations...
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Canberra, float, float, float, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Canberra, double, double, double, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Canberra, float, float, float, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Canberra, double, double, double, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::CorrelationExpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::CorrelationExpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::CorrelationExpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::CorrelationExpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::CosineExpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::CosineExpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::CosineExpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::CosineExpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::HammingUnexpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::HammingUnexpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::HammingUnexpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::HammingUnexpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::HellingerExpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::HellingerExpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::HellingerExpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::HellingerExpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::InnerProduct, float, float, float, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::InnerProduct,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::InnerProduct, float, float, float, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::InnerProduct,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::JensenShannon, float, float, float, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::JensenShannon,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::JensenShannon, float, float, float, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::JensenShannon,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::KLDivergence, float, float, float, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::KLDivergence,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::KLDivergence, float, float, float, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::KLDivergence,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L1, float, float, float, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L1, double, double, double, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L1, float, float, float, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L1, double, double, double, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Expanded, float, float, float, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Expanded, double, double, double, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Expanded, float, float, float, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Expanded, double, double, double, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2SqrtExpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2SqrtExpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2SqrtExpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2SqrtExpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2SqrtUnexpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2SqrtUnexpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2SqrtUnexpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2SqrtUnexpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Unexpanded, float, float, float, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2Unexpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::L2Unexpanded, float, float, float, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2Unexpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Linf, float, float, float, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Linf, double, double, double, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Linf, float, float, float, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::Linf, double, double, double, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::LpUnexpanded, float, float, float, raft::layout_c_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::LpUnexpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(
-  cuvs::distance::DistanceType::LpUnexpanded, float, float, float, raft::layout_f_contiguous, int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::LpUnexpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::RusselRaoExpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::RusselRaoExpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_c_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::RusselRaoExpanded,
-                                   float,
-                                   float,
-                                   float,
-                                   raft::layout_f_contiguous,
-                                   int);
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::RusselRaoExpanded,
-                                   double,
-                                   double,
-                                   double,
-                                   raft::layout_f_contiguous,
-                                   int);
-
-#undef instantiate_raft_distance_distance
-
-#define instantiate_raft_distance_pairwise_distance(DataT, layout, IdxT) \
-  template void cuvs::distance::pairwise_distance(                       \
-    raft::resources const& handle,                                       \
-    raft::device_matrix_view<DataT, IdxT, layout> const x,               \
-    raft::device_matrix_view<DataT, IdxT, layout> const y,               \
-    raft::device_matrix_view<DataT, IdxT, layout> dist,                  \
-    cuvs::distance::DistanceType metric,                                 \
-    DataT metric_arg)
-
-instantiate_raft_distance_pairwise_distance(float, raft::layout_c_contiguous, int);
-instantiate_raft_distance_pairwise_distance(float, raft::layout_f_contiguous, int);
-instantiate_raft_distance_pairwise_distance(double, raft::layout_c_contiguous, int);
-instantiate_raft_distance_pairwise_distance(double, raft::layout_f_contiguous, int);
-
-#undef instantiate_raft_distance_pairwise_distance
diff --git a/cpp/src/distance/fused_l2_nn.cu b/cpp/src/distance/fused_l2_nn.cu
deleted file mode 100644
index 251a954d3..000000000
--- a/cpp/src/distance/fused_l2_nn.cu
+++ /dev/null
@@ -1,54 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/distance/fused_l2_nn-inl.cuh>
-#include <raft/core/kvp.hpp>  // raft::KeyValuePair
-
-#define instantiate_raft_distance_fusedL2NNMinReduce(DataT, OutT, IdxT)                   \
-  template void cuvs::distance::fusedL2NNMinReduce<DataT, OutT, IdxT>(OutT * min,         \
-                                                                      const DataT* x,     \
-                                                                      const DataT* y,     \
-                                                                      const DataT* xn,    \
-                                                                      const DataT* yn,    \
-                                                                      IdxT m,             \
-                                                                      IdxT n,             \
-                                                                      IdxT k,             \
-                                                                      void* workspace,    \
-                                                                      bool sqrt,          \
-                                                                      bool initOutBuffer, \
-                                                                      cudaStream_t stream)
-
-instantiate_raft_distance_fusedL2NNMinReduce(double, double, int);
-instantiate_raft_distance_fusedL2NNMinReduce(double, double, int64_t);
-instantiate_raft_distance_fusedL2NNMinReduce(float, float, int);
-instantiate_raft_distance_fusedL2NNMinReduce(float, float, int64_t);
-
-// We can't have comma's in the macro expansion, so we use the COMMA macro:
-#define COMMA ,
-
-instantiate_raft_distance_fusedL2NNMinReduce(double, raft::KeyValuePair<int COMMA double>, int);
-instantiate_raft_distance_fusedL2NNMinReduce(double,
-                                             raft::KeyValuePair<int64_t COMMA double>,
-                                             int64_t);
-instantiate_raft_distance_fusedL2NNMinReduce(float, raft::KeyValuePair<int COMMA float>, int);
-instantiate_raft_distance_fusedL2NNMinReduce(float,
-                                             raft::KeyValuePair<int64_t COMMA float>,
-                                             int64_t);
-
-#undef COMMA
-
-#undef instantiate_raft_distance_fusedL2NNMinReduce
diff --git a/cpp/src/matrix/detail/select_k_double_int64_t.cu b/cpp/src/matrix/detail/select_k_double_int64_t.cu
deleted file mode 100644
index c3949149f..000000000
--- a/cpp/src/matrix/detail/select_k_double_int64_t.cu
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <raft/matrix/detail/select_k-inl.cuh>
-
-#define instantiate_cuvs_matrix_detail_select_k(T, IdxT)                            \
-  template void raft::matrix::detail::select_k(raft::resources const& handle,       \
-                                               const T* in_val,                     \
-                                               const IdxT* in_idx,                  \
-                                               size_t batch_size,                   \
-                                               size_t len,                          \
-                                               int k,                               \
-                                               T* out_val,                          \
-                                               IdxT* out_idx,                       \
-                                               bool select_min,                     \
-                                               rmm::mr::device_memory_resource* mr, \
-                                               bool sorted)
-
-instantiate_cuvs_matrix_detail_select_k(double, int64_t);
-
-#undef instantiate_cuvs_matrix_detail_select_k
diff --git a/cpp/src/matrix/detail/select_k_double_uint32_t.cu b/cpp/src/matrix/detail/select_k_double_uint32_t.cu
deleted file mode 100644
index 171c8a1ae..000000000
--- a/cpp/src/matrix/detail/select_k_double_uint32_t.cu
+++ /dev/null
@@ -1,35 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cstdint>  // uint32_t
-#include <raft/matrix/detail/select_k-inl.cuh>
-
-#define instantiate_raft_matrix_detail_select_k(T, IdxT)                            \
-  template void raft::matrix::detail::select_k(raft::resources const& handle,       \
-                                               const T* in_val,                     \
-                                               const IdxT* in_idx,                  \
-                                               size_t batch_size,                   \
-                                               size_t len,                          \
-                                               int k,                               \
-                                               T* out_val,                          \
-                                               IdxT* out_idx,                       \
-                                               bool select_min,                     \
-                                               rmm::mr::device_memory_resource* mr, \
-                                               bool sorted)
-
-instantiate_raft_matrix_detail_select_k(double, uint32_t);
-
-#undef instantiate_raft_matrix_detail_select_k
diff --git a/cpp/src/matrix/detail/select_k_float_int32.cu b/cpp/src/matrix/detail/select_k_float_int32.cu
deleted file mode 100644
index a21444dc0..000000000
--- a/cpp/src/matrix/detail/select_k_float_int32.cu
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <raft/matrix/detail/select_k-inl.cuh>
-
-#define instantiate_raft_matrix_detail_select_k(T, IdxT)                            \
-  template void raft::matrix::detail::select_k(raft::resources const& handle,       \
-                                               const T* in_val,                     \
-                                               const IdxT* in_idx,                  \
-                                               size_t batch_size,                   \
-                                               size_t len,                          \
-                                               int k,                               \
-                                               T* out_val,                          \
-                                               IdxT* out_idx,                       \
-                                               bool select_min,                     \
-                                               rmm::mr::device_memory_resource* mr, \
-                                               bool sorted)
-
-instantiate_raft_matrix_detail_select_k(float, int);
-
-#undef instantiate_raft_matrix_detail_select_k
diff --git a/cpp/src/matrix/detail/select_k_float_int64_t.cu b/cpp/src/matrix/detail/select_k_float_int64_t.cu
deleted file mode 100644
index 9542874ec..000000000
--- a/cpp/src/matrix/detail/select_k_float_int64_t.cu
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <raft/matrix/detail/select_k-inl.cuh>
-
-#define instantiate_raft_matrix_detail_select_k(T, IdxT)                            \
-  template void raft::matrix::detail::select_k(raft::resources const& handle,       \
-                                               const T* in_val,                     \
-                                               const IdxT* in_idx,                  \
-                                               size_t batch_size,                   \
-                                               size_t len,                          \
-                                               int k,                               \
-                                               T* out_val,                          \
-                                               IdxT* out_idx,                       \
-                                               bool select_min,                     \
-                                               rmm::mr::device_memory_resource* mr, \
-                                               bool sorted)
-
-instantiate_raft_matrix_detail_select_k(float, int64_t);
-
-#undef instantiate_raft_matrix_detail_select_k
diff --git a/cpp/src/matrix/detail/select_k_float_uint32_t.cu b/cpp/src/matrix/detail/select_k_float_uint32_t.cu
deleted file mode 100644
index fbf311d9b..000000000
--- a/cpp/src/matrix/detail/select_k_float_uint32_t.cu
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <raft/matrix/detail/select_k-inl.cuh>
-
-#define instantiate_raft_matrix_detail_select_k(T, IdxT)                            \
-  template void raft::matrix::detail::select_k(raft::resources const& handle,       \
-                                               const T* in_val,                     \
-                                               const IdxT* in_idx,                  \
-                                               size_t batch_size,                   \
-                                               size_t len,                          \
-                                               int k,                               \
-                                               T* out_val,                          \
-                                               IdxT* out_idx,                       \
-                                               bool select_min,                     \
-                                               rmm::mr::device_memory_resource* mr, \
-                                               bool sorted)
-
-instantiate_raft_matrix_detail_select_k(float, uint32_t);
-
-#undef instantiate_raft_matrix_detail_select_k
diff --git a/cpp/src/matrix/detail/select_k_half_int64_t.cu b/cpp/src/matrix/detail/select_k_half_int64_t.cu
deleted file mode 100644
index fdbfd66c4..000000000
--- a/cpp/src/matrix/detail/select_k_half_int64_t.cu
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <raft/matrix/detail/select_k-inl.cuh>
-
-#define instantiate_raft_matrix_detail_select_k(T, IdxT)                            \
-  template void raft::matrix::detail::select_k(raft::resources const& handle,       \
-                                               const T* in_val,                     \
-                                               const IdxT* in_idx,                  \
-                                               size_t batch_size,                   \
-                                               size_t len,                          \
-                                               int k,                               \
-                                               T* out_val,                          \
-                                               IdxT* out_idx,                       \
-                                               bool select_min,                     \
-                                               rmm::mr::device_memory_resource* mr, \
-                                               bool sorted)
-
-instantiate_raft_matrix_detail_select_k(__half, int64_t);
-
-#undef instantiate_raft_matrix_detail_select_k
diff --git a/cpp/src/matrix/detail/select_k_half_uint32_t.cu b/cpp/src/matrix/detail/select_k_half_uint32_t.cu
deleted file mode 100644
index 48a3e91f9..000000000
--- a/cpp/src/matrix/detail/select_k_half_uint32_t.cu
+++ /dev/null
@@ -1,34 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <raft/matrix/detail/select_k-inl.cuh>
-
-#define instantiate_raft_matrix_detail_select_k(T, IdxT)                            \
-  template void raft::matrix::detail::select_k(raft::resources const& handle,       \
-                                               const T* in_val,                     \
-                                               const IdxT* in_idx,                  \
-                                               size_t batch_size,                   \
-                                               size_t len,                          \
-                                               int k,                               \
-                                               T* out_val,                          \
-                                               IdxT* out_idx,                       \
-                                               bool select_min,                     \
-                                               rmm::mr::device_memory_resource* mr, \
-                                               bool sorted)
-
-instantiate_raft_matrix_detail_select_k(__half, uint32_t);
-
-#undef instantiate_raft_matrix_detail_select_k
diff --git a/cpp/src/neighbors/ball_cover.cu b/cpp/src/neighbors/ball_cover.cu
deleted file mode 100644
index c9a1e9763..000000000
--- a/cpp/src/neighbors/ball_cover.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cstdint>
-#include <cuvs/neighbors/ball_cover-inl.cuh>
-
-#define instantiate_raft_neighbors_ball_cover(idx_t, value_t, int_t, matrix_idx_t)                 \
-  template void cuvs::neighbors::ball_cover::build_index<idx_t, value_t, int_t, matrix_idx_t>(     \
-    raft::resources const& handle,                                                                 \
-    cuvs::neighbors::ball_cover::BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index);      \
-                                                                                                   \
-  template void cuvs::neighbors::ball_cover::all_knn_query<idx_t, value_t, int_t, matrix_idx_t>(   \
-    raft::resources const& handle,                                                                 \
-    cuvs::neighbors::ball_cover::BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index,       \
-    int_t k,                                                                                       \
-    idx_t* inds,                                                                                   \
-    value_t* dists,                                                                                \
-    bool perform_post_filtering,                                                                   \
-    float weight);                                                                                 \
-                                                                                                   \
-  template void cuvs::neighbors::ball_cover::all_knn_query<idx_t, value_t, int_t, matrix_idx_t>(   \
-    raft::resources const& handle,                                                                 \
-    cuvs::neighbors::ball_cover::BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index,       \
-    raft::device_matrix_view<idx_t, matrix_idx_t, raft::row_major> inds,                           \
-    raft::device_matrix_view<value_t, matrix_idx_t, raft::row_major> dists,                        \
-    int_t k,                                                                                       \
-    bool perform_post_filtering,                                                                   \
-    float weight);                                                                                 \
-                                                                                                   \
-  template void cuvs::neighbors::ball_cover::knn_query<idx_t, value_t, int_t>(                     \
-    raft::resources const& handle,                                                                 \
-    const cuvs::neighbors::ball_cover::BallCoverIndex<idx_t, value_t, int_t>& index,               \
-    int_t k,                                                                                       \
-    const value_t* query,                                                                          \
-    int_t n_query_pts,                                                                             \
-    idx_t* inds,                                                                                   \
-    value_t* dists,                                                                                \
-    bool perform_post_filtering,                                                                   \
-    float weight);                                                                                 \
-                                                                                                   \
-  template void cuvs::neighbors::ball_cover::knn_query<idx_t, value_t, int_t, matrix_idx_t>(       \
-    raft::resources const& handle,                                                                 \
-    const cuvs::neighbors::ball_cover::BallCoverIndex<idx_t, value_t, int_t, matrix_idx_t>& index, \
-    raft::device_matrix_view<const value_t, matrix_idx_t, raft::row_major> query,                  \
-    raft::device_matrix_view<idx_t, matrix_idx_t, raft::row_major> inds,                           \
-    raft::device_matrix_view<value_t, matrix_idx_t, raft::row_major> dists,                        \
-    int_t k,                                                                                       \
-    bool perform_post_filtering,                                                                   \
-    float weight);
-
-instantiate_raft_neighbors_ball_cover(int64_t, float, uint32_t, uint32_t);
-
-#undef instantiate_raft_neighbors_ball_cover
diff --git a/cpp/src/neighbors/brute_force_00_generate.py b/cpp/src/neighbors/brute_force_00_generate.py
deleted file mode 100644
index 21f984502..000000000
--- a/cpp/src/neighbors/brute_force_00_generate.py
+++ /dev/null
@@ -1,106 +0,0 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-header = """
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by brute_force_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python brute_force_00_generate.py
- *
- */
-
-#include <cstdint>
-#include <cuvs/neighbors/brute_force-inl.cuh>
-
-"""
-
-knn_macro = """
-#define instantiate_raft_neighbors_brute_force_knn(idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op) \\
-    template void cuvs::neighbors::brute_force::knn<idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op>( \\
-        raft::resources const& handle,                           \\
-        std::vector<raft::device_matrix_view<const value_t, matrix_idx, index_layout>> index, \\
-        raft::device_matrix_view<const value_t, matrix_idx, search_layout> search, \\
-        raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices, \\
-        raft::device_matrix_view<value_t, matrix_idx, raft::row_major> distances, \\
-        cuvs::distance::DistanceType metric,                            \\
-        std::optional<float> metric_arg,                                \\
-        std::optional<idx_t> global_id_offset,                          \\
-        epilogue_op distance_epilogue);
-
-"""
-
-fused_l2_knn_macro = """
-#define instantiate_raft_neighbors_brute_force_fused_l2_knn(value_t, idx_t, idx_layout, query_layout) \\
-    template void cuvs::neighbors::brute_force::fused_l2_knn(    \\
-        raft::resources const& handle,                           \\
-        raft::device_matrix_view<const value_t, idx_t, idx_layout> index, \\
-        raft::device_matrix_view<const value_t, idx_t, query_layout> query, \\
-        raft::device_matrix_view<idx_t, idx_t, raft::row_major> out_inds,     \\
-        raft::device_matrix_view<value_t, idx_t, raft::row_major> out_dists,  \\
-        cuvs::distance::DistanceType metric);
-
-"""
-
-knn_types = dict(
-    int64_t_float_uint32_t=("int64_t","float","uint32_t"),
-    int64_t_float_int64_t=("int64_t","float","int64_t"),
-    int_float_int=("int","float","int"),
-    uint32_t_float_uint32_t=("uint32_t","float","uint32_t"),
-)
-
-fused_l2_knn_types = dict(
-    float_int64_t=("float", "int64_t"),
-)
-
-# knn
-for type_path, (idx_t, value_t, matrix_idx) in knn_types.items():
-    path = f"brute_force_knn_{type_path}.cu"
-    with open(path, "w") as f:
-        f.write(header)
-        f.write(knn_macro)
-        f.write(f"instantiate_raft_neighbors_brute_force_knn({idx_t},{value_t},{matrix_idx},raft::row_major,raft::row_major,raft::identity_op);\n\n")
-        f.write("#undef instantiate_raft_neighbors_brute_force_knn\n")
-
-    # For pasting into CMakeLists.txt
-    print(f"src/neighbors/{path}")
-
-#fused_l2_knn
-for type_path, (value_t, idx_t) in fused_l2_knn_types.items():
-    path = f"brute_force_fused_l2_knn_{type_path}.cu"
-    with open(path, "w") as f:
-        f.write(header)
-        f.write(fused_l2_knn_macro)
-        f.write(f"instantiate_raft_neighbors_brute_force_fused_l2_knn({value_t},{idx_t},raft::row_major,raft::row_major);\n\n")
-        f.write("#undef instantiate_raft_neighbors_brute_force_fused_l2_knn\n")
-
-    # For pasting into CMakeLists.txt
-    print(f"src/neighbors/{path}")
diff --git a/cpp/src/neighbors/brute_force_fused_l2_knn_float_int64_t.cu b/cpp/src/neighbors/brute_force_fused_l2_knn_float_int64_t.cu
deleted file mode 100644
index 5e61f59df..000000000
--- a/cpp/src/neighbors/brute_force_fused_l2_knn_float_int64_t.cu
+++ /dev/null
@@ -1,45 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by brute_force_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python brute_force_00_generate.py
- *
- */
-
-#include <cstdint>
-#include <cuvs/neighbors/brute_force-inl.cuh>
-
-#define instantiate_raft_neighbors_brute_force_fused_l2_knn(             \
-  value_t, idx_t, idx_layout, query_layout)                              \
-  template void cuvs::neighbors::brute_force::fused_l2_knn(              \
-    raft::resources const& handle,                                       \
-    raft::device_matrix_view<const value_t, idx_t, idx_layout> index,    \
-    raft::device_matrix_view<const value_t, idx_t, query_layout> query,  \
-    raft::device_matrix_view<idx_t, idx_t, raft::row_major> out_inds,    \
-    raft::device_matrix_view<value_t, idx_t, raft::row_major> out_dists, \
-    cuvs::distance::DistanceType metric);
-
-instantiate_raft_neighbors_brute_force_fused_l2_knn(float,
-                                                    int64_t,
-                                                    raft::row_major,
-                                                    raft::row_major);
-
-#undef instantiate_raft_neighbors_brute_force_fused_l2_knn
diff --git a/cpp/src/neighbors/brute_force_knn_index_float.cu b/cpp/src/neighbors/brute_force_knn_index_float.cu
deleted file mode 100644
index 1b98a37d0..000000000
--- a/cpp/src/neighbors/brute_force_knn_index_float.cu
+++ /dev/null
@@ -1,39 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cstdint>
-#include <cuvs/neighbors/brute_force-inl.cuh>
-
-template void cuvs::neighbors::brute_force::search<float, int>(
-  raft::resources const& res,
-  const cuvs::neighbors::brute_force::index<float>& idx,
-  raft::device_matrix_view<const float, int64_t, raft::row_major> queries,
-  raft::device_matrix_view<int, int64_t, raft::row_major> neighbors,
-  raft::device_matrix_view<float, int64_t, raft::row_major> distances);
-
-template void cuvs::neighbors::brute_force::search<float, int64_t>(
-  raft::resources const& res,
-  const cuvs::neighbors::brute_force::index<float>& idx,
-  raft::device_matrix_view<const float, int64_t, raft::row_major> queries,
-  raft::device_matrix_view<int64_t, int64_t, raft::row_major> neighbors,
-  raft::device_matrix_view<float, int64_t, raft::row_major> distances);
-
-template cuvs::neighbors::brute_force::index<float> cuvs::neighbors::brute_force::build<float>(
-  raft::resources const& res,
-  raft::device_matrix_view<const float, int64_t, raft::row_major> dataset,
-  cuvs::distance::DistanceType metric,
-  float metric_arg);
diff --git a/cpp/src/neighbors/brute_force_knn_int64_t_float_int64_t.cu b/cpp/src/neighbors/brute_force_knn_int64_t_float_int64_t.cu
deleted file mode 100644
index fbcabc642..000000000
--- a/cpp/src/neighbors/brute_force_knn_int64_t_float_int64_t.cu
+++ /dev/null
@@ -1,47 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by brute_force_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python brute_force_00_generate.py
- *
- */
-
-#include <cstdint>
-#include <cuvs/neighbors/brute_force-inl.cuh>
-
-#define instantiate_raft_neighbors_brute_force_knn(                                         \
-  idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op)                     \
-  template void cuvs::neighbors::brute_force::                                              \
-    knn<idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op>(              \
-      raft::resources const& handle,                                                        \
-      std::vector<raft::device_matrix_view<const value_t, matrix_idx, index_layout>> index, \
-      raft::device_matrix_view<const value_t, matrix_idx, search_layout> search,            \
-      raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                 \
-      raft::device_matrix_view<value_t, matrix_idx, raft::row_major> distances,             \
-      cuvs::distance::DistanceType metric,                                                  \
-      std::optional<float> metric_arg,                                                      \
-      std::optional<idx_t> global_id_offset,                                                \
-      epilogue_op distance_epilogue);
-
-instantiate_raft_neighbors_brute_force_knn(
-  int64_t, float, int64_t, raft::row_major, raft::row_major, raft::identity_op);
-
-#undef instantiate_raft_neighbors_brute_force_knn
diff --git a/cpp/src/neighbors/brute_force_knn_int64_t_float_uint32_t.cu b/cpp/src/neighbors/brute_force_knn_int64_t_float_uint32_t.cu
deleted file mode 100644
index f9ddf2ddd..000000000
--- a/cpp/src/neighbors/brute_force_knn_int64_t_float_uint32_t.cu
+++ /dev/null
@@ -1,47 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by brute_force_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python brute_force_00_generate.py
- *
- */
-
-#include <cstdint>
-#include <cuvs/neighbors/brute_force-inl.cuh>
-
-#define instantiate_raft_neighbors_brute_force_knn(                                         \
-  idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op)                     \
-  template void cuvs::neighbors::brute_force::                                              \
-    knn<idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op>(              \
-      raft::resources const& handle,                                                        \
-      std::vector<raft::device_matrix_view<const value_t, matrix_idx, index_layout>> index, \
-      raft::device_matrix_view<const value_t, matrix_idx, search_layout> search,            \
-      raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                 \
-      raft::device_matrix_view<value_t, matrix_idx, raft::row_major> distances,             \
-      cuvs::distance::DistanceType metric,                                                  \
-      std::optional<float> metric_arg,                                                      \
-      std::optional<idx_t> global_id_offset,                                                \
-      epilogue_op distance_epilogue);
-
-instantiate_raft_neighbors_brute_force_knn(
-  int64_t, float, uint32_t, raft::row_major, raft::row_major, raft::identity_op);
-
-#undef instantiate_raft_neighbors_brute_force_knn
diff --git a/cpp/src/neighbors/brute_force_knn_int_float_int.cu b/cpp/src/neighbors/brute_force_knn_int_float_int.cu
deleted file mode 100644
index bf24dd9df..000000000
--- a/cpp/src/neighbors/brute_force_knn_int_float_int.cu
+++ /dev/null
@@ -1,47 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by brute_force_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python brute_force_00_generate.py
- *
- */
-
-#include <cstdint>
-#include <cuvs/neighbors/brute_force-inl.cuh>
-
-#define instantiate_raft_neighbors_brute_force_knn(                                         \
-  idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op)                     \
-  template void cuvs::neighbors::brute_force::                                              \
-    knn<idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op>(              \
-      raft::resources const& handle,                                                        \
-      std::vector<raft::device_matrix_view<const value_t, matrix_idx, index_layout>> index, \
-      raft::device_matrix_view<const value_t, matrix_idx, search_layout> search,            \
-      raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                 \
-      raft::device_matrix_view<value_t, matrix_idx, raft::row_major> distances,             \
-      cuvs::distance::DistanceType metric,                                                  \
-      std::optional<float> metric_arg,                                                      \
-      std::optional<idx_t> global_id_offset,                                                \
-      epilogue_op distance_epilogue);
-
-instantiate_raft_neighbors_brute_force_knn(
-  int, float, int, raft::row_major, raft::row_major, raft::identity_op);
-
-#undef instantiate_raft_neighbors_brute_force_knn
diff --git a/cpp/src/neighbors/brute_force_knn_uint32_t_float_uint32_t.cu b/cpp/src/neighbors/brute_force_knn_uint32_t_float_uint32_t.cu
deleted file mode 100644
index 2514352a0..000000000
--- a/cpp/src/neighbors/brute_force_knn_uint32_t_float_uint32_t.cu
+++ /dev/null
@@ -1,47 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by brute_force_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python brute_force_00_generate.py
- *
- */
-
-#include <cstdint>
-#include <cuvs/neighbors/brute_force-inl.cuh>
-
-#define instantiate_raft_neighbors_brute_force_knn(                                         \
-  idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op)                     \
-  template void cuvs::neighbors::brute_force::                                              \
-    knn<idx_t, value_t, matrix_idx, index_layout, search_layout, epilogue_op>(              \
-      raft::resources const& handle,                                                        \
-      std::vector<raft::device_matrix_view<const value_t, matrix_idx, index_layout>> index, \
-      raft::device_matrix_view<const value_t, matrix_idx, search_layout> search,            \
-      raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                 \
-      raft::device_matrix_view<value_t, matrix_idx, raft::row_major> distances,             \
-      cuvs::distance::DistanceType metric,                                                  \
-      std::optional<float> metric_arg,                                                      \
-      std::optional<idx_t> global_id_offset,                                                \
-      epilogue_op distance_epilogue);
-
-instantiate_raft_neighbors_brute_force_knn(
-  uint32_t, float, uint32_t, raft::row_major, raft::row_major, raft::identity_op);
-
-#undef instantiate_raft_neighbors_brute_force_knn
diff --git a/cpp/src/neighbors/cagra_build_float.cpp b/cpp/src/neighbors/cagra_build_float.cpp
index 426a811f3..f66ae5cfb 100644
--- a/cpp/src/neighbors/cagra_build_float.cpp
+++ b/cpp/src/neighbors/cagra_build_float.cpp
@@ -19,41 +19,39 @@
 
 namespace cuvs::neighbors::cagra {
 
-#define CUVS_INST_CAGRA_BUILD(T, IdxT)                                                   \
-  auto build(raft::resources const& handle,                                              \
-             const cuvs::neighbors::cagra::index_params& params,                         \
-             raft::device_matrix_view<const T, int64_t, raft::row_major> dataset)        \
-    ->cuvs::neighbors::cagra::index<T, IdxT>                                             \
-  {                                                                                      \
-    return cuvs::neighbors::cagra::index<T, IdxT>(                                       \
-      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));  \
-  }                                                                                      \
-                                                                                         \
-  auto build(raft::resources const& handle,                                              \
-             const cuvs::neighbors::cagra::index_params& params,                         \
-             raft::host_matrix_view<const T, int64_t, raft::row_major> dataset)          \
-    ->cuvs::neighbors::cagra::index<T, IdxT>                                             \
-  {                                                                                      \
-    return cuvs::neighbors::cagra::index<T, IdxT>(                                       \
-      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));  \
-  }                                                                                      \
-                                                                                         \
-  void build_device(raft::resources const& handle,                                       \
-                    const cuvs::neighbors::cagra::index_params& params,                  \
-                    raft::device_matrix_view<const T, int64_t, raft::row_major> dataset, \
-                    cuvs::neighbors::cagra::index<T, IdxT>& idx)                         \
-  {                                                                                      \
-    raft::runtime::neighbors::cagra::build_device(                                       \
-      handle, params, dataset, *idx.get_raft_index());                              \
-  }                                                                                      \
-                                                                                         \
-  void build_host(raft::resources const& handle,                                         \
-                  const cuvs::neighbors::cagra::index_params& params,                    \
-                  raft::host_matrix_view<const T, int64_t, raft::row_major> dataset,     \
-                  cuvs::neighbors::cagra::index<T, IdxT>& idx)                           \
-  {                                                                                      \
-    raft::runtime::neighbors::cagra::build_host(                                         \
-      handle, params, dataset, *idx.get_raft_index());                              \
+#define CUVS_INST_CAGRA_BUILD(T, IdxT)                                                             \
+  auto build(raft::resources const& handle,                                                        \
+             const cuvs::neighbors::cagra::index_params& params,                                   \
+             raft::device_matrix_view<const T, int64_t, raft::row_major> dataset)                  \
+    ->cuvs::neighbors::cagra::index<T, IdxT>                                                       \
+  {                                                                                                \
+    return cuvs::neighbors::cagra::index<T, IdxT>(                                                 \
+      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));                 \
+  }                                                                                                \
+                                                                                                   \
+  auto build(raft::resources const& handle,                                                        \
+             const cuvs::neighbors::cagra::index_params& params,                                   \
+             raft::host_matrix_view<const T, int64_t, raft::row_major> dataset)                    \
+    ->cuvs::neighbors::cagra::index<T, IdxT>                                                       \
+  {                                                                                                \
+    return cuvs::neighbors::cagra::index<T, IdxT>(                                                 \
+      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));                 \
+  }                                                                                                \
+                                                                                                   \
+  void build_device(raft::resources const& handle,                                                 \
+                    const cuvs::neighbors::cagra::index_params& params,                            \
+                    raft::device_matrix_view<const T, int64_t, raft::row_major> dataset,           \
+                    cuvs::neighbors::cagra::index<T, IdxT>& idx)                                   \
+  {                                                                                                \
+    raft::runtime::neighbors::cagra::build_device(handle, params, dataset, *idx.get_raft_index()); \
+  }                                                                                                \
+                                                                                                   \
+  void build_host(raft::resources const& handle,                                                   \
+                  const cuvs::neighbors::cagra::index_params& params,                              \
+                  raft::host_matrix_view<const T, int64_t, raft::row_major> dataset,               \
+                  cuvs::neighbors::cagra::index<T, IdxT>& idx)                                     \
+  {                                                                                                \
+    raft::runtime::neighbors::cagra::build_host(handle, params, dataset, *idx.get_raft_index());   \
   }
 
 CUVS_INST_CAGRA_BUILD(float, uint32_t);
diff --git a/cpp/src/neighbors/cagra_build_int8.cpp b/cpp/src/neighbors/cagra_build_int8.cpp
index deff5d6c6..4123a4293 100644
--- a/cpp/src/neighbors/cagra_build_int8.cpp
+++ b/cpp/src/neighbors/cagra_build_int8.cpp
@@ -19,41 +19,39 @@
 
 namespace cuvs::neighbors::cagra {
 
-#define CUVS_INST_CAGRA_BUILD(T, IdxT)                                                   \
-  auto build(raft::resources const& handle,                                              \
-             const cuvs::neighbors::cagra::index_params& params,                         \
-             raft::device_matrix_view<const T, int64_t, raft::row_major> dataset)        \
-    ->cuvs::neighbors::cagra::index<T, IdxT>                                             \
-  {                                                                                      \
-    return cuvs::neighbors::cagra::index<T, IdxT>(                                       \
-      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));  \
-  }                                                                                      \
-                                                                                         \
-  auto build(raft::resources const& handle,                                              \
-             const cuvs::neighbors::cagra::index_params& params,                         \
-             raft::host_matrix_view<const T, int64_t, raft::row_major> dataset)          \
-    ->cuvs::neighbors::cagra::index<T, IdxT>                                             \
-  {                                                                                      \
-    return cuvs::neighbors::cagra::index<T, IdxT>(                                       \
-      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));  \
-  }                                                                                      \
-                                                                                         \
-  void build_device(raft::resources const& handle,                                       \
-                    const cuvs::neighbors::cagra::index_params& params,                  \
-                    raft::device_matrix_view<const T, int64_t, raft::row_major> dataset, \
-                    cuvs::neighbors::cagra::index<T, IdxT>& idx)                         \
-  {                                                                                      \
-    raft::runtime::neighbors::cagra::build_device(                                       \
-      handle, params, dataset, *idx.get_raft_index());                              \
-  }                                                                                      \
-                                                                                         \
-  void build_host(raft::resources const& handle,                                         \
-                  const cuvs::neighbors::cagra::index_params& params,                    \
-                  raft::host_matrix_view<const T, int64_t, raft::row_major> dataset,     \
-                  cuvs::neighbors::cagra::index<T, IdxT>& idx)                           \
-  {                                                                                      \
-    raft::runtime::neighbors::cagra::build_host(                                         \
-      handle, params, dataset, *idx.get_raft_index());                              \
+#define CUVS_INST_CAGRA_BUILD(T, IdxT)                                                             \
+  auto build(raft::resources const& handle,                                                        \
+             const cuvs::neighbors::cagra::index_params& params,                                   \
+             raft::device_matrix_view<const T, int64_t, raft::row_major> dataset)                  \
+    ->cuvs::neighbors::cagra::index<T, IdxT>                                                       \
+  {                                                                                                \
+    return cuvs::neighbors::cagra::index<T, IdxT>(                                                 \
+      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));                 \
+  }                                                                                                \
+                                                                                                   \
+  auto build(raft::resources const& handle,                                                        \
+             const cuvs::neighbors::cagra::index_params& params,                                   \
+             raft::host_matrix_view<const T, int64_t, raft::row_major> dataset)                    \
+    ->cuvs::neighbors::cagra::index<T, IdxT>                                                       \
+  {                                                                                                \
+    return cuvs::neighbors::cagra::index<T, IdxT>(                                                 \
+      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));                 \
+  }                                                                                                \
+                                                                                                   \
+  void build_device(raft::resources const& handle,                                                 \
+                    const cuvs::neighbors::cagra::index_params& params,                            \
+                    raft::device_matrix_view<const T, int64_t, raft::row_major> dataset,           \
+                    cuvs::neighbors::cagra::index<T, IdxT>& idx)                                   \
+  {                                                                                                \
+    raft::runtime::neighbors::cagra::build_device(handle, params, dataset, *idx.get_raft_index()); \
+  }                                                                                                \
+                                                                                                   \
+  void build_host(raft::resources const& handle,                                                   \
+                  const cuvs::neighbors::cagra::index_params& params,                              \
+                  raft::host_matrix_view<const T, int64_t, raft::row_major> dataset,               \
+                  cuvs::neighbors::cagra::index<T, IdxT>& idx)                                     \
+  {                                                                                                \
+    raft::runtime::neighbors::cagra::build_host(handle, params, dataset, *idx.get_raft_index());   \
   }
 
 CUVS_INST_CAGRA_BUILD(int8_t, uint32_t);
diff --git a/cpp/src/neighbors/cagra_build_uint8.cpp b/cpp/src/neighbors/cagra_build_uint8.cpp
index 1e6a29ca2..fcdcaa0b6 100644
--- a/cpp/src/neighbors/cagra_build_uint8.cpp
+++ b/cpp/src/neighbors/cagra_build_uint8.cpp
@@ -19,41 +19,39 @@
 
 namespace cuvs::neighbors::cagra {
 
-#define CUVS_INST_CAGRA_BUILD(T, IdxT)                                                   \
-  auto build(raft::resources const& handle,                                              \
-             const cuvs::neighbors::cagra::index_params& params,                         \
-             raft::device_matrix_view<const T, int64_t, raft::row_major> dataset)        \
-    ->cuvs::neighbors::cagra::index<T, IdxT>                                             \
-  {                                                                                      \
-    return cuvs::neighbors::cagra::index<T, IdxT>(                                       \
-      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));  \
-  }                                                                                      \
-                                                                                         \
-  auto build(raft::resources const& handle,                                              \
-             const cuvs::neighbors::cagra::index_params& params,                         \
-             raft::host_matrix_view<const T, int64_t, raft::row_major> dataset)          \
-    ->cuvs::neighbors::cagra::index<T, IdxT>                                             \
-  {                                                                                      \
-    return cuvs::neighbors::cagra::index<T, IdxT>(                                       \
-      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));  \
-  }                                                                                      \
-                                                                                         \
-  void build_device(raft::resources const& handle,                                       \
-                    const cuvs::neighbors::cagra::index_params& params,                  \
-                    raft::device_matrix_view<const T, int64_t, raft::row_major> dataset, \
-                    cuvs::neighbors::cagra::index<T, IdxT>& idx)                         \
-  {                                                                                      \
-    raft::runtime::neighbors::cagra::build_device(                                       \
-      handle, params, dataset, *idx.get_raft_index());                              \
-  }                                                                                      \
-                                                                                         \
-  void build_host(raft::resources const& handle,                                         \
-                  const cuvs::neighbors::cagra::index_params& params,                    \
-                  raft::host_matrix_view<const T, int64_t, raft::row_major> dataset,     \
-                  cuvs::neighbors::cagra::index<T, IdxT>& idx)                           \
-  {                                                                                      \
-    raft::runtime::neighbors::cagra::build_host(                                         \
-      handle, params, dataset, *idx.get_raft_index());                              \
+#define CUVS_INST_CAGRA_BUILD(T, IdxT)                                                             \
+  auto build(raft::resources const& handle,                                                        \
+             const cuvs::neighbors::cagra::index_params& params,                                   \
+             raft::device_matrix_view<const T, int64_t, raft::row_major> dataset)                  \
+    ->cuvs::neighbors::cagra::index<T, IdxT>                                                       \
+  {                                                                                                \
+    return cuvs::neighbors::cagra::index<T, IdxT>(                                                 \
+      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));                 \
+  }                                                                                                \
+                                                                                                   \
+  auto build(raft::resources const& handle,                                                        \
+             const cuvs::neighbors::cagra::index_params& params,                                   \
+             raft::host_matrix_view<const T, int64_t, raft::row_major> dataset)                    \
+    ->cuvs::neighbors::cagra::index<T, IdxT>                                                       \
+  {                                                                                                \
+    return cuvs::neighbors::cagra::index<T, IdxT>(                                                 \
+      std::move(raft::runtime::neighbors::cagra::build(handle, params, dataset)));                 \
+  }                                                                                                \
+                                                                                                   \
+  void build_device(raft::resources const& handle,                                                 \
+                    const cuvs::neighbors::cagra::index_params& params,                            \
+                    raft::device_matrix_view<const T, int64_t, raft::row_major> dataset,           \
+                    cuvs::neighbors::cagra::index<T, IdxT>& idx)                                   \
+  {                                                                                                \
+    raft::runtime::neighbors::cagra::build_device(handle, params, dataset, *idx.get_raft_index()); \
+  }                                                                                                \
+                                                                                                   \
+  void build_host(raft::resources const& handle,                                                   \
+                  const cuvs::neighbors::cagra::index_params& params,                              \
+                  raft::host_matrix_view<const T, int64_t, raft::row_major> dataset,               \
+                  cuvs::neighbors::cagra::index<T, IdxT>& idx)                                     \
+  {                                                                                                \
+    raft::runtime::neighbors::cagra::build_host(handle, params, dataset, *idx.get_raft_index());   \
   }
 
 CUVS_INST_CAGRA_BUILD(uint8_t, uint32_t);
diff --git a/cpp/src/neighbors/cagra_optimize.cpp b/cpp/src/neighbors/cagra_optimize.cpp
index 6b2b5adc4..5ec53c043 100644
--- a/cpp/src/neighbors/cagra_optimize.cpp
+++ b/cpp/src/neighbors/cagra_optimize.cpp
@@ -19,17 +19,17 @@
 
 namespace cuvs::neighbors::cagra {
 
-void optimize_device(raft::resources const& handle,                                      
-                     raft::device_matrix_view<uint32_t, int64_t, raft::row_major> knn_graph, 
-                     raft::host_matrix_view<uint32_t, int64_t, raft::row_major> new_graph)   
-{                                                                                        
-  raft::runtime::neighbors::cagra::optimize_device(handle, knn_graph, new_graph);        
-}                                                                                        
-void optimize_host(raft::resources const& handle,                                        
-                   raft::host_matrix_view<uint32_t, int64_t, raft::row_major> knn_graph,     
-                   raft::host_matrix_view<uint32_t, int64_t, raft::row_major> new_graph)     
-{                                                                                        
-  raft::runtime::neighbors::cagra::optimize_host(handle, knn_graph, new_graph);          
+void optimize_device(raft::resources const& handle,
+                     raft::device_matrix_view<uint32_t, int64_t, raft::row_major> knn_graph,
+                     raft::host_matrix_view<uint32_t, int64_t, raft::row_major> new_graph)
+{
+  raft::runtime::neighbors::cagra::optimize_device(handle, knn_graph, new_graph);
+}
+void optimize_host(raft::resources const& handle,
+                   raft::host_matrix_view<uint32_t, int64_t, raft::row_major> knn_graph,
+                   raft::host_matrix_view<uint32_t, int64_t, raft::row_major> new_graph)
+{
+  raft::runtime::neighbors::cagra::optimize_host(handle, knn_graph, new_graph);
 }
 
 }  // namespace cuvs::neighbors::cagra
\ No newline at end of file
diff --git a/cpp/src/neighbors/cagra_search_float.cpp b/cpp/src/neighbors/cagra_search_float.cpp
index b20c5cc37..f05f1e690 100644
--- a/cpp/src/neighbors/cagra_search_float.cpp
+++ b/cpp/src/neighbors/cagra_search_float.cpp
@@ -19,16 +19,16 @@
 
 namespace cuvs::neighbors::cagra {
 
-#define CUVS_INST_CAGRA_SEARCH(T, IdxT)                                                       \
-  void search(raft::resources const& handle,                                                  \
-              cuvs::neighbors::cagra::search_params const& params,                            \
-              const cuvs::neighbors::cagra::index<T, IdxT>& index,                            \
-              raft::device_matrix_view<const T, int64_t, raft::row_major> queries,            \
-              raft::device_matrix_view<IdxT, int64_t, raft::row_major> neighbors,             \
-              raft::device_matrix_view<float, int64_t, raft::row_major> distances)            \
-  {                                                                                           \
-    raft::runtime::neighbors::cagra::search(                                                  \
-      handle, params, *index.get_raft_index(), queries, neighbors, distances);    \
+#define CUVS_INST_CAGRA_SEARCH(T, IdxT)                                            \
+  void search(raft::resources const& handle,                                       \
+              cuvs::neighbors::cagra::search_params const& params,                 \
+              const cuvs::neighbors::cagra::index<T, IdxT>& index,                 \
+              raft::device_matrix_view<const T, int64_t, raft::row_major> queries, \
+              raft::device_matrix_view<IdxT, int64_t, raft::row_major> neighbors,  \
+              raft::device_matrix_view<float, int64_t, raft::row_major> distances) \
+  {                                                                                \
+    raft::runtime::neighbors::cagra::search(                                       \
+      handle, params, *index.get_raft_index(), queries, neighbors, distances);     \
   }
 
 CUVS_INST_CAGRA_SEARCH(float, uint32_t);
diff --git a/cpp/src/neighbors/cagra_search_int8.cpp b/cpp/src/neighbors/cagra_search_int8.cpp
index 04d37107c..56b722f7b 100644
--- a/cpp/src/neighbors/cagra_search_int8.cpp
+++ b/cpp/src/neighbors/cagra_search_int8.cpp
@@ -19,16 +19,16 @@
 
 namespace cuvs::neighbors::cagra {
 
-#define CUVS_INST_CAGRA_SEARCH(T, IdxT)                                                       \
-  void search(raft::resources const& handle,                                                  \
-              cuvs::neighbors::cagra::search_params const& params,                            \
-              const cuvs::neighbors::cagra::index<T, IdxT>& index,                            \
-              raft::device_matrix_view<const T, int64_t, raft::row_major> queries,            \
-              raft::device_matrix_view<IdxT, int64_t, raft::row_major> neighbors,             \
-              raft::device_matrix_view<float, int64_t, raft::row_major> distances)            \
-  {                                                                                           \
-    raft::runtime::neighbors::cagra::search(                                                  \
-      handle, params, *index.get_raft_index(), queries, neighbors, distances);    \
+#define CUVS_INST_CAGRA_SEARCH(T, IdxT)                                            \
+  void search(raft::resources const& handle,                                       \
+              cuvs::neighbors::cagra::search_params const& params,                 \
+              const cuvs::neighbors::cagra::index<T, IdxT>& index,                 \
+              raft::device_matrix_view<const T, int64_t, raft::row_major> queries, \
+              raft::device_matrix_view<IdxT, int64_t, raft::row_major> neighbors,  \
+              raft::device_matrix_view<float, int64_t, raft::row_major> distances) \
+  {                                                                                \
+    raft::runtime::neighbors::cagra::search(                                       \
+      handle, params, *index.get_raft_index(), queries, neighbors, distances);     \
   }
 
 CUVS_INST_CAGRA_SEARCH(int8_t, uint32_t);
diff --git a/cpp/src/neighbors/cagra_search_uint8.cpp b/cpp/src/neighbors/cagra_search_uint8.cpp
index 65a74dabf..3b980cf47 100644
--- a/cpp/src/neighbors/cagra_search_uint8.cpp
+++ b/cpp/src/neighbors/cagra_search_uint8.cpp
@@ -19,16 +19,16 @@
 
 namespace cuvs::neighbors::cagra {
 
-#define CUVS_INST_CAGRA_SEARCH(T, IdxT)                                                       \
-  void search(raft::resources const& handle,                                                  \
-              cuvs::neighbors::cagra::search_params const& params,                            \
-              const cuvs::neighbors::cagra::index<T, IdxT>& index,                            \
-              raft::device_matrix_view<const T, int64_t, raft::row_major> queries,            \
-              raft::device_matrix_view<IdxT, int64_t, raft::row_major> neighbors,             \
-              raft::device_matrix_view<float, int64_t, raft::row_major> distances)            \
-  {                                                                                           \
-    raft::runtime::neighbors::cagra::search(                                                  \
-      handle, params, *index.get_raft_index(), queries, neighbors, distances);    \
+#define CUVS_INST_CAGRA_SEARCH(T, IdxT)                                            \
+  void search(raft::resources const& handle,                                       \
+              cuvs::neighbors::cagra::search_params const& params,                 \
+              const cuvs::neighbors::cagra::index<T, IdxT>& index,                 \
+              raft::device_matrix_view<const T, int64_t, raft::row_major> queries, \
+              raft::device_matrix_view<IdxT, int64_t, raft::row_major> neighbors,  \
+              raft::device_matrix_view<float, int64_t, raft::row_major> distances) \
+  {                                                                                \
+    raft::runtime::neighbors::cagra::search(                                       \
+      handle, params, *index.get_raft_index(), queries, neighbors, distances);     \
   }
 
 CUVS_INST_CAGRA_SEARCH(uint8_t, uint32_t);
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_00_generate.py b/cpp/src/neighbors/detail/cagra/search_multi_cta_00_generate.py
deleted file mode 100644
index 8826439c4..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_00_generate.py
+++ /dev/null
@@ -1,108 +0,0 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-header = """
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \\
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \\
-  template void                                                                             \\
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \\
-    raft::device_matrix_view<const DATA_T, int64_t,raft::layout_stride> dataset,                 \\
-    raft::device_matrix_view<const INDEX_T, int64_t,raft::row_major> graph,                      \\
-    INDEX_T* const topk_indices_ptr,                                                        \\
-    DISTANCE_T* const topk_distances_ptr,                                                   \\
-    const DATA_T* const queries_ptr,                                                        \\
-    const uint32_t num_queries,                                                             \\
-    const INDEX_T* dev_seed_ptr,                                                            \\
-    uint32_t* const num_executed_iterations,                                                \\
-    uint32_t topk,                                                                          \\
-    uint32_t block_size,                                                                    \\
-    uint32_t result_buffer_size,                                                            \\
-    uint32_t smem_size,                                                                     \\
-    int64_t hash_bitlen,                                                                    \\
-    INDEX_T* hashmap_ptr,                                                                   \\
-    uint32_t num_cta_per_query,                                                             \\
-    uint32_t num_random_samplings,                                                          \\
-    uint64_t rand_xor_mask,                                                                 \\
-    uint32_t num_seeds,                                                                     \\
-    size_t itopk_size,                                                                      \\
-    size_t search_width,                                                                    \\
-    size_t min_iterations,                                                                  \\
-    size_t max_iterations,                                                                  \\
-    SAMPLE_FILTER_T sample_filter,                                                          \\
-    cudaStream_t stream);
-
-"""
-
-trailer = """
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
-"""
-
-mxdim_team = [(128, 8), (256, 16), (512, 32), (1024, 32)]
-# block = [(64, 16), (128, 8), (256, 4), (512, 2), (1024, 1)]
-# mxelem = [64, 128, 256]
-load_types = ["uint4"]
-search_types = dict(
-    float_uint32=(
-        "float",
-        "uint32_t",
-        "float",
-    ),  # data_t, vec_idx_t, distance_t
-    int8_uint32=("int8_t", "uint32_t", "float"),
-    uint8_uint32=("uint8_t", "uint32_t", "float"),
-    float_uint64=("float", "uint64_t", "float"),
-)
-# knn
-for type_path, (data_t, idx_t, distance_t) in search_types.items():
-    for (mxdim, team) in mxdim_team:
-        path = f"search_multi_cta_{type_path}_dim{mxdim}_t{team}.cu"
-        with open(path, "w") as f:
-            f.write(header)
-            f.write(
-                f"instantiate_kernel_selection(\n  {team}, {mxdim}, {data_t}, {idx_t}, {distance_t}, cuvs::neighbors::filtering::none_cagra_sample_filter);\n"
-            )
-            f.write(trailer)
-            # For pasting into CMakeLists.txt
-        print(f"src/neighbors/detail/cagra/{path}")
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim1024_t32.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim1024_t32.cu
deleted file mode 100644
index c457e87c1..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim1024_t32.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  32, 1024, float, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim128_t8.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim128_t8.cu
deleted file mode 100644
index ab5e2821d..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim128_t8.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  8, 128, float, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim256_t16.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim256_t16.cu
deleted file mode 100644
index bb517f6bc..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim256_t16.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  16, 256, float, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim512_t32.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim512_t32.cu
deleted file mode 100644
index f99ac9340..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint32_dim512_t32.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  32, 512, float, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim1024_t32.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim1024_t32.cu
deleted file mode 100644
index 766f65c08..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim1024_t32.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  32, 1024, float, uint64_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim128_t8.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim128_t8.cu
deleted file mode 100644
index d9d5ee97b..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim128_t8.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  8, 128, float, uint64_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim256_t16.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim256_t16.cu
deleted file mode 100644
index 062817e63..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim256_t16.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  16, 256, float, uint64_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim512_t32.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim512_t32.cu
deleted file mode 100644
index b6c0cd07b..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_float_uint64_dim512_t32.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  32, 512, float, uint64_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim1024_t32.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim1024_t32.cu
deleted file mode 100644
index d71f0bfbc..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim1024_t32.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  32, 1024, int8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim128_t8.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim128_t8.cu
deleted file mode 100644
index 041e162f6..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim128_t8.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  8, 128, int8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim256_t16.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim256_t16.cu
deleted file mode 100644
index 6e38154b5..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim256_t16.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  16, 256, int8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim512_t32.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim512_t32.cu
deleted file mode 100644
index 7663ea28d..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_int8_uint32_dim512_t32.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  32, 512, int8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim1024_t32.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim1024_t32.cu
deleted file mode 100644
index dbbf3438e..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim1024_t32.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  32, 1024, uint8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim128_t8.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim128_t8.cu
deleted file mode 100644
index f842d6eef..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim128_t8.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  8, 128, uint8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim256_t16.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim256_t16.cu
deleted file mode 100644
index ca0770533..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim256_t16.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  16, 256, uint8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim512_t32.cu b/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim512_t32.cu
deleted file mode 100644
index 4d8376946..000000000
--- a/cpp/src/neighbors/detail/cagra/search_multi_cta_uint8_uint32_dim512_t32.cu
+++ /dev/null
@@ -1,66 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_multi_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_multi_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_multi_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::multi_cta_search {
-
-#define instantiate_kernel_selection(                                                       \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t result_buffer_size,                                                            \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    uint32_t num_cta_per_query,                                                             \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_kernel_selection(
-  32, 512, uint8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_kernel_selection
-
-}  // namespace cuvs::neighbors::cagra::detail::multi_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_00_generate.py b/cpp/src/neighbors/detail/cagra/search_single_cta_00_generate.py
deleted file mode 100644
index 524500055..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_00_generate.py
+++ /dev/null
@@ -1,113 +0,0 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-header = """
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \\
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \\
-  template void                                                                             \\
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \\
-    raft::device_matrix_view<const DATA_T, int64_t,raft::layout_stride> dataset,                 \\
-    raft::device_matrix_view<const INDEX_T, int64_t,raft::row_major> graph,                      \\
-    INDEX_T* const topk_indices_ptr,                                                        \\
-    DISTANCE_T* const topk_distances_ptr,                                                   \\
-    const DATA_T* const queries_ptr,                                                        \\
-    const uint32_t num_queries,                                                             \\
-    const INDEX_T* dev_seed_ptr,                                                            \\
-    uint32_t* const num_executed_iterations,                                                \\
-    uint32_t topk,                                                                          \\
-    uint32_t num_itopk_candidates,                                                          \\
-    uint32_t block_size,                                                                    \\
-    uint32_t smem_size,                                                                     \\
-    int64_t hash_bitlen,                                                                    \\
-    INDEX_T* hashmap_ptr,                                                                   \\
-    size_t small_hash_bitlen,                                                               \\
-    size_t small_hash_reset_interval,                                                       \\
-    uint32_t num_random_samplings,                                                          \\
-    uint64_t rand_xor_mask,                                                                 \\
-    uint32_t num_seeds,                                                                     \\
-    size_t itopk_size,                                                                      \\
-    size_t search_width,                                                                    \\
-    size_t min_iterations,                                                                  \\
-    size_t max_iterations,                                                                  \\
-    SAMPLE_FILTER_T sample_filter,                                                          \\
-    cudaStream_t stream);
-
-"""
-
-trailer = """
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
-"""
-
-mxdim_team = [(128, 8), (256, 16), (512, 32), (1024, 32)]
-# block = [(64, 16), (128, 8), (256, 4), (512, 2), (1024, 1)]
-# itopk_candidates = [64, 128, 256]
-# itopk_size = [64, 128, 256, 512]
-# mxelem = [64, 128, 256]
-
-# rblock = [(256, 4), (512, 2), (1024, 1)]
-# rcandidates = [32]
-# rsize = [256, 512]
-
-search_types = dict(
-    float_uint32=("float", "uint32_t", "float"),  # data_t, idx_t, distance_t
-    int8_uint32=("int8_t", "uint32_t", "float"),
-    uint8_uint32=("uint8_t", "uint32_t", "float"),
-    float_uint64=("float", "uint64_t", "float"),
-)
-
-# knn
-for type_path, (data_t, idx_t, distance_t) in search_types.items():
-    for (mxdim, team) in mxdim_team:
-        path = f"search_single_cta_{type_path}_dim{mxdim}_t{team}.cu"
-        with open(path, "w") as f:
-            f.write(header)
-            f.write(
-                f"instantiate_single_cta_select_and_run(\n  {team}, {mxdim}, {data_t}, {idx_t}, {distance_t}, cuvs::neighbors::filtering::none_cagra_sample_filter);\n"
-            )
-
-            f.write(trailer)
-            # For pasting into CMakeLists.txt
-            print(f"src/neighbors/detail/cagra/{path}")
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim1024_t32.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim1024_t32.cu
deleted file mode 100644
index 0534c84fc..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim1024_t32.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  32, 1024, float, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim128_t8.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim128_t8.cu
deleted file mode 100644
index c29a1ca66..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim128_t8.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  8, 128, float, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim256_t16.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim256_t16.cu
deleted file mode 100644
index 1e905e732..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim256_t16.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  16, 256, float, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim512_t32.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim512_t32.cu
deleted file mode 100644
index 40f34dba7..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint32_dim512_t32.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  32, 512, float, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim1024_t32.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim1024_t32.cu
deleted file mode 100644
index b0cfd20a1..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim1024_t32.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  32, 1024, float, uint64_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim128_t8.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim128_t8.cu
deleted file mode 100644
index 52c86a856..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim128_t8.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  8, 128, float, uint64_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim256_t16.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim256_t16.cu
deleted file mode 100644
index 755e82e3c..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim256_t16.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  16, 256, float, uint64_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim512_t32.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim512_t32.cu
deleted file mode 100644
index fd49b7a82..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_float_uint64_dim512_t32.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  32, 512, float, uint64_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim1024_t32.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim1024_t32.cu
deleted file mode 100644
index 4dd934945..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim1024_t32.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  32, 1024, int8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim128_t8.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim128_t8.cu
deleted file mode 100644
index eb41a6940..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim128_t8.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  8, 128, int8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim256_t16.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim256_t16.cu
deleted file mode 100644
index 3a27a6b70..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim256_t16.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  16, 256, int8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim512_t32.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim512_t32.cu
deleted file mode 100644
index 5b5a6311d..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_int8_uint32_dim512_t32.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  32, 512, int8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim1024_t32.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim1024_t32.cu
deleted file mode 100644
index 741ac5306..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim1024_t32.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  32, 1024, uint8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim128_t8.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim128_t8.cu
deleted file mode 100644
index 0951ced54..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim128_t8.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  8, 128, uint8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim256_t16.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim256_t16.cu
deleted file mode 100644
index 15b2cbb24..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim256_t16.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  16, 256, uint8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim512_t32.cu b/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim512_t32.cu
deleted file mode 100644
index 425d04ec8..000000000
--- a/cpp/src/neighbors/detail/cagra/search_single_cta_uint8_uint32_dim512_t32.cu
+++ /dev/null
@@ -1,67 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by search_single_cta_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python search_single_cta_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-namespace cuvs::neighbors::cagra::detail::single_cta_search {
-
-#define instantiate_single_cta_select_and_run(                                              \
-  TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T)                 \
-  template void                                                                             \
-  select_and_run<TEAM_SIZE, MAX_DATASET_DIM, DATA_T, INDEX_T, DISTANCE_T, SAMPLE_FILTER_T>( \
-    raft::device_matrix_view<const DATA_T, int64_t, raft::layout_stride> dataset,           \
-    raft::device_matrix_view<const INDEX_T, int64_t, raft::row_major> graph,                \
-    INDEX_T* const topk_indices_ptr,                                                        \
-    DISTANCE_T* const topk_distances_ptr,                                                   \
-    const DATA_T* const queries_ptr,                                                        \
-    const uint32_t num_queries,                                                             \
-    const INDEX_T* dev_seed_ptr,                                                            \
-    uint32_t* const num_executed_iterations,                                                \
-    uint32_t topk,                                                                          \
-    uint32_t num_itopk_candidates,                                                          \
-    uint32_t block_size,                                                                    \
-    uint32_t smem_size,                                                                     \
-    int64_t hash_bitlen,                                                                    \
-    INDEX_T* hashmap_ptr,                                                                   \
-    size_t small_hash_bitlen,                                                               \
-    size_t small_hash_reset_interval,                                                       \
-    uint32_t num_random_samplings,                                                          \
-    uint64_t rand_xor_mask,                                                                 \
-    uint32_t num_seeds,                                                                     \
-    size_t itopk_size,                                                                      \
-    size_t search_width,                                                                    \
-    size_t min_iterations,                                                                  \
-    size_t max_iterations,                                                                  \
-    SAMPLE_FILTER_T sample_filter,                                                          \
-    cudaStream_t stream);
-
-instantiate_single_cta_select_and_run(
-  32, 512, uint8_t, uint32_t, float, cuvs::neighbors::filtering::none_cagra_sample_filter);
-
-#undef instantiate_single_cta_search_kernel
-
-}  // namespace cuvs::neighbors::cagra::detail::single_cta_search
diff --git a/cpp/src/neighbors/detail/ivf_flat_interleaved_scan_float_float_int64_t.cu b/cpp/src/neighbors/detail/ivf_flat_interleaved_scan_float_float_int64_t.cu
deleted file mode 100644
index 2f0660e99..000000000
--- a/cpp/src/neighbors/detail/ivf_flat_interleaved_scan_float_float_int64_t.cu
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/detail/ivf_flat_interleaved_scan-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-#define instantiate_raft_neighbors_ivf_flat_detail_ivfflat_interleaved_scan(                    \
-  T, AccT, IdxT, IvfSampleFilterT)                                                              \
-  template void                                                                                 \
-  cuvs::neighbors::ivf_flat::detail::ivfflat_interleaved_scan<T, AccT, IdxT, IvfSampleFilterT>( \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,                                     \
-    const T* queries,                                                                           \
-    const uint32_t* coarse_query_results,                                                       \
-    const uint32_t n_queries,                                                                   \
-    const uint32_t queries_offset,                                                              \
-    const cuvs::distance::DistanceType metric,                                                  \
-    const uint32_t n_probes,                                                                    \
-    const uint32_t k,                                                                           \
-    const bool select_min,                                                                      \
-    IvfSampleFilterT sample_filter,                                                             \
-    IdxT* neighbors,                                                                            \
-    float* distances,                                                                           \
-    uint32_t& grid_dim_x,                                                                       \
-    rmm::cuda_stream_view stream)
-
-instantiate_raft_neighbors_ivf_flat_detail_ivfflat_interleaved_scan(
-  float, float, int64_t, cuvs::neighbors::filtering::none_ivf_sample_filter);
-
-#undef instantiate_raft_neighbors_ivf_flat_detail_ivfflat_interleaved_scan
diff --git a/cpp/src/neighbors/detail/ivf_flat_interleaved_scan_int8_t_int32_t_int64_t.cu b/cpp/src/neighbors/detail/ivf_flat_interleaved_scan_int8_t_int32_t_int64_t.cu
deleted file mode 100644
index 4dbf64c1b..000000000
--- a/cpp/src/neighbors/detail/ivf_flat_interleaved_scan_int8_t_int32_t_int64_t.cu
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/detail/ivf_flat_interleaved_scan-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-#define instantiate_raft_neighbors_ivf_flat_detail_ivfflat_interleaved_scan(                    \
-  T, AccT, IdxT, IvfSampleFilterT)                                                              \
-  template void                                                                                 \
-  cuvs::neighbors::ivf_flat::detail::ivfflat_interleaved_scan<T, AccT, IdxT, IvfSampleFilterT>( \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,                                     \
-    const T* queries,                                                                           \
-    const uint32_t* coarse_query_results,                                                       \
-    const uint32_t n_queries,                                                                   \
-    const uint32_t queries_offset,                                                              \
-    const cuvs::distance::DistanceType metric,                                                  \
-    const uint32_t n_probes,                                                                    \
-    const uint32_t k,                                                                           \
-    const bool select_min,                                                                      \
-    IvfSampleFilterT sample_filter,                                                             \
-    IdxT* neighbors,                                                                            \
-    float* distances,                                                                           \
-    uint32_t& grid_dim_x,                                                                       \
-    rmm::cuda_stream_view stream)
-
-instantiate_raft_neighbors_ivf_flat_detail_ivfflat_interleaved_scan(
-  int8_t, int32_t, int64_t, cuvs::neighbors::filtering::none_ivf_sample_filter);
-
-#undef instantiate_raft_neighbors_ivf_flat_detail_ivfflat_interleaved_scan
diff --git a/cpp/src/neighbors/detail/ivf_flat_interleaved_scan_uint8_t_uint32_t_int64_t.cu b/cpp/src/neighbors/detail/ivf_flat_interleaved_scan_uint8_t_uint32_t_int64_t.cu
deleted file mode 100644
index 933fa07b3..000000000
--- a/cpp/src/neighbors/detail/ivf_flat_interleaved_scan_uint8_t_uint32_t_int64_t.cu
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/detail/ivf_flat_interleaved_scan-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-#define instantiate_raft_neighbors_ivf_flat_detail_ivfflat_interleaved_scan(                    \
-  T, AccT, IdxT, IvfSampleFilterT)                                                              \
-  template void                                                                                 \
-  cuvs::neighbors::ivf_flat::detail::ivfflat_interleaved_scan<T, AccT, IdxT, IvfSampleFilterT>( \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,                                     \
-    const T* queries,                                                                           \
-    const uint32_t* coarse_query_results,                                                       \
-    const uint32_t n_queries,                                                                   \
-    const uint32_t queries_offset,                                                              \
-    const cuvs::distance::DistanceType metric,                                                  \
-    const uint32_t n_probes,                                                                    \
-    const uint32_t k,                                                                           \
-    const bool select_min,                                                                      \
-    IvfSampleFilterT sample_filter,                                                             \
-    IdxT* neighbors,                                                                            \
-    float* distances,                                                                           \
-    uint32_t& grid_dim_x,                                                                       \
-    rmm::cuda_stream_view stream)
-
-instantiate_raft_neighbors_ivf_flat_detail_ivfflat_interleaved_scan(
-  uint8_t, uint32_t, int64_t, cuvs::neighbors::filtering::none_ivf_sample_filter);
-
-#undef instantiate_raft_neighbors_ivf_flat_detail_ivfflat_interleaved_scan
diff --git a/cpp/src/neighbors/detail/ivf_flat_search.cu b/cpp/src/neighbors/detail/ivf_flat_search.cu
deleted file mode 100644
index ada611e9f..000000000
--- a/cpp/src/neighbors/detail/ivf_flat_search.cu
+++ /dev/null
@@ -1,40 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/detail/ivf_flat_search-inl.cuh>
-#include <cuvs/neighbors/sample_filter_types.hpp>
-
-#define instantiate_raft_neighbors_ivf_flat_detail_search(T, IdxT, IvfSampleFilterT)  \
-  template void cuvs::neighbors::ivf_flat::detail::search<T, IdxT, IvfSampleFilterT>( \
-    raft::resources const& handle,                                                    \
-    const search_params& params,                                                      \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,                           \
-    const T* queries,                                                                 \
-    uint32_t n_queries,                                                               \
-    uint32_t k,                                                                       \
-    IdxT* neighbors,                                                                  \
-    float* distances,                                                                 \
-    rmm::mr::device_memory_resource* mr,                                              \
-    IvfSampleFilterT sample_filter)
-
-instantiate_raft_neighbors_ivf_flat_detail_search(
-  float, int64_t, cuvs::neighbors::filtering::none_ivf_sample_filter);
-instantiate_raft_neighbors_ivf_flat_detail_search(
-  int8_t, int64_t, cuvs::neighbors::filtering::none_ivf_sample_filter);
-instantiate_raft_neighbors_ivf_flat_detail_search(
-  uint8_t, int64_t, cuvs::neighbors::filtering::none_ivf_sample_filter);
-
-#undef instantiate_raft_neighbors_ivf_flat_detail_search
diff --git a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_00_generate.py b/cpp/src/neighbors/detail/ivf_pq_compute_similarity_00_generate.py
deleted file mode 100644
index 822198aa3..000000000
--- a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_00_generate.py
+++ /dev/null
@@ -1,108 +0,0 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-header = """
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_pq_compute_similarity_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_pq_compute_similarity_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/ivf_pq_compute_similarity-inl.cuh>
-#include <cuvs/neighbors/detail/ivf_pq_fp_8bit.cuh>
-
-#define instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(OutT, LutT, IvfSampleFilterT) \\
-    template auto cuvs::neighbors::ivf_pq::detail::compute_similarity_select<OutT, LutT, IvfSampleFilterT>( \\
-        const cudaDeviceProp& dev_props,                                \\
-        bool manage_local_topk,                                         \\
-        int locality_hint,                                              \\
-        double preferred_shmem_carveout,                                \\
-        uint32_t pq_bits,                                               \\
-        uint32_t pq_dim,                                                \\
-        uint32_t precomp_data_count,                                    \\
-        uint32_t n_queries,                                             \\
-        uint32_t n_probes,                                              \\
-        uint32_t topk) -> cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT>; \\
-\\
-    template void cuvs::neighbors::ivf_pq::detail::compute_similarity_run<OutT, LutT, IvfSampleFilterT>( \\
-        cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT> s,        \\
-        rmm::cuda_stream_view stream,                                   \\
-        uint32_t dim,                                                   \\
-        uint32_t n_probes,                                              \\
-        uint32_t pq_dim,                                                \\
-        uint32_t n_queries,                                             \\
-        uint32_t queries_offset,                                        \\
-        cuvs::distance::DistanceType metric,                                  \\
-        cuvs::neighbors::ivf_pq::codebook_gen codebook_kind,            \\
-        uint32_t topk,                                                  \\
-        uint32_t max_samples,                                           \\
-        const float* cluster_centers,                                   \\
-        const float* pq_centers,                                        \\
-        const uint8_t* const* pq_dataset,                               \\
-        const uint32_t* cluster_labels,                                 \\
-        const uint32_t* _chunk_indices,                                 \\
-        const float* queries,                                           \\
-        const uint32_t* index_list,                                     \\
-        float* query_kths,                                              \\
-        IvfSampleFilterT sample_filter,                                    \\
-        LutT* lut_scores,                                               \\
-        OutT* _out_scores,                                              \\
-        uint32_t* _out_indices);
-
-
-#define COMMA ,
-"""
-
-trailer = """
-#undef COMMA
-
-#undef instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select
-"""
-
-types = dict(
-    half_fp8_false=("half", "cuvs::neighbors::ivf_pq::detail::fp_8bit<5u COMMA false>"),
-    half_fp8_true=("half", "cuvs::neighbors::ivf_pq::detail::fp_8bit<5u COMMA true>"),
-    half_half=("half", "half"),
-    float_half=("float", "half"),
-    float_float= ("float", "float"),
-    float_fp8_false=("float", "cuvs::neighbors::ivf_pq::detail::fp_8bit<5u COMMA false>"),
-    float_fp8_true=("float", "cuvs::neighbors::ivf_pq::detail::fp_8bit<5u COMMA true>"),
-)
-
-for path_key, (OutT, LutT) in types.items():
-    path = f"ivf_pq_compute_similarity_{path_key}.cu"
-    with open(path, "w") as f:
-        f.write(header)
-        f.write(f"instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select({OutT}, {LutT}, cuvs::neighbors::filtering::ivf_to_sample_filter<int64_t COMMA cuvs::neighbors::filtering::none_ivf_sample_filter>);\n")
-        f.write(trailer)
-    print(f"src/neighbors/detail/{path}")
diff --git a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_float.cu b/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_float.cu
deleted file mode 100644
index a97110c73..000000000
--- a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_float.cu
+++ /dev/null
@@ -1,81 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_pq_compute_similarity_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_pq_compute_similarity_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/ivf_pq_compute_similarity-inl.cuh>
-#include <cuvs/neighbors/detail/ivf_pq_fp_8bit.cuh>
-
-#define instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(                 \
-  OutT, LutT, IvfSampleFilterT)                                                             \
-  template auto                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_select<OutT, LutT, IvfSampleFilterT>( \
-    const cudaDeviceProp& dev_props,                                                        \
-    bool manage_local_topk,                                                                 \
-    int locality_hint,                                                                      \
-    double preferred_shmem_carveout,                                                        \
-    uint32_t pq_bits,                                                                       \
-    uint32_t pq_dim,                                                                        \
-    uint32_t precomp_data_count,                                                            \
-    uint32_t n_queries,                                                                     \
-    uint32_t n_probes,                                                                      \
-    uint32_t topk)                                                                          \
-    ->cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT>;              \
-                                                                                            \
-  template void                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_run<OutT, LutT, IvfSampleFilterT>(    \
-    cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT> s,              \
-    rmm::cuda_stream_view stream,                                                           \
-    uint32_t dim,                                                                           \
-    uint32_t n_probes,                                                                      \
-    uint32_t pq_dim,                                                                        \
-    uint32_t n_queries,                                                                     \
-    uint32_t queries_offset,                                                                \
-    cuvs::distance::DistanceType metric,                                                    \
-    cuvs::neighbors::ivf_pq::codebook_gen codebook_kind,                                    \
-    uint32_t topk,                                                                          \
-    uint32_t max_samples,                                                                   \
-    const float* cluster_centers,                                                           \
-    const float* pq_centers,                                                                \
-    const uint8_t* const* pq_dataset,                                                       \
-    const uint32_t* cluster_labels,                                                         \
-    const uint32_t* _chunk_indices,                                                         \
-    const float* queries,                                                                   \
-    const uint32_t* index_list,                                                             \
-    float* query_kths,                                                                      \
-    IvfSampleFilterT sample_filter,                                                         \
-    LutT* lut_scores,                                                                       \
-    OutT* _out_scores,                                                                      \
-    uint32_t* _out_indices);
-
-#define COMMA ,
-instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(
-  float,
-  float,
-  cuvs::neighbors::filtering::ivf_to_sample_filter<
-    int64_t COMMA cuvs::neighbors::filtering::none_ivf_sample_filter>);
-
-#undef COMMA
-
-#undef instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select
diff --git a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_fp8_false.cu b/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_fp8_false.cu
deleted file mode 100644
index 8b8bd4dce..000000000
--- a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_fp8_false.cu
+++ /dev/null
@@ -1,81 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_pq_compute_similarity_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_pq_compute_similarity_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/ivf_pq_compute_similarity-inl.cuh>
-#include <cuvs/neighbors/detail/ivf_pq_fp_8bit.cuh>
-
-#define instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(                 \
-  OutT, LutT, IvfSampleFilterT)                                                             \
-  template auto                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_select<OutT, LutT, IvfSampleFilterT>( \
-    const cudaDeviceProp& dev_props,                                                        \
-    bool manage_local_topk,                                                                 \
-    int locality_hint,                                                                      \
-    double preferred_shmem_carveout,                                                        \
-    uint32_t pq_bits,                                                                       \
-    uint32_t pq_dim,                                                                        \
-    uint32_t precomp_data_count,                                                            \
-    uint32_t n_queries,                                                                     \
-    uint32_t n_probes,                                                                      \
-    uint32_t topk)                                                                          \
-    ->cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT>;              \
-                                                                                            \
-  template void                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_run<OutT, LutT, IvfSampleFilterT>(    \
-    cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT> s,              \
-    rmm::cuda_stream_view stream,                                                           \
-    uint32_t dim,                                                                           \
-    uint32_t n_probes,                                                                      \
-    uint32_t pq_dim,                                                                        \
-    uint32_t n_queries,                                                                     \
-    uint32_t queries_offset,                                                                \
-    cuvs::distance::DistanceType metric,                                                    \
-    cuvs::neighbors::ivf_pq::codebook_gen codebook_kind,                                    \
-    uint32_t topk,                                                                          \
-    uint32_t max_samples,                                                                   \
-    const float* cluster_centers,                                                           \
-    const float* pq_centers,                                                                \
-    const uint8_t* const* pq_dataset,                                                       \
-    const uint32_t* cluster_labels,                                                         \
-    const uint32_t* _chunk_indices,                                                         \
-    const float* queries,                                                                   \
-    const uint32_t* index_list,                                                             \
-    float* query_kths,                                                                      \
-    IvfSampleFilterT sample_filter,                                                         \
-    LutT* lut_scores,                                                                       \
-    OutT* _out_scores,                                                                      \
-    uint32_t* _out_indices);
-
-#define COMMA ,
-instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(
-  float,
-  cuvs::neighbors::ivf_pq::detail::fp_8bit<5u COMMA false>,
-  cuvs::neighbors::filtering::ivf_to_sample_filter<
-    int64_t COMMA cuvs::neighbors::filtering::none_ivf_sample_filter>);
-
-#undef COMMA
-
-#undef instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select
diff --git a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_fp8_true.cu b/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_fp8_true.cu
deleted file mode 100644
index a35b5da37..000000000
--- a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_fp8_true.cu
+++ /dev/null
@@ -1,81 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_pq_compute_similarity_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_pq_compute_similarity_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/ivf_pq_compute_similarity-inl.cuh>
-#include <cuvs/neighbors/detail/ivf_pq_fp_8bit.cuh>
-
-#define instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(                 \
-  OutT, LutT, IvfSampleFilterT)                                                             \
-  template auto                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_select<OutT, LutT, IvfSampleFilterT>( \
-    const cudaDeviceProp& dev_props,                                                        \
-    bool manage_local_topk,                                                                 \
-    int locality_hint,                                                                      \
-    double preferred_shmem_carveout,                                                        \
-    uint32_t pq_bits,                                                                       \
-    uint32_t pq_dim,                                                                        \
-    uint32_t precomp_data_count,                                                            \
-    uint32_t n_queries,                                                                     \
-    uint32_t n_probes,                                                                      \
-    uint32_t topk)                                                                          \
-    ->cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT>;              \
-                                                                                            \
-  template void                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_run<OutT, LutT, IvfSampleFilterT>(    \
-    cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT> s,              \
-    rmm::cuda_stream_view stream,                                                           \
-    uint32_t dim,                                                                           \
-    uint32_t n_probes,                                                                      \
-    uint32_t pq_dim,                                                                        \
-    uint32_t n_queries,                                                                     \
-    uint32_t queries_offset,                                                                \
-    cuvs::distance::DistanceType metric,                                                    \
-    cuvs::neighbors::ivf_pq::codebook_gen codebook_kind,                                    \
-    uint32_t topk,                                                                          \
-    uint32_t max_samples,                                                                   \
-    const float* cluster_centers,                                                           \
-    const float* pq_centers,                                                                \
-    const uint8_t* const* pq_dataset,                                                       \
-    const uint32_t* cluster_labels,                                                         \
-    const uint32_t* _chunk_indices,                                                         \
-    const float* queries,                                                                   \
-    const uint32_t* index_list,                                                             \
-    float* query_kths,                                                                      \
-    IvfSampleFilterT sample_filter,                                                         \
-    LutT* lut_scores,                                                                       \
-    OutT* _out_scores,                                                                      \
-    uint32_t* _out_indices);
-
-#define COMMA ,
-instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(
-  float,
-  cuvs::neighbors::ivf_pq::detail::fp_8bit<5u COMMA true>,
-  cuvs::neighbors::filtering::ivf_to_sample_filter<
-    int64_t COMMA cuvs::neighbors::filtering::none_ivf_sample_filter>);
-
-#undef COMMA
-
-#undef instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select
diff --git a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_half.cu b/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_half.cu
deleted file mode 100644
index 7300e38a3..000000000
--- a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_float_half.cu
+++ /dev/null
@@ -1,81 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_pq_compute_similarity_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_pq_compute_similarity_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/ivf_pq_compute_similarity-inl.cuh>
-#include <cuvs/neighbors/detail/ivf_pq_fp_8bit.cuh>
-
-#define instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(                 \
-  OutT, LutT, IvfSampleFilterT)                                                             \
-  template auto                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_select<OutT, LutT, IvfSampleFilterT>( \
-    const cudaDeviceProp& dev_props,                                                        \
-    bool manage_local_topk,                                                                 \
-    int locality_hint,                                                                      \
-    double preferred_shmem_carveout,                                                        \
-    uint32_t pq_bits,                                                                       \
-    uint32_t pq_dim,                                                                        \
-    uint32_t precomp_data_count,                                                            \
-    uint32_t n_queries,                                                                     \
-    uint32_t n_probes,                                                                      \
-    uint32_t topk)                                                                          \
-    ->cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT>;              \
-                                                                                            \
-  template void                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_run<OutT, LutT, IvfSampleFilterT>(    \
-    cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT> s,              \
-    rmm::cuda_stream_view stream,                                                           \
-    uint32_t dim,                                                                           \
-    uint32_t n_probes,                                                                      \
-    uint32_t pq_dim,                                                                        \
-    uint32_t n_queries,                                                                     \
-    uint32_t queries_offset,                                                                \
-    cuvs::distance::DistanceType metric,                                                    \
-    cuvs::neighbors::ivf_pq::codebook_gen codebook_kind,                                    \
-    uint32_t topk,                                                                          \
-    uint32_t max_samples,                                                                   \
-    const float* cluster_centers,                                                           \
-    const float* pq_centers,                                                                \
-    const uint8_t* const* pq_dataset,                                                       \
-    const uint32_t* cluster_labels,                                                         \
-    const uint32_t* _chunk_indices,                                                         \
-    const float* queries,                                                                   \
-    const uint32_t* index_list,                                                             \
-    float* query_kths,                                                                      \
-    IvfSampleFilterT sample_filter,                                                         \
-    LutT* lut_scores,                                                                       \
-    OutT* _out_scores,                                                                      \
-    uint32_t* _out_indices);
-
-#define COMMA ,
-instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(
-  float,
-  half,
-  cuvs::neighbors::filtering::ivf_to_sample_filter<
-    int64_t COMMA cuvs::neighbors::filtering::none_ivf_sample_filter>);
-
-#undef COMMA
-
-#undef instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select
diff --git a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_fp8_false.cu b/cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_fp8_false.cu
deleted file mode 100644
index c74e0fa25..000000000
--- a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_fp8_false.cu
+++ /dev/null
@@ -1,81 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_pq_compute_similarity_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_pq_compute_similarity_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/ivf_pq_compute_similarity-inl.cuh>
-#include <cuvs/neighbors/detail/ivf_pq_fp_8bit.cuh>
-
-#define instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(                 \
-  OutT, LutT, IvfSampleFilterT)                                                             \
-  template auto                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_select<OutT, LutT, IvfSampleFilterT>( \
-    const cudaDeviceProp& dev_props,                                                        \
-    bool manage_local_topk,                                                                 \
-    int locality_hint,                                                                      \
-    double preferred_shmem_carveout,                                                        \
-    uint32_t pq_bits,                                                                       \
-    uint32_t pq_dim,                                                                        \
-    uint32_t precomp_data_count,                                                            \
-    uint32_t n_queries,                                                                     \
-    uint32_t n_probes,                                                                      \
-    uint32_t topk)                                                                          \
-    ->cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT>;              \
-                                                                                            \
-  template void                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_run<OutT, LutT, IvfSampleFilterT>(    \
-    cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT> s,              \
-    rmm::cuda_stream_view stream,                                                           \
-    uint32_t dim,                                                                           \
-    uint32_t n_probes,                                                                      \
-    uint32_t pq_dim,                                                                        \
-    uint32_t n_queries,                                                                     \
-    uint32_t queries_offset,                                                                \
-    cuvs::distance::DistanceType metric,                                                    \
-    cuvs::neighbors::ivf_pq::codebook_gen codebook_kind,                                    \
-    uint32_t topk,                                                                          \
-    uint32_t max_samples,                                                                   \
-    const float* cluster_centers,                                                           \
-    const float* pq_centers,                                                                \
-    const uint8_t* const* pq_dataset,                                                       \
-    const uint32_t* cluster_labels,                                                         \
-    const uint32_t* _chunk_indices,                                                         \
-    const float* queries,                                                                   \
-    const uint32_t* index_list,                                                             \
-    float* query_kths,                                                                      \
-    IvfSampleFilterT sample_filter,                                                         \
-    LutT* lut_scores,                                                                       \
-    OutT* _out_scores,                                                                      \
-    uint32_t* _out_indices);
-
-#define COMMA ,
-instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(
-  half,
-  cuvs::neighbors::ivf_pq::detail::fp_8bit<5u COMMA false>,
-  cuvs::neighbors::filtering::ivf_to_sample_filter<
-    int64_t COMMA cuvs::neighbors::filtering::none_ivf_sample_filter>);
-
-#undef COMMA
-
-#undef instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select
diff --git a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_fp8_true.cu b/cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_fp8_true.cu
deleted file mode 100644
index 1556cf31e..000000000
--- a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_fp8_true.cu
+++ /dev/null
@@ -1,81 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_pq_compute_similarity_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_pq_compute_similarity_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/ivf_pq_compute_similarity-inl.cuh>
-#include <cuvs/neighbors/detail/ivf_pq_fp_8bit.cuh>
-
-#define instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(                 \
-  OutT, LutT, IvfSampleFilterT)                                                             \
-  template auto                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_select<OutT, LutT, IvfSampleFilterT>( \
-    const cudaDeviceProp& dev_props,                                                        \
-    bool manage_local_topk,                                                                 \
-    int locality_hint,                                                                      \
-    double preferred_shmem_carveout,                                                        \
-    uint32_t pq_bits,                                                                       \
-    uint32_t pq_dim,                                                                        \
-    uint32_t precomp_data_count,                                                            \
-    uint32_t n_queries,                                                                     \
-    uint32_t n_probes,                                                                      \
-    uint32_t topk)                                                                          \
-    ->cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT>;              \
-                                                                                            \
-  template void                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_run<OutT, LutT, IvfSampleFilterT>(    \
-    cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT> s,              \
-    rmm::cuda_stream_view stream,                                                           \
-    uint32_t dim,                                                                           \
-    uint32_t n_probes,                                                                      \
-    uint32_t pq_dim,                                                                        \
-    uint32_t n_queries,                                                                     \
-    uint32_t queries_offset,                                                                \
-    cuvs::distance::DistanceType metric,                                                    \
-    cuvs::neighbors::ivf_pq::codebook_gen codebook_kind,                                    \
-    uint32_t topk,                                                                          \
-    uint32_t max_samples,                                                                   \
-    const float* cluster_centers,                                                           \
-    const float* pq_centers,                                                                \
-    const uint8_t* const* pq_dataset,                                                       \
-    const uint32_t* cluster_labels,                                                         \
-    const uint32_t* _chunk_indices,                                                         \
-    const float* queries,                                                                   \
-    const uint32_t* index_list,                                                             \
-    float* query_kths,                                                                      \
-    IvfSampleFilterT sample_filter,                                                         \
-    LutT* lut_scores,                                                                       \
-    OutT* _out_scores,                                                                      \
-    uint32_t* _out_indices);
-
-#define COMMA ,
-instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(
-  half,
-  cuvs::neighbors::ivf_pq::detail::fp_8bit<5u COMMA true>,
-  cuvs::neighbors::filtering::ivf_to_sample_filter<
-    int64_t COMMA cuvs::neighbors::filtering::none_ivf_sample_filter>);
-
-#undef COMMA
-
-#undef instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select
diff --git a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_half.cu b/cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_half.cu
deleted file mode 100644
index ab9454f5f..000000000
--- a/cpp/src/neighbors/detail/ivf_pq_compute_similarity_half_half.cu
+++ /dev/null
@@ -1,81 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_pq_compute_similarity_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_pq_compute_similarity_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/ivf_pq_compute_similarity-inl.cuh>
-#include <cuvs/neighbors/detail/ivf_pq_fp_8bit.cuh>
-
-#define instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(                 \
-  OutT, LutT, IvfSampleFilterT)                                                             \
-  template auto                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_select<OutT, LutT, IvfSampleFilterT>( \
-    const cudaDeviceProp& dev_props,                                                        \
-    bool manage_local_topk,                                                                 \
-    int locality_hint,                                                                      \
-    double preferred_shmem_carveout,                                                        \
-    uint32_t pq_bits,                                                                       \
-    uint32_t pq_dim,                                                                        \
-    uint32_t precomp_data_count,                                                            \
-    uint32_t n_queries,                                                                     \
-    uint32_t n_probes,                                                                      \
-    uint32_t topk)                                                                          \
-    ->cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT>;              \
-                                                                                            \
-  template void                                                                             \
-  cuvs::neighbors::ivf_pq::detail::compute_similarity_run<OutT, LutT, IvfSampleFilterT>(    \
-    cuvs::neighbors::ivf_pq::detail::selected<OutT, LutT, IvfSampleFilterT> s,              \
-    rmm::cuda_stream_view stream,                                                           \
-    uint32_t dim,                                                                           \
-    uint32_t n_probes,                                                                      \
-    uint32_t pq_dim,                                                                        \
-    uint32_t n_queries,                                                                     \
-    uint32_t queries_offset,                                                                \
-    cuvs::distance::DistanceType metric,                                                    \
-    cuvs::neighbors::ivf_pq::codebook_gen codebook_kind,                                    \
-    uint32_t topk,                                                                          \
-    uint32_t max_samples,                                                                   \
-    const float* cluster_centers,                                                           \
-    const float* pq_centers,                                                                \
-    const uint8_t* const* pq_dataset,                                                       \
-    const uint32_t* cluster_labels,                                                         \
-    const uint32_t* _chunk_indices,                                                         \
-    const float* queries,                                                                   \
-    const uint32_t* index_list,                                                             \
-    float* query_kths,                                                                      \
-    IvfSampleFilterT sample_filter,                                                         \
-    LutT* lut_scores,                                                                       \
-    OutT* _out_scores,                                                                      \
-    uint32_t* _out_indices);
-
-#define COMMA ,
-instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select(
-  half,
-  half,
-  cuvs::neighbors::filtering::ivf_to_sample_filter<
-    int64_t COMMA cuvs::neighbors::filtering::none_ivf_sample_filter>);
-
-#undef COMMA
-
-#undef instantiate_raft_neighbors_ivf_pq_detail_compute_similarity_select
diff --git a/cpp/src/neighbors/detail/refine_host_float_float.cpp b/cpp/src/neighbors/detail/refine_host_float_float.cpp
deleted file mode 100644
index ceea544a0..000000000
--- a/cpp/src/neighbors/detail/refine_host_float_float.cpp
+++ /dev/null
@@ -1,29 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include <cuvs/neighbors/detail/refine_host-inl.hpp>
-
-#define instantiate_raft_neighbors_refine(IdxT, DataT, DistanceT, ExtentsT)             \
-  template void cuvs::neighbors::detail::refine_host<IdxT, DataT, DistanceT, ExtentsT>( \
-    raft::host_matrix_view<const DataT, ExtentsT, raft::row_major> dataset,             \
-    raft::host_matrix_view<const DataT, ExtentsT, raft::row_major> queries,             \
-    raft::host_matrix_view<const IdxT, ExtentsT, raft::row_major> neighbor_candidates,  \
-    raft::host_matrix_view<IdxT, ExtentsT, raft::row_major> indices,                    \
-    raft::host_matrix_view<DistanceT, ExtentsT, raft::row_major> distances,             \
-    distance::DistanceType metric);
-
-instantiate_raft_neighbors_refine(int64_t, float, float, int64_t);
-
-#undef instantiate_raft_neighbors_refine
diff --git a/cpp/src/neighbors/detail/refine_host_int8_t_float.cpp b/cpp/src/neighbors/detail/refine_host_int8_t_float.cpp
deleted file mode 100644
index c236740b4..000000000
--- a/cpp/src/neighbors/detail/refine_host_int8_t_float.cpp
+++ /dev/null
@@ -1,29 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/detail/refine_host-inl.hpp>
-
-#define instantiate_raft_neighbors_refine(IdxT, DataT, DistanceT, ExtentsT)             \
-  template void cuvs::neighbors::detail::refine_host<IdxT, DataT, DistanceT, ExtentsT>( \
-    raft::host_matrix_view<const DataT, ExtentsT, raft::row_major> dataset,             \
-    raft::host_matrix_view<const DataT, ExtentsT, raft::row_major> queries,             \
-    raft::host_matrix_view<const IdxT, ExtentsT, raft::row_major> neighbor_candidates,  \
-    raft::host_matrix_view<IdxT, ExtentsT, raft::row_major> indices,                    \
-    raft::host_matrix_view<DistanceT, ExtentsT, raft::row_major> distances,             \
-    distance::DistanceType metric);
-instantiate_raft_neighbors_refine(int64_t, int8_t, float, int64_t);
-
-#undef instantiate_raft_neighbors_refine
diff --git a/cpp/src/neighbors/detail/refine_host_uint8_t_float.cpp b/cpp/src/neighbors/detail/refine_host_uint8_t_float.cpp
deleted file mode 100644
index 5aee0f29e..000000000
--- a/cpp/src/neighbors/detail/refine_host_uint8_t_float.cpp
+++ /dev/null
@@ -1,30 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/detail/refine_host-inl.hpp>
-
-#define instantiate_raft_neighbors_refine(IdxT, DataT, DistanceT, ExtentsT)             \
-  template void cuvs::neighbors::detail::refine_host<IdxT, DataT, DistanceT, ExtentsT>( \
-    raft::host_matrix_view<const DataT, ExtentsT, raft::row_major> dataset,             \
-    raft::host_matrix_view<const DataT, ExtentsT, raft::row_major> queries,             \
-    raft::host_matrix_view<const IdxT, ExtentsT, raft::row_major> neighbor_candidates,  \
-    raft::host_matrix_view<IdxT, ExtentsT, raft::row_major> indices,                    \
-    raft::host_matrix_view<DistanceT, ExtentsT, raft::row_major> distances,             \
-    distance::DistanceType metric);
-
-instantiate_raft_neighbors_refine(int64_t, uint8_t, float, int64_t);
-
-#undef instantiate_raft_neighbors_refine
diff --git a/cpp/src/neighbors/detail/selection_faiss_00_generate.py b/cpp/src/neighbors/detail/selection_faiss_00_generate.py
deleted file mode 100644
index b6a410d34..000000000
--- a/cpp/src/neighbors/detail/selection_faiss_00_generate.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-header = """
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by selection_faiss_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python selection_faiss_00_generate.py
- *
- */
-
-#include <cstddef>  // size_t
-#include <cstdint>  // uint32_t
-#include <cuvs/neighbors/detail/selection_faiss-inl.cuh>
-
-#define instantiate_raft_neighbors_detail_select_k(payload_t, key_t)    \\
-  template void cuvs::neighbors::detail::select_k(const key_t* inK,     \\
-                                                  const payload_t* inV, \\
-                                                  size_t n_rows,        \\
-                                                  size_t n_cols,        \\
-                                                  key_t* outK,          \\
-                                                  payload_t* outV,      \\
-                                                  bool select_min,      \\
-                                                  int k,                \\
-                                                  cudaStream_t stream)
-
-"""
-
-types = dict(
-    uint32_t_float=("uint32_t", "float"),
-    uint32_t_double=("uint32_t", "double"),
-    uint32_t_half=("uint32_t", "half"),
-    int64_t_double=("int64_t", "double"),
-    int64_t_half=("int64_t", "half"),
-    int32_t_float=("int32_t", "float"),
-    long_float=("long", "float"),
-    size_t_double=("size_t", "double"),
-    int_double=("int", "double"),
-    size_t_float=("size_t", "float"),
-)
-
-for type_path, (payload_t, key_t) in types.items():
-    path = f"selection_faiss_{type_path}.cu"
-    with open(path, "w") as f:
-        f.write(header)
-        f.write(f"instantiate_raft_neighbors_detail_select_k({payload_t}, {key_t});\n\n")
-        f.write(f"#undef instantiate_raft_neighbors_detail_select_k\n")
-
-    # for pasting into CMakeLists.txt
-    print(f"src/neighbors/detail/{path}")
diff --git a/cpp/src/neighbors/detail/selection_faiss_int32_t_float.cu b/cpp/src/neighbors/detail/selection_faiss_int32_t_float.cu
deleted file mode 100644
index 69b039f46..000000000
--- a/cpp/src/neighbors/detail/selection_faiss_int32_t_float.cu
+++ /dev/null
@@ -1,44 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by selection_faiss_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python selection_faiss_00_generate.py
- *
- */
-
-#include <cstddef>  // size_t
-#include <cstdint>  // uint32_t
-#include <cuvs/neighbors/detail/selection_faiss-inl.cuh>
-
-#define instantiate_raft_neighbors_detail_select_k(payload_t, key_t)    \
-  template void cuvs::neighbors::detail::select_k(const key_t* inK,     \
-                                                  const payload_t* inV, \
-                                                  size_t n_rows,        \
-                                                  size_t n_cols,        \
-                                                  key_t* outK,          \
-                                                  payload_t* outV,      \
-                                                  bool select_min,      \
-                                                  int k,                \
-                                                  cudaStream_t stream)
-
-instantiate_raft_neighbors_detail_select_k(int32_t, float);
-
-#undef instantiate_raft_neighbors_detail_select_k
diff --git a/cpp/src/neighbors/detail/selection_faiss_int64_t_double.cu b/cpp/src/neighbors/detail/selection_faiss_int64_t_double.cu
deleted file mode 100644
index cf11f65cd..000000000
--- a/cpp/src/neighbors/detail/selection_faiss_int64_t_double.cu
+++ /dev/null
@@ -1,44 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by selection_faiss_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python selection_faiss_00_generate.py
- *
- */
-
-#include <cstddef>  // size_t
-#include <cstdint>  // uint32_t
-#include <cuvs/neighbors/detail/selection_faiss-inl.cuh>
-
-#define instantiate_raft_neighbors_detail_select_k(payload_t, key_t)    \
-  template void cuvs::neighbors::detail::select_k(const key_t* inK,     \
-                                                  const payload_t* inV, \
-                                                  size_t n_rows,        \
-                                                  size_t n_cols,        \
-                                                  key_t* outK,          \
-                                                  payload_t* outV,      \
-                                                  bool select_min,      \
-                                                  int k,                \
-                                                  cudaStream_t stream)
-
-instantiate_raft_neighbors_detail_select_k(int64_t, double);
-
-#undef instantiate_raft_neighbors_detail_select_k
diff --git a/cpp/src/neighbors/detail/selection_faiss_int64_t_half.cu b/cpp/src/neighbors/detail/selection_faiss_int64_t_half.cu
deleted file mode 100644
index d6cc4826b..000000000
--- a/cpp/src/neighbors/detail/selection_faiss_int64_t_half.cu
+++ /dev/null
@@ -1,44 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by selection_faiss_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python selection_faiss_00_generate.py
- *
- */
-
-#include <cstddef>  // size_t
-#include <cstdint>  // uint32_t
-#include <cuvs/neighbors/detail/selection_faiss-inl.cuh>
-
-#define instantiate_raft_neighbors_detail_select_k(payload_t, key_t)    \
-  template void cuvs::neighbors::detail::select_k(const key_t* inK,     \
-                                                  const payload_t* inV, \
-                                                  size_t n_rows,        \
-                                                  size_t n_cols,        \
-                                                  key_t* outK,          \
-                                                  payload_t* outV,      \
-                                                  bool select_min,      \
-                                                  int k,                \
-                                                  cudaStream_t stream)
-
-instantiate_raft_neighbors_detail_select_k(int64_t, half);
-
-#undef instantiate_raft_neighbors_detail_select_k
diff --git a/cpp/src/neighbors/detail/selection_faiss_int_double.cu b/cpp/src/neighbors/detail/selection_faiss_int_double.cu
deleted file mode 100644
index 6819eff13..000000000
--- a/cpp/src/neighbors/detail/selection_faiss_int_double.cu
+++ /dev/null
@@ -1,44 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by selection_faiss_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python selection_faiss_00_generate.py
- *
- */
-
-#include <cstddef>  // size_t
-#include <cstdint>  // uint32_t
-#include <cuvs/neighbors/detail/selection_faiss-inl.cuh>
-
-#define instantiate_raft_neighbors_detail_select_k(payload_t, key_t)    \
-  template void cuvs::neighbors::detail::select_k(const key_t* inK,     \
-                                                  const payload_t* inV, \
-                                                  size_t n_rows,        \
-                                                  size_t n_cols,        \
-                                                  key_t* outK,          \
-                                                  payload_t* outV,      \
-                                                  bool select_min,      \
-                                                  int k,                \
-                                                  cudaStream_t stream)
-
-instantiate_raft_neighbors_detail_select_k(int, double);
-
-#undef instantiate_raft_neighbors_detail_select_k
diff --git a/cpp/src/neighbors/detail/selection_faiss_long_float.cu b/cpp/src/neighbors/detail/selection_faiss_long_float.cu
deleted file mode 100644
index 512ee028c..000000000
--- a/cpp/src/neighbors/detail/selection_faiss_long_float.cu
+++ /dev/null
@@ -1,44 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by selection_faiss_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python selection_faiss_00_generate.py
- *
- */
-
-#include <cstddef>  // size_t
-#include <cstdint>  // uint32_t
-#include <cuvs/neighbors/detail/selection_faiss-inl.cuh>
-
-#define instantiate_raft_neighbors_detail_select_k(payload_t, key_t)    \
-  template void cuvs::neighbors::detail::select_k(const key_t* inK,     \
-                                                  const payload_t* inV, \
-                                                  size_t n_rows,        \
-                                                  size_t n_cols,        \
-                                                  key_t* outK,          \
-                                                  payload_t* outV,      \
-                                                  bool select_min,      \
-                                                  int k,                \
-                                                  cudaStream_t stream)
-
-instantiate_raft_neighbors_detail_select_k(long, float);
-
-#undef instantiate_raft_neighbors_detail_select_k
diff --git a/cpp/src/neighbors/detail/selection_faiss_size_t_double.cu b/cpp/src/neighbors/detail/selection_faiss_size_t_double.cu
deleted file mode 100644
index 5311ca084..000000000
--- a/cpp/src/neighbors/detail/selection_faiss_size_t_double.cu
+++ /dev/null
@@ -1,44 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by selection_faiss_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python selection_faiss_00_generate.py
- *
- */
-
-#include <cstddef>  // size_t
-#include <cstdint>  // uint32_t
-#include <cuvs/neighbors/detail/selection_faiss-inl.cuh>
-
-#define instantiate_raft_neighbors_detail_select_k(payload_t, key_t)    \
-  template void cuvs::neighbors::detail::select_k(const key_t* inK,     \
-                                                  const payload_t* inV, \
-                                                  size_t n_rows,        \
-                                                  size_t n_cols,        \
-                                                  key_t* outK,          \
-                                                  payload_t* outV,      \
-                                                  bool select_min,      \
-                                                  int k,                \
-                                                  cudaStream_t stream)
-
-instantiate_raft_neighbors_detail_select_k(size_t, double);
-
-#undef instantiate_raft_neighbors_detail_select_k
diff --git a/cpp/src/neighbors/detail/selection_faiss_size_t_float.cu b/cpp/src/neighbors/detail/selection_faiss_size_t_float.cu
deleted file mode 100644
index d69826848..000000000
--- a/cpp/src/neighbors/detail/selection_faiss_size_t_float.cu
+++ /dev/null
@@ -1,44 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by selection_faiss_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python selection_faiss_00_generate.py
- *
- */
-
-#include <cstddef>  // size_t
-#include <cstdint>  // uint32_t
-#include <cuvs/neighbors/detail/selection_faiss-inl.cuh>
-
-#define instantiate_raft_neighbors_detail_select_k(payload_t, key_t)    \
-  template void cuvs::neighbors::detail::select_k(const key_t* inK,     \
-                                                  const payload_t* inV, \
-                                                  size_t n_rows,        \
-                                                  size_t n_cols,        \
-                                                  key_t* outK,          \
-                                                  payload_t* outV,      \
-                                                  bool select_min,      \
-                                                  int k,                \
-                                                  cudaStream_t stream)
-
-instantiate_raft_neighbors_detail_select_k(size_t, float);
-
-#undef instantiate_raft_neighbors_detail_select_k
diff --git a/cpp/src/neighbors/detail/selection_faiss_uint32_t_double.cu b/cpp/src/neighbors/detail/selection_faiss_uint32_t_double.cu
deleted file mode 100644
index 3cb34cabe..000000000
--- a/cpp/src/neighbors/detail/selection_faiss_uint32_t_double.cu
+++ /dev/null
@@ -1,44 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by selection_faiss_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python selection_faiss_00_generate.py
- *
- */
-
-#include <cstddef>  // size_t
-#include <cstdint>  // uint32_t
-#include <cuvs/neighbors/detail/selection_faiss-inl.cuh>
-
-#define instantiate_raft_neighbors_detail_select_k(payload_t, key_t)    \
-  template void cuvs::neighbors::detail::select_k(const key_t* inK,     \
-                                                  const payload_t* inV, \
-                                                  size_t n_rows,        \
-                                                  size_t n_cols,        \
-                                                  key_t* outK,          \
-                                                  payload_t* outV,      \
-                                                  bool select_min,      \
-                                                  int k,                \
-                                                  cudaStream_t stream)
-
-instantiate_raft_neighbors_detail_select_k(uint32_t, double);
-
-#undef instantiate_raft_neighbors_detail_select_k
diff --git a/cpp/src/neighbors/detail/selection_faiss_uint32_t_float.cu b/cpp/src/neighbors/detail/selection_faiss_uint32_t_float.cu
deleted file mode 100644
index e50b852f5..000000000
--- a/cpp/src/neighbors/detail/selection_faiss_uint32_t_float.cu
+++ /dev/null
@@ -1,44 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by selection_faiss_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python selection_faiss_00_generate.py
- *
- */
-
-#include <cstddef>  // size_t
-#include <cstdint>  // uint32_t
-#include <cuvs/neighbors/detail/selection_faiss-inl.cuh>
-
-#define instantiate_raft_neighbors_detail_select_k(payload_t, key_t)    \
-  template void cuvs::neighbors::detail::select_k(const key_t* inK,     \
-                                                  const payload_t* inV, \
-                                                  size_t n_rows,        \
-                                                  size_t n_cols,        \
-                                                  key_t* outK,          \
-                                                  payload_t* outV,      \
-                                                  bool select_min,      \
-                                                  int k,                \
-                                                  cudaStream_t stream)
-
-instantiate_raft_neighbors_detail_select_k(uint32_t, float);
-
-#undef instantiate_raft_neighbors_detail_select_k
diff --git a/cpp/src/neighbors/detail/selection_faiss_uint32_t_half.cu b/cpp/src/neighbors/detail/selection_faiss_uint32_t_half.cu
deleted file mode 100644
index ae85b97ae..000000000
--- a/cpp/src/neighbors/detail/selection_faiss_uint32_t_half.cu
+++ /dev/null
@@ -1,44 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by selection_faiss_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python selection_faiss_00_generate.py
- *
- */
-
-#include <cstddef>  // size_t
-#include <cstdint>  // uint32_t
-#include <cuvs/neighbors/detail/selection_faiss-inl.cuh>
-
-#define instantiate_raft_neighbors_detail_select_k(payload_t, key_t)    \
-  template void cuvs::neighbors::detail::select_k(const key_t* inK,     \
-                                                  const payload_t* inV, \
-                                                  size_t n_rows,        \
-                                                  size_t n_cols,        \
-                                                  key_t* outK,          \
-                                                  payload_t* outV,      \
-                                                  bool select_min,      \
-                                                  int k,                \
-                                                  cudaStream_t stream)
-
-instantiate_raft_neighbors_detail_select_k(uint32_t, half);
-
-#undef instantiate_raft_neighbors_detail_select_k
diff --git a/cpp/src/neighbors/ivf_flat_00_generate.py b/cpp/src/neighbors/ivf_flat_00_generate.py
deleted file mode 100644
index dc7d7374f..000000000
--- a/cpp/src/neighbors/ivf_flat_00_generate.py
+++ /dev/null
@@ -1,148 +0,0 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-header = """/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_flat_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_flat_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_flat-inl.cuh>
-"""
-
-types = dict(
-    float_int64_t= ("float", "int64_t"),
-    int8_t_int64_t=("int8_t", "int64_t"),
-    uint8_t_int64_t=("uint8_t", "int64_t"),
-)
-
-build_macro = """
-#define instantiate_raft_neighbors_ivf_flat_build(T, IdxT)        \\
-  template auto cuvs::neighbors::ivf_flat::build<T, IdxT>( \\
-    raft::resources const& handle,                         \\
-    const cuvs::neighbors::ivf_flat::index_params& params,        \\
-    const T* dataset,                                             \\
-    IdxT n_rows,                                                  \\
-    uint32_t dim)                                                 \\
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                  \\
-                                                                  \\
-  template auto cuvs::neighbors::ivf_flat::build<T, IdxT>( \\
-    raft::resources const& handle,                         \\
-    const cuvs::neighbors::ivf_flat::index_params& params,        \\
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset)   \\
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                  \\
-                                                                  \\
-  template void cuvs::neighbors::ivf_flat::build<T, IdxT>( \\
-    raft::resources const& handle,                         \\
-    const cuvs::neighbors::ivf_flat::index_params& params,        \\
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset,   \\
-    cuvs::neighbors::ivf_flat::index<T, IdxT>& idx);
-"""
-
-extend_macro = """
-#define instantiate_raft_neighbors_ivf_flat_extend(T, IdxT)                \\
-  template auto cuvs::neighbors::ivf_flat::extend<T, IdxT>(         \\
-    raft::resources const& handle,                                  \\
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index,           \\
-    const T* new_vectors,                                                  \\
-    const IdxT* new_indices,                                               \\
-    IdxT n_rows)                                                           \\
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                           \\
-                                                                           \\
-  template auto cuvs::neighbors::ivf_flat::extend<T, IdxT>(         \\
-    raft::resources const& handle,                                  \\
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,        \\
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \\
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index)           \\
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                           \\
-                                                                           \\
-  template void cuvs::neighbors::ivf_flat::extend<T, IdxT>(         \\
-    raft::resources const& handle,                                  \\
-    cuvs::neighbors::ivf_flat::index<T, IdxT>* index,                      \\
-    const T* new_vectors,                                                  \\
-    const IdxT* new_indices,                                               \\
-    IdxT n_rows);                                                          \\
-                                                                           \\
-  template void cuvs::neighbors::ivf_flat::extend<T, IdxT>(         \\
-    raft::resources const& handle,                                  \\
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,        \\
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \\
-    cuvs::neighbors::ivf_flat::index<T, IdxT>* index);
-"""
-
-search_macro = """
-#define instantiate_raft_neighbors_ivf_flat_search(T, IdxT)        \\
-  template void cuvs::neighbors::ivf_flat::search<T, IdxT>( \\
-    raft::resources const& handle,                          \\
-    const cuvs::neighbors::ivf_flat::search_params& params,        \\
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,        \\
-    const T* queries,                                              \\
-    uint32_t n_queries,                                            \\
-    uint32_t k,                                                    \\
-    IdxT* neighbors,                                               \\
-    float* distances,                                              \\
-    rmm::mr::device_memory_resource* mr );                         \\
-                                                                   \\
-  template void cuvs::neighbors::ivf_flat::search<T, IdxT>( \\
-    raft::resources const& handle,                          \\
-    const cuvs::neighbors::ivf_flat::search_params& params,        \\
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,        \\
-    raft::device_matrix_view<const T, IdxT, raft::row_major> queries,    \\
-    raft::device_matrix_view<IdxT, IdxT, raft::row_major> neighbors,     \\
-    raft::device_matrix_view<float, IdxT, raft::row_major> distances);
-"""
-
-macros = dict(
-    build=dict(
-        definition=build_macro,
-        name="instantiate_raft_neighbors_ivf_flat_build"),
-    extend=dict(
-        definition=extend_macro,
-        name="instantiate_raft_neighbors_ivf_flat_extend"),
-    search=dict(
-        definition=search_macro,
-        name="instantiate_raft_neighbors_ivf_flat_search"),
-)
-
-for type_path, (T, IdxT) in types.items():
-    for macro_path, macro in macros.items():
-        path = f"ivf_flat_{macro_path}_{type_path}.cu"
-        with open(path, "w") as f:
-            f.write(header)
-            f.write(macro['definition'])
-
-
-            f.write(f"{macro['name']}({T}, {IdxT});\n\n")
-            f.write(f"#undef {macro['name']}\n")
-
-        print(f"src/neighbors/{path}")
diff --git a/cpp/src/neighbors/ivf_flat_build_float_int64_t.cu b/cpp/src/neighbors/ivf_flat_build_float_int64_t.cu
deleted file mode 100644
index 794e435bd..000000000
--- a/cpp/src/neighbors/ivf_flat_build_float_int64_t.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_flat_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_flat_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_flat-inl.cuh>
-
-#define instantiate_raft_neighbors_ivf_flat_build(T, IdxT)            \
-  template auto cuvs::neighbors::ivf_flat::build<T, IdxT>(            \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    const T* dataset,                                                 \
-    IdxT n_rows,                                                      \
-    uint32_t dim)                                                     \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                      \
-                                                                      \
-  template auto cuvs::neighbors::ivf_flat::build<T, IdxT>(            \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset) \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                      \
-                                                                      \
-  template void cuvs::neighbors::ivf_flat::build<T, IdxT>(            \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset, \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>& idx);
-instantiate_raft_neighbors_ivf_flat_build(float, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_flat_build
diff --git a/cpp/src/neighbors/ivf_flat_build_int8_t_int64_t.cu b/cpp/src/neighbors/ivf_flat_build_int8_t_int64_t.cu
deleted file mode 100644
index a2e9bd828..000000000
--- a/cpp/src/neighbors/ivf_flat_build_int8_t_int64_t.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_flat_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_flat_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_flat-inl.cuh>
-
-#define instantiate_raft_neighbors_ivf_flat_build(T, IdxT)            \
-  template auto cuvs::neighbors::ivf_flat::build<T, IdxT>(            \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    const T* dataset,                                                 \
-    IdxT n_rows,                                                      \
-    uint32_t dim)                                                     \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                      \
-                                                                      \
-  template auto cuvs::neighbors::ivf_flat::build<T, IdxT>(            \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset) \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                      \
-                                                                      \
-  template void cuvs::neighbors::ivf_flat::build<T, IdxT>(            \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset, \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>& idx);
-instantiate_raft_neighbors_ivf_flat_build(int8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_flat_build
diff --git a/cpp/src/neighbors/ivf_flat_build_uint8_t_int64_t.cu b/cpp/src/neighbors/ivf_flat_build_uint8_t_int64_t.cu
deleted file mode 100644
index 51d855374..000000000
--- a/cpp/src/neighbors/ivf_flat_build_uint8_t_int64_t.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_flat_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_flat_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_flat-inl.cuh>
-
-#define instantiate_raft_neighbors_ivf_flat_build(T, IdxT)            \
-  template auto cuvs::neighbors::ivf_flat::build<T, IdxT>(            \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    const T* dataset,                                                 \
-    IdxT n_rows,                                                      \
-    uint32_t dim)                                                     \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                      \
-                                                                      \
-  template auto cuvs::neighbors::ivf_flat::build<T, IdxT>(            \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset) \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                      \
-                                                                      \
-  template void cuvs::neighbors::ivf_flat::build<T, IdxT>(            \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::index_params& params,            \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset, \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>& idx);
-instantiate_raft_neighbors_ivf_flat_build(uint8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_flat_build
diff --git a/cpp/src/neighbors/ivf_flat_extend_float_int64_t.cu b/cpp/src/neighbors/ivf_flat_extend_float_int64_t.cu
deleted file mode 100644
index 2e825938c..000000000
--- a/cpp/src/neighbors/ivf_flat_extend_float_int64_t.cu
+++ /dev/null
@@ -1,58 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_flat_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_flat_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_flat-inl.cuh>
-
-#define instantiate_raft_neighbors_ivf_flat_extend(T, IdxT)                \
-  template auto cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index,           \
-    const T* new_vectors,                                                  \
-    const IdxT* new_indices,                                               \
-    IdxT n_rows)                                                           \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                           \
-                                                                           \
-  template auto cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,  \
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index)           \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                           \
-                                                                           \
-  template void cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>* index,                      \
-    const T* new_vectors,                                                  \
-    const IdxT* new_indices,                                               \
-    IdxT n_rows);                                                          \
-                                                                           \
-  template void cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,  \
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>* index);
-instantiate_raft_neighbors_ivf_flat_extend(float, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_flat_extend
diff --git a/cpp/src/neighbors/ivf_flat_extend_int8_t_int64_t.cu b/cpp/src/neighbors/ivf_flat_extend_int8_t_int64_t.cu
deleted file mode 100644
index fa11b4472..000000000
--- a/cpp/src/neighbors/ivf_flat_extend_int8_t_int64_t.cu
+++ /dev/null
@@ -1,58 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_flat_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_flat_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_flat-inl.cuh>
-
-#define instantiate_raft_neighbors_ivf_flat_extend(T, IdxT)                \
-  template auto cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index,           \
-    const T* new_vectors,                                                  \
-    const IdxT* new_indices,                                               \
-    IdxT n_rows)                                                           \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                           \
-                                                                           \
-  template auto cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,  \
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index)           \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                           \
-                                                                           \
-  template void cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>* index,                      \
-    const T* new_vectors,                                                  \
-    const IdxT* new_indices,                                               \
-    IdxT n_rows);                                                          \
-                                                                           \
-  template void cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,  \
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>* index);
-instantiate_raft_neighbors_ivf_flat_extend(int8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_flat_extend
diff --git a/cpp/src/neighbors/ivf_flat_extend_uint8_t_int64_t.cu b/cpp/src/neighbors/ivf_flat_extend_uint8_t_int64_t.cu
deleted file mode 100644
index 8e4e24349..000000000
--- a/cpp/src/neighbors/ivf_flat_extend_uint8_t_int64_t.cu
+++ /dev/null
@@ -1,58 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_flat_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_flat_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_flat-inl.cuh>
-
-#define instantiate_raft_neighbors_ivf_flat_extend(T, IdxT)                \
-  template auto cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index,           \
-    const T* new_vectors,                                                  \
-    const IdxT* new_indices,                                               \
-    IdxT n_rows)                                                           \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                           \
-                                                                           \
-  template auto cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,  \
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& orig_index)           \
-    ->cuvs::neighbors::ivf_flat::index<T, IdxT>;                           \
-                                                                           \
-  template void cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>* index,                      \
-    const T* new_vectors,                                                  \
-    const IdxT* new_indices,                                               \
-    IdxT n_rows);                                                          \
-                                                                           \
-  template void cuvs::neighbors::ivf_flat::extend<T, IdxT>(                \
-    raft::resources const& handle,                                         \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,  \
-    std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, \
-    cuvs::neighbors::ivf_flat::index<T, IdxT>* index);
-instantiate_raft_neighbors_ivf_flat_extend(uint8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_flat_extend
diff --git a/cpp/src/neighbors/ivf_flat_search_float_int64_t.cu b/cpp/src/neighbors/ivf_flat_search_float_int64_t.cu
deleted file mode 100644
index d0072c0ff..000000000
--- a/cpp/src/neighbors/ivf_flat_search_float_int64_t.cu
+++ /dev/null
@@ -1,49 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_flat_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_flat_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_flat-inl.cuh>
-
-#define instantiate_raft_neighbors_ivf_flat_search(T, IdxT)           \
-  template void cuvs::neighbors::ivf_flat::search<T, IdxT>(           \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::search_params& params,           \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,           \
-    const T* queries,                                                 \
-    uint32_t n_queries,                                               \
-    uint32_t k,                                                       \
-    IdxT* neighbors,                                                  \
-    float* distances,                                                 \
-    rmm::mr::device_memory_resource* mr);                             \
-                                                                      \
-  template void cuvs::neighbors::ivf_flat::search<T, IdxT>(           \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::search_params& params,           \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,           \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> queries, \
-    raft::device_matrix_view<IdxT, IdxT, raft::row_major> neighbors,  \
-    raft::device_matrix_view<float, IdxT, raft::row_major> distances);
-instantiate_raft_neighbors_ivf_flat_search(float, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_flat_search
diff --git a/cpp/src/neighbors/ivf_flat_search_int8_t_int64_t.cu b/cpp/src/neighbors/ivf_flat_search_int8_t_int64_t.cu
deleted file mode 100644
index ecdc2a2ed..000000000
--- a/cpp/src/neighbors/ivf_flat_search_int8_t_int64_t.cu
+++ /dev/null
@@ -1,49 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_flat_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_flat_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_flat-inl.cuh>
-
-#define instantiate_raft_neighbors_ivf_flat_search(T, IdxT)           \
-  template void cuvs::neighbors::ivf_flat::search<T, IdxT>(           \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::search_params& params,           \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,           \
-    const T* queries,                                                 \
-    uint32_t n_queries,                                               \
-    uint32_t k,                                                       \
-    IdxT* neighbors,                                                  \
-    float* distances,                                                 \
-    rmm::mr::device_memory_resource* mr);                             \
-                                                                      \
-  template void cuvs::neighbors::ivf_flat::search<T, IdxT>(           \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::search_params& params,           \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,           \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> queries, \
-    raft::device_matrix_view<IdxT, IdxT, raft::row_major> neighbors,  \
-    raft::device_matrix_view<float, IdxT, raft::row_major> distances);
-instantiate_raft_neighbors_ivf_flat_search(int8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_flat_search
diff --git a/cpp/src/neighbors/ivf_flat_search_uint8_t_int64_t.cu b/cpp/src/neighbors/ivf_flat_search_uint8_t_int64_t.cu
deleted file mode 100644
index a87486004..000000000
--- a/cpp/src/neighbors/ivf_flat_search_uint8_t_int64_t.cu
+++ /dev/null
@@ -1,49 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by ivf_flat_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python ivf_flat_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_flat-inl.cuh>
-
-#define instantiate_raft_neighbors_ivf_flat_search(T, IdxT)           \
-  template void cuvs::neighbors::ivf_flat::search<T, IdxT>(           \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::search_params& params,           \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,           \
-    const T* queries,                                                 \
-    uint32_t n_queries,                                               \
-    uint32_t k,                                                       \
-    IdxT* neighbors,                                                  \
-    float* distances,                                                 \
-    rmm::mr::device_memory_resource* mr);                             \
-                                                                      \
-  template void cuvs::neighbors::ivf_flat::search<T, IdxT>(           \
-    raft::resources const& handle,                                    \
-    const cuvs::neighbors::ivf_flat::search_params& params,           \
-    const cuvs::neighbors::ivf_flat::index<T, IdxT>& index,           \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> queries, \
-    raft::device_matrix_view<IdxT, IdxT, raft::row_major> neighbors,  \
-    raft::device_matrix_view<float, IdxT, raft::row_major> distances);
-instantiate_raft_neighbors_ivf_flat_search(uint8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_flat_search
diff --git a/cpp/src/neighbors/ivfpq_build_float_int64_t.cu b/cpp/src/neighbors/ivfpq_build_float_int64_t.cu
deleted file mode 100644
index 672bbe732..000000000
--- a/cpp/src/neighbors/ivfpq_build_float_int64_t.cu
+++ /dev/null
@@ -1,36 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq-inl.cuh>
-#include <cuvs/neighbors/ivf_pq_types.hpp>  // cuvs::neighbors::ivf_pq::index
-
-#define instantiate_raft_neighbors_ivf_pq_build(T, IdxT)                                 \
-  template cuvs::neighbors::ivf_pq::index<IdxT> cuvs::neighbors::ivf_pq::build<T, IdxT>( \
-    raft::resources const& handle,                                                       \
-    const cuvs::neighbors::ivf_pq::index_params& params,                                 \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset);                   \
-                                                                                         \
-  template auto cuvs::neighbors::ivf_pq::build(                                          \
-    raft::resources const& handle,                                                       \
-    const cuvs::neighbors::ivf_pq::index_params& params,                                 \
-    const T* dataset,                                                                    \
-    IdxT n_rows,                                                                         \
-    uint32_t dim)                                                                        \
-    ->cuvs::neighbors::ivf_pq::index<IdxT>;
-
-instantiate_raft_neighbors_ivf_pq_build(float, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_build
diff --git a/cpp/src/neighbors/ivfpq_build_int8_t_int64_t.cu b/cpp/src/neighbors/ivfpq_build_int8_t_int64_t.cu
deleted file mode 100644
index f3d3a4e5f..000000000
--- a/cpp/src/neighbors/ivfpq_build_int8_t_int64_t.cu
+++ /dev/null
@@ -1,36 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq-inl.cuh>
-#include <cuvs/neighbors/ivf_pq_types.hpp>  // cuvs::neighbors::ivf_pq::index
-
-#define instantiate_raft_neighbors_ivf_pq_build(T, IdxT)                                 \
-  template cuvs::neighbors::ivf_pq::index<IdxT> cuvs::neighbors::ivf_pq::build<T, IdxT>( \
-    raft::resources const& handle,                                                       \
-    const cuvs::neighbors::ivf_pq::index_params& params,                                 \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset);                   \
-                                                                                         \
-  template auto cuvs::neighbors::ivf_pq::build(                                          \
-    raft::resources const& handle,                                                       \
-    const cuvs::neighbors::ivf_pq::index_params& params,                                 \
-    const T* dataset,                                                                    \
-    IdxT n_rows,                                                                         \
-    uint32_t dim)                                                                        \
-    ->cuvs::neighbors::ivf_pq::index<IdxT>;
-
-instantiate_raft_neighbors_ivf_pq_build(int8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_build
diff --git a/cpp/src/neighbors/ivfpq_build_uint8_t_int64_t.cu b/cpp/src/neighbors/ivfpq_build_uint8_t_int64_t.cu
deleted file mode 100644
index ffd630b1e..000000000
--- a/cpp/src/neighbors/ivfpq_build_uint8_t_int64_t.cu
+++ /dev/null
@@ -1,36 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq-inl.cuh>
-#include <cuvs/neighbors/ivf_pq_types.hpp>  // cuvs::neighbors::ivf_pq::index
-
-#define instantiate_raft_neighbors_ivf_pq_build(T, IdxT)                                 \
-  template cuvs::neighbors::ivf_pq::index<IdxT> cuvs::neighbors::ivf_pq::build<T, IdxT>( \
-    raft::resources const& handle,                                                       \
-    const cuvs::neighbors::ivf_pq::index_params& params,                                 \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> dataset);                   \
-                                                                                         \
-  template auto cuvs::neighbors::ivf_pq::build(                                          \
-    raft::resources const& handle,                                                       \
-    const cuvs::neighbors::ivf_pq::index_params& params,                                 \
-    const T* dataset,                                                                    \
-    IdxT n_rows,                                                                         \
-    uint32_t dim)                                                                        \
-    ->cuvs::neighbors::ivf_pq::index<IdxT>;
-
-instantiate_raft_neighbors_ivf_pq_build(uint8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_build
diff --git a/cpp/src/neighbors/ivfpq_extend_float_int64_t.cu b/cpp/src/neighbors/ivfpq_extend_float_int64_t.cu
deleted file mode 100644
index 59385c8b7..000000000
--- a/cpp/src/neighbors/ivfpq_extend_float_int64_t.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq-inl.cuh>
-#include <cuvs/neighbors/ivf_pq_types.hpp>  // cuvs::neighbors::ivf_pq::index
-
-#define instantiate_raft_neighbors_ivf_pq_extend(T, IdxT)                                   \
-  template cuvs::neighbors::ivf_pq::index<IdxT> cuvs::neighbors::ivf_pq::extend<T, IdxT>(   \
-    raft::resources const& handle,                                                          \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,                   \
-    std::optional<raft::device_vector_view<const IdxT, IdxT, raft::row_major>> new_indices, \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx);                                       \
-                                                                                            \
-  template void cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                   \
-    raft::resources const& handle,                                                          \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,                   \
-    std::optional<raft::device_vector_view<const IdxT, IdxT, raft::row_major>> new_indices, \
-    cuvs::neighbors::ivf_pq::index<IdxT>* idx);                                             \
-                                                                                            \
-  template auto cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                   \
-    raft::resources const& handle,                                                          \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                                        \
-    const T* new_vectors,                                                                   \
-    const IdxT* new_indices,                                                                \
-    IdxT n_rows)                                                                            \
-    ->cuvs::neighbors::ivf_pq::index<IdxT>;                                                 \
-                                                                                            \
-  template void cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                   \
-    raft::resources const& handle,                                                          \
-    cuvs::neighbors::ivf_pq::index<IdxT>* idx,                                              \
-    const T* new_vectors,                                                                   \
-    const IdxT* new_indices,                                                                \
-    IdxT n_rows);
-
-instantiate_raft_neighbors_ivf_pq_extend(float, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_extend
diff --git a/cpp/src/neighbors/ivfpq_extend_int8_t_int64_t.cu b/cpp/src/neighbors/ivfpq_extend_int8_t_int64_t.cu
deleted file mode 100644
index 7fad247d1..000000000
--- a/cpp/src/neighbors/ivfpq_extend_int8_t_int64_t.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq-inl.cuh>
-#include <cuvs/neighbors/ivf_pq_types.hpp>  // cuvs::neighbors::ivf_pq::index
-
-#define instantiate_raft_neighbors_ivf_pq_extend(T, IdxT)                                   \
-  template cuvs::neighbors::ivf_pq::index<IdxT> cuvs::neighbors::ivf_pq::extend<T, IdxT>(   \
-    raft::resources const& handle,                                                          \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,                   \
-    std::optional<raft::device_vector_view<const IdxT, IdxT, raft::row_major>> new_indices, \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx);                                       \
-                                                                                            \
-  template void cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                   \
-    raft::resources const& handle,                                                          \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,                   \
-    std::optional<raft::device_vector_view<const IdxT, IdxT, raft::row_major>> new_indices, \
-    cuvs::neighbors::ivf_pq::index<IdxT>* idx);                                             \
-                                                                                            \
-  template auto cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                   \
-    raft::resources const& handle,                                                          \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                                        \
-    const T* new_vectors,                                                                   \
-    const IdxT* new_indices,                                                                \
-    IdxT n_rows)                                                                            \
-    ->cuvs::neighbors::ivf_pq::index<IdxT>;                                                 \
-                                                                                            \
-  template void cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                   \
-    raft::resources const& handle,                                                          \
-    cuvs::neighbors::ivf_pq::index<IdxT>* idx,                                              \
-    const T* new_vectors,                                                                   \
-    const IdxT* new_indices,                                                                \
-    IdxT n_rows);
-
-instantiate_raft_neighbors_ivf_pq_extend(int8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_extend
diff --git a/cpp/src/neighbors/ivfpq_extend_uint8_t_int64_t.cu b/cpp/src/neighbors/ivfpq_extend_uint8_t_int64_t.cu
deleted file mode 100644
index 7bbdd5ffd..000000000
--- a/cpp/src/neighbors/ivfpq_extend_uint8_t_int64_t.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq-inl.cuh>
-#include <cuvs/neighbors/ivf_pq_types.hpp>  // cuvs::neighbors::ivf_pq::index
-
-#define instantiate_raft_neighbors_ivf_pq_extend(T, IdxT)                                   \
-  template cuvs::neighbors::ivf_pq::index<IdxT> cuvs::neighbors::ivf_pq::extend<T, IdxT>(   \
-    raft::resources const& handle,                                                          \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,                   \
-    std::optional<raft::device_vector_view<const IdxT, IdxT, raft::row_major>> new_indices, \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx);                                       \
-                                                                                            \
-  template void cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                   \
-    raft::resources const& handle,                                                          \
-    raft::device_matrix_view<const T, IdxT, raft::row_major> new_vectors,                   \
-    std::optional<raft::device_vector_view<const IdxT, IdxT, raft::row_major>> new_indices, \
-    cuvs::neighbors::ivf_pq::index<IdxT>* idx);                                             \
-                                                                                            \
-  template auto cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                   \
-    raft::resources const& handle,                                                          \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                                        \
-    const T* new_vectors,                                                                   \
-    const IdxT* new_indices,                                                                \
-    IdxT n_rows)                                                                            \
-    ->cuvs::neighbors::ivf_pq::index<IdxT>;                                                 \
-                                                                                            \
-  template void cuvs::neighbors::ivf_pq::extend<T, IdxT>(                                   \
-    raft::resources const& handle,                                                          \
-    cuvs::neighbors::ivf_pq::index<IdxT>* idx,                                              \
-    const T* new_vectors,                                                                   \
-    const IdxT* new_indices,                                                                \
-    IdxT n_rows);
-
-instantiate_raft_neighbors_ivf_pq_extend(uint8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_extend
diff --git a/cpp/src/neighbors/ivfpq_search_float_int64_t.cu b/cpp/src/neighbors/ivfpq_search_float_int64_t.cu
deleted file mode 100644
index 31ce6e8df..000000000
--- a/cpp/src/neighbors/ivfpq_search_float_int64_t.cu
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq-inl.cuh>
-#include <cuvs/neighbors/ivf_pq_types.hpp>  // cuvs::neighbors::ivf_pq::index
-
-#define instantiate_raft_neighbors_ivf_pq_search(T, IdxT)                  \
-  template void cuvs::neighbors::ivf_pq::search<T, IdxT>(                  \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_pq::search_params& params,                  \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                       \
-    raft::device_matrix_view<const T, uint32_t, raft::row_major> queries,  \
-    raft::device_matrix_view<IdxT, uint32_t, raft::row_major> neighbors,   \
-    raft::device_matrix_view<float, uint32_t, raft::row_major> distances); \
-                                                                           \
-  template void cuvs::neighbors::ivf_pq::search<T, IdxT>(                  \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_pq::search_params& params,                  \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                       \
-    const T* queries,                                                      \
-    uint32_t n_queries,                                                    \
-    uint32_t k,                                                            \
-    IdxT* neighbors,                                                       \
-    float* distances,                                                      \
-    rmm::mr::device_memory_resource* mr)
-
-instantiate_raft_neighbors_ivf_pq_search(float, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_search
diff --git a/cpp/src/neighbors/ivfpq_search_int8_t_int64_t.cu b/cpp/src/neighbors/ivfpq_search_int8_t_int64_t.cu
deleted file mode 100644
index 5b50b3b19..000000000
--- a/cpp/src/neighbors/ivfpq_search_int8_t_int64_t.cu
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq-inl.cuh>
-#include <cuvs/neighbors/ivf_pq_types.hpp>  // cuvs::neighbors::ivf_pq::index
-
-#define instantiate_raft_neighbors_ivf_pq_search(T, IdxT)                  \
-  template void cuvs::neighbors::ivf_pq::search<T, IdxT>(                  \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_pq::search_params& params,                  \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                       \
-    raft::device_matrix_view<const T, uint32_t, raft::row_major> queries,  \
-    raft::device_matrix_view<IdxT, uint32_t, raft::row_major> neighbors,   \
-    raft::device_matrix_view<float, uint32_t, raft::row_major> distances); \
-                                                                           \
-  template void cuvs::neighbors::ivf_pq::search<T, IdxT>(                  \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_pq::search_params& params,                  \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                       \
-    const T* queries,                                                      \
-    uint32_t n_queries,                                                    \
-    uint32_t k,                                                            \
-    IdxT* neighbors,                                                       \
-    float* distances,                                                      \
-    rmm::mr::device_memory_resource* mr)
-
-instantiate_raft_neighbors_ivf_pq_search(int8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_search
diff --git a/cpp/src/neighbors/ivfpq_search_uint8_t_int64_t.cu b/cpp/src/neighbors/ivfpq_search_uint8_t_int64_t.cu
deleted file mode 100644
index 100c3b49e..000000000
--- a/cpp/src/neighbors/ivfpq_search_uint8_t_int64_t.cu
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/neighbors/ivf_pq-inl.cuh>
-#include <cuvs/neighbors/ivf_pq_types.hpp>  // cuvs::neighbors::ivf_pq::index
-
-#define instantiate_raft_neighbors_ivf_pq_search(T, IdxT)                  \
-  template void cuvs::neighbors::ivf_pq::search<T, IdxT>(                  \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_pq::search_params& params,                  \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                       \
-    raft::device_matrix_view<const T, uint32_t, raft::row_major> queries,  \
-    raft::device_matrix_view<IdxT, uint32_t, raft::row_major> neighbors,   \
-    raft::device_matrix_view<float, uint32_t, raft::row_major> distances); \
-                                                                           \
-  template void cuvs::neighbors::ivf_pq::search<T, IdxT>(                  \
-    raft::resources const& handle,                                         \
-    const cuvs::neighbors::ivf_pq::search_params& params,                  \
-    const cuvs::neighbors::ivf_pq::index<IdxT>& idx,                       \
-    const T* queries,                                                      \
-    uint32_t n_queries,                                                    \
-    uint32_t k,                                                            \
-    IdxT* neighbors,                                                       \
-    float* distances,                                                      \
-    rmm::mr::device_memory_resource* mr)
-
-instantiate_raft_neighbors_ivf_pq_search(uint8_t, int64_t);
-
-#undef instantiate_raft_neighbors_ivf_pq_search
diff --git a/cpp/src/neighbors/refine_00_generate.py b/cpp/src/neighbors/refine_00_generate.py
deleted file mode 100644
index 45cfc473e..000000000
--- a/cpp/src/neighbors/refine_00_generate.py
+++ /dev/null
@@ -1,78 +0,0 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-header = """
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by refine_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python refine_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/refine-inl.cuh>
-
-#define instantiate_raft_neighbors_refine(idx_t, data_t, distance_t, matrix_idx)       \\
-  template void cuvs::neighbors::refine<idx_t, data_t, distance_t, matrix_idx>(        \\
-    raft::resources const& handle,                                              \\
-    raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,             \\
-    raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> queries,             \\
-    raft::device_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates,  \\
-    raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                    \\
-    raft::device_matrix_view<distance_t, matrix_idx, raft::row_major> distances,             \\
-    cuvs::distance::DistanceType metric);                                              \\
-                                                                                       \\
-  template void cuvs::neighbors::refine<idx_t, data_t, distance_t, matrix_idx>(        \\
-    raft::resources const& handle,                                              \\
-    raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,               \\
-    raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> queries,               \\
-    raft::host_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates,    \\
-    raft::host_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                      \\
-    raft::host_matrix_view<distance_t, matrix_idx, raft::row_major> distances,               \\
-    cuvs::distance::DistanceType metric);
-
-"""
-
-types = dict(
-    float_float= ("float", "float"),
-    int8_t_float=("int8_t", "float"),
-    uint8_t_float=("uint8_t", "float"),
-)
-
-for type_path, (data_t, distance_t) in types.items():
-    path = f"refine_{type_path}.cu"
-    with open(path, "w") as f:
-        f.write(header)
-        f.write(f"instantiate_raft_neighbors_refine(int64_t, {data_t}, {distance_t}, int64_t);\n\n")
-        f.write(f"#undef instantiate_raft_neighbors_refine\n")
-
-    # for pasting into CMakeLists.txt
-    print(f"src/neighbors/{path}")
diff --git a/cpp/src/neighbors/refine_float_float.cu b/cpp/src/neighbors/refine_float_float.cu
deleted file mode 100644
index 39d51c1bc..000000000
--- a/cpp/src/neighbors/refine_float_float.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by refine_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python refine_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/refine-inl.cuh>
-
-#define instantiate_raft_neighbors_refine(idx_t, data_t, distance_t, matrix_idx)            \
-  template void cuvs::neighbors::refine<idx_t, data_t, distance_t, matrix_idx>(             \
-    raft::resources const& handle,                                                          \
-    raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,            \
-    raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> queries,            \
-    raft::device_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates, \
-    raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                   \
-    raft::device_matrix_view<distance_t, matrix_idx, raft::row_major> distances,            \
-    cuvs::distance::DistanceType metric);                                                   \
-                                                                                            \
-  template void cuvs::neighbors::refine<idx_t, data_t, distance_t, matrix_idx>(             \
-    raft::resources const& handle,                                                          \
-    raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,              \
-    raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> queries,              \
-    raft::host_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates,   \
-    raft::host_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                     \
-    raft::host_matrix_view<distance_t, matrix_idx, raft::row_major> distances,              \
-    cuvs::distance::DistanceType metric);
-
-instantiate_raft_neighbors_refine(int64_t, float, float, int64_t);
-
-#undef instantiate_raft_neighbors_refine
diff --git a/cpp/src/neighbors/refine_int8_t_float.cu b/cpp/src/neighbors/refine_int8_t_float.cu
deleted file mode 100644
index f65590a4f..000000000
--- a/cpp/src/neighbors/refine_int8_t_float.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by refine_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python refine_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/refine-inl.cuh>
-
-#define instantiate_raft_neighbors_refine(idx_t, data_t, distance_t, matrix_idx)            \
-  template void cuvs::neighbors::refine<idx_t, data_t, distance_t, matrix_idx>(             \
-    raft::resources const& handle,                                                          \
-    raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,            \
-    raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> queries,            \
-    raft::device_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates, \
-    raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                   \
-    raft::device_matrix_view<distance_t, matrix_idx, raft::row_major> distances,            \
-    cuvs::distance::DistanceType metric);                                                   \
-                                                                                            \
-  template void cuvs::neighbors::refine<idx_t, data_t, distance_t, matrix_idx>(             \
-    raft::resources const& handle,                                                          \
-    raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,              \
-    raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> queries,              \
-    raft::host_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates,   \
-    raft::host_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                     \
-    raft::host_matrix_view<distance_t, matrix_idx, raft::row_major> distances,              \
-    cuvs::distance::DistanceType metric);
-
-instantiate_raft_neighbors_refine(int64_t, int8_t, float, int64_t);
-
-#undef instantiate_raft_neighbors_refine
diff --git a/cpp/src/neighbors/refine_uint8_t_float.cu b/cpp/src/neighbors/refine_uint8_t_float.cu
deleted file mode 100644
index ba76e4904..000000000
--- a/cpp/src/neighbors/refine_uint8_t_float.cu
+++ /dev/null
@@ -1,50 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by refine_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python refine_00_generate.py
- *
- */
-
-#include <cuvs/neighbors/refine-inl.cuh>
-
-#define instantiate_raft_neighbors_refine(idx_t, data_t, distance_t, matrix_idx)            \
-  template void cuvs::neighbors::refine<idx_t, data_t, distance_t, matrix_idx>(             \
-    raft::resources const& handle,                                                          \
-    raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,            \
-    raft::device_matrix_view<const data_t, matrix_idx, raft::row_major> queries,            \
-    raft::device_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates, \
-    raft::device_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                   \
-    raft::device_matrix_view<distance_t, matrix_idx, raft::row_major> distances,            \
-    cuvs::distance::DistanceType metric);                                                   \
-                                                                                            \
-  template void cuvs::neighbors::refine<idx_t, data_t, distance_t, matrix_idx>(             \
-    raft::resources const& handle,                                                          \
-    raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> dataset,              \
-    raft::host_matrix_view<const data_t, matrix_idx, raft::row_major> queries,              \
-    raft::host_matrix_view<const idx_t, matrix_idx, raft::row_major> neighbor_candidates,   \
-    raft::host_matrix_view<idx_t, matrix_idx, raft::row_major> indices,                     \
-    raft::host_matrix_view<distance_t, matrix_idx, raft::row_major> distances,              \
-    cuvs::distance::DistanceType metric);
-
-instantiate_raft_neighbors_refine(int64_t, uint8_t, float, int64_t);
-
-#undef instantiate_raft_neighbors_refine
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers.cu b/cpp/src/spatial/knn/detail/ball_cover/registers.cu
deleted file mode 100644
index b01c9014d..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers.cu
+++ /dev/null
@@ -1,60 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims)                                                   \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_one<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    cuvs::spatial::knn::detail::DistFunc<Mvalue_t, Mvalue_int>& dfunc,                       \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims)                                                   \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_two<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    cuvs::spatial::knn::detail::DistFunc<Mvalue_t, Mvalue_int>& dfunc,                       \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(std::int64_t, float, std::uint32_t, 2);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(std::int64_t, float, std::uint32_t, 3);
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(std::int64_t, float, std::uint32_t, 2);
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(std::int64_t, float, std::uint32_t, 3);
-
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_00_generate.py b/cpp/src/spatial/knn/detail/ball_cover/registers_00_generate.py
deleted file mode 100644
index 807c6a32d..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_00_generate.py
+++ /dev/null
@@ -1,112 +0,0 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-header = """/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint> // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-"""
-
-
-macro_pass_one = """
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(                            \\
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \\
-  template void                                                                       \\
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_one<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \\
-    raft::resources const& handle,                                                    \\
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \\
-    const Mvalue_t* query,                                                                   \\
-    const Mvalue_int n_query_rows,                                                           \\
-    Mvalue_int k,                                                                            \\
-    const Mvalue_idx* R_knn_inds,                                                            \\
-    const Mvalue_t* R_knn_dists,                                                             \\
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \\
-    Mvalue_idx* inds,                                                                        \\
-    Mvalue_t* dists,                                                                         \\
-    float weight,                                                                            \\
-    Mvalue_int* dists_counter)
-
-"""
-
-macro_pass_two = """
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(                            \\
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \\
-  template void                                                                       \\
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_two<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \\
-    raft::resources const& handle,                                                    \\
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \\
-    const Mvalue_t* query,                                                                   \\
-    const Mvalue_int n_query_rows,                                                           \\
-    Mvalue_int k,                                                                            \\
-    const Mvalue_idx* R_knn_inds,                                                            \\
-    const Mvalue_t* R_knn_dists,                                                             \\
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \\
-    Mvalue_idx* inds,                                                                        \\
-    Mvalue_t* dists,                                                                         \\
-    float weight,                                                                            \\
-    Mvalue_int* dists_counter)
-
-"""
-
-distances = dict(
-    haversine="cuvs::spatial::knn::detail::HaversineFunc",
-    euclidean="cuvs::spatial::knn::detail::EuclideanFunc",
-    dist="cuvs::spatial::knn::detail::DistFunc",
-)
-
-for k, v in distances.items():
-    for dim in [2, 3]:
-        path = f"registers_pass_one_{dim}d_{k}.cu"
-        with open(path, "w") as f:
-            f.write(header)
-            f.write(macro_pass_one)
-            f.write(f"instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(\n")
-            f.write(f"  std::int64_t, float, std::uint32_t, {dim}, {v});\n")
-            f.write("#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one\n")
-        print(f"src/spatial/knn/detail/ball_cover/{path}")
-
-for k, v in distances.items():
-    for dim in [2, 3]:
-        path = f"registers_pass_two_{dim}d_{k}.cu"
-        with open(path, "w") as f:
-            f.write(header)
-            f.write(macro_pass_two)
-            f.write(f"instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(\n")
-            f.write(f"  std::int64_t, float, std::uint32_t, {dim}, {v});\n")
-            f.write("#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two\n")
-        print(f"src/spatial/knn/detail/ball_cover/{path}")
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_dist.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_dist.cu
deleted file mode 100644
index ac7230a4b..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_dist.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_one<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::DistFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_euclidean.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_euclidean.cu
deleted file mode 100644
index da5c5604c..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_euclidean.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_one<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::EuclideanFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_haversine.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_haversine.cu
deleted file mode 100644
index 8af080bcf..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_2d_haversine.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_one<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::HaversineFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_dist.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_dist.cu
deleted file mode 100644
index 74a555f15..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_dist.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_one<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::DistFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_euclidean.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_euclidean.cu
deleted file mode 100644
index 274a5241a..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_euclidean.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_one<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::EuclideanFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_haversine.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_haversine.cu
deleted file mode 100644
index e881822c4..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_one_3d_haversine.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_one<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::HaversineFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_one
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_dist.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_dist.cu
deleted file mode 100644
index 988aa01ac..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_dist.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_two<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::DistFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_euclidean.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_euclidean.cu
deleted file mode 100644
index a2d0a423f..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_euclidean.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_two<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::EuclideanFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_haversine.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_haversine.cu
deleted file mode 100644
index 8a331a43c..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_2d_haversine.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_two<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 2, cuvs::spatial::knn::detail::HaversineFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_dist.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_dist.cu
deleted file mode 100644
index 861ca8a92..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_dist.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_two<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::DistFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_euclidean.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_euclidean.cu
deleted file mode 100644
index d41512189..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_euclidean.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_two<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::EuclideanFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two
diff --git a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_haversine.cu b/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_haversine.cu
deleted file mode 100644
index 98d00f022..000000000
--- a/cpp/src/spatial/knn/detail/ball_cover/registers_pass_two_3d_haversine.cu
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by registers_00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python registers_00_generate.py
- *
- */
-
-#include <cstdint>  // int64_t
-#include <cuvs/spatial/knn/detail/ball_cover/registers-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(                            \
-  Mvalue_idx, Mvalue_t, Mvalue_int, Mdims, Mdist_func)                                       \
-  template void                                                                              \
-  cuvs::spatial::knn::detail::rbc_low_dim_pass_two<Mvalue_idx, Mvalue_t, Mvalue_int, Mdims>( \
-    raft::resources const& handle,                                                           \
-    const BallCoverIndex<Mvalue_idx, Mvalue_t, Mvalue_int>& index,                           \
-    const Mvalue_t* query,                                                                   \
-    const Mvalue_int n_query_rows,                                                           \
-    Mvalue_int k,                                                                            \
-    const Mvalue_idx* R_knn_inds,                                                            \
-    const Mvalue_t* R_knn_dists,                                                             \
-    Mdist_func<Mvalue_t, Mvalue_int>& dfunc,                                                 \
-    Mvalue_idx* inds,                                                                        \
-    Mvalue_t* dists,                                                                         \
-    float weight,                                                                            \
-    Mvalue_int* dists_counter)
-
-instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two(
-  std::int64_t, float, std::uint32_t, 3, cuvs::spatial::knn::detail::HaversineFunc);
-#undef instantiate_raft_spatial_knn_detail_rbc_low_dim_pass_two
diff --git a/cpp/src/spatial/knn/detail/fused_l2_knn_int32_t_float.cu b/cpp/src/spatial/knn/detail/fused_l2_knn_int32_t_float.cu
deleted file mode 100644
index 456d2fac5..000000000
--- a/cpp/src/spatial/knn/detail/fused_l2_knn_int32_t_float.cu
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cstddef>                           // size_t
-#include <cstdint>                           // int_Xt
-#include <cuvs/distance/distance_types.hpp>  // DistanceType
-#include <cuvs/spatial/knn/detail/fused_l2_knn-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_fusedL2Knn(Mvalue_idx, Mvalue_t, MusePrevTopKs)  \
-  template void cuvs::spatial::knn::detail::fusedL2Knn<Mvalue_idx, Mvalue_t, MusePrevTopKs>( \
-    size_t D,                                                                                \
-    Mvalue_idx * out_inds,                                                                   \
-    Mvalue_t * out_dists,                                                                    \
-    const Mvalue_t* index,                                                                   \
-    const Mvalue_t* query,                                                                   \
-    size_t n_index_rows,                                                                     \
-    size_t n_query_rows,                                                                     \
-    int k,                                                                                   \
-    bool rowMajorIndex,                                                                      \
-    bool rowMajorQuery,                                                                      \
-    cudaStream_t stream,                                                                     \
-    cuvs::distance::DistanceType metric,                                                     \
-    const Mvalue_t* index_norms,                                                             \
-    const Mvalue_t* query_norms)
-
-instantiate_raft_spatial_knn_detail_fusedL2Knn(int32_t, float, true);
-instantiate_raft_spatial_knn_detail_fusedL2Knn(int32_t, float, false);
-
-#undef instantiate_raft_spatial_knn_detail_fusedL2Knn
diff --git a/cpp/src/spatial/knn/detail/fused_l2_knn_int64_t_float.cu b/cpp/src/spatial/knn/detail/fused_l2_knn_int64_t_float.cu
deleted file mode 100644
index 332428efb..000000000
--- a/cpp/src/spatial/knn/detail/fused_l2_knn_int64_t_float.cu
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cstddef>                           // size_t
-#include <cstdint>                           // int_Xt
-#include <cuvs/distance/distance_types.hpp>  // DistanceType
-#include <cuvs/spatial/knn/detail/fused_l2_knn-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_fusedL2Knn(Mvalue_idx, Mvalue_t, MusePrevTopKs)  \
-  template void cuvs::spatial::knn::detail::fusedL2Knn<Mvalue_idx, Mvalue_t, MusePrevTopKs>( \
-    size_t D,                                                                                \
-    Mvalue_idx * out_inds,                                                                   \
-    Mvalue_t * out_dists,                                                                    \
-    const Mvalue_t* index,                                                                   \
-    const Mvalue_t* query,                                                                   \
-    size_t n_index_rows,                                                                     \
-    size_t n_query_rows,                                                                     \
-    int k,                                                                                   \
-    bool rowMajorIndex,                                                                      \
-    bool rowMajorQuery,                                                                      \
-    cudaStream_t stream,                                                                     \
-    cuvs::distance::DistanceType metric,                                                     \
-    const Mvalue_t* index_norms,                                                             \
-    const Mvalue_t* query_norms)
-
-instantiate_raft_spatial_knn_detail_fusedL2Knn(int64_t, float, true);
-instantiate_raft_spatial_knn_detail_fusedL2Knn(int64_t, float, false);
-
-#undef instantiate_raft_spatial_knn_detail_fusedL2Knn
diff --git a/cpp/src/spatial/knn/detail/fused_l2_knn_uint32_t_float.cu b/cpp/src/spatial/knn/detail/fused_l2_knn_uint32_t_float.cu
deleted file mode 100644
index f4e4d7a49..000000000
--- a/cpp/src/spatial/knn/detail/fused_l2_knn_uint32_t_float.cu
+++ /dev/null
@@ -1,43 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cstddef>                           // size_t
-#include <cstdint>                           // int_Xt
-#include <cuvs/distance/distance_types.hpp>  // DistanceType
-#include <cuvs/spatial/knn/detail/fused_l2_knn-inl.cuh>
-
-#define instantiate_raft_spatial_knn_detail_fusedL2Knn(Mvalue_idx, Mvalue_t, MusePrevTopKs)  \
-  template void cuvs::spatial::knn::detail::fusedL2Knn<Mvalue_idx, Mvalue_t, MusePrevTopKs>( \
-    size_t D,                                                                                \
-    Mvalue_idx * out_inds,                                                                   \
-    Mvalue_t * out_dists,                                                                    \
-    const Mvalue_t* index,                                                                   \
-    const Mvalue_t* query,                                                                   \
-    size_t n_index_rows,                                                                     \
-    size_t n_query_rows,                                                                     \
-    int k,                                                                                   \
-    bool rowMajorIndex,                                                                      \
-    bool rowMajorQuery,                                                                      \
-    cudaStream_t stream,                                                                     \
-    cuvs::distance::DistanceType metric,                                                     \
-    const Mvalue_t* index_norms,                                                             \
-    const Mvalue_t* query_norms)
-
-// These are used by brute_force_knn:
-instantiate_raft_spatial_knn_detail_fusedL2Knn(uint32_t, float, true);
-instantiate_raft_spatial_knn_detail_fusedL2Knn(uint32_t, float, false);
-
-#undef instantiate_raft_spatial_knn_detail_fusedL2Knn
diff --git a/cpp/template/src/cagra_example.cu b/cpp/template/src/cagra_example.cu
index 7bf854aa9..da58a9259 100644
--- a/cpp/template/src/cagra_example.cu
+++ b/cpp/template/src/cagra_example.cu
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
+ * Copyright (c) 2022-2024, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
diff --git a/cpp/test/CMakeLists.txt b/cpp/test/CMakeLists.txt
index 1914dfc81..559dc7384 100644
--- a/cpp/test/CMakeLists.txt
+++ b/cpp/test/CMakeLists.txt
@@ -84,67 +84,7 @@ endfunction()
 # test sources ##################################################################################
 # ##################################################################################################
 
-# ##################################################################################################
-# * distance tests -------------------------------------------------------------------------
-
 if(BUILD_TESTS)
-  ConfigureTest(
-    NAME
-    CLUSTER_TEST
-    PATH
-    test/cluster/kmeans.cu
-    test/cluster/kmeans_balanced.cu
-    test/cluster/cluster_solvers.cu
-    test/cluster/linkage.cu
-    test/cluster/kmeans_find_k.cu
-    LIB
-    EXPLICIT_INSTANTIATE_ONLY
-  )
-
-  ConfigureTest(
-    NAME
-    DISTANCE_TEST
-    PATH
-    test/distance/dist_adj.cu
-    test/distance/dist_adj_distance_instance.cu
-    test/distance/dist_canberra.cu
-    test/distance/dist_correlation.cu
-    test/distance/dist_cos.cu
-    test/distance/dist_hamming.cu
-    test/distance/dist_hellinger.cu
-    test/distance/dist_inner_product.cu
-    test/distance/dist_jensen_shannon.cu
-    test/distance/dist_kl_divergence.cu
-    test/distance/dist_l1.cu
-    test/distance/dist_l2_exp.cu
-    test/distance/dist_l2_unexp.cu
-    test/distance/dist_l2_sqrt_exp.cu
-    test/distance/dist_l_inf.cu
-    test/distance/dist_lp_unexp.cu
-    test/distance/dist_russell_rao.cu
-    test/distance/masked_nn.cu
-    test/distance/masked_nn_compress_to_bits.cu
-    test/distance/fused_l2_nn.cu
-    test/distance/gram.cu
-    LIB
-    EXPLICIT_INSTANTIATE_ONLY
-  )
-
-  ConfigureTest(
-    NAME
-    NEIGHBORS_TEST
-    PATH
-    test/neighbors/knn.cu
-    test/neighbors/fused_l2_knn.cu
-    test/neighbors/tiled_knn.cu
-    test/neighbors/haversine.cu
-    test/neighbors/ball_cover.cu
-    test/neighbors/epsilon_neighborhood.cu
-    test/neighbors/refine.cu
-    LIB
-    EXPLICIT_INSTANTIATE_ONLY
-  )
-
   ConfigureTest(
     NAME
     NEIGHBORS_ANN_CAGRA_TEST
@@ -157,70 +97,6 @@ if(BUILD_TESTS)
     PERCENT
     100
   )
-
-  ConfigureTest(
-    NAME
-    NEIGHBORS_ANN_IVF_TEST
-    PATH
-    test/neighbors/ann_ivf_flat/test_filter_float_int64_t.cu
-    test/neighbors/ann_ivf_flat/test_float_int64_t.cu
-    test/neighbors/ann_ivf_flat/test_int8_t_int64_t.cu
-    test/neighbors/ann_ivf_flat/test_uint8_t_int64_t.cu
-    test/neighbors/ann_ivf_pq/test_float_uint32_t.cu
-    test/neighbors/ann_ivf_pq/test_float_int64_t.cu
-    test/neighbors/ann_ivf_pq/test_int8_t_int64_t.cu
-    test/neighbors/ann_ivf_pq/test_uint8_t_int64_t.cu
-    test/neighbors/ann_ivf_pq/test_filter_float_int64_t.cu
-    test/neighbors/ann_ivf_pq/test_filter_int8_t_int64_t.cu
-    LIB
-    EXPLICIT_INSTANTIATE_ONLY
-    GPUS
-    1
-    PERCENT
-    100
-  )
-
-  ConfigureTest(
-    NAME
-    NEIGHBORS_ANN_NN_DESCENT_TEST
-    PATH
-    test/neighbors/ann_nn_descent/test_float_uint32_t.cu
-    test/neighbors/ann_nn_descent/test_int8_t_uint32_t.cu
-    test/neighbors/ann_nn_descent/test_uint8_t_uint32_t.cu
-    LIB
-    EXPLICIT_INSTANTIATE_ONLY
-    GPUS
-    1
-    PERCENT
-    100
-  )
-
-  ConfigureTest(
-    NAME NEIGHBORS_SELECTION_TEST PATH test/neighbors/selection.cu LIB EXPLICIT_INSTANTIATE_ONLY
-    GPUS 1 PERCENT 50
-  )
-
-  ConfigureTest(
-    NAME
-    STATS_TEST
-    PATH
-    test/stats/contingencyMatrix.cu
-    test/stats/cov.cu
-    test/stats/dispersion.cu
-    test/stats/entropy.cu
-    test/stats/histogram.cu
-    test/stats/homogeneity_score.cu
-    test/stats/information_criterion.cu
-    test/stats/kl_divergence.cu
-    test/stats/mutual_info_score.cu
-    test/stats/neighborhood_recall.cu
-    test/stats/r2_score.cu
-    test/stats/rand_index.cu
-    test/stats/silhouette_score.cu
-    test/stats/trustworthiness.cu
-    LIB
-    EXPLICIT_INSTANTIATE_ONLY
-  )
 endif()
 
 # ##################################################################################################
diff --git a/cpp/test/cluster/cluster_solvers.cu b/cpp/test/cluster/cluster_solvers.cu
deleted file mode 100644
index b10a47a14..000000000
--- a/cpp/test/cluster/cluster_solvers.cu
+++ /dev/null
@@ -1,103 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <gtest/gtest.h>
-#include <iostream>
-#include <memory>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resource/device_id.hpp>
-#include <raft/core/resources.hpp>
-
-#include <cuvs/spectral/cluster_solvers.cuh>
-#include <cuvs/spectral/modularity_maximization.cuh>
-
-namespace cuvs {
-namespace spectral {
-
-TEST(Raft, ClusterSolvers)
-{
-  using namespace matrix;
-  using index_type = int;
-  using value_type = double;
-
-  raft::resources h;
-
-  index_type maxiter{100};
-  value_type tol{1.0e-10};
-  unsigned long long seed{100110021003};
-
-  auto stream = resource::get_cuda_stream(h);
-
-  index_type n{100};
-  index_type d{10};
-  index_type k{5};
-
-  // nullptr expected to trigger exceptions:
-  //
-  value_type* eigvecs{nullptr};
-  index_type* codes{nullptr};
-
-  cluster_solver_config_t<index_type, value_type> cfg{k, maxiter, tol, seed};
-
-  kmeans_solver_t<index_type, value_type> cluster_solver{cfg};
-
-  EXPECT_ANY_THROW(cluster_solver.solve(h, n, d, eigvecs, codes));
-}
-
-TEST(Raft, ModularitySolvers)
-{
-  using namespace matrix;
-  using index_type = int;
-  using value_type = double;
-
-  raft::resources h;
-  ASSERT_EQ(0, resource::get_device_id(h));
-
-  index_type neigvs{10};
-  index_type maxiter{100};
-  index_type restart_iter{10};
-  value_type tol{1.0e-10};
-  bool reorthog{true};
-
-  // nullptr expected to trigger exceptions:
-  //
-  index_type* clusters{nullptr};
-  value_type* eigvals{nullptr};
-  value_type* eigvecs{nullptr};
-
-  unsigned long long seed{100110021003};
-
-  eigen_solver_config_t<index_type, value_type> eig_cfg{
-    neigvs, maxiter, restart_iter, tol, reorthog, seed};
-  lanczos_solver_t<index_type, value_type> eig_solver{eig_cfg};
-
-  index_type k{5};
-
-  cluster_solver_config_t<index_type, value_type> clust_cfg{k, maxiter, tol, seed};
-  kmeans_solver_t<index_type, value_type> cluster_solver{clust_cfg};
-
-  auto stream = resource::get_cuda_stream(h);
-  sparse_matrix_t<index_type, value_type> sm{h, nullptr, nullptr, nullptr, 0, 0};
-
-  EXPECT_ANY_THROW(spectral::modularity_maximization(
-    h, sm, eig_solver, cluster_solver, clusters, eigvals, eigvecs));
-
-  value_type modularity{0};
-  EXPECT_ANY_THROW(spectral::analyzeModularity(h, sm, k, clusters, modularity));
-}
-
-}  // namespace spectral
-}  // namespace cuvs
diff --git a/cpp/test/cluster/kmeans.cu b/cpp/test/cluster/kmeans.cu
deleted file mode 100644
index a04d181a3..000000000
--- a/cpp/test/cluster/kmeans.cu
+++ /dev/null
@@ -1,359 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <optional>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <vector>
-
-#include <cuvs/cluster/kmeans.cuh>
-#include <cuvs/stats/adjusted_rand_index.cuh>
-#include <raft/core/cudart_utils.hpp>
-#include <raft/core/operators.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/random/make_blobs.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <rmm/device_uvector.hpp>
-#include <thrust/fill.h>
-
-namespace raft {
-
-template <typename T>
-struct KmeansInputs {
-  int n_row;
-  int n_col;
-  int n_clusters;
-  T tol;
-  bool weighted;
-};
-
-template <typename DataT, typename IndexT>
-void run_cluster_cost(const raft::resources& handle,
-                      raft::device_vector_view<DataT, IndexT> minClusterDistance,
-                      rmm::device_uvector<char>& workspace,
-                      raft::device_scalar_view<DataT> clusterCost)
-{
-  cuvs::cluster::kmeans::cluster_cost(
-    handle, minClusterDistance, workspace, clusterCost, raft::add_op{});
-}
-
-template <typename T>
-class KmeansTest : public ::testing::TestWithParam<KmeansInputs<T>> {
- protected:
-  KmeansTest()
-    : d_labels(0, resource::get_cuda_stream(handle)),
-      d_labels_ref(0, resource::get_cuda_stream(handle)),
-      d_centroids(0, resource::get_cuda_stream(handle)),
-      d_sample_weight(0, resource::get_cuda_stream(handle))
-  {
-  }
-
-  void apiTest()
-  {
-    testparams = ::testing::TestWithParam<KmeansInputs<T>>::GetParam();
-
-    auto stream                = resource::get_cuda_stream(handle);
-    int n_samples              = testparams.n_row;
-    int n_features             = testparams.n_col;
-    params.n_clusters          = testparams.n_clusters;
-    params.tol                 = testparams.tol;
-    params.n_init              = 1;
-    params.rng_state.seed      = 1;
-    params.oversampling_factor = 0;
-
-    raft::random::RngState rng(params.rng_state.seed, params.rng_state.type);
-
-    auto X      = raft::make_device_matrix<T, int>(handle, n_samples, n_features);
-    auto labels = raft::make_device_vector<int, int>(handle, n_samples);
-
-    raft::random::make_blobs<T, int>(X.data_handle(),
-                                     labels.data_handle(),
-                                     n_samples,
-                                     n_features,
-                                     params.n_clusters,
-                                     stream,
-                                     true,
-                                     nullptr,
-                                     nullptr,
-                                     T(1.0),
-                                     false,
-                                     (T)-10.0f,
-                                     (T)10.0f,
-                                     (uint64_t)1234);
-    d_labels.resize(n_samples, stream);
-    d_labels_ref.resize(n_samples, stream);
-    d_centroids.resize(params.n_clusters * n_features, stream);
-    raft::copy(d_labels_ref.data(), labels.data_handle(), n_samples, stream);
-    rmm::device_uvector<T> d_sample_weight(n_samples, stream);
-    thrust::fill(
-      thrust::cuda::par.on(stream), d_sample_weight.data(), d_sample_weight.data() + n_samples, 1);
-    auto weight_view =
-      raft::make_device_vector_view<const T, int>(d_sample_weight.data(), n_samples);
-
-    T inertia  = 0;
-    int n_iter = 0;
-    rmm::device_uvector<char> workspace(0, stream);
-    rmm::device_uvector<T> L2NormBuf_OR_DistBuf(0, stream);
-    rmm::device_uvector<T> inRankCp(0, stream);
-    auto X_view = raft::make_const_mdspan(X.view());
-    auto centroids_view =
-      raft::make_device_matrix_view<T, int>(d_centroids.data(), params.n_clusters, n_features);
-    auto miniX = raft::make_device_matrix<T, int>(handle, n_samples / 4, n_features);
-
-    // Initialize kmeans on a portion of X
-    cuvs::cluster::kmeans::shuffle_and_gather(
-      handle,
-      X_view,
-      raft::make_device_matrix_view<T, int>(miniX.data_handle(), miniX.extent(0), miniX.extent(1)),
-      miniX.extent(0),
-      params.rng_state.seed);
-
-    cuvs::cluster::kmeans::init_plus_plus(
-      handle, params, raft::make_const_mdspan(miniX.view()), centroids_view, workspace);
-
-    auto minClusterDistance = raft::make_device_vector<T, int>(handle, n_samples);
-    auto minClusterAndDistance =
-      raft::make_device_vector<raft::KeyValuePair<int, T>, int>(handle, n_samples);
-    auto L2NormX           = raft::make_device_vector<T, int>(handle, n_samples);
-    auto clusterCostBefore = raft::make_device_scalar<T>(handle, 0);
-    auto clusterCostAfter  = raft::make_device_scalar<T>(handle, 0);
-
-    raft::linalg::rowNorm(L2NormX.data_handle(),
-                          X.data_handle(),
-                          X.extent(1),
-                          X.extent(0),
-                          raft::linalg::L2Norm,
-                          true,
-                          stream);
-
-    cuvs::cluster::kmeans::min_cluster_distance(handle,
-                                                X_view,
-                                                centroids_view,
-                                                minClusterDistance.view(),
-                                                L2NormX.view(),
-                                                L2NormBuf_OR_DistBuf,
-                                                params.metric,
-                                                params.batch_samples,
-                                                params.batch_centroids,
-                                                workspace);
-
-    run_cluster_cost(handle, minClusterDistance.view(), workspace, clusterCostBefore.view());
-
-    // Run a fit of kmeans
-    cuvs::cluster::kmeans::fit_main(handle,
-                                    params,
-                                    X_view,
-                                    weight_view,
-                                    centroids_view,
-                                    raft::make_host_scalar_view(&inertia),
-                                    raft::make_host_scalar_view(&n_iter),
-                                    workspace);
-
-    // Check that the cluster cost decreased
-    cuvs::cluster::kmeans::min_cluster_distance(handle,
-                                                X_view,
-                                                centroids_view,
-                                                minClusterDistance.view(),
-                                                L2NormX.view(),
-                                                L2NormBuf_OR_DistBuf,
-                                                params.metric,
-                                                params.batch_samples,
-                                                params.batch_centroids,
-                                                workspace);
-
-    run_cluster_cost(handle, minClusterDistance.view(), workspace, clusterCostAfter.view());
-    T h_clusterCostBefore = T(0);
-    T h_clusterCostAfter  = T(0);
-    raft::update_host(&h_clusterCostBefore, clusterCostBefore.data_handle(), 1, stream);
-    raft::update_host(&h_clusterCostAfter, clusterCostAfter.data_handle(), 1, stream);
-    ASSERT_TRUE(h_clusterCostAfter < h_clusterCostBefore);
-
-    // Count samples in clusters using 2 methods and compare them
-    // Fill minClusterAndDistance
-    cuvs::cluster::kmeans::min_cluster_and_distance(
-      handle,
-      X_view,
-      raft::make_device_matrix_view<const T, int>(
-        d_centroids.data(), params.n_clusters, n_features),
-      minClusterAndDistance.view(),
-      L2NormX.view(),
-      L2NormBuf_OR_DistBuf,
-      params.metric,
-      params.batch_samples,
-      params.batch_centroids,
-      workspace);
-    cuvs::cluster::kmeans::KeyValueIndexOp<int, T> conversion_op;
-    cub::TransformInputIterator<int,
-                                cuvs::cluster::kmeans::KeyValueIndexOp<int, T>,
-                                raft::KeyValuePair<int, T>*>
-      itr(minClusterAndDistance.data_handle(), conversion_op);
-
-    auto sampleCountInCluster = raft::make_device_vector<T, int>(handle, params.n_clusters);
-    auto weigthInCluster      = raft::make_device_vector<T, int>(handle, params.n_clusters);
-    auto newCentroids = raft::make_device_matrix<T, int>(handle, params.n_clusters, n_features);
-    cuvs::cluster::kmeans::update_centroids(handle,
-                                            X_view,
-                                            weight_view,
-                                            raft::make_device_matrix_view<const T, int>(
-                                              d_centroids.data(), params.n_clusters, n_features),
-                                            itr,
-                                            weigthInCluster.view(),
-                                            newCentroids.view());
-    cuvs::cluster::kmeans::count_samples_in_cluster(handle,
-                                                    params,
-                                                    X_view,
-                                                    L2NormX.view(),
-                                                    newCentroids.view(),
-                                                    workspace,
-                                                    sampleCountInCluster.view());
-
-    ASSERT_TRUE(devArrMatch(sampleCountInCluster.data_handle(),
-                            weigthInCluster.data_handle(),
-                            params.n_clusters,
-                            CompareApprox<T>(params.tol)));
-  }
-
-  void basicTest()
-  {
-    testparams = ::testing::TestWithParam<KmeansInputs<T>>::GetParam();
-
-    int n_samples              = testparams.n_row;
-    int n_features             = testparams.n_col;
-    params.n_clusters          = testparams.n_clusters;
-    params.tol                 = testparams.tol;
-    params.n_init              = 5;
-    params.rng_state.seed      = 1;
-    params.oversampling_factor = 0;
-
-    auto X      = raft::make_device_matrix<T, int>(handle, n_samples, n_features);
-    auto labels = raft::make_device_vector<int, int>(handle, n_samples);
-    auto stream = resource::get_cuda_stream(handle);
-
-    raft::random::make_blobs<T, int>(X.data_handle(),
-                                     labels.data_handle(),
-                                     n_samples,
-                                     n_features,
-                                     params.n_clusters,
-                                     stream,
-                                     true,
-                                     nullptr,
-                                     nullptr,
-                                     T(1.0),
-                                     false,
-                                     (T)-10.0f,
-                                     (T)10.0f,
-                                     (uint64_t)1234);
-
-    d_labels.resize(n_samples, stream);
-    d_labels_ref.resize(n_samples, stream);
-    d_centroids.resize(params.n_clusters * n_features, stream);
-
-    std::optional<raft::device_vector_view<const T>> d_sw = std::nullopt;
-    auto d_centroids_view =
-      raft::make_device_matrix_view<T, int>(d_centroids.data(), params.n_clusters, n_features);
-    if (testparams.weighted) {
-      d_sample_weight.resize(n_samples, stream);
-      d_sw = std::make_optional(
-        raft::make_device_vector_view<const T, int>(d_sample_weight.data(), n_samples));
-      thrust::fill(thrust::cuda::par.on(stream),
-                   d_sample_weight.data(),
-                   d_sample_weight.data() + n_samples,
-                   1);
-    }
-
-    raft::copy(d_labels_ref.data(), labels.data_handle(), n_samples, stream);
-
-    T inertia   = 0;
-    int n_iter  = 0;
-    auto X_view = raft::make_const_mdspan(X.view());
-
-    cuvs::cluster::kmeans_fit_predict<T, int>(
-      handle,
-      params,
-      X_view,
-      d_sw,
-      d_centroids_view,
-      raft::make_device_vector_view<int, int>(d_labels.data(), n_samples),
-      raft::make_host_scalar_view<T>(&inertia),
-      raft::make_host_scalar_view<int>(&n_iter));
-
-    resource::sync_stream(handle, stream);
-
-    score = cuvs::stats::adjusted_rand_index(
-      d_labels_ref.data(), d_labels.data(), n_samples, resource::get_cuda_stream(handle));
-
-    if (score < 1.0) {
-      std::stringstream ss;
-      ss << "Expected: " << raft::arr2Str(d_labels_ref.data(), 25, "d_labels_ref", stream);
-      std::cout << (ss.str().c_str()) << '\n';
-      ss.str(std::string());
-      ss << "Actual: " << raft::arr2Str(d_labels.data(), 25, "d_labels", stream);
-      std::cout << (ss.str().c_str()) << '\n';
-      std::cout << "Score = " << score << '\n';
-    }
-  }
-
-  void SetUp() override
-  {
-    basicTest();
-    apiTest();
-  }
-
- protected:
-  raft::resources handle;
-  KmeansInputs<T> testparams;
-  rmm::device_uvector<int> d_labels;
-  rmm::device_uvector<int> d_labels_ref;
-  rmm::device_uvector<T> d_centroids;
-  rmm::device_uvector<T> d_sample_weight;
-  double score;
-  cuvs::cluster::KMeansParams params;
-};
-
-const std::vector<KmeansInputs<float>> inputsf2 = {{1000, 32, 5, 0.0001f, true},
-                                                   {1000, 32, 5, 0.0001f, false},
-                                                   {1000, 100, 20, 0.0001f, true},
-                                                   {1000, 100, 20, 0.0001f, false},
-                                                   {10000, 32, 10, 0.0001f, true},
-                                                   {10000, 32, 10, 0.0001f, false},
-                                                   {10000, 100, 50, 0.0001f, true},
-                                                   {10000, 100, 50, 0.0001f, false},
-                                                   {10000, 500, 100, 0.0001f, true},
-                                                   {10000, 500, 100, 0.0001f, false}};
-
-const std::vector<KmeansInputs<double>> inputsd2 = {{1000, 32, 5, 0.0001, true},
-                                                    {1000, 32, 5, 0.0001, false},
-                                                    {1000, 100, 20, 0.0001, true},
-                                                    {1000, 100, 20, 0.0001, false},
-                                                    {10000, 32, 10, 0.0001, true},
-                                                    {10000, 32, 10, 0.0001, false},
-                                                    {10000, 100, 50, 0.0001, true},
-                                                    {10000, 100, 50, 0.0001, false},
-                                                    {10000, 500, 100, 0.0001, true},
-                                                    {10000, 500, 100, 0.0001, false}};
-
-typedef KmeansTest<float> KmeansTestF;
-TEST_P(KmeansTestF, Result) { ASSERT_TRUE(score == 1.0); }
-
-typedef KmeansTest<double> KmeansTestD;
-TEST_P(KmeansTestD, Result) { ASSERT_TRUE(score == 1.0); }
-
-INSTANTIATE_TEST_CASE_P(KmeansTests, KmeansTestF, ::testing::ValuesIn(inputsf2));
-
-INSTANTIATE_TEST_CASE_P(KmeansTests, KmeansTestD, ::testing::ValuesIn(inputsd2));
-
-}  // namespace raft
diff --git a/cpp/test/cluster/kmeans_balanced.cu b/cpp/test/cluster/kmeans_balanced.cu
deleted file mode 100644
index 1af6dbb7f..000000000
--- a/cpp/test/cluster/kmeans_balanced.cu
+++ /dev/null
@@ -1,236 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.h"
-#include <gtest/gtest.h>
-#include <optional>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <vector>
-
-#include <cuvs/cluster/kmeans_balanced.cuh>
-#include <cuvs/stats/adjusted_rand_index.cuh>
-#include <raft/core/cudart_utils.hpp>
-#include <raft/core/handle.hpp>
-#include <raft/core/operators.hpp>
-#include <raft/linalg/unary_op.cuh>
-#include <raft/random/make_blobs.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <rmm/device_uvector.hpp>
-#include <thrust/fill.h>
-
-/* This test takes advantage of the fact that make_blobs generates balanced clusters.
- * It doesn't currently test whether the algorithm can make balanced clusters with an imbalanced
- * dataset.
- */
-
-namespace cuvs {
-
-template <typename MathT, typename IdxT>
-struct KmeansBalancedInputs {
-  IdxT n_rows;
-  IdxT n_cols;
-  IdxT n_clusters;
-  cuvs::cluster::kmeans_balanced_params kb_params;
-  MathT tol;
-};
-
-template <typename MathT, typename IdxT>
-::std::ostream& operator<<(::std::ostream& os, const KmeansBalancedInputs<MathT, IdxT>& p)
-{
-  os << "{ " << p.n_rows << ", " << p.n_cols << ", " << p.n_clusters << ", " << p.kb_params.n_iters
-     << static_cast<int>(p.kb_params.metric) << '}' << std::endl;
-  return os;
-}
-
-template <typename DataT, typename MathT, typename LabelT, typename IdxT, typename MappingOpT>
-class KmeansBalancedTest : public ::testing::TestWithParam<KmeansBalancedInputs<MathT, IdxT>> {
- protected:
-  KmeansBalancedTest()
-    : stream(resource::get_cuda_stream(handle)),
-      d_labels(0, stream),
-      d_labels_ref(0, stream),
-      d_centroids(0, stream)
-  {
-  }
-
-  void basicTest()
-  {
-    MappingOpT op{};
-
-    auto p = ::testing::TestWithParam<KmeansBalancedInputs<MathT, IdxT>>::GetParam();
-
-    auto X           = raft::make_device_matrix<DataT, IdxT>(handle, p.n_rows, p.n_cols);
-    auto blob_labels = raft::make_device_vector<IdxT, IdxT>(handle, p.n_rows);
-
-    MathT* blobs_ptr;
-    rmm::device_uvector<MathT> blobs(0, stream);
-    if constexpr (!std::is_same_v<DataT, MathT>) {
-      blobs.resize(p.n_rows * p.n_cols, stream);
-      blobs_ptr = blobs.data();
-    } else {
-      blobs_ptr = X.data_handle();
-    }
-
-    raft::random::make_blobs<MathT, IdxT>(blobs_ptr,
-                                          blob_labels.data_handle(),
-                                          p.n_rows,
-                                          p.n_cols,
-                                          p.n_clusters,
-                                          stream,
-                                          true,
-                                          nullptr,
-                                          nullptr,
-                                          MathT{0.1},
-                                          true,
-                                          MathT{-1},
-                                          MathT{1},
-                                          (uint64_t)1234);
-
-    // Convert blobs dataset to DataT if necessary
-    if constexpr (!std::is_same_v<DataT, MathT>) {
-      raft::linalg::unaryOp(
-        X.data_handle(), blobs.data(), p.n_rows * p.n_cols, op.reverse_op, stream);
-    }
-
-    d_labels.resize(p.n_rows, stream);
-    d_labels_ref.resize(p.n_rows, stream);
-    d_centroids.resize(p.n_clusters * p.n_cols, stream);
-
-    raft::linalg::unaryOp(
-      d_labels_ref.data(), blob_labels.data_handle(), p.n_rows, raft::cast_op<LabelT>(), stream);
-
-    auto X_view =
-      raft::make_device_matrix_view<const DataT, IdxT>(X.data_handle(), X.extent(0), X.extent(1));
-    auto d_centroids_view =
-      raft::make_device_matrix_view<MathT, IdxT>(d_centroids.data(), p.n_clusters, p.n_cols);
-    auto d_labels_view = raft::make_device_vector_view<LabelT, IdxT>(d_labels.data(), p.n_rows);
-
-    cuvs::cluster::kmeans_balanced::fit_predict(
-      handle, p.kb_params, X_view, d_centroids_view, d_labels_view, op);
-
-    resource::sync_stream(handle, stream);
-
-    score = cuvs::stats::adjusted_rand_index(
-      d_labels_ref.data(), d_labels.data(), p.n_rows, resource::get_cuda_stream(handle));
-
-    if (score < 1.0) {
-      std::stringstream ss;
-      ss << "Expected: " << raft::arr2Str(d_labels_ref.data(), 25, "d_labels_ref", stream);
-      std::cout << (ss.str().c_str()) << '\n';
-      ss.str(std::string());
-      ss << "Actual: " << raft::arr2Str(d_labels.data(), 25, "d_labels", stream);
-      std::cout << (ss.str().c_str()) << '\n';
-      std::cout << "Score = " << score << '\n';
-    }
-  }
-
-  void SetUp() override { basicTest(); }
-
- protected:
-  raft::handle_t handle;
-  cudaStream_t stream;
-  rmm::device_uvector<LabelT> d_labels;
-  rmm::device_uvector<LabelT> d_labels_ref;
-  rmm::device_uvector<MathT> d_centroids;
-  double score;
-};
-
-template <typename MathT, typename IdxT>
-std::vector<KmeansBalancedInputs<MathT, IdxT>> get_kmeans_balanced_inputs()
-{
-  std::vector<KmeansBalancedInputs<MathT, IdxT>> out;
-  KmeansBalancedInputs<MathT, IdxT> p;
-  p.kb_params.n_iters = 20;
-  p.kb_params.metric  = cuvs::distance::DistanceType::L2Expanded;
-  p.tol               = MathT{0.0001};
-  std::vector<std::tuple<size_t, size_t, size_t>> row_cols_k = {{1000, 32, 5},
-                                                                {1000, 100, 20},
-                                                                {10000, 32, 10},
-                                                                {10000, 100, 50},
-                                                                {10000, 500, 100},
-                                                                {1000000, 128, 10}};
-  for (auto& rck : row_cols_k) {
-    p.n_rows     = static_cast<IdxT>(std::get<0>(rck));
-    p.n_cols     = static_cast<IdxT>(std::get<1>(rck));
-    p.n_clusters = static_cast<IdxT>(std::get<2>(rck));
-    out.push_back(p);
-  }
-  return out;
-}
-
-const auto inputsf_i32 = get_kmeans_balanced_inputs<float, int>();
-const auto inputsd_i32 = get_kmeans_balanced_inputs<double, int>();
-const auto inputsf_i64 = get_kmeans_balanced_inputs<float, int64_t>();
-const auto inputsd_i64 = get_kmeans_balanced_inputs<double, int64_t>();
-
-#define KB_TEST(test_type, test_name, test_inputs)         \
-  typedef RAFT_DEPAREN(test_type) test_name;               \
-  TEST_P(test_name, Result) { ASSERT_TRUE(score == 1.0); } \
-  INSTANTIATE_TEST_CASE_P(KmeansBalancedTests, test_name, ::testing::ValuesIn(test_inputs))
-
-/*
- * First set of tests: no conversion
- */
-
-KB_TEST((KmeansBalancedTest<float, float, uint32_t, int, raft::identity_op>),
-        KmeansBalancedTestFFU32I32,
-        inputsf_i32);
-KB_TEST((KmeansBalancedTest<double, double, uint32_t, int, raft::identity_op>),
-        KmeansBalancedTestDDU32I32,
-        inputsd_i32);
-KB_TEST((KmeansBalancedTest<float, float, uint32_t, int64_t, raft::identity_op>),
-        KmeansBalancedTestFFU32I64,
-        inputsf_i64);
-KB_TEST((KmeansBalancedTest<double, double, uint32_t, int64_t, raft::identity_op>),
-        KmeansBalancedTestDDU32I64,
-        inputsd_i64);
-KB_TEST((KmeansBalancedTest<float, float, int, int, raft::identity_op>),
-        KmeansBalancedTestFFI32I32,
-        inputsf_i32);
-KB_TEST((KmeansBalancedTest<float, float, int, int64_t, raft::identity_op>),
-        KmeansBalancedTestFFI32I64,
-        inputsf_i64);
-KB_TEST((KmeansBalancedTest<float, float, int64_t, int, raft::identity_op>),
-        KmeansBalancedTestFFI64I32,
-        inputsf_i32);
-KB_TEST((KmeansBalancedTest<float, float, int64_t, int64_t, raft::identity_op>),
-        KmeansBalancedTestFFI64I64,
-        inputsf_i64);
-
-/*
- * Second set of tests: integer dataset with conversion
- */
-
-template <typename DataT, typename MathT>
-struct i2f_scaler {
-  // Note: with a scaling factor of 42, and generating blobs with centers between -1 and 1 with a
-  // standard deviation of 0.1, it's statistically very unlikely that we'd overflow
-  const raft::compose_op<raft::div_const_op<MathT>, raft::cast_op<MathT>> op{
-    raft::div_const_op<MathT>{42}, raft::cast_op<MathT>{}};
-  const raft::compose_op<raft::cast_op<DataT>, raft::mul_const_op<MathT>> reverse_op{
-    raft::cast_op<DataT>{}, raft::mul_const_op<MathT>{42}};
-
-  RAFT_INLINE_FUNCTION auto operator()(const DataT& x) const { return op(x); };
-};
-
-KB_TEST((KmeansBalancedTest<int8_t, float, uint32_t, int, i2f_scaler<int8_t, float>>),
-        KmeansBalancedTestFI8U32I32,
-        inputsf_i32);
-KB_TEST((KmeansBalancedTest<int8_t, double, uint32_t, int, i2f_scaler<int8_t, double>>),
-        KmeansBalancedTestDI8U32I32,
-        inputsd_i32);
-
-}  // namespace cuvs
diff --git a/cpp/test/cluster/kmeans_find_k.cu b/cpp/test/cluster/kmeans_find_k.cu
deleted file mode 100644
index 0c2a34ece..000000000
--- a/cpp/test/cluster/kmeans_find_k.cu
+++ /dev/null
@@ -1,140 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.h"
-#include <gtest/gtest.h>
-#include <optional>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <vector>
-
-#include <cuvs/cluster/kmeans.cuh>
-#include <raft/core/cudart_utils.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/random/make_blobs.cuh>
-#include <raft/util/cuda_utils.cuh>
-
-namespace cuvs {
-
-template <typename T>
-struct KmeansFindKInputs {
-  int n_row;
-  int n_col;
-  int n_clusters;
-  T tol;
-  bool weighted;
-};
-
-template <typename T>
-class KmeansFindKTest : public ::testing::TestWithParam<KmeansFindKInputs<T>> {
- protected:
-  KmeansFindKTest()
-    : stream(resource::get_cuda_stream(handle)), best_k(raft::make_host_scalar<int>(0))
-  {
-  }
-
-  void basicTest()
-  {
-    testparams = ::testing::TestWithParam<KmeansFindKInputs<T>>::GetParam();
-
-    int n_samples  = testparams.n_row;
-    int n_features = testparams.n_col;
-    int n_clusters = testparams.n_clusters;
-
-    auto X      = raft::make_device_matrix<T, int>(handle, n_samples, n_features);
-    auto labels = raft::make_device_vector<int, int>(handle, n_samples);
-
-    raft::random::make_blobs<T, int>(X.data_handle(),
-                                     labels.data_handle(),
-                                     n_samples,
-                                     n_features,
-                                     n_clusters,
-                                     stream,
-                                     true,
-                                     nullptr,
-                                     nullptr,
-                                     T(.001),
-                                     false,
-                                     (T)-10.0f,
-                                     (T)10.0f,
-                                     (uint64_t)1234);
-
-    auto inertia = raft::make_host_scalar<T>(0);
-    auto n_iter  = raft::make_host_scalar<int>(0);
-
-    auto X_view =
-      raft::make_device_matrix_view<const T, int>(X.data_handle(), X.extent(0), X.extent(1));
-
-    cuvs::cluster::kmeans::find_k<int, T>(
-      handle, X_view, best_k.view(), inertia.view(), n_iter.view(), n_clusters);
-
-    resource::sync_stream(handle, stream);
-  }
-
-  void SetUp() override { basicTest(); }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream;
-  KmeansFindKInputs<T> testparams;
-  raft::host_scalar<int> best_k;
-};
-
-const std::vector<KmeansFindKInputs<float>> inputsf2 = {{1000, 32, 8, 0.001f, true},
-                                                        {1000, 32, 8, 0.001f, false},
-                                                        {1000, 100, 20, 0.001f, true},
-                                                        {1000, 100, 20, 0.001f, false},
-                                                        {10000, 32, 10, 0.001f, true},
-                                                        {10000, 32, 10, 0.001f, false},
-                                                        {10000, 100, 50, 0.001f, true},
-                                                        {10000, 100, 50, 0.001f, false},
-                                                        {10000, 500, 100, 0.001f, true},
-                                                        {10000, 500, 100, 0.001f, false}};
-
-const std::vector<KmeansFindKInputs<double>> inputsd2 = {{1000, 32, 5, 0.0001, true},
-                                                         {1000, 32, 5, 0.0001, false},
-                                                         {1000, 100, 20, 0.0001, true},
-                                                         {1000, 100, 20, 0.0001, false},
-                                                         {10000, 32, 10, 0.0001, true},
-                                                         {10000, 32, 10, 0.0001, false},
-                                                         {10000, 100, 50, 0.0001, true},
-                                                         {10000, 100, 50, 0.0001, false},
-                                                         {10000, 500, 100, 0.0001, true},
-                                                         {10000, 500, 100, 0.0001, false}};
-
-typedef KmeansFindKTest<float> KmeansFindKTestF;
-TEST_P(KmeansFindKTestF, Result)
-{
-  if (best_k.view()[0] != testparams.n_clusters) {
-    std::cout << best_k.view()[0] << " " << testparams.n_clusters << std::endl;
-  }
-  ASSERT_TRUE(best_k.view()[0] == testparams.n_clusters);
-}
-
-typedef KmeansFindKTest<double> KmeansFindKTestD;
-TEST_P(KmeansFindKTestD, Result)
-{
-  if (best_k.view()[0] != testparams.n_clusters) {
-    std::cout << best_k.view()[0] << " " << testparams.n_clusters << std::endl;
-  }
-
-  ASSERT_TRUE(best_k.view()[0] == testparams.n_clusters);
-}
-
-INSTANTIATE_TEST_CASE_P(KmeansFindKTests, KmeansFindKTestF, ::testing::ValuesIn(inputsf2));
-
-INSTANTIATE_TEST_CASE_P(KmeansFindKTests, KmeansFindKTestD, ::testing::ValuesIn(inputsd2));
-
-}  // namespace cuvs
diff --git a/cpp/test/cluster/linkage.cu b/cpp/test/cluster/linkage.cu
deleted file mode 100644
index ce4187ff3..000000000
--- a/cpp/test/cluster/linkage.cu
+++ /dev/null
@@ -1,675 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-// XXX: We allow the instantiation of masked_l2_nn here:
-// raft::linkage::FixConnectivitiesRedOp<value_idx, value_t> red_op(params.n_row);
-// raft::linkage::cross_component_nn<value_idx, value_t>(
-//   handle, out_edges, data.data(), colors.data(), params.n_row, params.n_col, red_op);
-//
-// TODO: consider adding this to libraft.so or creating an instance in a
-// separate translation unit for this test.
-#undef RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-#include "../test_utils.cuh"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/linalg/transpose.cuh>
-#include <raft/sparse/coo.hpp>
-
-#include <cuvs/cluster/single_linkage.cuh>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/util/cudart_utils.hpp>
-
-#include <rmm/device_uvector.hpp>
-
-#include <gtest/gtest.h>
-
-#include <vector>
-
-namespace cuvs {
-
-using namespace std;
-
-template <typename T, typename IdxT>
-struct LinkageInputs {
-  IdxT n_row;
-  IdxT n_col;
-
-  std::vector<T> data;
-
-  std::vector<IdxT> expected_labels;
-
-  int n_clusters;
-
-  bool use_knn;
-
-  int c;
-};
-
-/**
- * @brief kernel to calculate the values of a and b
- * @param firstClusterArray: the array of classes of type T
- * @param secondClusterArray: the array of classes of type T
- * @param size: the size of the data points
- * @param a: number of pairs of points that both the clusters have classified the same
- * @param b: number of pairs of points that both the clusters have classified differently
- */
-template <typename T, int BLOCK_DIM_X, int BLOCK_DIM_Y>
-RAFT_KERNEL computeTheNumerator(
-  const T* firstClusterArray, const T* secondClusterArray, uint64_t size, uint64_t* a, uint64_t* b)
-{
-  // calculating the indices of pairs of datapoints compared by the current thread
-  uint64_t j = threadIdx.x + blockIdx.x * blockDim.x;
-  uint64_t i = threadIdx.y + blockIdx.y * blockDim.y;
-
-  // thread-local variables to count a and b
-  uint64_t myA = 0, myB = 0;
-
-  if (i < size && j < size && j < i) {
-    // checking if the pair have been classified the same by both the clusters
-    if (firstClusterArray[i] == firstClusterArray[j] &&
-        secondClusterArray[i] == secondClusterArray[j]) {
-      ++myA;
-    }
-
-    // checking if the pair have been classified differently by both the clusters
-    else if (firstClusterArray[i] != firstClusterArray[j] &&
-             secondClusterArray[i] != secondClusterArray[j]) {
-      ++myB;
-    }
-  }
-
-  // specialize blockReduce for a 2D block of 1024 threads of type uint64_t
-  typedef cub::BlockReduce<uint64_t, BLOCK_DIM_X, cub::BLOCK_REDUCE_WARP_REDUCTIONS, BLOCK_DIM_Y>
-    BlockReduce;
-
-  // Allocate shared memory for blockReduce
-  __shared__ typename BlockReduce::TempStorage temp_storage;
-
-  // summing up thread-local counts specific to a block
-  myA = BlockReduce(temp_storage).Sum(myA);
-  __syncthreads();
-  myB = BlockReduce(temp_storage).Sum(myB);
-  __syncthreads();
-
-  // executed once per block
-  if (threadIdx.x == 0 && threadIdx.y == 0) {
-    raft::myAtomicAdd<unsigned long long int>((unsigned long long int*)a, myA);
-    raft::myAtomicAdd<unsigned long long int>((unsigned long long int*)b, myB);
-  }
-}
-
-/**
- * @brief Function to calculate RandIndex
- * <a href="https://en.wikipedia.org/wiki/Rand_index">more info on rand index</a>
- * @param firstClusterArray: the array of classes of type T
- * @param secondClusterArray: the array of classes of type T
- * @param size: the size of the data points of type uint64_t
- * @param stream: the cudaStream object
- */
-template <typename T>
-double compute_rand_index(T* firstClusterArray,
-                          T* secondClusterArray,
-                          uint64_t size,
-                          cudaStream_t stream)
-{
-  // rand index for size less than 2 is not defined
-  ASSERT(size >= 2, "Rand Index for size less than 2 not defined!");
-
-  // allocating and initializing memory for a and b in the GPU
-  rmm::device_uvector<uint64_t> arr_buf(2, stream);
-  RAFT_CUDA_TRY(cudaMemsetAsync(arr_buf.data(), 0, 2 * sizeof(uint64_t), stream));
-
-  // kernel configuration
-  static const int BLOCK_DIM_Y = 16, BLOCK_DIM_X = 16;
-  dim3 numThreadsPerBlock(BLOCK_DIM_X, BLOCK_DIM_Y);
-  dim3 numBlocks(raft::ceildiv<int>(size, numThreadsPerBlock.x),
-                 raft::ceildiv<int>(size, numThreadsPerBlock.y));
-
-  // calling the kernel
-  computeTheNumerator<T, BLOCK_DIM_X, BLOCK_DIM_Y><<<numBlocks, numThreadsPerBlock, 0, stream>>>(
-    firstClusterArray, secondClusterArray, size, arr_buf.data(), arr_buf.data() + 1);
-
-  // synchronizing and updating the calculated values of a and b from device to host
-  uint64_t ab_host[2] = {0};
-  raft::update_host(ab_host, arr_buf.data(), 2, stream);
-  RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
-
-  // error handling
-  RAFT_CUDA_TRY(cudaGetLastError());
-
-  // denominator
-  uint64_t nChooseTwo = size * (size - 1) / 2;
-
-  // calculating the rand_index
-  return (double)(((double)(ab_host[0] + ab_host[1])) / (double)nChooseTwo);
-}
-
-template <typename T, typename IdxT>
-::std::ostream& operator<<(::std::ostream& os, const LinkageInputs<T, IdxT>& dims)
-{
-  return os;
-}
-
-template <typename T, typename IdxT>
-class LinkageTest : public ::testing::TestWithParam<LinkageInputs<T, IdxT>> {
- public:
-  LinkageTest()
-    : params(::testing::TestWithParam<LinkageInputs<T, IdxT>>::GetParam()),
-      labels(0, resource::get_cuda_stream(handle)),
-      labels_ref(0, resource::get_cuda_stream(handle))
-  {
-  }
-
- protected:
-  void basicTest()
-  {
-    auto stream = resource::get_cuda_stream(handle);
-
-    labels.resize(params.n_row, stream);
-    labels_ref.resize(params.n_row, stream);
-    rmm::device_uvector<T> data(params.n_row * params.n_col, stream);
-
-    raft::copy(data.data(), params.data.data(), data.size(), stream);
-    raft::copy(labels_ref.data(), params.expected_labels.data(), params.n_row, stream);
-
-    rmm::device_uvector<IdxT> out_children(params.n_row * 2, stream);
-
-    auto data_view = raft::make_device_matrix_view<const T, IdxT, row_major>(
-      data.data(), params.n_row, params.n_col);
-    auto dendrogram_view =
-      raft::make_device_matrix_view<IdxT, IdxT, row_major>(out_children.data(), params.n_row, 2);
-    auto labels_view = raft::make_device_vector_view<IdxT, IdxT>(labels.data(), params.n_row);
-
-    if (params.use_knn) {
-      cuvs::cluster::hierarchy::
-        single_linkage<T, IdxT, cuvs::cluster::hierarchy::LinkageDistance::KNN_GRAPH>(
-          handle,
-          data_view,
-          dendrogram_view,
-          labels_view,
-          cuvs::distance::DistanceType::L2SqrtExpanded,
-          params.n_clusters,
-          std::make_optional<int>(params.c));
-
-    } else {
-      cuvs::cluster::hierarchy::
-        single_linkage<T, IdxT, cuvs::cluster::hierarchy::LinkageDistance::PAIRWISE>(
-          handle,
-          data_view,
-          dendrogram_view,
-          labels_view,
-          cuvs::distance::DistanceType::L2SqrtExpanded,
-          params.n_clusters,
-          std::make_optional<int>(params.c));
-    }
-
-    resource::sync_stream(handle, stream);
-
-    score = compute_rand_index(labels.data(), labels_ref.data(), params.n_row, stream);
-  }
-
-  void SetUp() override { basicTest(); }
-
- protected:
-  raft::resources handle;
-
-  LinkageInputs<T, IdxT> params;
-  rmm::device_uvector<IdxT> labels, labels_ref;
-  double score;
-};
-
-const std::vector<LinkageInputs<float, int>> linkage_inputsf2 = {
-  // Test n_clusters == n_points
-  {10,
-   5,
-   {0.21390334, 0.50261639, 0.91036676, 0.59166485, 0.71162682, 0.10248392, 0.77782677, 0.43772379,
-    0.4035871,  0.3282796,  0.47544681, 0.59862974, 0.12319357, 0.06239463, 0.28200272, 0.1345717,
-    0.50498218, 0.5113505,  0.16233086, 0.62165332, 0.42281548, 0.933117,   0.41386077, 0.23264562,
-    0.73325968, 0.37537541, 0.70719873, 0.14522645, 0.73279625, 0.9126674,  0.84854131, 0.28890216,
-    0.85267903, 0.74703138, 0.83842071, 0.34942792, 0.27864171, 0.70911132, 0.21338564, 0.32035554,
-    0.73788331, 0.46926692, 0.57570162, 0.42559178, 0.87120209, 0.22734951, 0.01847905, 0.75549396,
-    0.76166195, 0.66613745},
-   {9, 8, 7, 6, 5, 4, 3, 2, 1, 0},
-   10,
-   true,
-   -1},
-  //  // Test outlier points
-  {9,
-   2,
-   {-1, -50, 3, 4, 5000, 10000, 1, 3, 4, 5, 0.000005, 0.00002, 2000000, 500000, 10, 50, 30, 5},
-   {6, 0, 5, 0, 0, 4, 3, 2, 1},
-   7,
-   true,
-   -1},
-
-  // Test n_clusters == (n_points / 2)
-  {10,
-   5,
-   {0.21390334, 0.50261639, 0.91036676, 0.59166485, 0.71162682, 0.10248392, 0.77782677, 0.43772379,
-    0.4035871,  0.3282796,  0.47544681, 0.59862974, 0.12319357, 0.06239463, 0.28200272, 0.1345717,
-    0.50498218, 0.5113505,  0.16233086, 0.62165332, 0.42281548, 0.933117,   0.41386077, 0.23264562,
-    0.73325968, 0.37537541, 0.70719873, 0.14522645, 0.73279625, 0.9126674,  0.84854131, 0.28890216,
-    0.85267903, 0.74703138, 0.83842071, 0.34942792, 0.27864171, 0.70911132, 0.21338564, 0.32035554,
-    0.73788331, 0.46926692, 0.57570162, 0.42559178, 0.87120209, 0.22734951, 0.01847905, 0.75549396,
-    0.76166195, 0.66613745},
-   {1, 0, 4, 0, 0, 3, 2, 0, 2, 1},
-   5,
-   true,
-   -1},
-
-  // Test n_points == 100
-  {100,
-   10,
-   {6.26168372e-01, 9.30437651e-01, 6.02450208e-01, 2.73025296e-01, 9.53050619e-01, 3.32164396e-01,
-    6.88942598e-01, 5.79163537e-01, 6.70341547e-01, 2.70140602e-02, 9.30429671e-01, 7.17721157e-01,
-    9.89948537e-01, 7.75253347e-01, 1.34491522e-02, 2.48522428e-02, 3.51413378e-01, 7.64405834e-01,
-    7.86373507e-01, 7.18748577e-01, 8.66998621e-01, 6.80316582e-01, 2.51288712e-01, 4.91078420e-01,
-    3.76246281e-01, 4.86828710e-01, 5.67464772e-01, 5.30734742e-01, 8.99478296e-01, 7.66699088e-01,
-    9.49339111e-01, 3.55248484e-01, 9.06046929e-01, 4.48407772e-01, 6.96395305e-01, 2.44277335e-01,
-    7.74840000e-01, 5.21046603e-01, 4.66423971e-02, 5.12019638e-02, 8.95019614e-01, 5.28956953e-01,
-    4.31536306e-01, 5.83857744e-01, 4.41787364e-01, 4.68656523e-01, 5.73971433e-01, 6.79989654e-01,
-    3.19650588e-01, 6.12579596e-01, 6.49126442e-02, 8.39131142e-01, 2.85252117e-01, 5.84848929e-01,
-    9.46507115e-01, 8.58440748e-01, 3.61528940e-01, 2.44215959e-01, 3.80101125e-01, 4.57128957e-02,
-    8.82216988e-01, 8.31498633e-01, 7.23474381e-01, 7.75788607e-01, 1.40864146e-01, 6.62092382e-01,
-    5.13985168e-01, 3.00686418e-01, 8.70109949e-01, 2.43187753e-01, 2.89391938e-01, 2.84214238e-01,
-    8.70985521e-01, 8.77491176e-01, 6.72537226e-01, 3.30929686e-01, 1.85934324e-01, 9.16222614e-01,
-    6.18239142e-01, 2.64768597e-01, 5.76145451e-01, 8.62961369e-01, 6.84757925e-01, 7.60549082e-01,
-    1.27645356e-01, 4.51004673e-01, 3.92292980e-01, 4.63170803e-01, 4.35449330e-02, 2.17583404e-01,
-    5.71832605e-02, 2.06763039e-01, 3.70116249e-01, 2.09750028e-01, 6.17283019e-01, 8.62549231e-01,
-    9.84156240e-02, 2.66249156e-01, 3.87635103e-01, 2.85591012e-02, 4.24826068e-01, 4.45795088e-01,
-    6.86227676e-01, 1.08848960e-01, 5.96731841e-02, 3.71770228e-01, 1.91548833e-01, 6.95136078e-01,
-    9.00700636e-01, 8.76363105e-01, 2.67334632e-01, 1.80619709e-01, 7.94060419e-01, 1.42854171e-02,
-    1.09372387e-01, 8.74028108e-01, 6.46403232e-01, 4.86588834e-01, 5.93446175e-02, 6.11886291e-01,
-    8.83865057e-01, 3.15879821e-01, 2.27043992e-01, 9.76764951e-01, 6.15620336e-01, 9.76199360e-01,
-    2.40548962e-01, 3.21795663e-01, 8.75087904e-02, 8.11234663e-01, 6.96070480e-01, 8.12062321e-01,
-    1.21958818e-01, 3.44348628e-02, 8.72630414e-01, 3.06162776e-01, 1.76043529e-02, 9.45894971e-01,
-    5.33896401e-01, 6.21642973e-01, 4.93062535e-01, 4.48984262e-01, 2.24560379e-01, 4.24052195e-02,
-    4.43447610e-01, 8.95646149e-01, 6.05220676e-01, 1.81840491e-01, 9.70831206e-01, 2.12563586e-02,
-    6.92582693e-01, 7.55946922e-01, 7.95086143e-01, 6.05328941e-01, 3.99350764e-01, 4.32846636e-01,
-    9.81114529e-01, 4.98266428e-01, 6.37127930e-03, 1.59085889e-01, 6.34682067e-05, 5.59429440e-01,
-    7.38827633e-01, 8.93214770e-01, 2.16494306e-01, 9.35430573e-02, 4.75665868e-02, 7.80503518e-01,
-    7.86240041e-01, 7.06854594e-01, 2.13725879e-02, 7.68246091e-01, 4.50234808e-01, 5.21231104e-01,
-    5.01989826e-03, 4.22081572e-02, 1.65337732e-01, 8.54134740e-01, 4.99430262e-01, 8.94525601e-01,
-    1.14028379e-01, 3.69739861e-01, 1.32955599e-01, 2.65563824e-01, 2.52811151e-01, 1.44792843e-01,
-    6.88449594e-01, 4.44921417e-01, 8.23296587e-01, 1.93266317e-01, 1.19033309e-01, 1.36368966e-01,
-    3.42600285e-01, 5.64505195e-01, 5.57594559e-01, 7.44257892e-01, 8.38231569e-02, 4.11548847e-01,
-    3.21010077e-01, 8.55081359e-01, 4.30105779e-01, 1.16229135e-01, 9.87731964e-02, 3.14712335e-01,
-    4.50880592e-01, 2.72289598e-01, 6.31615256e-01, 8.97432958e-01, 4.44764250e-01, 8.03776440e-01,
-    2.68767748e-02, 2.43374608e-01, 4.02141103e-01, 4.98881209e-01, 5.33173003e-01, 8.82890436e-01,
-    7.16149148e-01, 4.19664401e-01, 2.29335357e-01, 2.88637806e-01, 3.44696803e-01, 6.78171906e-01,
-    5.69849716e-01, 5.86454477e-01, 3.54474989e-01, 9.03876540e-01, 6.45980000e-01, 6.34887593e-01,
-    7.88039746e-02, 2.04814126e-01, 7.82251754e-01, 2.43147074e-01, 7.50951808e-01, 1.72799092e-02,
-    2.95349590e-01, 6.57991826e-01, 8.81214312e-01, 5.73970708e-01, 2.77610881e-01, 1.82155097e-01,
-    7.69797417e-02, 6.44792402e-01, 9.46950998e-01, 7.73064845e-01, 6.04733624e-01, 5.80094567e-01,
-    1.67498426e-01, 2.66514296e-01, 6.50140368e-01, 1.91170299e-01, 2.08752199e-01, 3.01664091e-01,
-    9.85033484e-01, 2.92909152e-01, 8.65816607e-01, 1.85222119e-01, 2.28814559e-01, 1.34286382e-02,
-    2.89234322e-01, 8.18668708e-01, 4.71706924e-01, 9.23199803e-01, 2.80879188e-01, 1.47319284e-01,
-    4.13915748e-01, 9.31274932e-02, 6.66322195e-01, 9.66953974e-01, 3.19405786e-01, 6.69486551e-01,
-    5.03096313e-02, 6.95225201e-01, 5.78469859e-01, 6.29481655e-01, 1.39252534e-01, 1.22564968e-01,
-    6.80663678e-01, 6.34607157e-01, 6.42765834e-01, 1.57127410e-02, 2.92132086e-01, 5.24423878e-01,
-    4.68676824e-01, 2.86003928e-01, 7.18608322e-01, 8.95617933e-01, 5.48844309e-01, 1.74517278e-01,
-    5.24379196e-01, 2.13526524e-01, 5.88375435e-01, 9.88560185e-01, 4.17435771e-01, 6.14438688e-01,
-    9.53760881e-01, 5.27151288e-01, 7.03017278e-01, 3.44448559e-01, 4.47059676e-01, 2.83414901e-01,
-    1.98979011e-01, 4.24917361e-01, 5.73172761e-01, 2.32398853e-02, 1.65887230e-01, 4.05552785e-01,
-    9.29665524e-01, 2.26135696e-01, 9.20563384e-01, 7.65259963e-01, 4.54820075e-01, 8.97710267e-01,
-    3.78559302e-03, 9.15219382e-01, 3.55705698e-01, 6.94905124e-01, 8.58540202e-01, 3.89790666e-01,
-    2.49478206e-01, 7.93679304e-01, 4.75830027e-01, 4.40425353e-01, 3.70579459e-01, 1.40578049e-01,
-    1.70386675e-01, 7.04056121e-01, 4.85963102e-01, 9.68450060e-01, 6.77178001e-01, 2.65934654e-01,
-    2.58915007e-01, 6.70052890e-01, 2.61945109e-01, 8.46207759e-01, 1.01928951e-01, 2.85611334e-01,
-    2.45776933e-01, 2.66658783e-01, 3.71724077e-01, 4.34319025e-01, 4.24407347e-01, 7.15417683e-01,
-    8.07997684e-01, 1.64296275e-01, 6.01638065e-01, 8.60606804e-02, 2.68719187e-01, 5.11764101e-01,
-    9.75844338e-01, 7.81226782e-01, 2.20925515e-01, 7.18135040e-01, 9.82395577e-01, 8.39160243e-01,
-    9.08058083e-01, 6.88010677e-01, 8.14271847e-01, 5.12460821e-01, 1.17311345e-01, 5.96075228e-01,
-    9.17455497e-01, 2.12052706e-01, 7.04074603e-01, 8.72872565e-02, 8.76047818e-01, 6.96235046e-01,
-    8.54801557e-01, 2.49729159e-01, 9.76594604e-01, 2.87386363e-01, 2.36461559e-02, 9.94075254e-01,
-    4.25193986e-01, 7.61869994e-01, 5.13334255e-01, 6.44711165e-02, 8.92156689e-01, 3.55235167e-01,
-    1.08154647e-01, 8.78446825e-01, 2.43833016e-01, 9.23071293e-01, 2.72724115e-01, 9.46631338e-01,
-    3.74510294e-01, 4.08451278e-02, 9.78392777e-01, 3.65079221e-01, 6.37199516e-01, 5.51144906e-01,
-    5.25978080e-01, 1.42803678e-01, 4.05451674e-01, 7.79788219e-01, 6.26009784e-01, 3.35249497e-01,
-    1.43159543e-02, 1.80363779e-01, 5.05096904e-01, 2.82619947e-01, 5.83561392e-01, 3.10951324e-01,
-    8.73223968e-01, 4.38545619e-01, 4.81348800e-01, 6.68497085e-01, 3.79345401e-01, 9.58832501e-01,
-    1.89869550e-01, 2.34083070e-01, 2.94066207e-01, 5.74892667e-02, 6.92106828e-02, 9.61127686e-02,
-    6.72650672e-02, 8.47345378e-01, 2.80916761e-01, 7.32177357e-03, 9.80785961e-01, 5.73192225e-02,
-    8.48781331e-01, 8.83225408e-01, 7.34398275e-01, 7.70381941e-01, 6.20778343e-01, 8.96822048e-01,
-    5.40732486e-01, 3.69704071e-01, 5.77305837e-01, 2.08221827e-01, 7.34275341e-01, 1.06110900e-01,
-    3.49496706e-01, 8.34948910e-01, 1.56403291e-02, 6.78576376e-01, 8.96141268e-01, 5.94835119e-01,
-    1.43943153e-01, 3.49618530e-01, 2.10440392e-01, 3.46585620e-01, 1.05153093e-01, 3.45446174e-01,
-    2.72177079e-01, 7.07946300e-01, 4.33717726e-02, 3.31232203e-01, 3.91874320e-01, 4.76338141e-01,
-    6.22777789e-01, 2.95989228e-02, 4.32855769e-01, 7.61049310e-01, 3.63279149e-01, 9.47210350e-01,
-    6.43721247e-01, 6.58025802e-01, 1.05247633e-02, 5.29974442e-01, 7.30675767e-01, 4.30041079e-01,
-    6.62634841e-01, 8.25936616e-01, 9.91253704e-01, 6.79399281e-01, 5.44177006e-01, 7.52876048e-01,
-    3.32139049e-01, 7.98732398e-01, 7.38865223e-01, 9.16055132e-01, 6.11736493e-01, 9.63672879e-01,
-    1.83778839e-01, 7.27558919e-02, 5.91602822e-01, 3.25235484e-01, 2.34741217e-01, 9.52346277e-01,
-    9.18556407e-01, 9.35373324e-01, 6.89209070e-01, 2.56049054e-01, 6.17975395e-01, 7.82285691e-01,
-    9.84983432e-01, 6.62322741e-01, 2.04144457e-01, 3.98446577e-01, 1.38918297e-01, 3.05919921e-01,
-    3.14043787e-01, 5.91072666e-01, 7.44703771e-01, 8.92272567e-01, 9.78017873e-01, 9.01203161e-01,
-    1.41526372e-01, 4.14878484e-01, 6.80683651e-01, 5.01733152e-02, 8.14635389e-01, 2.27926375e-01,
-    9.03269815e-01, 8.68443745e-01, 9.86939190e-01, 7.40779486e-01, 2.61005311e-01, 3.19276232e-01,
-    9.69509248e-01, 1.11908818e-01, 4.49198556e-01, 1.27056715e-01, 3.84064823e-01, 5.14591811e-01,
-    2.10747488e-01, 9.53884090e-01, 8.43167950e-01, 4.51187972e-01, 3.75331782e-01, 6.23566461e-01,
-    3.55290379e-01, 2.95705968e-01, 1.69622690e-01, 1.42981830e-01, 2.72180991e-01, 9.46468040e-01,
-    3.70932500e-01, 9.94292830e-01, 4.62587505e-01, 7.14817405e-01, 2.45370540e-02, 3.00906377e-01,
-    5.75768304e-01, 9.71448393e-01, 6.95574827e-02, 3.93693854e-01, 5.29306116e-01, 5.04694554e-01,
-    6.73797120e-02, 6.76596969e-01, 5.50948898e-01, 3.24909641e-01, 7.70337719e-01, 6.51842631e-03,
-    3.03264879e-01, 7.61037886e-03, 2.72289601e-01, 1.50502041e-01, 6.71103888e-02, 7.41503703e-01,
-    1.92088941e-01, 2.19043977e-01, 9.09320161e-01, 2.37993569e-01, 6.18107973e-02, 8.31447852e-01,
-    2.23355609e-01, 1.84789435e-01, 4.16104518e-01, 4.21573859e-01, 8.72446305e-02, 2.97294197e-01,
-    4.50328256e-01, 8.72199917e-01, 2.51279916e-01, 4.86219272e-01, 7.57071329e-01, 4.85655942e-01,
-    1.06187277e-01, 4.92341327e-01, 1.46017513e-01, 5.25421017e-01, 4.22637906e-01, 2.24685018e-01,
-    8.72648431e-01, 5.54051490e-01, 1.80745062e-01, 2.12756336e-01, 5.20883169e-01, 7.60363654e-01,
-    8.30254678e-01, 5.00003328e-01, 4.69017439e-01, 6.38105527e-01, 3.50638261e-02, 5.22217353e-02,
-    9.06516882e-02, 8.52975842e-01, 1.19985883e-01, 3.74926753e-01, 6.50302066e-01, 1.98875727e-01,
-    6.28362507e-02, 4.32693501e-01, 3.10500685e-01, 6.20732833e-01, 4.58503272e-01, 3.20790034e-01,
-    7.91284868e-01, 7.93054570e-01, 2.93406765e-01, 8.95399023e-01, 1.06441034e-01, 7.53085241e-02,
-    8.67523104e-01, 1.47963482e-01, 1.25584706e-01, 3.81545040e-02, 6.34338619e-01, 1.76368938e-02,
-    5.75553531e-02, 5.31607516e-01, 2.63869588e-01, 9.41945823e-01, 9.24028838e-02, 5.21496463e-01,
-    7.74866558e-01, 5.65210610e-01, 7.28015327e-02, 6.51963790e-01, 8.94727453e-01, 4.49571590e-01,
-    1.29932405e-01, 8.64026259e-01, 9.92599934e-01, 7.43721560e-01, 8.87300215e-01, 1.06369925e-01,
-    8.11335531e-01, 7.87734900e-01, 9.87344678e-01, 5.32502820e-01, 4.42612382e-01, 9.64041183e-01,
-    1.66085871e-01, 1.12937664e-01, 5.24423470e-01, 6.54689333e-01, 4.59119726e-01, 5.22774091e-01,
-    3.08722276e-02, 6.26979315e-01, 4.49754105e-01, 8.07495757e-01, 2.34199499e-01, 1.67765675e-01,
-    9.22168418e-01, 3.73210378e-01, 8.04432575e-01, 5.61890354e-01, 4.47025593e-01, 6.43155678e-01,
-    2.40407640e-01, 5.91631279e-01, 1.59369206e-01, 7.75799090e-01, 8.32067212e-01, 5.59791576e-02,
-    6.39105224e-01, 4.85274738e-01, 2.12630838e-01, 2.81431312e-02, 7.16205363e-01, 6.83885011e-01,
-    5.23869697e-01, 9.99418314e-01, 8.35331599e-01, 4.69877463e-02, 6.74712562e-01, 7.99273684e-01,
-    2.77001890e-02, 5.75809742e-01, 2.78513031e-01, 8.36209905e-01, 7.25472379e-01, 4.87173943e-01,
-    7.88311357e-01, 9.64676177e-01, 1.75752651e-01, 4.98112580e-01, 8.08850418e-02, 6.40981131e-01,
-    4.06647450e-01, 8.46539387e-01, 2.12620694e-01, 9.11012851e-01, 8.25041445e-01, 8.90065575e-01,
-    9.63626055e-01, 5.96689242e-01, 1.63372670e-01, 4.51640148e-01, 3.43026542e-01, 5.80658851e-01,
-    2.82327625e-01, 4.75535418e-01, 6.27760926e-01, 8.46314115e-01, 9.61961932e-01, 3.19806094e-01,
-    5.05508062e-01, 5.28102944e-01, 6.13045057e-01, 7.44714938e-01, 1.50586073e-01, 7.91878033e-01,
-    4.89839179e-01, 3.10496849e-01, 8.82309038e-01, 2.86922314e-01, 4.84687559e-01, 5.20838630e-01,
-    4.62955493e-01, 2.38185305e-01, 5.47259907e-02, 7.10916137e-01, 7.31887202e-01, 6.25602317e-01,
-    8.77741168e-01, 4.19881322e-01, 4.81222328e-01, 1.28224501e-01, 2.46034010e-01, 3.34971854e-01,
-    7.37216484e-01, 5.62134821e-02, 7.14089724e-01, 9.85549393e-01, 4.66295827e-01, 3.08722434e-03,
-    4.70237690e-01, 2.66524167e-01, 7.93875484e-01, 4.54795911e-02, 8.09702944e-01, 1.47709735e-02,
-    1.70082405e-01, 6.35905179e-01, 3.75379109e-01, 4.30315011e-01, 3.15788760e-01, 5.58065230e-01,
-    2.24643800e-01, 2.42142981e-01, 6.57283636e-01, 3.34921891e-01, 1.26588975e-01, 7.68064155e-01,
-    9.43856291e-01, 4.47518596e-01, 5.44453573e-01, 9.95764932e-01, 7.16444391e-01, 8.51019765e-01,
-    1.01179183e-01, 4.45473958e-01, 4.60327322e-01, 4.96895844e-02, 4.72907738e-01, 5.58987444e-01,
-    3.41027487e-01, 1.56175026e-01, 7.58283148e-01, 6.83600909e-01, 2.14623396e-01, 3.27348880e-01,
-    3.92517893e-01, 6.70418431e-01, 5.16440832e-01, 8.63140348e-01, 5.73277464e-01, 3.46608058e-01,
-    7.39396341e-01, 7.20852434e-01, 2.35653246e-02, 3.89935659e-01, 7.53783745e-01, 6.34563528e-01,
-    8.79339335e-01, 7.41599159e-02, 5.62433904e-01, 6.15553852e-01, 4.56956324e-01, 5.20047447e-01,
-    5.26845015e-02, 5.58471266e-01, 1.63632233e-01, 5.38936665e-02, 6.49593683e-01, 2.56838748e-01,
-    8.99035326e-01, 7.20847756e-01, 5.68954684e-01, 7.43684755e-01, 5.70924238e-01, 3.82318724e-01,
-    4.89328290e-01, 5.62208561e-01, 4.97540804e-02, 4.18011085e-01, 6.88041565e-01, 2.16234653e-01,
-    7.89548214e-01, 8.46136387e-01, 8.46816189e-01, 1.73842353e-01, 6.11627842e-02, 8.44440559e-01,
-    4.50646654e-01, 3.74785037e-01, 4.87196697e-01, 4.56276448e-01, 9.13284391e-01, 4.15715464e-01,
-    7.13597697e-01, 1.23641270e-02, 5.10031271e-01, 4.74601930e-02, 2.55731159e-01, 3.22090006e-01,
-    1.91165703e-01, 4.51170940e-01, 7.50843157e-01, 4.42420576e-01, 4.25380660e-01, 4.50667257e-01,
-    6.55689206e-01, 9.68257670e-02, 1.96528793e-01, 8.97343028e-01, 4.99940904e-01, 6.65504083e-01,
-    9.41828079e-01, 4.54397338e-01, 5.61893331e-01, 5.09839880e-01, 4.53117514e-01, 8.96804127e-02,
-    1.74888861e-01, 6.65641378e-01, 2.81668336e-01, 1.89532742e-01, 5.61668382e-01, 8.68330157e-02,
-    8.25092797e-01, 5.18106324e-01, 1.71904024e-01, 3.68385523e-01, 1.62005436e-01, 7.48507399e-01,
-    9.30274827e-01, 2.38198517e-01, 9.52222901e-01, 5.23587800e-01, 6.94384557e-01, 1.09338652e-01,
-    4.83356794e-01, 2.73050402e-01, 3.68027050e-01, 5.92366466e-01, 1.83192289e-01, 8.60376029e-01,
-    7.13926203e-01, 8.16750052e-01, 1.57890291e-01, 6.25691951e-01, 5.24831646e-01, 1.73873797e-01,
-    1.02429784e-01, 9.17488471e-01, 4.03584434e-01, 9.31170884e-01, 2.79386137e-01, 8.77745206e-01,
-    2.45200576e-01, 1.28896951e-01, 3.15713052e-01, 5.27874291e-01, 2.16444335e-01, 7.03883817e-01,
-    7.74738919e-02, 8.42422142e-01, 3.75598924e-01, 3.51002411e-01, 6.22752776e-01, 4.82407943e-01,
-    7.43107867e-01, 9.46182666e-01, 9.44344819e-01, 3.28124763e-01, 1.06147431e-01, 1.65102684e-01,
-    3.84060507e-01, 2.91057722e-01, 7.68173662e-02, 1.03543651e-01, 6.76698940e-01, 1.43141994e-01,
-    7.21342202e-01, 6.69471294e-03, 9.07298311e-01, 5.57080171e-01, 8.10954489e-01, 4.11120526e-01,
-    2.06407453e-01, 2.59590556e-01, 7.58512718e-01, 5.79873897e-01, 2.92875650e-01, 2.83686529e-01,
-    2.42829343e-01, 9.19323719e-01, 3.46832864e-01, 3.58238858e-01, 7.42827585e-01, 2.05760059e-01,
-    9.58438860e-01, 5.66326411e-01, 6.60292846e-01, 5.61095078e-02, 6.79465531e-01, 7.05118513e-01,
-    4.44713264e-01, 2.09732933e-01, 5.22732436e-01, 1.74396512e-01, 5.29356748e-01, 4.38475687e-01,
-    4.94036404e-01, 4.09785794e-01, 6.40025507e-01, 5.79371821e-01, 1.57726118e-01, 6.04572263e-01,
-    5.41072639e-01, 5.18847173e-01, 1.97093284e-01, 8.91767002e-01, 4.29050835e-01, 8.25490570e-01,
-    3.87699807e-01, 4.50705808e-01, 2.49371643e-01, 3.36074898e-01, 9.29925118e-01, 6.65393649e-01,
-    9.07275994e-01, 3.73075859e-01, 4.14044139e-03, 2.37463702e-01, 2.25893784e-01, 2.46900245e-01,
-    4.50350196e-01, 3.48618117e-01, 5.07193932e-01, 5.23435142e-01, 8.13611417e-01, 8.92715622e-01,
-    1.02623450e-01, 3.06088345e-01, 7.80461650e-01, 2.21453645e-01, 2.01419652e-01, 2.84254457e-01,
-    3.68286735e-01, 7.39358243e-01, 8.97879394e-01, 9.81599566e-01, 7.56526442e-01, 7.37645545e-01,
-    4.23976657e-02, 8.25922012e-01, 2.60956996e-01, 2.90702065e-01, 8.98388344e-01, 3.03733299e-01,
-    8.49071471e-01, 3.45835425e-01, 7.65458276e-01, 5.68094872e-01, 8.93770930e-01, 9.93161641e-01,
-    5.63368667e-02, 4.26548945e-01, 5.46745780e-01, 5.75674571e-01, 7.94599487e-01, 7.18935553e-02,
-    4.46492976e-01, 6.40240123e-01, 2.73246969e-01, 2.00465968e-01, 1.30718835e-01, 1.92492005e-01,
-    1.96617189e-01, 6.61271644e-01, 8.12687657e-01, 8.66342445e-01
-
-   },
-   {0, 9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-    7, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 6, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-    0, 0, 0, 0, 0, 5, 0, 0, 0, 0, 4, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0,
-    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
-   10,
-   true,
-   -4},
-  {10,
-   5,
-   {0.21390334, 0.50261639, 0.91036676, 0.59166485, 0.71162682, 0.10248392, 0.77782677, 0.43772379,
-    0.4035871,  0.3282796,  0.47544681, 0.59862974, 0.12319357, 0.06239463, 0.28200272, 0.1345717,
-    0.50498218, 0.5113505,  0.16233086, 0.62165332, 0.42281548, 0.933117,   0.41386077, 0.23264562,
-    0.73325968, 0.37537541, 0.70719873, 0.14522645, 0.73279625, 0.9126674,  0.84854131, 0.28890216,
-    0.85267903, 0.74703138, 0.83842071, 0.34942792, 0.27864171, 0.70911132, 0.21338564, 0.32035554,
-    0.73788331, 0.46926692, 0.57570162, 0.42559178, 0.87120209, 0.22734951, 0.01847905, 0.75549396,
-    0.76166195, 0.66613745},
-   {9, 8, 7, 6, 5, 4, 3, 2, 1, 0},
-   10,
-   false,
-   5},
-  // Test outlier points
-  {9,
-   2,
-   {-1, -50, 3, 4, 5000, 10000, 1, 3, 4, 5, 0.000005, 0.00002, 2000000, 500000, 10, 50, 30, 5},
-   {6, 0, 5, 0, 0, 4, 3, 2, 1},
-   7,
-   false,
-   5},
-
-  // Test n_clusters == (n_points / 2)
-  {10,
-   5,
-   {0.21390334, 0.50261639, 0.91036676, 0.59166485, 0.71162682, 0.10248392, 0.77782677, 0.43772379,
-    0.4035871,  0.3282796,  0.47544681, 0.59862974, 0.12319357, 0.06239463, 0.28200272, 0.1345717,
-    0.50498218, 0.5113505,  0.16233086, 0.62165332, 0.42281548, 0.933117,   0.41386077, 0.23264562,
-    0.73325968, 0.37537541, 0.70719873, 0.14522645, 0.73279625, 0.9126674,  0.84854131, 0.28890216,
-    0.85267903, 0.74703138, 0.83842071, 0.34942792, 0.27864171, 0.70911132, 0.21338564, 0.32035554,
-    0.73788331, 0.46926692, 0.57570162, 0.42559178, 0.87120209, 0.22734951, 0.01847905, 0.75549396,
-    0.76166195, 0.66613745},
-   {1, 0, 4, 0, 0, 3, 2, 0, 2, 1},
-   5,
-   false,
-   10},
-
-  // Test n_points == 100
-  {100,
-   10,
-   {6.26168372e-01, 9.30437651e-01, 6.02450208e-01, 2.73025296e-01, 9.53050619e-01, 3.32164396e-01,
-    6.88942598e-01, 5.79163537e-01, 6.70341547e-01, 2.70140602e-02, 9.30429671e-01, 7.17721157e-01,
-    9.89948537e-01, 7.75253347e-01, 1.34491522e-02, 2.48522428e-02, 3.51413378e-01, 7.64405834e-01,
-    7.86373507e-01, 7.18748577e-01, 8.66998621e-01, 6.80316582e-01, 2.51288712e-01, 4.91078420e-01,
-    3.76246281e-01, 4.86828710e-01, 5.67464772e-01, 5.30734742e-01, 8.99478296e-01, 7.66699088e-01,
-    9.49339111e-01, 3.55248484e-01, 9.06046929e-01, 4.48407772e-01, 6.96395305e-01, 2.44277335e-01,
-    7.74840000e-01, 5.21046603e-01, 4.66423971e-02, 5.12019638e-02, 8.95019614e-01, 5.28956953e-01,
-    4.31536306e-01, 5.83857744e-01, 4.41787364e-01, 4.68656523e-01, 5.73971433e-01, 6.79989654e-01,
-    3.19650588e-01, 6.12579596e-01, 6.49126442e-02, 8.39131142e-01, 2.85252117e-01, 5.84848929e-01,
-    9.46507115e-01, 8.58440748e-01, 3.61528940e-01, 2.44215959e-01, 3.80101125e-01, 4.57128957e-02,
-    8.82216988e-01, 8.31498633e-01, 7.23474381e-01, 7.75788607e-01, 1.40864146e-01, 6.62092382e-01,
-    5.13985168e-01, 3.00686418e-01, 8.70109949e-01, 2.43187753e-01, 2.89391938e-01, 2.84214238e-01,
-    8.70985521e-01, 8.77491176e-01, 6.72537226e-01, 3.30929686e-01, 1.85934324e-01, 9.16222614e-01,
-    6.18239142e-01, 2.64768597e-01, 5.76145451e-01, 8.62961369e-01, 6.84757925e-01, 7.60549082e-01,
-    1.27645356e-01, 4.51004673e-01, 3.92292980e-01, 4.63170803e-01, 4.35449330e-02, 2.17583404e-01,
-    5.71832605e-02, 2.06763039e-01, 3.70116249e-01, 2.09750028e-01, 6.17283019e-01, 8.62549231e-01,
-    9.84156240e-02, 2.66249156e-01, 3.87635103e-01, 2.85591012e-02, 4.24826068e-01, 4.45795088e-01,
-    6.86227676e-01, 1.08848960e-01, 5.96731841e-02, 3.71770228e-01, 1.91548833e-01, 6.95136078e-01,
-    9.00700636e-01, 8.76363105e-01, 2.67334632e-01, 1.80619709e-01, 7.94060419e-01, 1.42854171e-02,
-    1.09372387e-01, 8.74028108e-01, 6.46403232e-01, 4.86588834e-01, 5.93446175e-02, 6.11886291e-01,
-    8.83865057e-01, 3.15879821e-01, 2.27043992e-01, 9.76764951e-01, 6.15620336e-01, 9.76199360e-01,
-    2.40548962e-01, 3.21795663e-01, 8.75087904e-02, 8.11234663e-01, 6.96070480e-01, 8.12062321e-01,
-    1.21958818e-01, 3.44348628e-02, 8.72630414e-01, 3.06162776e-01, 1.76043529e-02, 9.45894971e-01,
-    5.33896401e-01, 6.21642973e-01, 4.93062535e-01, 4.48984262e-01, 2.24560379e-01, 4.24052195e-02,
-    4.43447610e-01, 8.95646149e-01, 6.05220676e-01, 1.81840491e-01, 9.70831206e-01, 2.12563586e-02,
-    6.92582693e-01, 7.55946922e-01, 7.95086143e-01, 6.05328941e-01, 3.99350764e-01, 4.32846636e-01,
-    9.81114529e-01, 4.98266428e-01, 6.37127930e-03, 1.59085889e-01, 6.34682067e-05, 5.59429440e-01,
-    7.38827633e-01, 8.93214770e-01, 2.16494306e-01, 9.35430573e-02, 4.75665868e-02, 7.80503518e-01,
-    7.86240041e-01, 7.06854594e-01, 2.13725879e-02, 7.68246091e-01, 4.50234808e-01, 5.21231104e-01,
-    5.01989826e-03, 4.22081572e-02, 1.65337732e-01, 8.54134740e-01, 4.99430262e-01, 8.94525601e-01,
-    1.14028379e-01, 3.69739861e-01, 1.32955599e-01, 2.65563824e-01, 2.52811151e-01, 1.44792843e-01,
-    6.88449594e-01, 4.44921417e-01, 8.23296587e-01, 1.93266317e-01, 1.19033309e-01, 1.36368966e-01,
-    3.42600285e-01, 5.64505195e-01, 5.57594559e-01, 7.44257892e-01, 8.38231569e-02, 4.11548847e-01,
-    3.21010077e-01, 8.55081359e-01, 4.30105779e-01, 1.16229135e-01, 9.87731964e-02, 3.14712335e-01,
-    4.50880592e-01, 2.72289598e-01, 6.31615256e-01, 8.97432958e-01, 4.44764250e-01, 8.03776440e-01,
-    2.68767748e-02, 2.43374608e-01, 4.02141103e-01, 4.98881209e-01, 5.33173003e-01, 8.82890436e-01,
-    7.16149148e-01, 4.19664401e-01, 2.29335357e-01, 2.88637806e-01, 3.44696803e-01, 6.78171906e-01,
-    5.69849716e-01, 5.86454477e-01, 3.54474989e-01, 9.03876540e-01, 6.45980000e-01, 6.34887593e-01,
-    7.88039746e-02, 2.04814126e-01, 7.82251754e-01, 2.43147074e-01, 7.50951808e-01, 1.72799092e-02,
-    2.95349590e-01, 6.57991826e-01, 8.81214312e-01, 5.73970708e-01, 2.77610881e-01, 1.82155097e-01,
-    7.69797417e-02, 6.44792402e-01, 9.46950998e-01, 7.73064845e-01, 6.04733624e-01, 5.80094567e-01,
-    1.67498426e-01, 2.66514296e-01, 6.50140368e-01, 1.91170299e-01, 2.08752199e-01, 3.01664091e-01,
-    9.85033484e-01, 2.92909152e-01, 8.65816607e-01, 1.85222119e-01, 2.28814559e-01, 1.34286382e-02,
-    2.89234322e-01, 8.18668708e-01, 4.71706924e-01, 9.23199803e-01, 2.80879188e-01, 1.47319284e-01,
-    4.13915748e-01, 9.31274932e-02, 6.66322195e-01, 9.66953974e-01, 3.19405786e-01, 6.69486551e-01,
-    5.03096313e-02, 6.95225201e-01, 5.78469859e-01, 6.29481655e-01, 1.39252534e-01, 1.22564968e-01,
-    6.80663678e-01, 6.34607157e-01, 6.42765834e-01, 1.57127410e-02, 2.92132086e-01, 5.24423878e-01,
-    4.68676824e-01, 2.86003928e-01, 7.18608322e-01, 8.95617933e-01, 5.48844309e-01, 1.74517278e-01,
-    5.24379196e-01, 2.13526524e-01, 5.88375435e-01, 9.88560185e-01, 4.17435771e-01, 6.14438688e-01,
-    9.53760881e-01, 5.27151288e-01, 7.03017278e-01, 3.44448559e-01, 4.47059676e-01, 2.83414901e-01,
-    1.98979011e-01, 4.24917361e-01, 5.73172761e-01, 2.32398853e-02, 1.65887230e-01, 4.05552785e-01,
-    9.29665524e-01, 2.26135696e-01, 9.20563384e-01, 7.65259963e-01, 4.54820075e-01, 8.97710267e-01,
-    3.78559302e-03, 9.15219382e-01, 3.55705698e-01, 6.94905124e-01, 8.58540202e-01, 3.89790666e-01,
-    2.49478206e-01, 7.93679304e-01, 4.75830027e-01, 4.40425353e-01, 3.70579459e-01, 1.40578049e-01,
-    1.70386675e-01, 7.04056121e-01, 4.85963102e-01, 9.68450060e-01, 6.77178001e-01, 2.65934654e-01,
-    2.58915007e-01, 6.70052890e-01, 2.61945109e-01, 8.46207759e-01, 1.01928951e-01, 2.85611334e-01,
-    2.45776933e-01, 2.66658783e-01, 3.71724077e-01, 4.34319025e-01, 4.24407347e-01, 7.15417683e-01,
-    8.07997684e-01, 1.64296275e-01, 6.01638065e-01, 8.60606804e-02, 2.68719187e-01, 5.11764101e-01,
-    9.75844338e-01, 7.81226782e-01, 2.20925515e-01, 7.18135040e-01, 9.82395577e-01, 8.39160243e-01,
-    9.08058083e-01, 6.88010677e-01, 8.14271847e-01, 5.12460821e-01, 1.17311345e-01, 5.96075228e-01,
-    9.17455497e-01, 2.12052706e-01, 7.04074603e-01, 8.72872565e-02, 8.76047818e-01, 6.96235046e-01,
-    8.54801557e-01, 2.49729159e-01, 9.76594604e-01, 2.87386363e-01, 2.36461559e-02, 9.94075254e-01,
-    4.25193986e-01, 7.61869994e-01, 5.13334255e-01, 6.44711165e-02, 8.92156689e-01, 3.55235167e-01,
-    1.08154647e-01, 8.78446825e-01, 2.43833016e-01, 9.23071293e-01, 2.72724115e-01, 9.46631338e-01,
-    3.74510294e-01, 4.08451278e-02, 9.78392777e-01, 3.65079221e-01, 6.37199516e-01, 5.51144906e-01,
-    5.25978080e-01, 1.42803678e-01, 4.05451674e-01, 7.79788219e-01, 6.26009784e-01, 3.35249497e-01,
-    1.43159543e-02, 1.80363779e-01, 5.05096904e-01, 2.82619947e-01, 5.83561392e-01, 3.10951324e-01,
-    8.73223968e-01, 4.38545619e-01, 4.81348800e-01, 6.68497085e-01, 3.79345401e-01, 9.58832501e-01,
-    1.89869550e-01, 2.34083070e-01, 2.94066207e-01, 5.74892667e-02, 6.92106828e-02, 9.61127686e-02,
-    6.72650672e-02, 8.47345378e-01, 2.80916761e-01, 7.32177357e-03, 9.80785961e-01, 5.73192225e-02,
-    8.48781331e-01, 8.83225408e-01, 7.34398275e-01, 7.70381941e-01, 6.20778343e-01, 8.96822048e-01,
-    5.40732486e-01, 3.69704071e-01, 5.77305837e-01, 2.08221827e-01, 7.34275341e-01, 1.06110900e-01,
-    3.49496706e-01, 8.34948910e-01, 1.56403291e-02, 6.78576376e-01, 8.96141268e-01, 5.94835119e-01,
-    1.43943153e-01, 3.49618530e-01, 2.10440392e-01, 3.46585620e-01, 1.05153093e-01, 3.45446174e-01,
-    2.72177079e-01, 7.07946300e-01, 4.33717726e-02, 3.31232203e-01, 3.91874320e-01, 4.76338141e-01,
-    6.22777789e-01, 2.95989228e-02, 4.32855769e-01, 7.61049310e-01, 3.63279149e-01, 9.47210350e-01,
-    6.43721247e-01, 6.58025802e-01, 1.05247633e-02, 5.29974442e-01, 7.30675767e-01, 4.30041079e-01,
-    6.62634841e-01, 8.25936616e-01, 9.91253704e-01, 6.79399281e-01, 5.44177006e-01, 7.52876048e-01,
-    3.32139049e-01, 7.98732398e-01, 7.38865223e-01, 9.16055132e-01, 6.11736493e-01, 9.63672879e-01,
-    1.83778839e-01, 7.27558919e-02, 5.91602822e-01, 3.25235484e-01, 2.34741217e-01, 9.52346277e-01,
-    9.18556407e-01, 9.35373324e-01, 6.89209070e-01, 2.56049054e-01, 6.17975395e-01, 7.82285691e-01,
-    9.84983432e-01, 6.62322741e-01, 2.04144457e-01, 3.98446577e-01, 1.38918297e-01, 3.05919921e-01,
-    3.14043787e-01, 5.91072666e-01, 7.44703771e-01, 8.92272567e-01, 9.78017873e-01, 9.01203161e-01,
-    1.41526372e-01, 4.14878484e-01, 6.80683651e-01, 5.01733152e-02, 8.14635389e-01, 2.27926375e-01,
-    9.03269815e-01, 8.68443745e-01, 9.86939190e-01, 7.40779486e-01, 2.61005311e-01, 3.19276232e-01,
-    9.69509248e-01, 1.11908818e-01, 4.49198556e-01, 1.27056715e-01, 3.84064823e-01, 5.14591811e-01,
-    2.10747488e-01, 9.53884090e-01, 8.43167950e-01, 4.51187972e-01, 3.75331782e-01, 6.23566461e-01,
-    3.55290379e-01, 2.95705968e-01, 1.69622690e-01, 1.42981830e-01, 2.72180991e-01, 9.46468040e-01,
-    3.70932500e-01, 9.94292830e-01, 4.62587505e-01, 7.14817405e-01, 2.45370540e-02, 3.00906377e-01,
-    5.75768304e-01, 9.71448393e-01, 6.95574827e-02, 3.93693854e-01, 5.29306116e-01, 5.04694554e-01,
-    6.73797120e-02, 6.76596969e-01, 5.50948898e-01, 3.24909641e-01, 7.70337719e-01, 6.51842631e-03,
-    3.03264879e-01, 7.61037886e-03, 2.72289601e-01, 1.50502041e-01, 6.71103888e-02, 7.41503703e-01,
-    1.92088941e-01, 2.19043977e-01, 9.09320161e-01, 2.37993569e-01, 6.18107973e-02, 8.31447852e-01,
-    2.23355609e-01, 1.84789435e-01, 4.16104518e-01, 4.21573859e-01, 8.72446305e-02, 2.97294197e-01,
-    4.50328256e-01, 8.72199917e-01, 2.51279916e-01, 4.86219272e-01, 7.57071329e-01, 4.85655942e-01,
-    1.06187277e-01, 4.92341327e-01, 1.46017513e-01, 5.25421017e-01, 4.22637906e-01, 2.24685018e-01,
-    8.72648431e-01, 5.54051490e-01, 1.80745062e-01, 2.12756336e-01, 5.20883169e-01, 7.60363654e-01,
-    8.30254678e-01, 5.00003328e-01, 4.69017439e-01, 6.38105527e-01, 3.50638261e-02, 5.22217353e-02,
-    9.06516882e-02, 8.52975842e-01, 1.19985883e-01, 3.74926753e-01, 6.50302066e-01, 1.98875727e-01,
-    6.28362507e-02, 4.32693501e-01, 3.10500685e-01, 6.20732833e-01, 4.58503272e-01, 3.20790034e-01,
-    7.91284868e-01, 7.93054570e-01, 2.93406765e-01, 8.95399023e-01, 1.06441034e-01, 7.53085241e-02,
-    8.67523104e-01, 1.47963482e-01, 1.25584706e-01, 3.81545040e-02, 6.34338619e-01, 1.76368938e-02,
-    5.75553531e-02, 5.31607516e-01, 2.63869588e-01, 9.41945823e-01, 9.24028838e-02, 5.21496463e-01,
-    7.74866558e-01, 5.65210610e-01, 7.28015327e-02, 6.51963790e-01, 8.94727453e-01, 4.49571590e-01,
-    1.29932405e-01, 8.64026259e-01, 9.92599934e-01, 7.43721560e-01, 8.87300215e-01, 1.06369925e-01,
-    8.11335531e-01, 7.87734900e-01, 9.87344678e-01, 5.32502820e-01, 4.42612382e-01, 9.64041183e-01,
-    1.66085871e-01, 1.12937664e-01, 5.24423470e-01, 6.54689333e-01, 4.59119726e-01, 5.22774091e-01,
-    3.08722276e-02, 6.26979315e-01, 4.49754105e-01, 8.07495757e-01, 2.34199499e-01, 1.67765675e-01,
-    9.22168418e-01, 3.73210378e-01, 8.04432575e-01, 5.61890354e-01, 4.47025593e-01, 6.43155678e-01,
-    2.40407640e-01, 5.91631279e-01, 1.59369206e-01, 7.75799090e-01, 8.32067212e-01, 5.59791576e-02,
-    6.39105224e-01, 4.85274738e-01, 2.12630838e-01, 2.81431312e-02, 7.16205363e-01, 6.83885011e-01,
-    5.23869697e-01, 9.99418314e-01, 8.35331599e-01, 4.69877463e-02, 6.74712562e-01, 7.99273684e-01,
-    2.77001890e-02, 5.75809742e-01, 2.78513031e-01, 8.36209905e-01, 7.25472379e-01, 4.87173943e-01,
-    7.88311357e-01, 9.64676177e-01, 1.75752651e-01, 4.98112580e-01, 8.08850418e-02, 6.40981131e-01,
-    4.06647450e-01, 8.46539387e-01, 2.12620694e-01, 9.11012851e-01, 8.25041445e-01, 8.90065575e-01,
-    9.63626055e-01, 5.96689242e-01, 1.63372670e-01, 4.51640148e-01, 3.43026542e-01, 5.80658851e-01,
-    2.82327625e-01, 4.75535418e-01, 6.27760926e-01, 8.46314115e-01, 9.61961932e-01, 3.19806094e-01,
-    5.05508062e-01, 5.28102944e-01, 6.13045057e-01, 7.44714938e-01, 1.50586073e-01, 7.91878033e-01,
-    4.89839179e-01, 3.10496849e-01, 8.82309038e-01, 2.86922314e-01, 4.84687559e-01, 5.20838630e-01,
-    4.62955493e-01, 2.38185305e-01, 5.47259907e-02, 7.10916137e-01, 7.31887202e-01, 6.25602317e-01,
-    8.77741168e-01, 4.19881322e-01, 4.81222328e-01, 1.28224501e-01, 2.46034010e-01, 3.34971854e-01,
-    7.37216484e-01, 5.62134821e-02, 7.14089724e-01, 9.85549393e-01, 4.66295827e-01, 3.08722434e-03,
-    4.70237690e-01, 2.66524167e-01, 7.93875484e-01, 4.54795911e-02, 8.09702944e-01, 1.47709735e-02,
-    1.70082405e-01, 6.35905179e-01, 3.75379109e-01, 4.30315011e-01, 3.15788760e-01, 5.58065230e-01,
-    2.24643800e-01, 2.42142981e-01, 6.57283636e-01, 3.34921891e-01, 1.26588975e-01, 7.68064155e-01,
-    9.43856291e-01, 4.47518596e-01, 5.44453573e-01, 9.95764932e-01, 7.16444391e-01, 8.51019765e-01,
-    1.01179183e-01, 4.45473958e-01, 4.60327322e-01, 4.96895844e-02, 4.72907738e-01, 5.58987444e-01,
-    3.41027487e-01, 1.56175026e-01, 7.58283148e-01, 6.83600909e-01, 2.14623396e-01, 3.27348880e-01,
-    3.92517893e-01, 6.70418431e-01, 5.16440832e-01, 8.63140348e-01, 5.73277464e-01, 3.46608058e-01,
-    7.39396341e-01, 7.20852434e-01, 2.35653246e-02, 3.89935659e-01, 7.53783745e-01, 6.34563528e-01,
-    8.79339335e-01, 7.41599159e-02, 5.62433904e-01, 6.15553852e-01, 4.56956324e-01, 5.20047447e-01,
-    5.26845015e-02, 5.58471266e-01, 1.63632233e-01, 5.38936665e-02, 6.49593683e-01, 2.56838748e-01,
-    8.99035326e-01, 7.20847756e-01, 5.68954684e-01, 7.43684755e-01, 5.70924238e-01, 3.82318724e-01,
-    4.89328290e-01, 5.62208561e-01, 4.97540804e-02, 4.18011085e-01, 6.88041565e-01, 2.16234653e-01,
-    7.89548214e-01, 8.46136387e-01, 8.46816189e-01, 1.73842353e-01, 6.11627842e-02, 8.44440559e-01,
-    4.50646654e-01, 3.74785037e-01, 4.87196697e-01, 4.56276448e-01, 9.13284391e-01, 4.15715464e-01,
-    7.13597697e-01, 1.23641270e-02, 5.10031271e-01, 4.74601930e-02, 2.55731159e-01, 3.22090006e-01,
-    1.91165703e-01, 4.51170940e-01, 7.50843157e-01, 4.42420576e-01, 4.25380660e-01, 4.50667257e-01,
-    6.55689206e-01, 9.68257670e-02, 1.96528793e-01, 8.97343028e-01, 4.99940904e-01, 6.65504083e-01,
-    9.41828079e-01, 4.54397338e-01, 5.61893331e-01, 5.09839880e-01, 4.53117514e-01, 8.96804127e-02,
-    1.74888861e-01, 6.65641378e-01, 2.81668336e-01, 1.89532742e-01, 5.61668382e-01, 8.68330157e-02,
-    8.25092797e-01, 5.18106324e-01, 1.71904024e-01, 3.68385523e-01, 1.62005436e-01, 7.48507399e-01,
-    9.30274827e-01, 2.38198517e-01, 9.52222901e-01, 5.23587800e-01, 6.94384557e-01, 1.09338652e-01,
-    4.83356794e-01, 2.73050402e-01, 3.68027050e-01, 5.92366466e-01, 1.83192289e-01, 8.60376029e-01,
-    7.13926203e-01, 8.16750052e-01, 1.57890291e-01, 6.25691951e-01, 5.24831646e-01, 1.73873797e-01,
-    1.02429784e-01, 9.17488471e-01, 4.03584434e-01, 9.31170884e-01, 2.79386137e-01, 8.77745206e-01,
-    2.45200576e-01, 1.28896951e-01, 3.15713052e-01, 5.27874291e-01, 2.16444335e-01, 7.03883817e-01,
-    7.74738919e-02, 8.42422142e-01, 3.75598924e-01, 3.51002411e-01, 6.22752776e-01, 4.82407943e-01,
-    7.43107867e-01, 9.46182666e-01, 9.44344819e-01, 3.28124763e-01, 1.06147431e-01, 1.65102684e-01,
-    3.84060507e-01, 2.91057722e-01, 7.68173662e-02, 1.03543651e-01, 6.76698940e-01, 1.43141994e-01,
-    7.21342202e-01, 6.69471294e-03, 9.07298311e-01, 5.57080171e-01, 8.10954489e-01, 4.11120526e-01,
-    2.06407453e-01, 2.59590556e-01, 7.58512718e-01, 5.79873897e-01, 2.92875650e-01, 2.83686529e-01,
-    2.42829343e-01, 9.19323719e-01, 3.46832864e-01, 3.58238858e-01, 7.42827585e-01, 2.05760059e-01,
-    9.58438860e-01, 5.66326411e-01, 6.60292846e-01, 5.61095078e-02, 6.79465531e-01, 7.05118513e-01,
-    4.44713264e-01, 2.09732933e-01, 5.22732436e-01, 1.74396512e-01, 5.29356748e-01, 4.38475687e-01,
-    4.94036404e-01, 4.09785794e-01, 6.40025507e-01, 5.79371821e-01, 1.57726118e-01, 6.04572263e-01,
-    5.41072639e-01, 5.18847173e-01, 1.97093284e-01, 8.91767002e-01, 4.29050835e-01, 8.25490570e-01,
-    3.87699807e-01, 4.50705808e-01, 2.49371643e-01, 3.36074898e-01, 9.29925118e-01, 6.65393649e-01,
-    9.07275994e-01, 3.73075859e-01, 4.14044139e-03, 2.37463702e-01, 2.25893784e-01, 2.46900245e-01,
-    4.50350196e-01, 3.48618117e-01, 5.07193932e-01, 5.23435142e-01, 8.13611417e-01, 8.92715622e-01,
-    1.02623450e-01, 3.06088345e-01, 7.80461650e-01, 2.21453645e-01, 2.01419652e-01, 2.84254457e-01,
-    3.68286735e-01, 7.39358243e-01, 8.97879394e-01, 9.81599566e-01, 7.56526442e-01, 7.37645545e-01,
-    4.23976657e-02, 8.25922012e-01, 2.60956996e-01, 2.90702065e-01, 8.98388344e-01, 3.03733299e-01,
-    8.49071471e-01, 3.45835425e-01, 7.65458276e-01, 5.68094872e-01, 8.93770930e-01, 9.93161641e-01,
-    5.63368667e-02, 4.26548945e-01, 5.46745780e-01, 5.75674571e-01, 7.94599487e-01, 7.18935553e-02,
-    4.46492976e-01, 6.40240123e-01, 2.73246969e-01, 2.00465968e-01, 1.30718835e-01, 1.92492005e-01,
-    1.96617189e-01, 6.61271644e-01, 8.12687657e-01, 8.66342445e-01
-
-   },
-   {0, 9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-    7, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 6, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-    0, 0, 0, 0, 0, 5, 0, 0, 0, 0, 4, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0,
-    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
-   10,
-   false,
-   5}};
-
-typedef LinkageTest<float, int> LinkageTestF_Int;
-TEST_P(LinkageTestF_Int, Result) { EXPECT_TRUE(score == 1.0); }
-
-INSTANTIATE_TEST_CASE_P(LinkageTest, LinkageTestF_Int, ::testing::ValuesIn(linkage_inputsf2));
-}  // namespace cuvs
diff --git a/cpp/test/distance/dist_adj.cu b/cpp/test/distance/dist_adj.cu
deleted file mode 100644
index 304744cb7..000000000
--- a/cpp/test/distance/dist_adj.cu
+++ /dev/null
@@ -1,194 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <cuvs/distance/distance.cuh>
-#include <gtest/gtest.h>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/random/rng.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <rmm/device_uvector.hpp>
-
-#include "dist_adj.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-RAFT_KERNEL naiveDistanceAdjKernel(uint8_t* dist,
-                                   const DataType* x,
-                                   const DataType* y,
-                                   int m,
-                                   int n,
-                                   int k,
-                                   DataType eps,
-                                   bool isRowMajor)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) return;
-  DataType acc = DataType(0);
-  for (int i = 0; i < k; ++i) {
-    int xidx  = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx  = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto diff = x[xidx] - y[yidx];
-    acc += diff * diff;
-  }
-  int outidx   = isRowMajor ? midx * n + nidx : midx + m * nidx;
-  dist[outidx] = acc <= eps;
-}
-
-template <typename DataType>
-void naiveDistanceAdj(uint8_t* dist,
-                      const DataType* x,
-                      const DataType* y,
-                      int m,
-                      int n,
-                      int k,
-                      DataType eps,
-                      bool isRowMajor,
-                      cudaStream_t stream)
-{
-  static const dim3 TPB(16, 32, 1);
-  dim3 nblks(raft::ceildiv(m, (int)TPB.x), raft::ceildiv(n, (int)TPB.y), 1);
-  naiveDistanceAdjKernel<DataType><<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, eps, isRowMajor);
-  RAFT_CUDA_TRY(cudaPeekAtLastError());
-}
-
-template <typename DataType>
-struct DistanceAdjInputs {
-  DataType eps;
-  int m, n, k;
-  bool isRowMajor;
-  unsigned long long int seed;
-};
-
-template <typename DataType>
-::std::ostream& operator<<(::std::ostream& os, const DistanceAdjInputs<DataType>& dims)
-{
-  return os;
-}
-
-template <typename DataType>
-class DistanceAdjTest : public ::testing::TestWithParam<DistanceAdjInputs<DataType>> {
- public:
-  DistanceAdjTest()
-    : params(::testing::TestWithParam<DistanceAdjInputs<DataType>>::GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      dist(params.m * params.n, stream),
-      dist_ref(params.m * params.n, stream)
-  {
-  }
-
-  void SetUp() override
-  {
-    raft::random::RngState r(params.seed);
-    int m           = params.m;
-    int n           = params.n;
-    int k           = params.k;
-    bool isRowMajor = params.isRowMajor;
-
-    rmm::device_uvector<DataType> x(m * k, stream);
-    rmm::device_uvector<DataType> y(n * k, stream);
-
-    uniform(handle, r, x.data(), m * k, DataType(-1.0), DataType(1.0));
-    uniform(handle, r, y.data(), n * k, DataType(-1.0), DataType(1.0));
-
-    DataType threshold = params.eps;
-
-    naiveDistanceAdj(dist_ref.data(), x.data(), y.data(), m, n, k, threshold, isRowMajor, stream);
-    size_t worksize = cuvs::distance::
-      getWorkspaceSize<cuvs::distance::DistanceType::L2Expanded, DataType, DataType, uint8_t>(
-        x.data(), y.data(), m, n, k);
-    rmm::device_uvector<char> workspace(worksize, stream);
-
-    using threshold_final_op_ = threshold_final_op<DataType, DataType, uint8_t, int>;
-    threshold_final_op_ threshold_op(threshold);
-
-    cuvs::distance::distance<cuvs::distance::DistanceType::L2Expanded,
-                             DataType,
-                             DataType,
-                             uint8_t,
-                             threshold_final_op_>(handle,
-                                                  x.data(),
-                                                  y.data(),
-                                                  dist.data(),
-                                                  m,
-                                                  n,
-                                                  k,
-                                                  workspace.data(),
-                                                  worksize,
-                                                  threshold_op,
-                                                  isRowMajor);
-    resource::sync_stream(handle, stream);
-  }
-
-  void TearDown() override {}
-
- protected:
-  DistanceAdjInputs<DataType> params;
-  // We use uint8_t even if the output in this test is a bool because
-  // cutlass doesn't support bool as output buffer yet. In cuda
-  // sizeof(bool) is 1 byte hence it doesn't increase
-  // memory consumption if we use uint8_t instead of bool.
-  rmm::device_uvector<uint8_t> dist_ref;
-  rmm::device_uvector<uint8_t> dist;
-  raft::resources handle;
-  cudaStream_t stream;
-};
-
-const std::vector<DistanceAdjInputs<float>> inputsf = {
-  {0.01f, 1024, 1024, 32, true, 1234ULL},
-  {0.1f, 1024, 1024, 32, true, 1234ULL},
-  {1.0f, 1024, 1024, 32, true, 1234ULL},
-  {10.0f, 1024, 1024, 32, true, 1234ULL},
-  {0.01f, 1024, 1024, 32, false, 1234ULL},
-  {0.1f, 1024, 1024, 32, false, 1234ULL},
-  {1.0f, 1024, 1024, 32, false, 1234ULL},
-  {10.0f, 1024, 1024, 32, false, 1234ULL},
-};
-typedef DistanceAdjTest<float> DistanceAdjTestF;
-TEST_P(DistanceAdjTestF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(devArrMatch(dist_ref.data(), dist.data(), m, n, raft::Compare<uint8_t>(), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceAdjTests, DistanceAdjTestF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceAdjInputs<double>> inputsd = {
-  {0.01, 1024, 1024, 32, true, 1234ULL},
-  {0.1, 1024, 1024, 32, true, 1234ULL},
-  {1.0, 1024, 1024, 32, true, 1234ULL},
-  {10.0, 1024, 1024, 32, true, 1234ULL},
-  {0.01, 1024, 1024, 32, false, 1234ULL},
-  {0.1, 1024, 1024, 32, false, 1234ULL},
-  {1.0, 1024, 1024, 32, false, 1234ULL},
-  {10.0, 1024, 1024, 32, false, 1234ULL},
-};
-typedef DistanceAdjTest<double> DistanceAdjTestD;
-TEST_P(DistanceAdjTestD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(devArrMatch(dist_ref.data(), dist.data(), m, n, raft::Compare<uint8_t>(), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceAdjTests, DistanceAdjTestD, ::testing::ValuesIn(inputsd));
-
-}  // namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_adj.cuh b/cpp/test/distance/dist_adj.cuh
deleted file mode 100644
index 0946c5b4e..000000000
--- a/cpp/test/distance/dist_adj.cuh
+++ /dev/null
@@ -1,71 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "dist_adj_threshold.cuh"
-#include <cuvs/distance/distance.cuh>
-
-#define instantiate_raft_distance_distance(DT, DataT, AccT, OutT, FinalLambda, IdxT)       \
-  extern template void cuvs::distance::distance<DT, DataT, AccT, OutT, FinalLambda, IdxT>( \
-    raft::resources const& handle,                                                         \
-    const DataT* x,                                                                        \
-    const DataT* y,                                                                        \
-    OutT* dist,                                                                            \
-    IdxT m,                                                                                \
-    IdxT n,                                                                                \
-    IdxT k,                                                                                \
-    void* workspace,                                                                       \
-    size_t worksize,                                                                       \
-    FinalLambda fin_op,                                                                    \
-    bool isRowMajor,                                                                       \
-    DataT metric_arg)
-
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2Expanded,
-                                   float,
-                                   float,
-                                   uint8_t,
-                                   cuvs::distance::threshold_float,
-                                   int);
-
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2Expanded,
-                                   double,
-                                   double,
-                                   uint8_t,
-                                   cuvs::distance::threshold_double,
-                                   int);
-
-#undef instantiate_raft_distance_distance
-
-#define instantiate_raft_distance_getWorkspaceSize(DistT, DataT, AccT, OutT, IdxT)         \
-  extern template size_t cuvs::distance::getWorkspaceSize<DistT, DataT, AccT, OutT, IdxT>( \
-    const DataT* x, const DataT* y, IdxT m, IdxT n, IdxT k)
-
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, float, float, uint8_t, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, double, double, uint8_t, int);
-
-#undef instantiate_raft_distance_getWorkspaceSize
-
-#define instantiate_raft_distance_getWorkspaceSize(DistT, DataT, AccT, OutT, IdxT)         \
-  extern template size_t cuvs::distance::getWorkspaceSize<DistT, DataT, AccT, OutT, IdxT>( \
-    const DataT* x, const DataT* y, IdxT m, IdxT n, IdxT k)
-
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, float, float, uint8_t, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, double, double, uint8_t, int);
-
-#undef instantiate_raft_distance_getWorkspaceSize
diff --git a/cpp/test/distance/dist_adj_distance_instance.cu b/cpp/test/distance/dist_adj_distance_instance.cu
deleted file mode 100644
index 84039db94..000000000
--- a/cpp/test/distance/dist_adj_distance_instance.cu
+++ /dev/null
@@ -1,63 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#undef RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-#include "dist_adj_threshold.cuh"
-#include <cstdint>
-#include <cuvs/distance/distance-inl.cuh>
-
-#define instantiate_raft_distance_distance(DT, DataT, AccT, OutT, FinalLambda, IdxT) \
-  template void cuvs::distance::distance<DT, DataT, AccT, OutT, FinalLambda, IdxT>(  \
-    raft::resources const& handle,                                                   \
-    const DataT* x,                                                                  \
-    const DataT* y,                                                                  \
-    OutT* dist,                                                                      \
-    IdxT m,                                                                          \
-    IdxT n,                                                                          \
-    IdxT k,                                                                          \
-    void* workspace,                                                                 \
-    size_t worksize,                                                                 \
-    FinalLambda fin_op,                                                              \
-    bool isRowMajor,                                                                 \
-    DataT metric_arg)
-
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2Expanded,
-                                   float,
-                                   float,
-                                   uint8_t,
-                                   cuvs::distance::threshold_float,
-                                   int);
-
-instantiate_raft_distance_distance(cuvs::distance::DistanceType::L2Expanded,
-                                   double,
-                                   double,
-                                   uint8_t,
-                                   cuvs::distance::threshold_double,
-                                   int);
-
-#undef instantiate_raft_distance_distance
-
-#define instantiate_raft_distance_getWorkspaceSize(DistT, DataT, AccT, OutT, IdxT)  \
-  template size_t cuvs::distance::getWorkspaceSize<DistT, DataT, AccT, OutT, IdxT>( \
-    const DataT* x, const DataT* y, IdxT m, IdxT n, IdxT k)
-
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, float, float, uint8_t, int);
-instantiate_raft_distance_getWorkspaceSize(
-  cuvs::distance::DistanceType::L2Expanded, double, double, uint8_t, int);
-
-#undef instantiate_raft_distance_getWorkspaceSize
diff --git a/cpp/test/distance/dist_adj_threshold.cuh b/cpp/test/distance/dist_adj_threshold.cuh
deleted file mode 100644
index a99c04895..000000000
--- a/cpp/test/distance/dist_adj_threshold.cuh
+++ /dev/null
@@ -1,36 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cstdint>  // uint8_t
-
-namespace cuvs::distance {
-
-template <typename AccT, typename DataT, typename OutT, typename Index>
-struct threshold_final_op {
-  DataT threshold_val;
-
-  __device__ __host__ threshold_final_op() noexcept : threshold_val(0.0) {}
-  __device__ __host__ threshold_final_op(DataT val) noexcept : threshold_val(val) {}
-  __device__ __host__ OutT operator()(AccT d_val, Index g_idx) const noexcept
-  {
-    return d_val <= threshold_val;
-  }
-};
-
-using threshold_float  = threshold_final_op<float, float, uint8_t, int>;
-using threshold_double = threshold_final_op<double, double, uint8_t, int>;
-
-}  // namespace cuvs::distance
diff --git a/cpp/test/distance/dist_canberra.cu b/cpp/test/distance/dist_canberra.cu
deleted file mode 100644
index b5a46bb9c..000000000
--- a/cpp/test/distance/dist_canberra.cu
+++ /dev/null
@@ -1,70 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceCanberra : public DistanceTest<cuvs::distance::DistanceType::Canberra, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceCanberra<float> DistanceCanberraF;
-TEST_P(DistanceCanberraF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceCanberraF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceCanberra<double> DistanceCanberraD;
-TEST_P(DistanceCanberraD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceCanberraD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixCanberra : public BigMatrixDistanceTest<cuvs::distance::DistanceType::Canberra> {};
-TEST_F(BigMatrixCanberra, Result) {}
-
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_correlation.cu b/cpp/test/distance/dist_correlation.cu
deleted file mode 100644
index dfef3f07a..000000000
--- a/cpp/test/distance/dist_correlation.cu
+++ /dev/null
@@ -1,94 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceCorrelation
-  : public DistanceTest<cuvs::distance::DistanceType::CorrelationExpanded, DataType> {};
-
-template <typename DataType>
-class DistanceCorrelationXequalY
-  : public DistanceTestSameBuffer<cuvs::distance::DistanceType::CorrelationExpanded, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceCorrelation<float> DistanceCorrelationF;
-TEST_P(DistanceCorrelationF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceCorrelationF, ::testing::ValuesIn(inputsf));
-
-typedef DistanceCorrelationXequalY<float> DistanceCorrelationXequalYF;
-TEST_P(DistanceCorrelationXequalYF, Result)
-{
-  int m = params.m;
-  ASSERT_TRUE(raft::devArrMatch(dist_ref[0].data(),
-                                dist[0].data(),
-                                m,
-                                m,
-                                raft::CompareApprox<float>(params.tolerance),
-                                stream));
-  ASSERT_TRUE(raft::devArrMatch(dist_ref[1].data(),
-                                dist[1].data(),
-                                m / 2,
-                                m,
-                                raft::CompareApprox<float>(params.tolerance),
-                                stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceCorrelationXequalYF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceCorrelation<double> DistanceCorrelationD;
-TEST_P(DistanceCorrelationD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceCorrelationD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixCorrelation
-  : public BigMatrixDistanceTest<cuvs::distance::DistanceType::CorrelationExpanded> {};
-TEST_F(BigMatrixCorrelation, Result) {}
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_cos.cu b/cpp/test/distance/dist_cos.cu
deleted file mode 100644
index 60d5c7a62..000000000
--- a/cpp/test/distance/dist_cos.cu
+++ /dev/null
@@ -1,110 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceExpCos : public DistanceTest<cuvs::distance::DistanceType::CosineExpanded, DataType> {
-};
-
-template <typename DataType>
-class DistanceExpCosXequalY
-  : public DistanceTestSameBuffer<cuvs::distance::DistanceType::CosineExpanded, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-
-const std::vector<DistanceInputs<float>> inputsXeqYf = {
-  {0.01f, 1024, 1024, 32, true, 1234ULL},
-  {0.01f, 1024, 32, 1024, true, 1234ULL},
-  {0.01f, 32, 1024, 1024, true, 1234ULL},
-  {0.03f, 1024, 1024, 1024, true, 1234ULL},
-  {0.01f, 1024, 1024, 32, false, 1234ULL},
-  {0.01f, 1024, 32, 1024, false, 1234ULL},
-  {0.01f, 32, 1024, 1024, false, 1234ULL},
-  {0.03f, 1024, 1024, 1024, false, 1234ULL},
-};
-
-typedef DistanceExpCos<float> DistanceExpCosF;
-TEST_P(DistanceExpCosF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceExpCosF, ::testing::ValuesIn(inputsf));
-
-typedef DistanceExpCosXequalY<float> DistanceExpCosXequalYF;
-TEST_P(DistanceExpCosXequalYF, Result)
-{
-  int m = params.m;
-  int n = params.m;
-  ASSERT_TRUE(raft::devArrMatch(dist_ref[0].data(),
-                                dist[0].data(),
-                                m,
-                                n,
-                                raft::CompareApprox<float>(params.tolerance),
-                                stream));
-  n = params.isRowMajor ? m : m / 2;
-  m = params.isRowMajor ? m / 2 : m;
-
-  ASSERT_TRUE(raft::devArrMatch(dist_ref[1].data(),
-                                dist[1].data(),
-                                m,
-                                n,
-                                raft::CompareApprox<float>(params.tolerance),
-                                stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceExpCosXequalYF, ::testing::ValuesIn(inputsXeqYf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceExpCos<double> DistanceExpCosD;
-TEST_P(DistanceExpCosD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceExpCosD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixCos : public BigMatrixDistanceTest<cuvs::distance::DistanceType::CosineExpanded> {};
-TEST_F(BigMatrixCos, Result) {}
-
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_hamming.cu b/cpp/test/distance/dist_hamming.cu
deleted file mode 100644
index a27d9acbc..000000000
--- a/cpp/test/distance/dist_hamming.cu
+++ /dev/null
@@ -1,71 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceHamming
-  : public DistanceTest<cuvs::distance::DistanceType::HammingUnexpanded, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceHamming<float> DistanceHammingF;
-TEST_P(DistanceHammingF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceHammingF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceHamming<double> DistanceHammingD;
-TEST_P(DistanceHammingD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceHammingD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixHamming
-  : public BigMatrixDistanceTest<cuvs::distance::DistanceType::HammingUnexpanded> {};
-TEST_F(BigMatrixHamming, Result) {}
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_hellinger.cu b/cpp/test/distance/dist_hellinger.cu
deleted file mode 100644
index 19cbf4643..000000000
--- a/cpp/test/distance/dist_hellinger.cu
+++ /dev/null
@@ -1,71 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceHellingerExp
-  : public DistanceTest<cuvs::distance::DistanceType::HellingerExpanded, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceHellingerExp<float> DistanceHellingerExpF;
-TEST_P(DistanceHellingerExpF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceHellingerExpF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceHellingerExp<double> DistanceHellingerExpD;
-TEST_P(DistanceHellingerExpD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceHellingerExpD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixHellingerExp
-  : public BigMatrixDistanceTest<cuvs::distance::DistanceType::HellingerExpanded> {};
-TEST_F(BigMatrixHellingerExp, Result) {}
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_inner_product.cu b/cpp/test/distance/dist_inner_product.cu
deleted file mode 100644
index 91a9f4508..000000000
--- a/cpp/test/distance/dist_inner_product.cu
+++ /dev/null
@@ -1,74 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceInnerProduct
-  : public DistanceTest<cuvs::distance::DistanceType::InnerProduct, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 10, 5, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceInnerProduct<float> DistanceInnerProductF;
-TEST_P(DistanceInnerProductF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-
-  ASSERT_TRUE(devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceInnerProductF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceInnerProduct<double> DistanceInnerProductD;
-TEST_P(DistanceInnerProductD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-
-  ASSERT_TRUE(devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceInnerProductD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixInnerProduct
-  : public BigMatrixDistanceTest<cuvs::distance::DistanceType::InnerProduct> {};
-TEST_F(BigMatrixInnerProduct, Result) {}
-
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_jensen_shannon.cu b/cpp/test/distance/dist_jensen_shannon.cu
deleted file mode 100644
index af0bdf5e0..000000000
--- a/cpp/test/distance/dist_jensen_shannon.cu
+++ /dev/null
@@ -1,71 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceJensenShannon
-  : public DistanceTest<cuvs::distance::DistanceType::JensenShannon, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceJensenShannon<float> DistanceJensenShannonF;
-TEST_P(DistanceJensenShannonF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceJensenShannonF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceJensenShannon<double> DistanceJensenShannonD;
-TEST_P(DistanceJensenShannonD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceJensenShannonD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixJensenShannon
-  : public BigMatrixDistanceTest<cuvs::distance::DistanceType::JensenShannon> {};
-TEST_F(BigMatrixJensenShannon, Result) {}
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_kl_divergence.cu b/cpp/test/distance/dist_kl_divergence.cu
deleted file mode 100644
index db566f1d3..000000000
--- a/cpp/test/distance/dist_kl_divergence.cu
+++ /dev/null
@@ -1,71 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceKLDivergence
-  : public DistanceTest<cuvs::distance::DistanceType::KLDivergence, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceKLDivergence<float> DistanceKLDivergenceF;
-TEST_P(DistanceKLDivergenceF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceKLDivergenceF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceKLDivergence<double> DistanceKLDivergenceD;
-TEST_P(DistanceKLDivergenceD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceKLDivergenceD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixKLDivergence
-  : public BigMatrixDistanceTest<cuvs::distance::DistanceType::KLDivergence> {};
-TEST_F(BigMatrixKLDivergence, Result) {}
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_l1.cu b/cpp/test/distance/dist_l1.cu
deleted file mode 100644
index 9c2b80160..000000000
--- a/cpp/test/distance/dist_l1.cu
+++ /dev/null
@@ -1,70 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceUnexpL1 : public DistanceTest<cuvs::distance::DistanceType::L1, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceUnexpL1<float> DistanceUnexpL1F;
-TEST_P(DistanceUnexpL1F, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceUnexpL1F, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceUnexpL1<double> DistanceUnexpL1D;
-TEST_P(DistanceUnexpL1D, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceUnexpL1D, ::testing::ValuesIn(inputsd));
-
-class BigMatrixUnexpL1 : public BigMatrixDistanceTest<cuvs::distance::DistanceType::L1> {};
-TEST_F(BigMatrixUnexpL1, Result) {}
-
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_l2_exp.cu b/cpp/test/distance/dist_l2_exp.cu
deleted file mode 100644
index 5fc3deed9..000000000
--- a/cpp/test/distance/dist_l2_exp.cu
+++ /dev/null
@@ -1,113 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceEucExpTest : public DistanceTest<cuvs::distance::DistanceType::L2Expanded, DataType> {
-};
-
-template <typename DataType>
-class DistanceEucExpTestXequalY
-  : public DistanceTestSameBuffer<cuvs::distance::DistanceType::L2Expanded, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 2048, 4096, 128, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.003f, 1021, 1021, 1021, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-  {0.003f, 1021, 1021, 1021, false, 1234ULL},
-};
-
-const std::vector<DistanceInputs<float>> inputsXeqYf = {
-  {0.01f, 2048, 4096, 128, true, 1234ULL},
-  {0.01f, 1024, 1024, 32, true, 1234ULL},
-  {0.01f, 1024, 32, 1024, true, 1234ULL},
-  {0.01f, 32, 1024, 1024, true, 1234ULL},
-  {0.03f, 1024, 1024, 1024, true, 1234ULL},
-  {0.03f, 1021, 1021, 1021, true, 1234ULL},
-  {0.01f, 1024, 1024, 32, false, 1234ULL},
-  {0.01f, 1024, 32, 1024, false, 1234ULL},
-  {0.01f, 32, 1024, 1024, false, 1234ULL},
-  {0.03f, 1024, 1024, 1024, false, 1234ULL},
-  {0.03f, 1021, 1021, 1021, false, 1234ULL},
-};
-
-typedef DistanceEucExpTest<float> DistanceEucExpTestF;
-TEST_P(DistanceEucExpTestF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceEucExpTestF, ::testing::ValuesIn(inputsf));
-
-typedef DistanceEucExpTestXequalY<float> DistanceEucExpTestXequalYF;
-TEST_P(DistanceEucExpTestXequalYF, Result)
-{
-  int m = params.m;
-  ASSERT_TRUE(raft::devArrMatch(dist_ref[0].data(),
-                                dist[0].data(),
-                                m,
-                                m,
-                                raft::CompareApprox<float>(params.tolerance),
-                                stream));
-  ASSERT_TRUE(raft::devArrMatch(dist_ref[1].data(),
-                                dist[1].data(),
-                                m / 2,
-                                m,
-                                raft::CompareApprox<float>(params.tolerance),
-                                stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests,
-                        DistanceEucExpTestXequalYF,
-                        ::testing::ValuesIn(inputsXeqYf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceEucExpTest<double> DistanceEucExpTestD;
-TEST_P(DistanceEucExpTestD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceEucExpTestD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixEucExp : public BigMatrixDistanceTest<cuvs::distance::DistanceType::L2Expanded> {};
-TEST_F(BigMatrixEucExp, Result) {}
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_l2_sqrt_exp.cu b/cpp/test/distance/dist_l2_sqrt_exp.cu
deleted file mode 100644
index e71828df9..000000000
--- a/cpp/test/distance/dist_l2_sqrt_exp.cu
+++ /dev/null
@@ -1,74 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceEucSqrtExpTest
-  : public DistanceTest<cuvs::distance::DistanceType::L2SqrtExpanded, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 2048, 4096, 128, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.003f, 1021, 1021, 1021, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-  {0.003f, 1021, 1021, 1021, false, 1234ULL},
-};
-typedef DistanceEucSqrtExpTest<float> DistanceEucSqrtExpTestF;
-TEST_P(DistanceEucSqrtExpTestF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceEucSqrtExpTestF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceEucSqrtExpTest<double> DistanceEucSqrtExpTestD;
-TEST_P(DistanceEucSqrtExpTestD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceEucSqrtExpTestD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixEucSqrtExp
-  : public BigMatrixDistanceTest<cuvs::distance::DistanceType::L2SqrtExpanded> {};
-TEST_F(BigMatrixEucSqrtExp, Result) {}
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_l2_unexp.cu b/cpp/test/distance/dist_l2_unexp.cu
deleted file mode 100644
index b832495ea..000000000
--- a/cpp/test/distance/dist_l2_unexp.cu
+++ /dev/null
@@ -1,71 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceEucUnexpTest
-  : public DistanceTest<cuvs::distance::DistanceType::L2Unexpanded, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceEucUnexpTest<float> DistanceEucUnexpTestF;
-TEST_P(DistanceEucUnexpTestF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceEucUnexpTestF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceEucUnexpTest<double> DistanceEucUnexpTestD;
-TEST_P(DistanceEucUnexpTestD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceEucUnexpTestD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixEucUnexp : public BigMatrixDistanceTest<cuvs::distance::DistanceType::L2Unexpanded> {
-};
-TEST_F(BigMatrixEucUnexp, Result) {}
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_l_inf.cu b/cpp/test/distance/dist_l_inf.cu
deleted file mode 100644
index f1999772f..000000000
--- a/cpp/test/distance/dist_l_inf.cu
+++ /dev/null
@@ -1,70 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceLinf : public DistanceTest<cuvs::distance::DistanceType::Linf, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceLinf<float> DistanceLinfF;
-TEST_P(DistanceLinfF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceLinfF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceLinf<double> DistanceLinfD;
-TEST_P(DistanceLinfD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceLinfD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixLinf : public BigMatrixDistanceTest<cuvs::distance::DistanceType::Linf> {};
-TEST_F(BigMatrixLinf, Result) {}
-
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_lp_unexp.cu b/cpp/test/distance/dist_lp_unexp.cu
deleted file mode 100644
index 12971f43d..000000000
--- a/cpp/test/distance/dist_lp_unexp.cu
+++ /dev/null
@@ -1,71 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceLpUnexp : public DistanceTest<cuvs::distance::DistanceType::LpUnexpanded, DataType> {
-};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL, 4.0f},
-  {0.001f, 1024, 32, 1024, true, 1234ULL, 3.0f},
-  {0.001f, 32, 1024, 1024, true, 1234ULL, 4.0f},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL, 3.0f},
-  {0.001f, 1024, 1024, 32, false, 1234ULL, 4.0f},
-  {0.001f, 1024, 32, 1024, false, 1234ULL, 3.0f},
-  {0.001f, 32, 1024, 1024, false, 1234ULL, 4.0f},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL, 3.0f},
-};
-typedef DistanceLpUnexp<float> DistanceLpUnexpF;
-TEST_P(DistanceLpUnexpF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceLpUnexpF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL, 4.0},
-  {0.001, 1024, 32, 1024, true, 1234ULL, 3.0},
-  {0.001, 32, 1024, 1024, true, 1234ULL, 4.0},
-  {0.003, 1024, 1024, 1024, true, 1234ULL, 3.0},
-  {0.001, 1024, 1024, 32, false, 1234ULL, 4.0},
-  {0.001, 1024, 32, 1024, false, 1234ULL, 3.0},
-  {0.001, 32, 1024, 1024, false, 1234ULL, 4.0},
-  {0.003, 1024, 1024, 1024, false, 1234ULL, 3.0},
-};
-typedef DistanceLpUnexp<double> DistanceLpUnexpD;
-TEST_P(DistanceLpUnexpD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceLpUnexpD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixLpUnexp : public BigMatrixDistanceTest<cuvs::distance::DistanceType::LpUnexpanded> {
-};
-TEST_F(BigMatrixLpUnexp, Result) {}
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/dist_russell_rao.cu b/cpp/test/distance/dist_russell_rao.cu
deleted file mode 100644
index 1693bb2e5..000000000
--- a/cpp/test/distance/dist_russell_rao.cu
+++ /dev/null
@@ -1,71 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "distance_base.cuh"
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-class DistanceRussellRao
-  : public DistanceTest<cuvs::distance::DistanceType::RusselRaoExpanded, DataType> {};
-
-const std::vector<DistanceInputs<float>> inputsf = {
-  {0.001f, 1024, 1024, 32, true, 1234ULL},
-  {0.001f, 1024, 32, 1024, true, 1234ULL},
-  {0.001f, 32, 1024, 1024, true, 1234ULL},
-  {0.003f, 1024, 1024, 1024, true, 1234ULL},
-  {0.001f, 1024, 1024, 32, false, 1234ULL},
-  {0.001f, 1024, 32, 1024, false, 1234ULL},
-  {0.001f, 32, 1024, 1024, false, 1234ULL},
-  {0.003f, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceRussellRao<float> DistanceRussellRaoF;
-TEST_P(DistanceRussellRaoF, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceRussellRaoF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DistanceInputs<double>> inputsd = {
-  {0.001, 1024, 1024, 32, true, 1234ULL},
-  {0.001, 1024, 32, 1024, true, 1234ULL},
-  {0.001, 32, 1024, 1024, true, 1234ULL},
-  {0.003, 1024, 1024, 1024, true, 1234ULL},
-  {0.001, 1024, 1024, 32, false, 1234ULL},
-  {0.001, 1024, 32, 1024, false, 1234ULL},
-  {0.001, 32, 1024, 1024, false, 1234ULL},
-  {0.003, 1024, 1024, 1024, false, 1234ULL},
-};
-typedef DistanceRussellRao<double> DistanceRussellRaoD;
-TEST_P(DistanceRussellRaoD, Result)
-{
-  int m = params.isRowMajor ? params.m : params.n;
-  int n = params.isRowMajor ? params.n : params.m;
-  ASSERT_TRUE(raft::devArrMatch(
-    dist_ref.data(), dist.data(), m, n, raft::CompareApprox<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(DistanceTests, DistanceRussellRaoD, ::testing::ValuesIn(inputsd));
-
-class BigMatrixRussellRao
-  : public BigMatrixDistanceTest<cuvs::distance::DistanceType::RusselRaoExpanded> {};
-TEST_F(BigMatrixRussellRao, Result) {}
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/distance_base.cuh b/cpp/test/distance/distance_base.cuh
deleted file mode 100644
index 72c6ebf7b..000000000
--- a/cpp/test/distance/distance_base.cuh
+++ /dev/null
@@ -1,673 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <raft/common/nvtx.hpp>  // common::nvtx::range
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cuvs/distance/distance.cuh>
-#include <cuvs/distance/distance_types.hpp>  // cuvs::distance::DistanceType
-#include <raft/core/device_mdspan.hpp>       // make_device_matrix_view
-#include <raft/core/operators.hpp>           // raft::sqrt
-#include <raft/core/resources.hpp>           // raft::resources
-#include <raft/random/rng.cuh>
-#include <rmm/device_uvector.hpp>  // rmm::device_uvector
-
-namespace raft {
-namespace distance {
-
-template <typename DataType>
-RAFT_KERNEL naiveDistanceKernel(DataType* dist,
-                                const DataType* x,
-                                const DataType* y,
-                                int m,
-                                int n,
-                                int k,
-                                cuvs::distance::DistanceType type,
-                                bool isRowMajor)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) return;
-  DataType acc = DataType(0);
-  for (int i = 0; i < k; ++i) {
-    int xidx  = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx  = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto diff = x[xidx] - y[yidx];
-    acc += diff * diff;
-  }
-  if (type == cuvs::distance::DistanceType::L2SqrtExpanded ||
-      type == cuvs::distance::DistanceType::L2SqrtUnexpanded)
-    acc = raft::sqrt(acc);
-  int outidx   = isRowMajor ? midx * n + nidx : midx + m * nidx;
-  dist[outidx] = acc;
-}
-
-template <typename DataType>
-RAFT_KERNEL naiveL1_Linf_CanberraDistanceKernel(DataType* dist,
-                                                const DataType* x,
-                                                const DataType* y,
-                                                int m,
-                                                int n,
-                                                int k,
-                                                cuvs::distance::DistanceType type,
-                                                bool isRowMajor)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) { return; }
-
-  DataType acc = DataType(0);
-  for (int i = 0; i < k; ++i) {
-    int xidx  = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx  = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto a    = x[xidx];
-    auto b    = y[yidx];
-    auto diff = (a > b) ? (a - b) : (b - a);
-    if (type == cuvs::distance::DistanceType::Linf) {
-      acc = raft::max(acc, diff);
-    } else if (type == cuvs::distance::DistanceType::Canberra) {
-      const auto add = raft::abs(a) + raft::abs(b);
-      // deal with potential for 0 in denominator by
-      // forcing 1/0 instead
-      acc += ((add != 0) * diff / (add + (add == 0)));
-    } else {
-      acc += diff;
-    }
-  }
-
-  int outidx   = isRowMajor ? midx * n + nidx : midx + m * nidx;
-  dist[outidx] = acc;
-}
-
-template <typename DataType>
-RAFT_KERNEL naiveCosineDistanceKernel(
-  DataType* dist, const DataType* x, const DataType* y, int m, int n, int k, bool isRowMajor)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) { return; }
-
-  DataType acc_a  = DataType(0);
-  DataType acc_b  = DataType(0);
-  DataType acc_ab = DataType(0);
-
-  for (int i = 0; i < k; ++i) {
-    int xidx = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto a   = x[xidx];
-    auto b   = y[yidx];
-    acc_a += a * a;
-    acc_b += b * b;
-    acc_ab += a * b;
-  }
-
-  int outidx = isRowMajor ? midx * n + nidx : midx + m * nidx;
-
-  // Use 1.0 - (cosine similarity) to calc the distance
-  dist[outidx] = (DataType)1.0 - acc_ab / (raft::sqrt(acc_a) * raft::sqrt(acc_b));
-}
-
-template <typename DataType>
-RAFT_KERNEL naiveInnerProductKernel(
-  DataType* dist, const DataType* x, const DataType* y, int m, int n, int k, bool isRowMajor)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) { return; }
-
-  DataType acc_ab = DataType(0);
-
-  for (int i = 0; i < k; ++i) {
-    int xidx = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto a   = x[xidx];
-    auto b   = y[yidx];
-    acc_ab += a * b;
-  }
-
-  int outidx   = isRowMajor ? midx * n + nidx : midx + m * nidx;
-  dist[outidx] = acc_ab;
-}
-
-template <typename DataType>
-RAFT_KERNEL naiveHellingerDistanceKernel(
-  DataType* dist, const DataType* x, const DataType* y, int m, int n, int k, bool isRowMajor)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) { return; }
-
-  DataType acc_ab = DataType(0);
-
-  for (int i = 0; i < k; ++i) {
-    int xidx = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto a   = x[xidx];
-    auto b   = y[yidx];
-    acc_ab += raft::sqrt(a) * raft::sqrt(b);
-  }
-
-  int outidx = isRowMajor ? midx * n + nidx : midx + m * nidx;
-
-  // Adjust to replace NaN in sqrt with 0 if input to sqrt is negative
-  acc_ab         = 1 - acc_ab;
-  auto rectifier = (!signbit(acc_ab));
-  dist[outidx]   = raft::sqrt(rectifier * acc_ab);
-}
-
-template <typename DataType>
-RAFT_KERNEL naiveLpUnexpDistanceKernel(DataType* dist,
-                                       const DataType* x,
-                                       const DataType* y,
-                                       int m,
-                                       int n,
-                                       int k,
-                                       bool isRowMajor,
-                                       DataType p)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) return;
-  DataType acc = DataType(0);
-  for (int i = 0; i < k; ++i) {
-    int xidx  = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx  = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto a    = x[xidx];
-    auto b    = y[yidx];
-    auto diff = raft::abs(a - b);
-    acc += raft::pow(diff, p);
-  }
-  auto one_over_p = 1 / p;
-  acc             = raft::pow(acc, one_over_p);
-  int outidx      = isRowMajor ? midx * n + nidx : midx + m * nidx;
-  dist[outidx]    = acc;
-}
-
-template <typename DataType>
-RAFT_KERNEL naiveHammingDistanceKernel(
-  DataType* dist, const DataType* x, const DataType* y, int m, int n, int k, bool isRowMajor)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) return;
-  DataType acc = DataType(0);
-  for (int i = 0; i < k; ++i) {
-    int xidx = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto a   = x[xidx];
-    auto b   = y[yidx];
-    acc += (a != b);
-  }
-  acc          = acc / k;
-  int outidx   = isRowMajor ? midx * n + nidx : midx + m * nidx;
-  dist[outidx] = acc;
-}
-
-template <typename DataType>
-RAFT_KERNEL naiveJensenShannonDistanceKernel(
-  DataType* dist, const DataType* x, const DataType* y, int m, int n, int k, bool isRowMajor)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) return;
-  DataType acc = DataType(0);
-  for (int i = 0; i < k; ++i) {
-    int xidx = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto a   = x[xidx];
-    auto b   = y[yidx];
-
-    DataType m  = 0.5f * (a + b);
-    bool a_zero = a == 0;
-    bool b_zero = b == 0;
-
-    DataType p = (!a_zero * m) / (a_zero + a);
-    DataType q = (!b_zero * m) / (b_zero + b);
-
-    bool p_zero = p == 0;
-    bool q_zero = q == 0;
-
-    acc += (-a * (!p_zero * log(p + p_zero))) + (-b * (!q_zero * log(q + q_zero)));
-  }
-  acc          = raft::sqrt(0.5f * acc);
-  int outidx   = isRowMajor ? midx * n + nidx : midx + m * nidx;
-  dist[outidx] = acc;
-}
-
-template <typename DataType, typename OutType>
-RAFT_KERNEL naiveRussellRaoDistanceKernel(
-  OutType* dist, const DataType* x, const DataType* y, int m, int n, int k, bool isRowMajor)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) return;
-  OutType acc = OutType(0);
-  for (int i = 0; i < k; ++i) {
-    int xidx = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto a   = x[xidx];
-    auto b   = y[yidx];
-    acc += (a * b);
-  }
-  acc          = (k - acc) / k;
-  int outidx   = isRowMajor ? midx * n + nidx : midx + m * nidx;
-  dist[outidx] = acc;
-}
-
-template <typename DataType, typename OutType>
-RAFT_KERNEL naiveKLDivergenceDistanceKernel(
-  OutType* dist, const DataType* x, const DataType* y, int m, int n, int k, bool isRowMajor)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) return;
-  OutType acc = OutType(0);
-  for (int i = 0; i < k; ++i) {
-    int xidx    = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx    = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto a      = x[xidx];
-    auto b      = y[yidx];
-    bool b_zero = (b == 0);
-    bool a_zero = (a == 0);
-    acc += a * (log(a + a_zero) - log(b + b_zero));
-  }
-  acc          = 0.5f * acc;
-  int outidx   = isRowMajor ? midx * n + nidx : midx + m * nidx;
-  dist[outidx] = acc;
-}
-
-template <typename DataType, typename OutType>
-RAFT_KERNEL naiveCorrelationDistanceKernel(
-  OutType* dist, const DataType* x, const DataType* y, int m, int n, int k, bool isRowMajor)
-{
-  int midx = threadIdx.x + blockIdx.x * blockDim.x;
-  int nidx = threadIdx.y + blockIdx.y * blockDim.y;
-  if (midx >= m || nidx >= n) return;
-  OutType acc    = OutType(0);
-  auto a_norm    = DataType(0);
-  auto b_norm    = DataType(0);
-  auto a_sq_norm = DataType(0);
-  auto b_sq_norm = DataType(0);
-  for (int i = 0; i < k; ++i) {
-    int xidx = isRowMajor ? i + midx * k : i * m + midx;
-    int yidx = isRowMajor ? i + nidx * k : i * n + nidx;
-    auto a   = x[xidx];
-    auto b   = y[yidx];
-    a_norm += a;
-    b_norm += b;
-    a_sq_norm += (a * a);
-    b_sq_norm += (b * b);
-    acc += (a * b);
-  }
-
-  auto numer   = k * acc - (a_norm * b_norm);
-  auto Q_denom = k * a_sq_norm - (a_norm * a_norm);
-  auto R_denom = k * b_sq_norm - (b_norm * b_norm);
-
-  acc = 1 - (numer / raft::sqrt(Q_denom * R_denom));
-
-  int outidx   = isRowMajor ? midx * n + nidx : midx + m * nidx;
-  dist[outidx] = acc;
-}
-
-template <typename DataType>
-void naiveDistance(DataType* dist,
-                   const DataType* x,
-                   const DataType* y,
-                   int m,
-                   int n,
-                   int k,
-                   cuvs::distance::DistanceType type,
-                   bool isRowMajor,
-                   DataType metric_arg = 2.0f,
-                   cudaStream_t stream = 0)
-{
-  static const dim3 TPB(16, 32, 1);
-  dim3 nblks(raft::ceildiv(m, (int)TPB.x), raft::ceildiv(n, (int)TPB.y), 1);
-
-  switch (type) {
-    case cuvs::distance::DistanceType::Canberra:
-    case cuvs::distance::DistanceType::Linf:
-    case cuvs::distance::DistanceType::L1:
-      naiveL1_Linf_CanberraDistanceKernel<DataType>
-        <<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, type, isRowMajor);
-      break;
-    case cuvs::distance::DistanceType::L2SqrtUnexpanded:
-    case cuvs::distance::DistanceType::L2Unexpanded:
-    case cuvs::distance::DistanceType::L2SqrtExpanded:
-    case cuvs::distance::DistanceType::L2Expanded:
-      naiveDistanceKernel<DataType>
-        <<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, type, isRowMajor);
-      break;
-    case cuvs::distance::DistanceType::CosineExpanded:
-      naiveCosineDistanceKernel<DataType>
-        <<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, isRowMajor);
-      break;
-    case cuvs::distance::DistanceType::HellingerExpanded:
-      naiveHellingerDistanceKernel<DataType>
-        <<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, isRowMajor);
-      break;
-    case cuvs::distance::DistanceType::LpUnexpanded:
-      naiveLpUnexpDistanceKernel<DataType>
-        <<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, isRowMajor, metric_arg);
-      break;
-    case cuvs::distance::DistanceType::HammingUnexpanded:
-      naiveHammingDistanceKernel<DataType>
-        <<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, isRowMajor);
-      break;
-    case cuvs::distance::DistanceType::InnerProduct:
-      naiveInnerProductKernel<DataType><<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, isRowMajor);
-      break;
-    case cuvs::distance::DistanceType::JensenShannon:
-      naiveJensenShannonDistanceKernel<DataType>
-        <<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, isRowMajor);
-      break;
-    case cuvs::distance::DistanceType::RusselRaoExpanded:
-      naiveRussellRaoDistanceKernel<DataType>
-        <<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, isRowMajor);
-      break;
-    case cuvs::distance::DistanceType::KLDivergence:
-      naiveKLDivergenceDistanceKernel<DataType>
-        <<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, isRowMajor);
-      break;
-    case cuvs::distance::DistanceType::CorrelationExpanded:
-      naiveCorrelationDistanceKernel<DataType>
-        <<<nblks, TPB, 0, stream>>>(dist, x, y, m, n, k, isRowMajor);
-      break;
-    default: FAIL() << "should be here\n";
-  }
-  RAFT_CUDA_TRY(cudaPeekAtLastError());
-}
-
-template <typename DataType>
-struct DistanceInputs {
-  DataType tolerance;
-  int m, n, k;
-  bool isRowMajor;
-  unsigned long long int seed;
-  DataType metric_arg = 2.0f;
-};
-
-template <typename DataType>
-::std::ostream& operator<<(::std::ostream& os, const DistanceInputs<DataType>& dims)
-{
-  return os;
-}
-
-// TODO: Remove when mdspan-based raft::runtime::distance::pairwise_distance is
-// implemented.
-//
-// Context:
-// https://github.com/rapidsai/raft/issues/1338
-template <typename layout>
-constexpr bool layout_to_row_major();
-
-template <>
-constexpr bool layout_to_row_major<layout_c_contiguous>()
-{
-  return true;
-}
-template <>
-constexpr bool layout_to_row_major<layout_f_contiguous>()
-{
-  return false;
-}
-
-template <cuvs::distance::DistanceType distanceType, typename DataType, typename layout>
-void distanceLauncher(raft::resources const& handle,
-                      DataType* x,
-                      DataType* y,
-                      DataType* dist,
-                      DataType* dist2,
-                      int m,
-                      int n,
-                      int k,
-                      DistanceInputs<DataType>& params,
-                      DataType threshold,
-                      DataType metric_arg = 2.0f)
-{
-  auto x_v    = make_device_matrix_view<DataType, int, layout>(x, m, k);
-  auto y_v    = make_device_matrix_view<DataType, int, layout>(y, n, k);
-  auto dist_v = make_device_matrix_view<DataType, int, layout>(dist, m, n);
-
-  cuvs::distance::distance<distanceType, DataType, DataType, DataType, layout>(
-    handle, x_v, y_v, dist_v, metric_arg);
-}
-
-template <cuvs::distance::DistanceType distanceType, typename DataType>
-class DistanceTest : public ::testing::TestWithParam<DistanceInputs<DataType>> {
- public:
-  DistanceTest()
-    : params(::testing::TestWithParam<DistanceInputs<DataType>>::GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      x(params.m * params.k, stream),
-      y(params.n * params.k, stream),
-      dist_ref(params.m * params.n, stream),
-      dist(params.m * params.n, stream),
-      dist2(params.m * params.n, stream)
-  {
-  }
-
-  void SetUp() override
-  {
-    auto testInfo = testing::UnitTest::GetInstance()->current_test_info();
-    common::nvtx::range fun_scope("test::%s/%s", testInfo->test_suite_name(), testInfo->name());
-
-    raft::random::RngState r(params.seed);
-    int m               = params.m;
-    int n               = params.n;
-    int k               = params.k;
-    DataType metric_arg = params.metric_arg;
-    bool isRowMajor     = params.isRowMajor;
-    if (distanceType == cuvs::distance::DistanceType::HellingerExpanded ||
-        distanceType == cuvs::distance::DistanceType::JensenShannon ||
-        distanceType == cuvs::distance::DistanceType::KLDivergence) {
-      // Hellinger works only on positive numbers
-      uniform(handle, r, x.data(), m * k, DataType(0.0), DataType(1.0));
-      uniform(handle, r, y.data(), n * k, DataType(0.0), DataType(1.0));
-    } else if (distanceType == cuvs::distance::DistanceType::RusselRaoExpanded) {
-      uniform(handle, r, x.data(), m * k, DataType(0.0), DataType(1.0));
-      uniform(handle, r, y.data(), n * k, DataType(0.0), DataType(1.0));
-      // Russel rao works on boolean values.
-      bernoulli(handle, r, x.data(), m * k, 0.5f);
-      bernoulli(handle, r, y.data(), n * k, 0.5f);
-    } else {
-      uniform(handle, r, x.data(), m * k, DataType(-1.0), DataType(1.0));
-      uniform(handle, r, y.data(), n * k, DataType(-1.0), DataType(1.0));
-    }
-    naiveDistance(
-      dist_ref.data(), x.data(), y.data(), m, n, k, distanceType, isRowMajor, metric_arg, stream);
-
-    DataType threshold = -10000.f;
-
-    if (isRowMajor) {
-      distanceLauncher<distanceType, DataType, layout_c_contiguous>(handle,
-                                                                    x.data(),
-                                                                    y.data(),
-                                                                    dist.data(),
-                                                                    dist2.data(),
-                                                                    m,
-                                                                    n,
-                                                                    k,
-                                                                    params,
-                                                                    threshold,
-                                                                    metric_arg);
-
-    } else {
-      distanceLauncher<distanceType, DataType, layout_f_contiguous>(handle,
-                                                                    x.data(),
-                                                                    y.data(),
-                                                                    dist.data(),
-                                                                    dist2.data(),
-                                                                    m,
-                                                                    n,
-                                                                    k,
-                                                                    params,
-                                                                    threshold,
-                                                                    metric_arg);
-    }
-    resource::sync_stream(handle, stream);
-  }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream;
-
-  DistanceInputs<DataType> params;
-  rmm::device_uvector<DataType> x, y, dist_ref, dist, dist2;
-};
-
-/*
- * This test suite verifies the path when X and Y are same buffer,
- * distance metrics which requires norms like L2 expanded/cosine/correlation
- * takes a more optimal path in such case to skip norm calculation for Y buffer.
- * It may happen that though both X and Y are same buffer but user passes
- * different dimensions for them like in case of tiled_brute_force_knn.
- */
-template <cuvs::distance::DistanceType distanceType, typename DataType>
-class DistanceTestSameBuffer : public ::testing::TestWithParam<DistanceInputs<DataType>> {
- public:
-  using dev_vector = rmm::device_uvector<DataType>;
-  DistanceTestSameBuffer()
-    : params(::testing::TestWithParam<DistanceInputs<DataType>>::GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      x(params.m * params.k, stream),
-      dist_ref({dev_vector(params.m * params.m, stream), dev_vector(params.m * params.m, stream)}),
-      dist({dev_vector(params.m * params.m, stream), dev_vector(params.m * params.m, stream)}),
-      dist2({dev_vector(params.m * params.m, stream), dev_vector(params.m * params.m, stream)})
-  {
-  }
-
-  void SetUp() override
-  {
-    auto testInfo = testing::UnitTest::GetInstance()->current_test_info();
-    common::nvtx::range fun_scope("test::%s/%s", testInfo->test_suite_name(), testInfo->name());
-
-    raft::random::RngState r(params.seed);
-    int m               = params.m;
-    int n               = params.m;
-    int k               = params.k;
-    DataType metric_arg = params.metric_arg;
-    bool isRowMajor     = params.isRowMajor;
-    if (distanceType == cuvs::distance::DistanceType::HellingerExpanded ||
-        distanceType == cuvs::distance::DistanceType::JensenShannon ||
-        distanceType == cuvs::distance::DistanceType::KLDivergence) {
-      // Hellinger works only on positive numbers
-      uniform(handle, r, x.data(), m * k, DataType(0.0), DataType(1.0));
-    } else if (distanceType == cuvs::distance::DistanceType::RusselRaoExpanded) {
-      uniform(handle, r, x.data(), m * k, DataType(0.0), DataType(1.0));
-      // Russel rao works on boolean values.
-      bernoulli(handle, r, x.data(), m * k, 0.5f);
-    } else {
-      uniform(handle, r, x.data(), m * k, DataType(-1.0), DataType(1.0));
-    }
-
-    for (int i = 0; i < 2; i++) {
-      // both X and Y are same buffer but when i = 1
-      // different dimensions for x & y is passed.
-      m = m / (i + 1);
-      naiveDistance(dist_ref[i].data(),
-                    x.data(),
-                    x.data(),
-                    m,
-                    n,
-                    k,
-                    distanceType,
-                    isRowMajor,
-                    metric_arg,
-                    stream);
-
-      DataType threshold = -10000.f;
-
-      if (isRowMajor) {
-        distanceLauncher<distanceType, DataType, layout_c_contiguous>(handle,
-                                                                      x.data(),
-                                                                      x.data(),
-                                                                      dist[i].data(),
-                                                                      dist2[i].data(),
-                                                                      m,
-                                                                      n,
-                                                                      k,
-                                                                      params,
-                                                                      threshold,
-                                                                      metric_arg);
-
-      } else {
-        distanceLauncher<distanceType, DataType, layout_f_contiguous>(handle,
-                                                                      x.data(),
-                                                                      x.data(),
-                                                                      dist[i].data(),
-                                                                      dist2[i].data(),
-                                                                      m,
-                                                                      n,
-                                                                      k,
-                                                                      params,
-                                                                      threshold,
-                                                                      metric_arg);
-      }
-    }
-    resource::sync_stream(handle, stream);
-  }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream;
-
-  DistanceInputs<DataType> params;
-  dev_vector x;
-  static const int N = 2;
-  std::array<dev_vector, N> dist_ref, dist, dist2;
-};
-
-template <cuvs::distance::DistanceType distanceType>
-class BigMatrixDistanceTest : public ::testing::Test {
- public:
-  BigMatrixDistanceTest()
-    : x(m * k, resource::get_cuda_stream(handle)),
-      dist(std::size_t(m) * m, resource::get_cuda_stream(handle)){};
-  void SetUp() override
-  {
-    auto testInfo = testing::UnitTest::GetInstance()->current_test_info();
-    common::nvtx::range fun_scope("test::%s/%s", testInfo->test_suite_name(), testInfo->name());
-
-    void pairwise_distance(raft::resources const& handle,
-                           float* x,
-                           float* y,
-                           float* dists,
-                           int m,
-                           int n,
-                           int k,
-                           cuvs::distance::DistanceType metric,
-                           bool isRowMajor,
-                           float metric_arg);
-    constexpr bool row_major   = true;
-    constexpr float metric_arg = 0.0f;
-    cuvs::distance::distance<distanceType, float, float, float>(
-      handle, x.data(), x.data(), dist.data(), m, n, k, row_major, metric_arg);
-    RAFT_CUDA_TRY(cudaStreamSynchronize(resource::get_cuda_stream(handle)));
-  }
-
- protected:
-  raft::resources handle;
-  int m = 48000;
-  int n = 48000;
-  int k = 1;
-  rmm::device_uvector<float> x, dist;
-};
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/fused_l2_nn.cu b/cpp/test/distance/fused_l2_nn.cu
deleted file mode 100644
index 39f7a8bfa..000000000
--- a/cpp/test/distance/fused_l2_nn.cu
+++ /dev/null
@@ -1,436 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <cuvs/distance/detail/fused_l2_nn.cuh>
-#include <cuvs/distance/fused_l2_nn.cuh>
-#include <gtest/gtest.h>
-#include <raft/core/kvp.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/linalg/norm.cuh>
-#include <raft/random/rng.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-
-namespace raft {
-namespace distance {
-
-template <typename LabelT, typename DataT>
-struct RaftKVPMinReduce {
-  typedef raft::KeyValuePair<LabelT, DataT> KVP;
-
-  DI KVP operator()(LabelT rit, const KVP& a, const KVP& b) { return b.value < a.value ? b : a; }
-
-  DI KVP operator()(const KVP& a, const KVP& b) { return b.value < a.value ? b : a; }
-
-};  // KVPMinReduce
-
-template <typename DataT, bool Sqrt, typename ReduceOpT, int NWARPS>
-RAFT_KERNEL naiveKernel(raft::KeyValuePair<int, DataT>* min,
-                        DataT* x,
-                        DataT* y,
-                        int m,
-                        int n,
-                        int k,
-                        int* workspace,
-                        DataT maxVal)
-{
-  int midx  = threadIdx.y + blockIdx.y * blockDim.y;
-  int nidx  = threadIdx.x + blockIdx.x * blockDim.x;
-  DataT acc = DataT(0);
-  for (int i = 0; i < k; ++i) {
-    int xidx  = i + midx * k;
-    int yidx  = i + nidx * k;
-    auto diff = midx >= m || nidx >= n ? DataT(0) : x[xidx] - y[yidx];
-    acc += diff * diff;
-  }
-
-  if (Sqrt) { acc = raft::sqrt(acc); }
-  ReduceOpT redOp;
-  typedef cub::WarpReduce<raft::KeyValuePair<int, DataT>> WarpReduce;
-  __shared__ typename WarpReduce::TempStorage temp[NWARPS];
-  int warpId = threadIdx.x / raft::WarpSize;
-  raft::KeyValuePair<int, DataT> tmp;
-  tmp.key   = nidx;
-  tmp.value = midx >= m || nidx >= n ? maxVal : acc;
-  tmp       = WarpReduce(temp[warpId]).Reduce(tmp, RaftKVPMinReduce<int, DataT>());
-  if (threadIdx.x % raft::WarpSize == 0 && midx < m) {
-    while (atomicCAS(workspace + midx, 0, 1) == 1)
-      ;
-    __threadfence();
-    redOp(midx, min + midx, tmp);
-    __threadfence();
-    atomicCAS(workspace + midx, 1, 0);
-  }
-}
-
-template <typename DataT, bool Sqrt>
-void naive(raft::KeyValuePair<int, DataT>* min,
-           DataT* x,
-           DataT* y,
-           int m,
-           int n,
-           int k,
-           int* workspace,
-           cudaStream_t stream)
-{
-  static const dim3 TPB(32, 16, 1);
-  dim3 nblks(raft::ceildiv(n, (int)TPB.x), raft::ceildiv(m, (int)TPB.y), 1);
-  RAFT_CUDA_TRY(cudaMemsetAsync(workspace, 0, sizeof(int) * m, stream));
-  auto blks = raft::ceildiv(m, 256);
-  MinAndDistanceReduceOp<int, DataT> op;
-  detail::initKernel<DataT, raft::KeyValuePair<int, DataT>, int>
-    <<<blks, 256, 0, stream>>>(min, m, std::numeric_limits<DataT>::max(), op);
-  RAFT_CUDA_TRY(cudaGetLastError());
-  naiveKernel<DataT, Sqrt, MinAndDistanceReduceOp<int, DataT>, 16>
-    <<<nblks, TPB, 0, stream>>>(min, x, y, m, n, k, workspace, std::numeric_limits<DataT>::max());
-  RAFT_CUDA_TRY(cudaGetLastError());
-}
-
-template <typename DataT>
-struct Inputs {
-  DataT tolerance;
-  int m, n, k;
-  unsigned long long int seed;
-
-  friend std::ostream& operator<<(std::ostream& os, const Inputs& p)
-  {
-    return os << "m: " << p.m
-              << ", "
-                 "n: "
-              << p.n
-              << ", "
-                 "k: "
-              << p.k
-              << ", "
-                 "seed: "
-              << p.seed
-              << ", "
-                 "tol: "
-              << p.tolerance;
-  }
-};
-
-template <typename DataT, bool Sqrt>
-class FusedL2NNTest : public ::testing::TestWithParam<Inputs<DataT>> {
- public:
-  FusedL2NNTest()
-    : params(::testing::TestWithParam<Inputs<DataT>>::GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      x(params.m * params.k, stream),
-      y(params.n * params.k, stream),
-      xn(params.m, stream),
-      yn(params.n, stream),
-      min(params.m, stream),
-      min_ref(params.m, stream),
-      workspace(params.m * sizeof(int), stream)
-  {
-  }
-
- protected:
-  void SetUp() override
-  {
-    raft::random::RngState r(params.seed);
-    int m = params.m;
-    int n = params.n;
-    int k = params.k;
-    uniform(handle, r, x.data(), m * k, DataT(-1.0), DataT(1.0));
-    uniform(handle, r, y.data(), n * k, DataT(-1.0), DataT(1.0));
-    generateGoldenResult();
-    raft::linalg::rowNorm(xn.data(), x.data(), k, m, raft::linalg::L2Norm, true, stream);
-    raft::linalg::rowNorm(yn.data(), y.data(), k, n, raft::linalg::L2Norm, true, stream);
-    RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
-  }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream;
-  Inputs<DataT> params;
-  rmm::device_uvector<DataT> x;
-  rmm::device_uvector<DataT> y;
-  rmm::device_uvector<DataT> xn;
-  rmm::device_uvector<DataT> yn;
-  rmm::device_uvector<raft::KeyValuePair<int, DataT>> min;
-  rmm::device_uvector<raft::KeyValuePair<int, DataT>> min_ref;
-  rmm::device_uvector<char> workspace;
-
-  virtual void generateGoldenResult()
-  {
-    int m = params.m;
-    int n = params.n;
-    int k = params.k;
-    naive<DataT, Sqrt>(min_ref.data(), x.data(), y.data(), m, n, k, (int*)workspace.data(), stream);
-  }
-
-  void runTest(raft::KeyValuePair<int, DataT>* out)
-  {
-    int m = params.m;
-    int n = params.n;
-    int k = params.k;
-
-    const bool init_out_buffer = true;
-    fusedL2NNMinReduce<DataT, raft::KeyValuePair<int, DataT>, int>(out,
-                                                                   x.data(),
-                                                                   y.data(),
-                                                                   xn.data(),
-                                                                   yn.data(),
-                                                                   m,
-                                                                   n,
-                                                                   k,
-                                                                   (void*)workspace.data(),
-                                                                   Sqrt,
-                                                                   init_out_buffer,
-                                                                   stream);
-    RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
-  }
-};
-
-template <typename T>
-struct CompareApproxAbsKVP {
-  typedef typename raft::KeyValuePair<int, T> KVP;
-  CompareApproxAbsKVP(T eps_) : eps(eps_) {}
-  bool operator()(const KVP& a, const KVP& b) const
-  {
-    T diff  = std::abs(std::abs(a.value) - std::abs(b.value));
-    T m     = std::max(std::abs(a.value), std::abs(b.value));
-    T ratio = m >= eps ? diff / m : diff;
-    return (ratio <= eps);
-  }
-
- private:
-  T eps;
-};
-
-template <typename T>
-struct CompareExactKVP {
-  typedef typename raft::KeyValuePair<int, T> KVP;
-  bool operator()(const KVP& a, const KVP& b) const
-  {
-    if (a.value != b.value) return false;
-    return true;
-  }
-};
-
-template <typename K, typename V, typename L>
-::testing::AssertionResult devArrMatch(const raft::KeyValuePair<K, V>* expected,
-                                       const raft::KeyValuePair<K, V>* actual,
-                                       size_t size,
-                                       L eq_compare,
-                                       cudaStream_t stream = 0)
-{
-  typedef typename raft::KeyValuePair<K, V> KVP;
-  std::shared_ptr<KVP> exp_h(new KVP[size]);
-  std::shared_ptr<KVP> act_h(new KVP[size]);
-  raft::update_host<KVP>(exp_h.get(), expected, size, stream);
-  raft::update_host<KVP>(act_h.get(), actual, size, stream);
-  RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
-  for (size_t i(0); i < size; ++i) {
-    auto exp = exp_h.get()[i];
-    auto act = act_h.get()[i];
-    if (!eq_compare(exp, act)) {
-      return ::testing::AssertionFailure()
-             << "actual=" << act.key << "," << act.value << " != expected=" << exp.key << ","
-             << exp.value << " @" << i;
-    }
-  }
-  return ::testing::AssertionSuccess();
-}
-
-const std::vector<Inputs<float>> inputsf = {
-  {0.001f, 32, 32, 32, 1234ULL},
-  {0.001f, 32, 64, 32, 1234ULL},
-  {0.001f, 64, 32, 32, 1234ULL},
-  {0.001f, 64, 64, 32, 1234ULL},
-  {0.001f, 128, 32, 32, 1234ULL},
-  {0.001f, 128, 64, 32, 1234ULL},
-  {0.001f, 128, 128, 64, 1234ULL},
-  {0.001f, 64, 128, 128, 1234ULL},
-
-  {0.001f, 32, 32, 34, 1234ULL},
-  {0.001f, 32, 64, 34, 1234ULL},
-  {0.001f, 64, 32, 34, 1234ULL},
-  {0.001f, 64, 64, 34, 1234ULL},
-  {0.001f, 128, 32, 34, 1234ULL},
-  {0.001f, 128, 64, 34, 1234ULL},
-  {0.001f, 128, 128, 66, 1234ULL},
-  {0.001f, 64, 128, 130, 1234ULL},
-
-  {0.001f, 32, 32, 33, 1234ULL},
-  {0.001f, 32, 64, 33, 1234ULL},
-  {0.001f, 64, 32, 33, 1234ULL},
-  {0.001f, 64, 64, 33, 1234ULL},
-  {0.001f, 128, 32, 33, 1234ULL},
-  {0.001f, 128, 64, 33, 1234ULL},
-  {0.001f, 128, 128, 65, 1234ULL},
-  {0.001f, 64, 128, 129, 1234ULL},
-  {0.006f, 1805, 134, 2, 1234ULL},
-  {0.006f, 8192, 1024, 64, 1234ULL},
-  {0.006f, 8192, 1025, 64, 1234ULL},
-
-  // Repeat with smaller values of k
-  {0.006f, 32, 32, 1, 1234ULL},
-  {0.001f, 32, 64, 2, 1234ULL},
-  {0.001f, 64, 32, 3, 1234ULL},
-  {0.001f, 64, 64, 4, 1234ULL},
-  {0.001f, 128, 32, 5, 1234ULL},
-  {0.001f, 128, 64, 6, 1234ULL},
-  {0.001f, 128, 128, 7, 1234ULL},
-  {0.001f, 64, 128, 8, 1234ULL},
-
-  {0.001f, 32, 32, 9, 1234ULL},
-  {0.001f, 32, 64, 10, 1234ULL},
-  {0.001f, 64, 32, 11, 1234ULL},
-  {0.001f, 64, 64, 12, 1234ULL},
-  {0.001f, 128, 32, 13, 1234ULL},
-  {0.001f, 128, 64, 14, 1234ULL},
-  {0.001f, 128, 128, 15, 1234ULL},
-  {0.001f, 64, 128, 16, 1234ULL},
-
-  {0.001f, 32, 32, 17, 1234ULL},
-  {0.001f, 32, 64, 18, 1234ULL},
-  {0.001f, 64, 32, 19, 1234ULL},
-  {0.001f, 64, 64, 20, 1234ULL},
-  {0.001f, 128, 32, 21, 1234ULL},
-  {0.001f, 128, 64, 22, 1234ULL},
-  {0.001f, 128, 128, 23, 1234ULL},
-  {0.00001, 64, 128, 24, 1234ULL},
-  {0.001f, 1805, 134, 25, 1234ULL},
-  {0.006f, 8192, 1024, 25, 1234ULL},
-  {0.006f, 8192, 1024, 66, 1234ULL},
-};
-typedef FusedL2NNTest<float, false> FusedL2NNTestF_Sq;
-TEST_P(FusedL2NNTestF_Sq, Result)
-{
-  runTest(min.data());
-  ASSERT_TRUE(devArrMatch(
-    min_ref.data(), min.data(), params.m, CompareApproxAbsKVP<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(FusedL2NNTests, FusedL2NNTestF_Sq, ::testing::ValuesIn(inputsf));
-typedef FusedL2NNTest<float, true> FusedL2NNTestF_Sqrt;
-TEST_P(FusedL2NNTestF_Sqrt, Result)
-{
-  runTest(min.data());
-  ASSERT_TRUE(devArrMatch(
-    min_ref.data(), min.data(), params.m, CompareApproxAbsKVP<float>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(FusedL2NNTests, FusedL2NNTestF_Sqrt, ::testing::ValuesIn(inputsf));
-
-const std::vector<Inputs<double>> inputsd = {
-  {0.00001, 32, 32, 32, 1234ULL},   {0.00001, 32, 64, 32, 1234ULL},
-  {0.00001, 64, 32, 32, 1234ULL},   {0.00001, 64, 64, 32, 1234ULL},
-  {0.00001, 128, 32, 32, 1234ULL},  {0.00001, 128, 64, 32, 1234ULL},
-  {0.00001, 128, 128, 64, 1234ULL}, {0.00001, 64, 128, 128, 1234ULL},
-
-  {0.00001, 32, 32, 34, 1234ULL},   {0.00001, 32, 64, 34, 1234ULL},
-  {0.00001, 64, 32, 34, 1234ULL},   {0.00001, 64, 64, 34, 1234ULL},
-  {0.00001, 128, 32, 34, 1234ULL},  {0.00001, 128, 64, 34, 1234ULL},
-  {0.00001, 128, 128, 66, 1234ULL}, {0.00001, 64, 128, 130, 1234ULL},
-
-  {0.00001, 32, 32, 33, 1234ULL},   {0.00001, 32, 64, 33, 1234ULL},
-  {0.00001, 64, 32, 33, 1234ULL},   {0.00001, 64, 64, 33, 1234ULL},
-  {0.00001, 128, 32, 33, 1234ULL},  {0.00001, 128, 64, 33, 1234ULL},
-  {0.00001, 128, 128, 65, 1234ULL}, {0.00001, 64, 128, 129, 1234ULL},
-
-  {0.00001, 1805, 134, 2, 1234ULL},  //{0.00001, 8192, 1024, 25, 1234ULL},
-};
-typedef FusedL2NNTest<double, false> FusedL2NNTestD_Sq;
-TEST_P(FusedL2NNTestD_Sq, Result)
-{
-  runTest(min.data());
-  ASSERT_TRUE(devArrMatch(
-    min_ref.data(), min.data(), params.m, CompareApproxAbsKVP<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(FusedL2NNTests, FusedL2NNTestD_Sq, ::testing::ValuesIn(inputsd));
-typedef FusedL2NNTest<double, true> FusedL2NNTestD_Sqrt;
-TEST_P(FusedL2NNTestD_Sqrt, Result)
-{
-  runTest(min.data());
-  ASSERT_TRUE(devArrMatch(
-    min_ref.data(), min.data(), params.m, CompareApproxAbsKVP<double>(params.tolerance), stream));
-}
-INSTANTIATE_TEST_CASE_P(FusedL2NNTests, FusedL2NNTestD_Sqrt, ::testing::ValuesIn(inputsd));
-
-/// This is to test output determinism of the prim
-template <typename DataT, bool Sqrt>
-class FusedL2NNDetTest : public FusedL2NNTest<DataT, Sqrt> {
- public:
-  FusedL2NNDetTest() : stream(resource::get_cuda_stream(handle)), min1(0, stream) {}
-
-  void SetUp() override
-  {
-    FusedL2NNTest<DataT, Sqrt>::SetUp();
-    int m = this->params.m;
-    min1.resize(m, stream);
-    RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
-  }
-
-  void TearDown() override { FusedL2NNTest<DataT, Sqrt>::TearDown(); }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream;
-
-  rmm::device_uvector<raft::KeyValuePair<int, DataT>> min1;
-
-  static const int NumRepeats = 3;
-
-  void generateGoldenResult() override {}
-};
-
-typedef FusedL2NNDetTest<float, false> FusedL2NNDetTestF_Sq;
-TEST_P(FusedL2NNDetTestF_Sq, Result)
-{
-  runTest(min.data());  // assumed to be golden
-  for (int i = 0; i < NumRepeats; ++i) {
-    runTest(min1.data());
-    ASSERT_TRUE(devArrMatch(min.data(), min1.data(), params.m, CompareExactKVP<float>(), stream));
-  }
-}
-INSTANTIATE_TEST_CASE_P(FusedL2NNDetTests, FusedL2NNDetTestF_Sq, ::testing::ValuesIn(inputsf));
-typedef FusedL2NNDetTest<float, true> FusedL2NNDetTestF_Sqrt;
-TEST_P(FusedL2NNDetTestF_Sqrt, Result)
-{
-  runTest(min.data());  // assumed to be golden
-  for (int i = 0; i < NumRepeats; ++i) {
-    runTest(min1.data());
-    ASSERT_TRUE(devArrMatch(min.data(), min1.data(), params.m, CompareExactKVP<float>(), stream));
-  }
-}
-INSTANTIATE_TEST_CASE_P(FusedL2NNDetTests, FusedL2NNDetTestF_Sqrt, ::testing::ValuesIn(inputsf));
-
-typedef FusedL2NNDetTest<double, false> FusedL2NNDetTestD_Sq;
-TEST_P(FusedL2NNDetTestD_Sq, Result)
-{
-  runTest(min.data());  // assumed to be golden
-  for (int i = 0; i < NumRepeats; ++i) {
-    runTest(min1.data());
-    ASSERT_TRUE(devArrMatch(min.data(), min1.data(), params.m, CompareExactKVP<double>(), stream));
-  }
-}
-INSTANTIATE_TEST_CASE_P(FusedL2NNDetTests, FusedL2NNDetTestD_Sq, ::testing::ValuesIn(inputsd));
-typedef FusedL2NNDetTest<double, true> FusedL2NNDetTestD_Sqrt;
-TEST_P(FusedL2NNDetTestD_Sqrt, Result)
-{
-  runTest(min.data());  // assumed to be golden
-  for (int i = 0; i < NumRepeats; ++i) {
-    runTest(min1.data());
-    ASSERT_TRUE(devArrMatch(min.data(), min1.data(), params.m, CompareExactKVP<double>(), stream));
-  }
-}
-INSTANTIATE_TEST_CASE_P(FusedL2NNDetTests, FusedL2NNDetTestD_Sqrt, ::testing::ValuesIn(inputsd));
-
-}  // end namespace distance
-}  // end namespace raft
diff --git a/cpp/test/distance/gram.cu b/cpp/test/distance/gram.cu
deleted file mode 100644
index 1d0c8c27a..000000000
--- a/cpp/test/distance/gram.cu
+++ /dev/null
@@ -1,170 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "gram_base.cuh"
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/distance/kernels.cuh>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <memory>
-#include <raft/random/rng.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <rmm/device_uvector.hpp>
-
-namespace cuvs::distance::kernels {
-
-struct GramMatrixInputs {
-  int n1;      // feature vectors in matrix 1
-  int n2;      // featuer vectors in matrix 2
-  int n_cols;  // number of elements in a feature vector
-  bool is_row_major;
-  KernelParams kernel;
-  int ld1;
-  int ld2;
-  int ld_out;
-  // We will generate random input using the dimensions given here.
-  // The reference output is calculated by a custom kernel.
-};
-
-std::ostream& operator<<(std::ostream& os, const GramMatrixInputs& p)
-{
-  std::vector<std::string> kernel_names{"linear", "poly", "rbf", "tanh"};
-  os << "/" << p.n1 << "x" << p.n2 << "x" << p.n_cols << "/"
-     << (p.is_row_major ? "RowMajor/" : "ColMajor/") << kernel_names[p.kernel.kernel] << "/ld_"
-     << p.ld1 << "x" << p.ld2 << "x" << p.ld_out;
-  return os;
-}
-
-const std::vector<GramMatrixInputs> inputs = {
-  {42, 137, 2, false, {KernelType::LINEAR}},
-  {42, 137, 2, true, {KernelType::LINEAR}},
-  {42, 137, 2, false, {KernelType::LINEAR}, 64, 179, 181},
-  {42, 137, 2, true, {KernelType::LINEAR}, 64, 179, 181},
-  {137, 42, 2, false, {KernelType::POLYNOMIAL, 2, 0.5, 2.4}},
-  {137, 42, 2, true, {KernelType::POLYNOMIAL, 2, 0.5, 2.4}},
-  {137, 42, 2, false, {KernelType::POLYNOMIAL, 2, 0.5, 2.4}, 159, 73, 144},
-  {137, 42, 2, true, {KernelType::POLYNOMIAL, 2, 0.5, 2.4}, 159, 73, 144},
-  {42, 137, 2, false, {KernelType::TANH, 0, 0.5, 2.4}},
-  {42, 137, 2, true, {KernelType::TANH, 0, 0.5, 2.4}},
-  {42, 137, 2, false, {KernelType::TANH, 0, 0.5, 2.4}, 64, 155, 49},
-  {42, 137, 2, true, {KernelType::TANH, 0, 0.5, 2.4}, 64, 155, 143},
-  {3, 4, 2, false, {KernelType::RBF, 0, 0.5}},
-  {42, 137, 2, false, {KernelType::RBF, 0, 0.5}},
-  {42, 137, 2, true, {KernelType::RBF, 0, 0.5}},
-  // Distance kernel does not support LD parameter yet.
-  //{42, 137, 2, false, {KernelType::RBF, 0, 0.5}, 64, 155, 49},
-  // {42, 137, 2, true, {KernelType::RBF, 0, 0.5}, 64, 155, 143},
-};
-
-template <typename math_t>
-class GramMatrixTest : public ::testing::TestWithParam<GramMatrixInputs> {
- protected:
-  GramMatrixTest()
-    : params(GetParam()),
-      handle(),
-      x1(0, resource::get_cuda_stream(handle)),
-      x2(0, resource::get_cuda_stream(handle)),
-      gram(0, resource::get_cuda_stream(handle)),
-      gram_host(0)
-  {
-    auto stream = resource::get_cuda_stream(handle);
-
-    if (params.ld1 == 0) { params.ld1 = params.is_row_major ? params.n_cols : params.n1; }
-    if (params.ld2 == 0) { params.ld2 = params.is_row_major ? params.n_cols : params.n2; }
-    if (params.ld_out == 0) { params.ld_out = params.is_row_major ? params.n2 : params.n1; }
-    // Derive the size of the output from the offset of the last element.
-    size_t size = get_offset(params.n1 - 1, params.n_cols - 1, params.ld1, params.is_row_major) + 1;
-    x1.resize(size, stream);
-    size = get_offset(params.n2 - 1, params.n_cols - 1, params.ld2, params.is_row_major) + 1;
-    x2.resize(size, stream);
-    size = get_offset(params.n1 - 1, params.n2 - 1, params.ld_out, params.is_row_major) + 1;
-
-    gram.resize(size, stream);
-    RAFT_CUDA_TRY(cudaMemsetAsync(gram.data(), 0, gram.size() * sizeof(math_t), stream));
-    gram_host.resize(gram.size());
-    std::fill(gram_host.begin(), gram_host.end(), 0);
-
-    raft::random::RngState rng(42137ULL);
-    raft::random::uniform(handle, rng, x1.data(), x1.size(), math_t(0), math_t(1));
-    raft::random::uniform(handle, rng, x2.data(), x2.size(), math_t(0), math_t(1));
-  }
-
-  ~GramMatrixTest() override {}
-
-  void runTest()
-  {
-    std::unique_ptr<GramMatrixBase<math_t>> kernel =
-      std::unique_ptr<GramMatrixBase<math_t>>(KernelFactory<math_t>::create(params.kernel));
-
-    auto x1_span =
-      params.is_row_major
-        ? raft::make_device_strided_matrix_view<const math_t, int, raft::layout_c_contiguous>(
-            x1.data(), params.n1, params.n_cols, params.ld1)
-        : raft::make_device_strided_matrix_view<const math_t, int, raft::layout_f_contiguous>(
-            x1.data(), params.n1, params.n_cols, params.ld1);
-    auto x2_span =
-      params.is_row_major
-        ? raft::make_device_strided_matrix_view<const math_t, int, raft::layout_c_contiguous>(
-            x2.data(), params.n2, params.n_cols, params.ld2)
-        : raft::make_device_strided_matrix_view<const math_t, int, raft::layout_f_contiguous>(
-            x2.data(), params.n2, params.n_cols, params.ld2);
-    auto out_span =
-      params.is_row_major
-        ? raft::make_device_strided_matrix_view<math_t, int, raft::layout_c_contiguous>(
-            gram.data(), params.n1, params.n2, params.ld_out)
-        : raft::make_device_strided_matrix_view<math_t, int, raft::layout_f_contiguous>(
-            gram.data(), params.n1, params.n2, params.ld_out);
-
-    (*kernel)(handle, x1_span, x2_span, out_span);
-
-    auto stream = resource::get_cuda_stream(handle);
-    naiveGramMatrixKernel(params.n1,
-                          params.n2,
-                          params.n_cols,
-                          x1,
-                          x2,
-                          gram_host.data(),
-                          params.ld1,
-                          params.ld2,
-                          params.ld_out,
-                          params.is_row_major,
-                          params.kernel,
-                          stream,
-                          handle);
-
-    ASSERT_TRUE(raft::devArrMatchHost(
-      gram_host.data(), gram.data(), gram.size(), raft::CompareApprox<math_t>(1e-6f), stream));
-  }
-
-  GramMatrixInputs params;
-  raft::resources handle;
-
-  rmm::device_uvector<math_t> x1;
-  rmm::device_uvector<math_t> x2;
-  rmm::device_uvector<math_t> gram;
-
-  std::vector<math_t> gram_host;
-};
-
-typedef GramMatrixTest<float> GramMatrixTestFloat;
-typedef GramMatrixTest<double> GramMatrixTestDouble;
-
-TEST_P(GramMatrixTestFloat, Gram) { runTest(); }
-
-INSTANTIATE_TEST_SUITE_P(GramMatrixTests, GramMatrixTestFloat, ::testing::ValuesIn(inputs));
-};  // end namespace cuvs::distance::kernels
diff --git a/cpp/test/distance/gram_base.cuh b/cpp/test/distance/gram_base.cuh
deleted file mode 100644
index cb98fa6fd..000000000
--- a/cpp/test/distance/gram_base.cuh
+++ /dev/null
@@ -1,88 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/distance/kernels.cuh>
-#include <iostream>
-#include <memory>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <rmm/device_uvector.hpp>
-
-namespace raft {
-namespace distance {
-namespace kernels {
-
-// Get the offset of element [i,k].
-HDI int get_offset(int i, int k, int ld, bool is_row_major)
-{
-  return is_row_major ? i * ld + k : i + k * ld;
-}
-
-// Calculate the Gram matrix on the host.
-template <typename math_t>
-void naiveGramMatrixKernel(int n1,
-                           int n2,
-                           int n_cols,
-                           const rmm::device_uvector<math_t>& x1,
-                           const rmm::device_uvector<math_t>& x2,
-                           math_t* gram_host,
-                           int ld1,
-                           int ld2,
-                           int ld_out,
-                           bool is_row_major,
-                           KernelParams kernel,
-                           cudaStream_t stream,
-                           const raft::resources& handle)
-{
-  std::vector<math_t> x1_host(x1.size());
-  raft::update_host(x1_host.data(), x1.data(), x1.size(), stream);
-  std::vector<math_t> x2_host(x2.size());
-  raft::update_host(x2_host.data(), x2.data(), x2.size(), stream);
-  resource::sync_stream(handle, stream);
-
-  for (int i = 0; i < n1; i++) {
-    for (int j = 0; j < n2; j++) {
-      float d = 0;
-      for (int k = 0; k < n_cols; k++) {
-        if (kernel.kernel == KernelType::RBF) {
-          math_t diff = x1_host[get_offset(i, k, ld1, is_row_major)] -
-                        x2_host[get_offset(j, k, ld2, is_row_major)];
-          d += diff * diff;
-        } else {
-          d += x1_host[get_offset(i, k, ld1, is_row_major)] *
-               x2_host[get_offset(j, k, ld2, is_row_major)];
-        }
-      }
-      int idx  = get_offset(i, j, ld_out, is_row_major);
-      math_t v = 0;
-      switch (kernel.kernel) {
-        case (KernelType::LINEAR): gram_host[idx] = d; break;
-        case (KernelType::POLYNOMIAL):
-          v              = kernel.gamma * d + kernel.coef0;
-          gram_host[idx] = std::pow(v, kernel.degree);
-          break;
-        case (KernelType::TANH): gram_host[idx] = std::tanh(kernel.gamma * d + kernel.coef0); break;
-        case (KernelType::RBF): gram_host[idx] = exp(-kernel.gamma * d); break;
-      }
-    }
-  }
-}
-
-}  // namespace kernels
-}  // namespace distance
-}  // namespace raft
diff --git a/cpp/test/distance/masked_nn.cu b/cpp/test/distance/masked_nn.cu
deleted file mode 100644
index 1c1ed225a..000000000
--- a/cpp/test/distance/masked_nn.cu
+++ /dev/null
@@ -1,435 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.h"
-#include <cuvs/distance/detail/masked_nn.cuh>
-#include <cuvs/distance/masked_nn.cuh>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/kvp.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/linalg/norm.cuh>
-#include <raft/random/rng.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <raft/util/itertools.hpp>
-
-namespace cuvs::distance::masked_nn {
-
-// The adjacency pattern determines what distances get computed.
-enum AdjacencyPattern {
-  checkerboard = 0,  // adjacency matrix looks like a checkerboard (half the distances are computed)
-  checkerboard_4  = 1,  // checkerboard with tiles of size 4x4
-  checkerboard_64 = 2,  // checkerboard with tiles of size 64x64
-  all_true        = 3,  // no distance computations can be skipped
-  all_false       = 4   // all distance computations can be skipped
-};
-
-// Kernels:
-// - init_adj: to initialize the adjacency kernel with a specific adjacency pattern
-// - referenceKernel: to produce the ground-truth output
-
-RAFT_KERNEL init_adj(AdjacencyPattern pattern,
-                     int n,
-                     raft::device_matrix_view<bool, int, raft::layout_c_contiguous> adj,
-                     raft::device_vector_view<int, int, raft::layout_c_contiguous> group_idxs)
-{
-  int m          = adj.extent(0);
-  int num_groups = adj.extent(1);
-
-  for (int idx_m = blockIdx.y * blockDim.y + threadIdx.y; idx_m < m;
-       idx_m += blockDim.y * gridDim.y) {
-    for (int idx_g = blockIdx.x * blockDim.x + threadIdx.x; idx_g < num_groups;
-         idx_g += blockDim.x * gridDim.x) {
-      switch (pattern) {
-        case checkerboard: adj(idx_m, idx_g) = (idx_m + idx_g) % 2; break;
-        case checkerboard_4: adj(idx_m, idx_g) = (idx_m / 4 + idx_g) % 2; break;
-        case checkerboard_64: adj(idx_m, idx_g) = (idx_m / 64 + idx_g) % 2; break;
-        case all_true: adj(idx_m, idx_g) = true; break;
-        case all_false: adj(idx_m, idx_g) = false; break;
-        default: assert(false && "unknown pattern");
-      }
-    }
-  }
-  // Each group is of size n / num_groups.
-  //
-  // - group_idxs[j] indicates the start of group j + 1 (i.e. is the inclusive
-  // scan of the group lengths)
-  //
-  // - The first group always starts at index zero, so we do not store it.
-  //
-  // - The group_idxs[num_groups - 1] should always equal n.
-
-  if (blockIdx.y == 0 && threadIdx.y == 0) {
-    const int g_stride = blockDim.x * gridDim.x;
-    for (int idx_g = blockIdx.x * blockDim.x + threadIdx.x; idx_g < num_groups; idx_g += g_stride) {
-      group_idxs(idx_g) = (idx_g + 1) * (n / num_groups);
-    }
-    group_idxs(num_groups - 1) = n;
-  }
-}
-
-template <typename DataT, typename ReduceOpT, int NWARPS>
-__launch_bounds__(32 * NWARPS, 2) RAFT_KERNEL referenceKernel(raft::KeyValuePair<int, DataT>* min,
-                                                              DataT* x,
-                                                              DataT* y,
-                                                              bool* adj,
-                                                              int* group_idxs,
-                                                              int m,
-                                                              int n,
-                                                              int k,
-                                                              int num_groups,
-                                                              bool sqrt,
-                                                              int* workspace,
-                                                              DataT maxVal)
-{
-  const int m_stride = blockDim.y * gridDim.y;
-  const int m_offset = threadIdx.y + blockIdx.y * blockDim.y;
-  const int n_stride = blockDim.x * gridDim.x;
-  const int n_offset = threadIdx.x + blockIdx.x * blockDim.x;
-
-  for (int m_grid = 0; m_grid < m; m_grid += m_stride) {
-    for (int n_grid = 0; n_grid < n; n_grid += n_stride) {
-      int midx = m_grid + m_offset;
-      int nidx = n_grid + n_offset;
-
-      // Do a reverse linear search to determine the group index.
-      int group_idx = 0;
-      for (int i = num_groups; 0 <= i; --i) {
-        if (nidx < group_idxs[i]) { group_idx = i; }
-      }
-      const bool include_dist = adj[midx * num_groups + group_idx] && midx < m && nidx < n;
-
-      // Compute L2 metric.
-      DataT acc = DataT(0);
-      for (int i = 0; i < k; ++i) {
-        int xidx  = i + midx * k;
-        int yidx  = i + nidx * k;
-        auto diff = x[xidx] - y[yidx];
-        acc += diff * diff;
-      }
-      if (sqrt) { acc = raft::sqrt(acc); }
-      ReduceOpT redOp;
-      typedef cub::WarpReduce<raft::KeyValuePair<int, DataT>> WarpReduce;
-      __shared__ typename WarpReduce::TempStorage temp[NWARPS];
-      int warpId = threadIdx.x / raft::WarpSize;
-      raft::KeyValuePair<int, DataT> tmp;
-      tmp.key   = include_dist ? nidx : -1;
-      tmp.value = include_dist ? acc : maxVal;
-      tmp       = WarpReduce(temp[warpId]).Reduce(tmp, cuvs::distance::KVPMinReduce<int, DataT>{});
-      if (threadIdx.x % raft::WarpSize == 0 && midx < m) {
-        while (atomicCAS(workspace + midx, 0, 1) == 1)
-          ;
-        __threadfence();
-        redOp(midx, min + midx, tmp);
-        __threadfence();
-        atomicCAS(workspace + midx, 1, 0);
-      }
-      __syncthreads();
-    }
-  }
-}
-
-// Structs
-// - Params: holds parameters for test case
-// - Inputs: holds the inputs to the functions under test (x, y, adj, group_idxs). Is generated from
-//   the inputs.
-struct Params {
-  double tolerance;
-  int m, n, k, num_groups;
-  bool sqrt;
-  unsigned long long int seed;
-  AdjacencyPattern pattern;
-};
-
-inline auto operator<<(std::ostream& os, const Params& p) -> std::ostream&
-{
-  os << "m: " << p.m << ", n: " << p.n << ", k: " << p.k << ", num_groups: " << p.num_groups
-     << ", sqrt: " << p.sqrt << ", seed: " << p.seed << ", tol: " << p.tolerance;
-  return os;
-}
-
-template <typename DataT>
-struct Inputs {
-  using IdxT = int;
-
-  raft::device_matrix<DataT, IdxT> x, y;
-  raft::device_matrix<bool, IdxT> adj;
-  raft::device_vector<IdxT, IdxT> group_idxs;
-
-  Inputs(const raft::handle_t& handle, const Params& p)
-    : x{raft::make_device_matrix<DataT, IdxT>(handle, p.m, p.k)},
-      y{raft::make_device_matrix<DataT, IdxT>(handle, p.n, p.k)},
-      adj{raft::make_device_matrix<bool, IdxT>(handle, p.m, p.num_groups)},
-      group_idxs{raft::make_device_vector<IdxT, IdxT>(handle, p.num_groups)}
-  {
-    // Initialize x, y
-    raft::random::RngState r(p.seed);
-    uniform(handle, r, x.data_handle(), p.m * p.k, DataT(-1.0), DataT(1.0));
-    uniform(handle, r, y.data_handle(), p.n * p.k, DataT(-1.0), DataT(1.0));
-
-    // Initialize adj, group_idxs.
-    dim3 block(32, 32);
-    dim3 grid(10, 10);
-    init_adj<<<grid, block, 0, resource::get_cuda_stream(handle)>>>(
-      p.pattern, p.n, adj.view(), group_idxs.view());
-    RAFT_CUDA_TRY(cudaGetLastError());
-  }
-};
-
-template <typename DataT, typename OutT = raft::KeyValuePair<int, DataT>>
-auto reference(const raft::handle_t& handle, Inputs<DataT> inp, const Params& p)
-  -> raft::device_vector<OutT, int>
-{
-  int m          = inp.x.extent(0);
-  int n          = inp.y.extent(0);
-  int k          = inp.x.extent(1);
-  int num_groups = inp.group_idxs.extent(0);
-
-  if (m == 0 || n == 0 || k == 0 || num_groups == 0) {
-    return raft::make_device_vector<OutT, int>(handle, 0);
-  }
-
-  // Initialize workspace
-  auto stream = resource::get_cuda_stream(handle);
-  rmm::device_uvector<char> workspace(p.m * sizeof(int), stream);
-  RAFT_CUDA_TRY(cudaMemsetAsync(workspace.data(), 0, sizeof(int) * m, stream));
-
-  // Initialize output
-  auto out  = raft::make_device_vector<OutT, int>(handle, m);
-  auto blks = raft::ceildiv(m, 256);
-  MinAndDistanceReduceOp<int, DataT> op;
-  cuvs::distance::detail::initKernel<DataT, raft::KeyValuePair<int, DataT>, int>
-    <<<blks, 256, 0, stream>>>(out.data_handle(), m, std::numeric_limits<DataT>::max(), op);
-  RAFT_CUDA_TRY(cudaGetLastError());
-
-  // Launch reference kernel
-  const int nwarps = 16;
-  static const dim3 TPB(32, nwarps, 1);
-  dim3 nblks(1, 200, 1);
-  referenceKernel<DataT, decltype(op), nwarps>
-    <<<nblks, TPB, 0, stream>>>(out.data_handle(),
-                                inp.x.data_handle(),
-                                inp.y.data_handle(),
-                                inp.adj.data_handle(),
-                                inp.group_idxs.data_handle(),
-                                m,
-                                n,
-                                k,
-                                num_groups,
-                                p.sqrt,
-                                (int*)workspace.data(),
-                                std::numeric_limits<DataT>::max());
-  RAFT_CUDA_TRY(cudaGetLastError());
-
-  return out;
-}
-
-template <typename DataT, typename OutT = raft::KeyValuePair<int, DataT>>
-auto run_masked_nn(const raft::handle_t& handle, Inputs<DataT> inp, const Params& p)
-  -> raft::device_vector<OutT, int>
-{
-  // Compute norms:
-  auto x_norm = raft::make_device_vector<DataT, int>(handle, p.m);
-  auto y_norm = raft::make_device_vector<DataT, int>(handle, p.n);
-
-  raft::linalg::norm(handle,
-                     std::as_const(inp.x).view(),
-                     x_norm.view(),
-                     raft::linalg::L2Norm,
-                     raft::linalg::Apply::ALONG_ROWS);
-  raft::linalg::norm(handle,
-                     std::as_const(inp.y).view(),
-                     y_norm.view(),
-                     raft::linalg::L2Norm,
-                     raft::linalg::Apply::ALONG_ROWS);
-
-  // Create parameters for masked_l2_nn
-  using IdxT       = int;
-  using RedOpT     = MinAndDistanceReduceOp<int, DataT>;
-  using PairRedOpT = cuvs::distance::KVPMinReduce<int, DataT>;
-  using ParamT     = cuvs::distance::masked_l2_nn_params<RedOpT, PairRedOpT>;
-
-  bool init_out = true;
-  ParamT masked_l2_params{RedOpT{}, PairRedOpT{}, p.sqrt, init_out};
-
-  // Create output
-  auto out = raft::make_device_vector<OutT, IdxT, raft::layout_c_contiguous>(handle, p.m);
-
-  // Launch kernel
-  cuvs::distance::masked_l2_nn<DataT, OutT, IdxT>(handle,
-                                                  masked_l2_params,
-                                                  inp.x.view(),
-                                                  inp.y.view(),
-                                                  x_norm.view(),
-                                                  y_norm.view(),
-                                                  inp.adj.view(),
-                                                  inp.group_idxs.view(),
-                                                  out.view());
-
-  resource::sync_stream(handle);
-
-  return out;
-}
-
-template <typename T>
-struct CompareApproxAbsKVP {
-  typedef typename raft::KeyValuePair<int, T> KVP;
-  CompareApproxAbsKVP(T eps_) : eps(eps_) {}
-  bool operator()(const KVP& a, const KVP& b) const
-  {
-    T diff  = raft::abs(raft::abs(a.value) - raft::abs(b.value));
-    T m     = std::max(raft::abs(a.value), raft::abs(b.value));
-    T ratio = m >= eps ? diff / m : diff;
-    return (ratio <= eps);
-  }
-
- private:
-  T eps;
-};
-
-template <typename K, typename V, typename L>
-::testing::AssertionResult devArrMatch(const raft::KeyValuePair<K, V>* expected,
-                                       const raft::KeyValuePair<K, V>* actual,
-                                       size_t size,
-                                       L eq_compare,
-                                       cudaStream_t stream = 0)
-{
-  typedef typename raft::KeyValuePair<K, V> KVP;
-  std::shared_ptr<KVP> exp_h(new KVP[size]);
-  std::shared_ptr<KVP> act_h(new KVP[size]);
-  raft::update_host<KVP>(exp_h.get(), expected, size, stream);
-  raft::update_host<KVP>(act_h.get(), actual, size, stream);
-  RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
-  for (size_t i(0); i < size; ++i) {
-    auto exp = exp_h.get()[i];
-    auto act = act_h.get()[i];
-    if (!eq_compare(exp, act)) {
-      return ::testing::AssertionFailure()
-             << "actual=" << act.key << "," << act.value << " != expected=" << exp.key << ","
-             << exp.value << " @" << i;
-    }
-  }
-  return ::testing::AssertionSuccess();
-}
-
-inline auto gen_params() -> std::vector<Params>
-{
-  // Regular powers of two
-  auto regular = raft::util::itertools::product<Params>({0.001f},       // tolerance
-                                                        {32, 64, 512},  // m
-                                                        {32, 64, 512},  // n
-                                                        {8, 32},        // k
-                                                        {2, 32},        // num_groups
-                                                        {true, false},  // sqrt
-                                                        {1234ULL},      // seed
-                                                        {AdjacencyPattern::all_true,
-                                                         AdjacencyPattern::checkerboard,
-                                                         AdjacencyPattern::checkerboard_64,
-                                                         AdjacencyPattern::all_false});
-
-  // Irregular sizes to check tiling and bounds checking
-  auto irregular = raft::util::itertools::product<Params>({0.001f},         // tolerance
-                                                          {511, 512, 513},  // m
-                                                          {127, 128, 129},  // n
-                                                          {5},              // k
-                                                          {3, 9},           // num_groups
-                                                          {true, false},    // sqrt
-                                                          {1234ULL},        // seed
-                                                          {AdjacencyPattern::all_true,
-                                                           AdjacencyPattern::checkerboard,
-                                                           AdjacencyPattern::checkerboard_64});
-
-  regular.insert(regular.end(), irregular.begin(), irregular.end());
-
-  return regular;
-}
-
-class MaskedL2NNTest : public ::testing::TestWithParam<Params> {
-  // Empty.
-};
-
-//
-TEST_P(MaskedL2NNTest, ReferenceCheckFloat)
-{
-  using DataT = float;
-
-  // Get parameters; create handle and input data.
-  Params p = GetParam();
-  raft::handle_t handle{};
-  Inputs<DataT> inputs{handle, p};
-
-  // Calculate reference and test output
-  auto out_reference = reference(handle, inputs, p);
-  auto out_fast      = run_masked_nn(handle, inputs, p);
-
-  // Check for differences.
-  ASSERT_TRUE(devArrMatch(out_reference.data_handle(),
-                          out_fast.data_handle(),
-                          p.m,
-                          CompareApproxAbsKVP<DataT>(p.tolerance),
-                          resource::get_cuda_stream(handle)));
-}
-
-// This test checks whether running the masked_l2_nn twice returns the same
-// output.
-TEST_P(MaskedL2NNTest, DeterminismCheck)
-{
-  using DataT = float;
-
-  // Get parameters; create handle and input data.
-  Params p = GetParam();
-  raft::handle_t handle{};
-  Inputs<DataT> inputs{handle, p};
-
-  // Calculate reference and test output
-  auto out1 = run_masked_nn(handle, inputs, p);
-  auto out2 = run_masked_nn(handle, inputs, p);
-
-  // Check for differences.
-  ASSERT_TRUE(devArrMatch(out1.data_handle(),
-                          out2.data_handle(),
-                          p.m,
-                          CompareApproxAbsKVP<DataT>(p.tolerance),
-                          resource::get_cuda_stream(handle)));
-}
-
-TEST_P(MaskedL2NNTest, ReferenceCheckDouble)
-{
-  using DataT = double;
-
-  // Get parameters; create handle and input data.
-  Params p = GetParam();
-  raft::handle_t handle{};
-  Inputs<DataT> inputs{handle, p};
-
-  // Calculate reference and test output
-  auto out_reference = reference(handle, inputs, p);
-  auto out_fast      = run_masked_nn(handle, inputs, p);
-
-  // Check for differences.
-  ASSERT_TRUE(devArrMatch(out_reference.data_handle(),
-                          out_fast.data_handle(),
-                          p.m,
-                          CompareApproxAbsKVP<DataT>(p.tolerance),
-                          resource::get_cuda_stream(handle)));
-}
-
-INSTANTIATE_TEST_CASE_P(MaskedL2NNTests, MaskedL2NNTest, ::testing::ValuesIn(gen_params()));
-
-}  // end namespace cuvs::distance::masked_nn
diff --git a/cpp/test/distance/masked_nn_compress_to_bits.cu b/cpp/test/distance/masked_nn_compress_to_bits.cu
deleted file mode 100644
index f761b6373..000000000
--- a/cpp/test/distance/masked_nn_compress_to_bits.cu
+++ /dev/null
@@ -1,217 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "../test_utils.h"
-#include <cstdio>
-#include <cuvs/distance/detail/compress_to_bits.cuh>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/handle.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/matrix/init.cuh>
-#include <raft/random/rng.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <raft/util/integer_utils.hpp>
-#include <raft/util/itertools.hpp>
-
-namespace cuvs::distance::masked_nn::compress_to_bits {
-
-/**
- * @brief Transpose and decompress 2D bitfield to boolean matrix
- *
- * Inverse operation of compress_to_bits
- *
- * @tparam T
- *
- * @parameter[in]  in       An `m x n` bitfield matrix. Row major.
- * @parameter      in_rows  The number of rows of `in`, i.e. `m`.
- * @parameter      in_cols  The number of cols of `in`, i.e. `n`.
- *
- * @parameter[out] out      An `(m * bits_per_elem) x n` boolean matrix.
- */
-template <typename T = uint64_t, typename = std::enable_if_t<std::is_integral<T>::value>>
-RAFT_KERNEL decompress_bits_kernel(const T* in, int in_rows, int in_cols, bool* out)
-{
-  constexpr int bits_per_element = 8 * sizeof(T);
-
-  const size_t i = threadIdx.y + blockIdx.y * blockDim.y;
-  const size_t j = threadIdx.x + blockIdx.x * blockDim.x;
-
-  if (in_rows <= i || in_cols <= j) { return; }
-
-  const size_t out_rows = in_rows * bits_per_element;
-  const size_t out_cols = in_cols;
-  const size_t out_i    = i * bits_per_element;
-  const size_t out_j    = j;
-
-  if (out_rows <= out_i && out_cols <= out_j) { return; }
-
-  T bitfield = in[i * in_cols + j];
-  for (int bitpos = 0; bitpos < bits_per_element; ++bitpos) {
-    bool bit                                 = ((T(1) << bitpos) & bitfield) != 0;
-    out[(out_i + bitpos) * out_cols + out_j] = bit;
-  }
-}
-
-/**
- * @brief Transpose and decompress 2D bitfield to boolean matrix
- *
- * Inverse operation of compress_to_bits
- *
- * @tparam T
- *
- * @parameter[in]  in       An `m x n` bitfield matrix. Row major.
- * @parameter      in_rows  The number of rows of `in`, i.e. `m`.
- * @parameter      in_cols  The number of cols of `in`, i.e. `n`.
- *
- * @parameter[out] out      An `n x (m * bits_per_elem)` boolean matrix.
- */
-template <typename T = uint64_t, typename = std::enable_if_t<std::is_integral<T>::value>>
-void decompress_bits(const raft::handle_t& handle, const T* in, int in_rows, int in_cols, bool* out)
-{
-  auto stream = resource::get_cuda_stream(handle);
-  dim3 grid(raft::ceildiv(in_cols, 32), raft::ceildiv(in_rows, 32));
-  dim3 block(32, 32);
-  decompress_bits_kernel<<<grid, block, 0, stream>>>(in, in_rows, in_cols, out);
-  RAFT_CUDA_TRY(cudaGetLastError());
-}
-
-// Params holds parameters for test case
-struct Params {
-  int m, n;
-};
-
-inline auto operator<<(std::ostream& os, const Params& p) -> std::ostream&
-{
-  return os << "m: " << p.m << ", n: " << p.n;
-}
-
-// Check that the following holds
-//
-//  decompress(compress(x)) == x
-//
-// for 2D boolean matrices x.
-template <typename T>
-void check_invertible(const Params& p)
-{
-  using cuvs::distance::detail::compress_to_bits;
-  constexpr int bits_per_elem = sizeof(T) * 8;
-
-  // Make m and n that are safe to ceildiv.
-  int m = raft::round_up_safe(p.m, bits_per_elem);
-  int n = p.n;
-
-  // Generate random input
-  raft::handle_t handle{};
-  raft::random::RngState r(1ULL);
-  auto in = raft::make_device_matrix<bool, int>(handle, m, n);
-  raft::random::bernoulli(handle, r, in.data_handle(), m * n, 0.5f);
-
-  int tmp_m = raft::ceildiv(m, bits_per_elem);
-  int out_m = tmp_m * bits_per_elem;
-
-  auto tmp = raft::make_device_matrix<T, int>(handle, tmp_m, n);
-  auto out = raft::make_device_matrix<bool, int>(handle, out_m, n);
-
-  resource::sync_stream(handle);
-  RAFT_CUDA_TRY(cudaGetLastError());
-
-  ASSERT_EQ(in.extent(0), out.extent(0)) << "M does not match";
-  ASSERT_EQ(in.extent(1), out.extent(1)) << "N does not match";
-
-  compress_to_bits(handle, in.view(), tmp.view());
-  resource::sync_stream(handle);
-  RAFT_CUDA_TRY(cudaGetLastError());
-
-  decompress_bits(handle, tmp.data_handle(), tmp.extent(0), tmp.extent(1), out.data_handle());
-  resource::sync_stream(handle);
-  RAFT_CUDA_TRY(cudaGetLastError());
-
-  // Check for differences.
-  ASSERT_TRUE(raft::devArrMatch(in.data_handle(),
-                                out.data_handle(),
-                                in.extent(0) * in.extent(1),
-                                raft::Compare<bool>(),
-                                resource::get_cuda_stream(handle)));
-  resource::sync_stream(handle);
-  RAFT_CUDA_TRY(cudaGetLastError());
-}
-
-void check_all_true(const Params& p)
-{
-  using cuvs::distance::detail::compress_to_bits;
-  using T                     = uint64_t;
-  constexpr int bits_per_elem = sizeof(T) * 8;
-
-  // Make m and n that are safe to ceildiv.
-  int m = raft::round_up_safe(p.m, bits_per_elem);
-  int n = p.n;
-
-  raft::handle_t handle{};
-  raft::random::RngState r(1ULL);
-  auto in = raft::make_device_matrix<bool, int>(handle, m, n);
-  raft::matrix::fill(handle, in.view(), true);
-
-  int tmp_m = raft::ceildiv(m, bits_per_elem);
-  auto tmp  = raft::make_device_matrix<T, int>(handle, tmp_m, n);
-  resource::sync_stream(handle);
-  RAFT_CUDA_TRY(cudaGetLastError());
-
-  compress_to_bits(handle, in.view(), tmp.view());
-  resource::sync_stream(handle);
-  RAFT_CUDA_TRY(cudaGetLastError());
-
-  auto expected = raft::make_device_matrix<T, int>(handle, tmp_m, n);
-  raft::matrix::fill(handle, expected.view(), ~T(0));
-
-  // Check for differences.
-  ASSERT_TRUE(raft::devArrMatch(expected.data_handle(),
-                                tmp.data_handle(),
-                                tmp.extent(0) * tmp.extent(1),
-                                raft::Compare<T>(),
-                                resource::get_cuda_stream(handle)));
-  resource::sync_stream(handle);
-  RAFT_CUDA_TRY(cudaGetLastError());
-}
-
-class CompressToBitsTest : public ::testing::TestWithParam<Params> {
-  // Empty.
-};
-
-TEST_P(CompressToBitsTest, CheckTrue64) { check_all_true(GetParam()); }
-
-TEST_P(CompressToBitsTest, CheckInvertible64)
-{
-  using T = uint64_t;
-  check_invertible<T>(GetParam());
-}
-
-TEST_P(CompressToBitsTest, CheckInvertible32)
-{
-  using T = uint32_t;
-  check_invertible<T>(GetParam());
-}
-
-std::vector<Params> params = raft::util::itertools::product<Params>(
-  {1, 3, 32, 33, 63, 64, 65, 128, 10013}, {1, 3, 32, 33, 63, 64, 65, 13001});
-
-INSTANTIATE_TEST_CASE_P(CompressToBits, CompressToBitsTest, ::testing::ValuesIn(params));
-
-}  // namespace cuvs::distance::masked_nn::compress_to_bits
\ No newline at end of file
diff --git a/cpp/test/ext_headers/00_generate.py b/cpp/test/ext_headers/00_generate.py
deleted file mode 100644
index 6100fba73..000000000
--- a/cpp/test/ext_headers/00_generate.py
+++ /dev/null
@@ -1,79 +0,0 @@
-# Copyright (c) 2023, NVIDIA CORPORATION.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-copyright_notice = """
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-"""
-
-ext_headers = [
-    "cuvs/neighbors/brute_force-ext.cuh",
-    "cuvs/distance/distance-ext.cuh",
-    "cuvs/distance/detail/pairwise_matrix/dispatch-ext.cuh",
-    "raft/matrix/detail/select_k-ext.cuh",
-    "cuvs/neighbors/ball_cover-ext.cuh",
-    "cuvs/spatial/knn/detail/fused_l2_knn-ext.cuh",
-    "cuvs/distance/fused_l2_nn-ext.cuh",
-    "cuvs/neighbors/ivf_pq-ext.cuh",
-    "raft/util/memory_pool-ext.hpp",
-    "cuvs/neighbors/ivf_flat-ext.cuh",
-    "raft/core/logger-ext.hpp",
-    "cuvs/neighbors/refine-ext.cuh",
-    "cuvs/neighbors/detail/ivf_flat_search-ext.cuh",
-    "cuvs/neighbors/detail/selection_faiss-ext.cuh",
-    "raft/linalg/detail/coalesced_reduction-ext.cuh",
-    "cuvs/spatial/knn/detail/ball_cover/registers-ext.cuh",
-    "cuvs/neighbors/detail/ivf_flat_interleaved_scan-ext.cuh",
-    "cuvs/neighbors/detail/ivf_pq_compute_similarity-ext.cuh",
-]
-
-for ext_header in ext_headers:
-    header = ext_header.replace("-ext", "")
-
-    path = (
-        header
-        .replace("/", "_")
-        .replace(".cuh", ".cu")
-        .replace(".hpp", ".cpp")
-    )
-
-    with open(path, "w") as f:
-        f.write(copyright_notice)
-        f.write(f"#include <{header}>\n")
-
-    # For in CMakeLists.txt
-    print(f"test/ext_headers/{path}")
diff --git a/cpp/test/ext_headers/raft_core_logger.cpp b/cpp/test/ext_headers/raft_core_logger.cpp
deleted file mode 100644
index 18ba9ef48..000000000
--- a/cpp/test/ext_headers/raft_core_logger.cpp
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <raft/core/logger.hpp>
diff --git a/cpp/test/ext_headers/raft_distance_detail_pairwise_matrix_dispatch.cu b/cpp/test/ext_headers/raft_distance_detail_pairwise_matrix_dispatch.cu
deleted file mode 100644
index 72816b760..000000000
--- a/cpp/test/ext_headers/raft_distance_detail_pairwise_matrix_dispatch.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/distance/detail/pairwise_matrix/dispatch.cuh>
diff --git a/cpp/test/ext_headers/raft_distance_distance.cu b/cpp/test/ext_headers/raft_distance_distance.cu
deleted file mode 100644
index c662267d6..000000000
--- a/cpp/test/ext_headers/raft_distance_distance.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/distance/distance.cuh>
diff --git a/cpp/test/ext_headers/raft_distance_fused_l2_nn.cu b/cpp/test/ext_headers/raft_distance_fused_l2_nn.cu
deleted file mode 100644
index 10ceedab5..000000000
--- a/cpp/test/ext_headers/raft_distance_fused_l2_nn.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/distance/fused_l2_nn.cuh>
diff --git a/cpp/test/ext_headers/raft_linalg_detail_coalesced_reduction.cu b/cpp/test/ext_headers/raft_linalg_detail_coalesced_reduction.cu
deleted file mode 100644
index 7f9482428..000000000
--- a/cpp/test/ext_headers/raft_linalg_detail_coalesced_reduction.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <raft/linalg/detail/coalesced_reduction.cuh>
diff --git a/cpp/test/ext_headers/raft_matrix_detail_select_k.cu b/cpp/test/ext_headers/raft_matrix_detail_select_k.cu
deleted file mode 100644
index adb10f5bb..000000000
--- a/cpp/test/ext_headers/raft_matrix_detail_select_k.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <raft/matrix/detail/select_k.cuh>
diff --git a/cpp/test/ext_headers/raft_neighbors_ball_cover.cu b/cpp/test/ext_headers/raft_neighbors_ball_cover.cu
deleted file mode 100644
index 3cc0d545b..000000000
--- a/cpp/test/ext_headers/raft_neighbors_ball_cover.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ball_cover.cuh>
diff --git a/cpp/test/ext_headers/raft_neighbors_brute_force.cu b/cpp/test/ext_headers/raft_neighbors_brute_force.cu
deleted file mode 100644
index da96897f8..000000000
--- a/cpp/test/ext_headers/raft_neighbors_brute_force.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/neighbors/brute_force.cuh>
diff --git a/cpp/test/ext_headers/raft_neighbors_detail_ivf_flat_interleaved_scan.cu b/cpp/test/ext_headers/raft_neighbors_detail_ivf_flat_interleaved_scan.cu
deleted file mode 100644
index 49a528747..000000000
--- a/cpp/test/ext_headers/raft_neighbors_detail_ivf_flat_interleaved_scan.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/ivf_flat_interleaved_scan.cuh>
diff --git a/cpp/test/ext_headers/raft_neighbors_detail_ivf_flat_search.cu b/cpp/test/ext_headers/raft_neighbors_detail_ivf_flat_search.cu
deleted file mode 100644
index d7fc13f10..000000000
--- a/cpp/test/ext_headers/raft_neighbors_detail_ivf_flat_search.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/ivf_flat_search.cuh>
diff --git a/cpp/test/ext_headers/raft_neighbors_detail_ivf_pq_compute_similarity.cu b/cpp/test/ext_headers/raft_neighbors_detail_ivf_pq_compute_similarity.cu
deleted file mode 100644
index 5e0f97294..000000000
--- a/cpp/test/ext_headers/raft_neighbors_detail_ivf_pq_compute_similarity.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/ivf_pq_compute_similarity.cuh>
diff --git a/cpp/test/ext_headers/raft_neighbors_detail_selection_faiss.cu b/cpp/test/ext_headers/raft_neighbors_detail_selection_faiss.cu
deleted file mode 100644
index 1db807b40..000000000
--- a/cpp/test/ext_headers/raft_neighbors_detail_selection_faiss.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/neighbors/detail/selection_faiss.cuh>
diff --git a/cpp/test/ext_headers/raft_neighbors_ivf_flat.cu b/cpp/test/ext_headers/raft_neighbors_ivf_flat.cu
deleted file mode 100644
index c42bb88ac..000000000
--- a/cpp/test/ext_headers/raft_neighbors_ivf_flat.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_flat.cuh>
diff --git a/cpp/test/ext_headers/raft_neighbors_ivf_pq.cu b/cpp/test/ext_headers/raft_neighbors_ivf_pq.cu
deleted file mode 100644
index a144e7bf2..000000000
--- a/cpp/test/ext_headers/raft_neighbors_ivf_pq.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/neighbors/ivf_pq.cuh>
diff --git a/cpp/test/ext_headers/raft_neighbors_refine.cu b/cpp/test/ext_headers/raft_neighbors_refine.cu
deleted file mode 100644
index 090f1e9df..000000000
--- a/cpp/test/ext_headers/raft_neighbors_refine.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/neighbors/refine.cuh>
diff --git a/cpp/test/ext_headers/raft_spatial_knn_detail_ball_cover_registers.cu b/cpp/test/ext_headers/raft_spatial_knn_detail_ball_cover_registers.cu
deleted file mode 100644
index e6f03c428..000000000
--- a/cpp/test/ext_headers/raft_spatial_knn_detail_ball_cover_registers.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/spatial/knn/detail/ball_cover/registers.cuh>
diff --git a/cpp/test/ext_headers/raft_spatial_knn_detail_fused_l2_knn.cu b/cpp/test/ext_headers/raft_spatial_knn_detail_fused_l2_knn.cu
deleted file mode 100644
index 633629bd3..000000000
--- a/cpp/test/ext_headers/raft_spatial_knn_detail_fused_l2_knn.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <cuvs/spatial/knn/detail/fused_l2_knn.cuh>
diff --git a/cpp/test/ext_headers/raft_util_memory_pool.cpp b/cpp/test/ext_headers/raft_util_memory_pool.cpp
deleted file mode 100644
index 11a024b95..000000000
--- a/cpp/test/ext_headers/raft_util_memory_pool.cpp
+++ /dev/null
@@ -1,27 +0,0 @@
-
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-/*
- * NOTE: this file is generated by 00_generate.py
- *
- * Make changes there and run in this directory:
- *
- * > python 00_generate.py
- *
- */
-
-#include <raft/util/memory_pool.hpp>
diff --git a/cpp/test/neighbors/ann_cagra.cuh b/cpp/test/neighbors/ann_cagra.cuh
index ee02581dc..9c58275c6 100644
--- a/cpp/test/neighbors/ann_cagra.cuh
+++ b/cpp/test/neighbors/ann_cagra.cuh
@@ -15,7 +15,6 @@
  */
 #pragma once
 
-
 #include "../test_utils.cuh"
 #include "ann_utils.cuh"
 #include <raft/core/resource/cuda_stream.hpp>
@@ -24,12 +23,11 @@
 
 #include <cuvs/distance/distance_types.hpp>
 #include <cuvs/neighbors/cagra.hpp>
-//#include <cuvs/neighbors/sample_filter.cuh>
-#include <raft/neighbors/cagra.cuh>
 #include <raft/core/device_mdspan.hpp>
 #include <raft/core/device_resources.hpp>
 #include <raft/core/logger.hpp>
 #include <raft/linalg/add.cuh>
+#include <raft/neighbors/cagra.cuh>
 #include <raft/random/rng.cuh>
 #include <raft/util/itertools.hpp>
 
@@ -184,15 +182,15 @@ class AnnCagraTest : public ::testing::TestWithParam<AnnCagraInputs> {
       rmm::device_uvector<DistanceT> distances_naive_dev(queries_size, stream_);
       rmm::device_uvector<IdxT> indices_naive_dev(queries_size, stream_);
       cuvs::neighbors::naive_knn<DistanceT, DataT, IdxT>(handle_,
-                                        distances_naive_dev.data(),
-                                        indices_naive_dev.data(),
-                                        search_queries.data(),
-                                        database.data(),
-                                        ps.n_queries,
-                                        ps.n_rows,
-                                        ps.dim,
-                                        ps.k,
-                                        ps.metric);
+                                                         distances_naive_dev.data(),
+                                                         indices_naive_dev.data(),
+                                                         search_queries.data(),
+                                                         database.data(),
+                                                         ps.n_queries,
+                                                         ps.n_rows,
+                                                         ps.dim,
+                                                         ps.k,
+                                                         ps.metric);
       raft::update_host(distances_naive.data(), distances_naive_dev.data(), queries_size, stream_);
       raft::update_host(indices_naive.data(), indices_naive_dev.data(), queries_size, stream_);
       raft::resource::sync_stream(handle_);
diff --git a/cpp/test/neighbors/ann_cagra/test_float_uint32_t.cu b/cpp/test/neighbors/ann_cagra/test_float_uint32_t.cu
index 500c10a11..c510dfd93 100644
--- a/cpp/test/neighbors/ann_cagra/test_float_uint32_t.cu
+++ b/cpp/test/neighbors/ann_cagra/test_float_uint32_t.cu
@@ -22,20 +22,7 @@ namespace cuvs::neighbors::cagra {
 
 typedef AnnCagraTest<float, float, std::uint32_t> AnnCagraTestF_U32;
 TEST_P(AnnCagraTestF_U32, AnnCagra) { this->testCagra(); }
-/*
-typedef AnnCagraSortTest<float, float, std::uint32_t> AnnCagraSortTestF_U32;
-TEST_P(AnnCagraSortTestF_U32, AnnCagraSort) { this->testCagraSort(); }
-
-typedef AnnCagraFilterTest<float, float, std::uint32_t> AnnCagraFilterTestF_U32;
-TEST_P(AnnCagraFilterTestF_U32, AnnCagraFilter)
-{
-  this->testCagraFilter();
-  this->testCagraRemoved();
-}
-*/
 
 INSTANTIATE_TEST_CASE_P(AnnCagraTest, AnnCagraTestF_U32, ::testing::ValuesIn(inputs));
-//INSTANTIATE_TEST_CASE_P(AnnCagraSortTest, AnnCagraSortTestF_U32, ::testing::ValuesIn(inputs));
-//INSTANTIATE_TEST_CASE_P(AnnCagraFilterTest, AnnCagraFilterTestF_U32, ::testing::ValuesIn(inputs));
 
 }  // namespace cuvs::neighbors::cagra
diff --git a/cpp/test/neighbors/ann_cagra/test_int8_t_uint32_t.cu b/cpp/test/neighbors/ann_cagra/test_int8_t_uint32_t.cu
index d0ac4b298..df5db45aa 100644
--- a/cpp/test/neighbors/ann_cagra/test_int8_t_uint32_t.cu
+++ b/cpp/test/neighbors/ann_cagra/test_int8_t_uint32_t.cu
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2023, NVIDIA CORPORATION.
+ * Copyright (c) 2023-2024, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -22,19 +22,7 @@ namespace cuvs::neighbors::cagra {
 
 typedef AnnCagraTest<float, std::int8_t, std::uint32_t> AnnCagraTestI8_U32;
 TEST_P(AnnCagraTestI8_U32, AnnCagra) { this->testCagra(); }
-/*
-typedef AnnCagraSortTest<float, std::int8_t, std::uint32_t> AnnCagraSortTestI8_U32;
-TEST_P(AnnCagraSortTestI8_U32, AnnCagraSort) { this->testCagraSort(); }
-typedef AnnCagraFilterTest<float, std::int8_t, std::uint32_t> AnnCagraFilterTestI8_U32;
-TEST_P(AnnCagraFilterTestI8_U32, AnnCagraFilter)
-{
-  this->testCagraFilter();
-  this->testCagraRemoved();
-}
-*/
 
 INSTANTIATE_TEST_CASE_P(AnnCagraTest, AnnCagraTestI8_U32, ::testing::ValuesIn(inputs));
-//INSTANTIATE_TEST_CASE_P(AnnCagraSortTest, AnnCagraSortTestI8_U32, ::testing::ValuesIn(inputs));
-//INSTANTIATE_TEST_CASE_P(AnnCagraFilterTest, AnnCagraFilterTestI8_U32, ::testing::ValuesIn(inputs));
 
 }  // namespace cuvs::neighbors::cagra
diff --git a/cpp/test/neighbors/ann_cagra/test_uint8_t_uint32_t.cu b/cpp/test/neighbors/ann_cagra/test_uint8_t_uint32_t.cu
index 07e26b773..27ffcd915 100644
--- a/cpp/test/neighbors/ann_cagra/test_uint8_t_uint32_t.cu
+++ b/cpp/test/neighbors/ann_cagra/test_uint8_t_uint32_t.cu
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2023, NVIDIA CORPORATION.
+ * Copyright (c) 2023-2024, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -22,19 +22,7 @@ namespace cuvs::neighbors::cagra {
 
 typedef AnnCagraTest<float, std::uint8_t, std::uint32_t> AnnCagraTestU8_U32;
 TEST_P(AnnCagraTestU8_U32, AnnCagra) { this->testCagra(); }
-/*
-typedef AnnCagraSortTest<float, std::uint8_t, std::uint32_t> AnnCagraSortTestU8_U32;
-TEST_P(AnnCagraSortTestU8_U32, AnnCagraSort) { this->testCagraSort(); }
 
-typedef AnnCagraFilterTest<float, std::uint8_t, std::uint32_t> AnnCagraFilterTestU8_U32;
-TEST_P(AnnCagraFilterTestU8_U32, AnnCagraSort)
-{
-  this->testCagraFilter();
-  this->testCagraRemoved();
-}
-*/
 INSTANTIATE_TEST_CASE_P(AnnCagraTest, AnnCagraTestU8_U32, ::testing::ValuesIn(inputs));
-//INSTANTIATE_TEST_CASE_P(AnnCagraSortTest, AnnCagraSortTestU8_U32, ::testing::ValuesIn(inputs));
-//INSTANTIATE_TEST_CASE_P(AnnCagraFilterTest, AnnCagraFilterTestU8_U32, ::testing::ValuesIn(inputs));
 
 }  // namespace cuvs::neighbors::cagra
diff --git a/cpp/test/neighbors/ann_ivf_flat.cuh b/cpp/test/neighbors/ann_ivf_flat.cuh
deleted file mode 100644
index 286f5c5fa..000000000
--- a/cpp/test/neighbors/ann_ivf_flat.cuh
+++ /dev/null
@@ -1,615 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#include "../test_utils.cuh"
-#include "ann_utils.cuh"
-#include <cuvs/neighbors/ivf_flat_types.hpp>
-#include <cuvs/neighbors/ivf_list.hpp>
-#include <cuvs/neighbors/sample_filter.cuh>
-#include <raft/core/device_mdarray.hpp>
-#include <raft/core/host_mdarray.hpp>
-#include <raft/core/mdspan.hpp>
-#include <raft/core/mdspan_types.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resource/thrust_policy.hpp>
-#include <raft/linalg/map.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <raft/util/fast_int_div.cuh>
-#include <thrust/functional.h>
-
-#include <cuvs_internal/neighbors/naive_knn.cuh>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/ivf_flat.cuh>
-#include <cuvs/neighbors/ivf_flat_helpers.cuh>
-#include <cuvs/spatial/knn/ann.cuh>
-#include <cuvs/spatial/knn/knn.cuh>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/logger.hpp>
-#include <raft/matrix/gather.cuh>
-#include <raft/random/rng.cuh>
-#include <raft/stats/mean.cuh>
-
-#include <rmm/cuda_stream_view.hpp>
-#include <rmm/device_buffer.hpp>
-
-#include <gtest/gtest.h>
-
-#include <rmm/device_uvector.hpp>
-#include <thrust/sequence.h>
-
-#include <cstddef>
-#include <iostream>
-#include <vector>
-
-namespace cuvs::neighbors::ivf_flat {
-
-struct test_ivf_sample_filter {
-  static constexpr unsigned offset = 300;
-};
-
-template <typename IdxT>
-struct AnnIvfFlatInputs {
-  IdxT num_queries;
-  IdxT num_db_vecs;
-  IdxT dim;
-  IdxT k;
-  IdxT nprobe;
-  IdxT nlist;
-  cuvs::distance::DistanceType metric;
-  bool adaptive_centers;
-};
-
-template <typename IdxT>
-::std::ostream& operator<<(::std::ostream& os, const AnnIvfFlatInputs<IdxT>& p)
-{
-  os << "{ " << p.num_queries << ", " << p.num_db_vecs << ", " << p.dim << ", " << p.k << ", "
-     << p.nprobe << ", " << p.nlist << ", " << static_cast<int>(p.metric) << ", "
-     << p.adaptive_centers << '}' << std::endl;
-  return os;
-}
-
-template <typename T, typename DataT, typename IdxT>
-class AnnIVFFlatTest : public ::testing::TestWithParam<AnnIvfFlatInputs<IdxT>> {
- public:
-  AnnIVFFlatTest()
-    : stream_(resource::get_cuda_stream(handle_)),
-      ps(::testing::TestWithParam<AnnIvfFlatInputs<IdxT>>::GetParam()),
-      database(0, stream_),
-      search_queries(0, stream_)
-  {
-  }
-
-  void testIVFFlat()
-  {
-    size_t queries_size = ps.num_queries * ps.k;
-    std::vector<IdxT> indices_ivfflat(queries_size);
-    std::vector<IdxT> indices_naive(queries_size);
-    std::vector<T> distances_ivfflat(queries_size);
-    std::vector<T> distances_naive(queries_size);
-
-    {
-      rmm::device_uvector<T> distances_naive_dev(queries_size, stream_);
-      rmm::device_uvector<IdxT> indices_naive_dev(queries_size, stream_);
-      naive_knn<T, DataT, IdxT>(handle_,
-                                distances_naive_dev.data(),
-                                indices_naive_dev.data(),
-                                search_queries.data(),
-                                database.data(),
-                                ps.num_queries,
-                                ps.num_db_vecs,
-                                ps.dim,
-                                ps.k,
-                                ps.metric);
-      update_host(distances_naive.data(), distances_naive_dev.data(), queries_size, stream_);
-      update_host(indices_naive.data(), indices_naive_dev.data(), queries_size, stream_);
-      resource::sync_stream(handle_);
-    }
-
-    {
-      // unless something is really wrong with clustering, this could serve as a lower bound on
-      // recall
-      double min_recall = static_cast<double>(ps.nprobe) / static_cast<double>(ps.nlist);
-
-      rmm::device_uvector<T> distances_ivfflat_dev(queries_size, stream_);
-      rmm::device_uvector<IdxT> indices_ivfflat_dev(queries_size, stream_);
-
-      {
-        // legacy interface
-        cuvs::spatial::knn::IVFFlatParam ivfParams;
-        ivfParams.nprobe = ps.nprobe;
-        ivfParams.nlist  = ps.nlist;
-        cuvs::spatial::knn::knnIndex index;
-
-        approx_knn_build_index(handle_,
-                               &index,
-                               dynamic_cast<cuvs::spatial::knn::knnIndexParam*>(&ivfParams),
-                               ps.metric,
-                               (IdxT)0,
-                               database.data(),
-                               ps.num_db_vecs,
-                               ps.dim);
-
-        resource::sync_stream(handle_);
-        approx_knn_search(handle_,
-                          distances_ivfflat_dev.data(),
-                          indices_ivfflat_dev.data(),
-                          &index,
-                          ps.k,
-                          search_queries.data(),
-                          ps.num_queries);
-
-        update_host(distances_ivfflat.data(), distances_ivfflat_dev.data(), queries_size, stream_);
-        update_host(indices_ivfflat.data(), indices_ivfflat_dev.data(), queries_size, stream_);
-        resource::sync_stream(handle_);
-      }
-
-      ASSERT_TRUE(eval_neighbours(indices_naive,
-                                  indices_ivfflat,
-                                  distances_naive,
-                                  distances_ivfflat,
-                                  ps.num_queries,
-                                  ps.k,
-                                  0.001,
-                                  min_recall));
-      {
-        ivf_flat::index_params index_params;
-        ivf_flat::search_params search_params;
-        index_params.n_lists          = ps.nlist;
-        index_params.metric           = ps.metric;
-        index_params.adaptive_centers = ps.adaptive_centers;
-        search_params.n_probes        = ps.nprobe;
-
-        index_params.add_data_on_build        = false;
-        index_params.kmeans_trainset_fraction = 0.5;
-        index_params.metric_arg               = 0;
-
-        auto database_view = raft::make_device_matrix_view<const DataT, IdxT>(
-          (const DataT*)database.data(), ps.num_db_vecs, ps.dim);
-
-        auto idx = ivf_flat::build(handle_, index_params, database_view);
-
-        rmm::device_uvector<IdxT> vector_indices(ps.num_db_vecs, stream_);
-        thrust::sequence(resource::get_thrust_policy(handle_),
-                         thrust::device_pointer_cast(vector_indices.data()),
-                         thrust::device_pointer_cast(vector_indices.data() + ps.num_db_vecs));
-        resource::sync_stream(handle_);
-
-        IdxT half_of_data = ps.num_db_vecs / 2;
-
-        auto half_of_data_view = raft::make_device_matrix_view<const DataT, IdxT>(
-          (const DataT*)database.data(), half_of_data, ps.dim);
-
-        const std::optional<raft::device_vector_view<const IdxT, IdxT>> no_opt = std::nullopt;
-        index<DataT, IdxT> index_2 = ivf_flat::extend(handle_, half_of_data_view, no_opt, idx);
-
-        auto new_half_of_data_view = raft::make_device_matrix_view<const DataT, IdxT>(
-          database.data() + half_of_data * ps.dim, IdxT(ps.num_db_vecs) - half_of_data, ps.dim);
-
-        auto new_half_of_data_indices_view = raft::make_device_vector_view<const IdxT, IdxT>(
-          vector_indices.data() + half_of_data, IdxT(ps.num_db_vecs) - half_of_data);
-
-        ivf_flat::extend(handle_,
-                         new_half_of_data_view,
-                         std::make_optional<raft::device_vector_view<const IdxT, IdxT>>(
-                           new_half_of_data_indices_view),
-                         &index_2);
-
-        auto search_queries_view = raft::make_device_matrix_view<const DataT, IdxT>(
-          search_queries.data(), ps.num_queries, ps.dim);
-        auto indices_out_view = raft::make_device_matrix_view<IdxT, IdxT>(
-          indices_ivfflat_dev.data(), ps.num_queries, ps.k);
-        auto dists_out_view = raft::make_device_matrix_view<T, IdxT>(
-          distances_ivfflat_dev.data(), ps.num_queries, ps.k);
-        ivf_flat::detail::serialize(handle_, "ivf_flat_index", index_2);
-
-        auto index_loaded = ivf_flat::detail::deserialize<DataT, IdxT>(handle_, "ivf_flat_index");
-        ASSERT_EQ(index_2.size(), index_loaded.size());
-
-        ivf_flat::search(handle_,
-                         search_params,
-                         index_loaded,
-                         search_queries_view,
-                         indices_out_view,
-                         dists_out_view);
-
-        update_host(distances_ivfflat.data(), distances_ivfflat_dev.data(), queries_size, stream_);
-        update_host(indices_ivfflat.data(), indices_ivfflat_dev.data(), queries_size, stream_);
-        resource::sync_stream(handle_);
-
-        // Test the centroid invariants
-        if (index_2.adaptive_centers()) {
-          // The centers must be up-to-date with the corresponding data
-          std::vector<uint32_t> list_sizes(index_2.n_lists());
-          std::vector<IdxT*> list_indices(index_2.n_lists());
-          rmm::device_uvector<float> centroid(ps.dim, stream_);
-          raft::copy(
-            list_sizes.data(), index_2.list_sizes().data_handle(), index_2.n_lists(), stream_);
-          raft::copy(
-            list_indices.data(), index_2.inds_ptrs().data_handle(), index_2.n_lists(), stream_);
-          resource::sync_stream(handle_);
-          for (uint32_t l = 0; l < index_2.n_lists(); l++) {
-            if (list_sizes[l] == 0) continue;
-            rmm::device_uvector<float> cluster_data(list_sizes[l] * ps.dim, stream_);
-            cuvs::spatial::knn::detail::utils::copy_selected<float>((IdxT)list_sizes[l],
-                                                                    (IdxT)ps.dim,
-                                                                    database.data(),
-                                                                    list_indices[l],
-                                                                    (IdxT)ps.dim,
-                                                                    cluster_data.data(),
-                                                                    (IdxT)ps.dim,
-                                                                    stream_);
-            raft::stats::mean<float, uint32_t>(
-              centroid.data(), cluster_data.data(), ps.dim, list_sizes[l], false, true, stream_);
-            ASSERT_TRUE(raft::devArrMatch(index_2.centers().data_handle() + ps.dim * l,
-                                          centroid.data(),
-                                          ps.dim,
-                                          raft::CompareApprox<float>(0.001),
-                                          stream_));
-          }
-        } else {
-          // The centers must be immutable
-          ASSERT_TRUE(raft::devArrMatch(index_2.centers().data_handle(),
-                                        idx.centers().data_handle(),
-                                        index_2.centers().size(),
-                                        raft::Compare<float>(),
-                                        stream_));
-        }
-      }
-      ASSERT_TRUE(eval_neighbours(indices_naive,
-                                  indices_ivfflat,
-                                  distances_naive,
-                                  distances_ivfflat,
-                                  ps.num_queries,
-                                  ps.k,
-                                  0.001,
-                                  min_recall));
-    }
-  }
-
-  void testPacker()
-  {
-    ivf_flat::index_params index_params;
-    ivf_flat::search_params search_params;
-    index_params.n_lists          = ps.nlist;
-    index_params.metric           = ps.metric;
-    index_params.adaptive_centers = false;
-    search_params.n_probes        = ps.nprobe;
-
-    index_params.add_data_on_build        = false;
-    index_params.kmeans_trainset_fraction = 1.0;
-    index_params.metric_arg               = 0;
-
-    auto database_view = raft::make_device_matrix_view<const DataT, IdxT>(
-      (const DataT*)database.data(), ps.num_db_vecs, ps.dim);
-
-    auto idx = ivf_flat::build(handle_, index_params, database_view);
-
-    const std::optional<raft::device_vector_view<const IdxT, IdxT>> no_opt = std::nullopt;
-    index<DataT, IdxT> extend_index = ivf_flat::extend(handle_, database_view, no_opt, idx);
-
-    auto list_sizes = raft::make_host_vector<uint32_t>(idx.n_lists());
-    update_host(list_sizes.data_handle(),
-                extend_index.list_sizes().data_handle(),
-                extend_index.n_lists(),
-                stream_);
-    resource::sync_stream(handle_);
-
-    auto& lists = idx.lists();
-
-    // conservative memory allocation for codepacking
-    auto list_device_spec = list_spec<uint32_t, DataT, IdxT>{idx.dim(), false};
-
-    for (uint32_t label = 0; label < idx.n_lists(); label++) {
-      uint32_t list_size = list_sizes.data_handle()[label];
-
-      ivf::resize_list(handle_, lists[label], list_device_spec, list_size, 0);
-    }
-
-    idx.recompute_internal_state(handle_);
-
-    using interleaved_group = Pow2<kIndexGroupSize>;
-
-    for (uint32_t label = 0; label < idx.n_lists(); label++) {
-      uint32_t list_size = list_sizes.data_handle()[label];
-
-      if (list_size > 0) {
-        uint32_t padded_list_size = interleaved_group::roundUp(list_size);
-        uint32_t n_elems          = padded_list_size * idx.dim();
-        auto list_data            = lists[label]->data;
-        auto list_inds            = extend_index.lists()[label]->indices;
-
-        // fetch the flat codes
-        auto flat_codes = make_device_matrix<DataT, uint32_t>(handle_, list_size, idx.dim());
-
-        matrix::gather(
-          handle_,
-          make_device_matrix_view<const DataT, uint32_t>(
-            (const DataT*)database.data(), static_cast<uint32_t>(ps.num_db_vecs), idx.dim()),
-          make_device_vector_view<const IdxT, uint32_t>((const IdxT*)list_inds.data_handle(),
-                                                        list_size),
-          flat_codes.view());
-
-        helpers::codepacker::pack<DataT, IdxT>(
-          handle_, make_const_mdspan(flat_codes.view()), idx.veclen(), 0, list_data.view());
-
-        {
-          auto mask = make_device_vector<bool>(handle_, n_elems);
-
-          linalg::map_offset(handle_,
-                             mask.view(),
-                             [dim = idx.dim(),
-                              list_size,
-                              padded_list_size,
-                              chunk_size = util::FastIntDiv(idx.veclen())] __device__(auto i) {
-                               uint32_t max_group_offset = interleaved_group::roundDown(list_size);
-                               if (i < max_group_offset * dim) { return true; }
-                               uint32_t surplus    = (i - max_group_offset * dim);
-                               uint32_t ingroup_id = interleaved_group::mod(surplus / chunk_size);
-                               return ingroup_id < (list_size - max_group_offset);
-                             });
-
-          // ensure that the correct number of indices are masked out
-          ASSERT_TRUE(thrust::reduce(resource::get_thrust_policy(handle_),
-                                     mask.data_handle(),
-                                     mask.data_handle() + n_elems,
-                                     0) == list_size * ps.dim);
-
-          auto packed_list_data = make_device_vector<DataT, uint32_t>(handle_, n_elems);
-
-          linalg::map_offset(handle_,
-                             packed_list_data.view(),
-                             [mask      = mask.data_handle(),
-                              list_data = list_data.data_handle()] __device__(uint32_t i) {
-                               if (mask[i]) return list_data[i];
-                               return DataT{0};
-                             });
-
-          auto extend_data          = extend_index.lists()[label]->data;
-          auto extend_data_filtered = make_device_vector<DataT, uint32_t>(handle_, n_elems);
-          linalg::map_offset(handle_,
-                             extend_data_filtered.view(),
-                             [mask        = mask.data_handle(),
-                              extend_data = extend_data.data_handle()] __device__(uint32_t i) {
-                               if (mask[i]) return extend_data[i];
-                               return DataT{0};
-                             });
-
-          ASSERT_TRUE(raft::devArrMatch(packed_list_data.data_handle(),
-                                        extend_data_filtered.data_handle(),
-                                        n_elems,
-                                        raft::Compare<DataT>(),
-                                        stream_));
-        }
-
-        auto unpacked_flat_codes =
-          make_device_matrix<DataT, uint32_t>(handle_, list_size, idx.dim());
-
-        helpers::codepacker::unpack<DataT, IdxT>(
-          handle_, list_data.view(), idx.veclen(), 0, unpacked_flat_codes.view());
-
-        ASSERT_TRUE(raft::devArrMatch(flat_codes.data_handle(),
-                                      unpacked_flat_codes.data_handle(),
-                                      list_size * ps.dim,
-                                      raft::Compare<DataT>(),
-                                      stream_));
-      }
-    }
-  }
-
-  void testFilter()
-  {
-    size_t queries_size = ps.num_queries * ps.k;
-    std::vector<IdxT> indices_ivfflat(queries_size);
-    std::vector<IdxT> indices_naive(queries_size);
-    std::vector<T> distances_ivfflat(queries_size);
-    std::vector<T> distances_naive(queries_size);
-
-    {
-      rmm::device_uvector<T> distances_naive_dev(queries_size, stream_);
-      rmm::device_uvector<IdxT> indices_naive_dev(queries_size, stream_);
-      auto* database_filtered_ptr = database.data() + test_ivf_sample_filter::offset * ps.dim;
-      naive_knn<T, DataT, IdxT>(handle_,
-                                distances_naive_dev.data(),
-                                indices_naive_dev.data(),
-                                search_queries.data(),
-                                database_filtered_ptr,
-                                ps.num_queries,
-                                ps.num_db_vecs - test_ivf_sample_filter::offset,
-                                ps.dim,
-                                ps.k,
-                                ps.metric);
-      raft::linalg::addScalar(indices_naive_dev.data(),
-                              indices_naive_dev.data(),
-                              IdxT(test_ivf_sample_filter::offset),
-                              queries_size,
-                              stream_);
-      update_host(distances_naive.data(), distances_naive_dev.data(), queries_size, stream_);
-      update_host(indices_naive.data(), indices_naive_dev.data(), queries_size, stream_);
-      resource::sync_stream(handle_);
-    }
-
-    {
-      // unless something is really wrong with clustering, this could serve as a lower bound on
-      // recall
-      double min_recall = static_cast<double>(ps.nprobe) / static_cast<double>(ps.nlist);
-
-      auto distances_ivfflat_dev = raft::make_device_matrix<T, IdxT>(handle_, ps.num_queries, ps.k);
-      auto indices_ivfflat_dev =
-        raft::make_device_matrix<IdxT, IdxT>(handle_, ps.num_queries, ps.k);
-
-      {
-        ivf_flat::index_params index_params;
-        ivf_flat::search_params search_params;
-        index_params.n_lists          = ps.nlist;
-        index_params.metric           = ps.metric;
-        index_params.adaptive_centers = ps.adaptive_centers;
-        search_params.n_probes        = ps.nprobe;
-
-        index_params.add_data_on_build        = true;
-        index_params.kmeans_trainset_fraction = 0.5;
-        index_params.metric_arg               = 0;
-
-        // Create IVF Flat index
-        auto database_view = raft::make_device_matrix_view<const DataT, IdxT>(
-          (const DataT*)database.data(), ps.num_db_vecs, ps.dim);
-        auto index = ivf_flat::build(handle_, index_params, database_view);
-
-        // Create Bitset filter
-        auto removed_indices =
-          raft::make_device_vector<IdxT, int64_t>(handle_, test_ivf_sample_filter::offset);
-        thrust::sequence(resource::get_thrust_policy(handle_),
-                         thrust::device_pointer_cast(removed_indices.data_handle()),
-                         thrust::device_pointer_cast(removed_indices.data_handle() +
-                                                     test_ivf_sample_filter::offset));
-        resource::sync_stream(handle_);
-
-        raft::core::bitset<std::uint32_t, IdxT> removed_indices_bitset(
-          handle_, removed_indices.view(), ps.num_db_vecs);
-
-        // Search with the filter
-        auto search_queries_view = raft::make_device_matrix_view<const DataT, IdxT>(
-          search_queries.data(), ps.num_queries, ps.dim);
-        ivf_flat::search_with_filtering(
-          handle_,
-          search_params,
-          index,
-          search_queries_view,
-          indices_ivfflat_dev.view(),
-          distances_ivfflat_dev.view(),
-          cuvs::neighbors::filtering::bitset_filter(removed_indices_bitset.view()));
-
-        update_host(
-          distances_ivfflat.data(), distances_ivfflat_dev.data_handle(), queries_size, stream_);
-        update_host(
-          indices_ivfflat.data(), indices_ivfflat_dev.data_handle(), queries_size, stream_);
-        resource::sync_stream(handle_);
-      }
-      ASSERT_TRUE(eval_neighbours(indices_naive,
-                                  indices_ivfflat,
-                                  distances_naive,
-                                  distances_ivfflat,
-                                  ps.num_queries,
-                                  ps.k,
-                                  0.001,
-                                  min_recall));
-    }
-  }
-
-  void SetUp() override
-  {
-    database.resize(ps.num_db_vecs * ps.dim, stream_);
-    search_queries.resize(ps.num_queries * ps.dim, stream_);
-
-    raft::random::RngState r(1234ULL);
-    if constexpr (std::is_same<DataT, float>{}) {
-      raft::random::uniform(
-        handle_, r, database.data(), ps.num_db_vecs * ps.dim, DataT(0.1), DataT(2.0));
-      raft::random::uniform(
-        handle_, r, search_queries.data(), ps.num_queries * ps.dim, DataT(0.1), DataT(2.0));
-    } else {
-      raft::random::uniformInt(
-        handle_, r, database.data(), ps.num_db_vecs * ps.dim, DataT(1), DataT(20));
-      raft::random::uniformInt(
-        handle_, r, search_queries.data(), ps.num_queries * ps.dim, DataT(1), DataT(20));
-    }
-    resource::sync_stream(handle_);
-  }
-
-  void TearDown() override
-  {
-    resource::sync_stream(handle_);
-    database.resize(0, stream_);
-    search_queries.resize(0, stream_);
-  }
-
- private:
-  raft::resources handle_;
-  rmm::cuda_stream_view stream_;
-  AnnIvfFlatInputs<IdxT> ps;
-  rmm::device_uvector<DataT> database;
-  rmm::device_uvector<DataT> search_queries;
-};
-
-const std::vector<AnnIvfFlatInputs<int64_t>> inputs = {
-  // test various dims (aligned and not aligned to vector sizes)
-  {1000, 10000, 1, 16, 40, 1024, cuvs::distance::DistanceType::L2Expanded, true},
-  {1000, 10000, 2, 16, 40, 1024, cuvs::distance::DistanceType::L2Expanded, false},
-  {1000, 10000, 3, 16, 40, 1024, cuvs::distance::DistanceType::L2Expanded, true},
-  {1000, 10000, 4, 16, 40, 1024, cuvs::distance::DistanceType::L2Expanded, false},
-  {1000, 10000, 5, 16, 40, 1024, cuvs::distance::DistanceType::InnerProduct, false},
-  {1000, 10000, 8, 16, 40, 1024, cuvs::distance::DistanceType::InnerProduct, true},
-  {1000, 10000, 5, 16, 40, 1024, cuvs::distance::DistanceType::L2SqrtExpanded, false},
-  {1000, 10000, 8, 16, 40, 1024, cuvs::distance::DistanceType::L2SqrtExpanded, true},
-
-  // test dims that do not fit into kernel shared memory limits
-  {1000, 10000, 2048, 16, 40, 1024, cuvs::distance::DistanceType::L2Expanded, false},
-  {1000, 10000, 2049, 16, 40, 1024, cuvs::distance::DistanceType::L2Expanded, false},
-  {1000, 10000, 2050, 16, 40, 1024, cuvs::distance::DistanceType::InnerProduct, false},
-  {1000, 10000, 2051, 16, 40, 1024, cuvs::distance::DistanceType::InnerProduct, true},
-  {1000, 10000, 2052, 16, 40, 1024, cuvs::distance::DistanceType::InnerProduct, false},
-  {1000, 10000, 2053, 16, 40, 1024, cuvs::distance::DistanceType::L2Expanded, true},
-  {1000, 10000, 2056, 16, 40, 1024, cuvs::distance::DistanceType::L2Expanded, true},
-
-  // various random combinations
-  {1000, 10000, 16, 10, 40, 1024, cuvs::distance::DistanceType::L2Expanded, false},
-  {1000, 10000, 16, 10, 50, 1024, cuvs::distance::DistanceType::L2Expanded, false},
-  {1000, 10000, 16, 10, 70, 1024, cuvs::distance::DistanceType::L2Expanded, false},
-  {100, 10000, 16, 10, 20, 512, cuvs::distance::DistanceType::L2Expanded, false},
-  {20, 100000, 16, 10, 20, 1024, cuvs::distance::DistanceType::L2Expanded, true},
-  {1000, 100000, 16, 10, 20, 1024, cuvs::distance::DistanceType::L2Expanded, true},
-  {10000, 131072, 8, 10, 20, 1024, cuvs::distance::DistanceType::L2Expanded, false},
-
-  {1000, 10000, 16, 10, 40, 1024, cuvs::distance::DistanceType::InnerProduct, true},
-  {1000, 10000, 16, 10, 50, 1024, cuvs::distance::DistanceType::InnerProduct, true},
-  {1000, 10000, 16, 10, 70, 1024, cuvs::distance::DistanceType::InnerProduct, false},
-  {100, 10000, 16, 10, 20, 512, cuvs::distance::DistanceType::InnerProduct, true},
-  {20, 100000, 16, 10, 20, 1024, cuvs::distance::DistanceType::InnerProduct, true},
-  {1000, 100000, 16, 10, 20, 1024, cuvs::distance::DistanceType::InnerProduct, false},
-  {10000, 131072, 8, 10, 50, 1024, cuvs::distance::DistanceType::InnerProduct, true},
-
-  {1000, 10000, 4096, 20, 50, 1024, cuvs::distance::DistanceType::InnerProduct, false},
-
-  // test splitting the big query batches  (> max gridDim.y) into smaller batches
-  {100000, 1024, 32, 10, 64, 64, cuvs::distance::DistanceType::InnerProduct, false},
-  {1000000, 1024, 32, 10, 256, 256, cuvs::distance::DistanceType::InnerProduct, false},
-  {98306, 1024, 32, 10, 64, 64, cuvs::distance::DistanceType::InnerProduct, true},
-
-  // test radix_sort for getting the cluster selection
-  {1000,
-   10000,
-   16,
-   10,
-   raft::matrix::detail::select::warpsort::kMaxCapacity * 2,
-   raft::matrix::detail::select::warpsort::kMaxCapacity * 4,
-   cuvs::distance::DistanceType::L2Expanded,
-   false},
-  {1000,
-   10000,
-   16,
-   10,
-   raft::matrix::detail::select::warpsort::kMaxCapacity * 4,
-   raft::matrix::detail::select::warpsort::kMaxCapacity * 4,
-   cuvs::distance::DistanceType::InnerProduct,
-   false},
-
-  // The following two test cases should show very similar recall.
-  // num_queries, num_db_vecs, dim, k, nprobe, nlist, metric, adaptive_centers
-  {20000, 8712, 3, 10, 51, 66, cuvs::distance::DistanceType::L2Expanded, false},
-  {100000, 8712, 3, 10, 51, 66, cuvs::distance::DistanceType::L2Expanded, false}};
-
-}  // namespace cuvs::neighbors::ivf_flat
diff --git a/cpp/test/neighbors/ann_ivf_flat/test_filter_float_int64_t.cu b/cpp/test/neighbors/ann_ivf_flat/test_filter_float_int64_t.cu
deleted file mode 100644
index b0f3f2443..000000000
--- a/cpp/test/neighbors/ann_ivf_flat/test_filter_float_int64_t.cu
+++ /dev/null
@@ -1,29 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <gtest/gtest.h>
-
-#undef RAFT_EXPLICIT_INSTANTIATE_ONLY  // Enable instantiation of search with filter
-#include "../ann_ivf_flat.cuh"
-
-namespace cuvs::neighbors::ivf_flat {
-
-typedef AnnIVFFlatTest<float, float, std::int64_t> AnnIVFFlatFilterTestF;
-TEST_P(AnnIVFFlatFilterTestF, AnnIVFFlatFilter) { this->testFilter(); }
-
-INSTANTIATE_TEST_CASE_P(AnnIVFFlatTest, AnnIVFFlatFilterTestF, ::testing::ValuesIn(inputs));
-
-}  // namespace cuvs::neighbors::ivf_flat
diff --git a/cpp/test/neighbors/ann_ivf_flat/test_float_int64_t.cu b/cpp/test/neighbors/ann_ivf_flat/test_float_int64_t.cu
deleted file mode 100644
index f4cc99da0..000000000
--- a/cpp/test/neighbors/ann_ivf_flat/test_float_int64_t.cu
+++ /dev/null
@@ -1,32 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <gtest/gtest.h>
-
-#include "../ann_ivf_flat.cuh"
-
-namespace cuvs::neighbors::ivf_flat {
-
-typedef AnnIVFFlatTest<float, float, std::int64_t> AnnIVFFlatTestF;
-TEST_P(AnnIVFFlatTestF, AnnIVFFlat)
-{
-  this->testIVFFlat();
-  this->testPacker();
-}
-
-INSTANTIATE_TEST_CASE_P(AnnIVFFlatTest, AnnIVFFlatTestF, ::testing::ValuesIn(inputs));
-
-}  // namespace cuvs::neighbors::ivf_flat
diff --git a/cpp/test/neighbors/ann_ivf_flat/test_int8_t_int64_t.cu b/cpp/test/neighbors/ann_ivf_flat/test_int8_t_int64_t.cu
deleted file mode 100644
index 445b30eaa..000000000
--- a/cpp/test/neighbors/ann_ivf_flat/test_int8_t_int64_t.cu
+++ /dev/null
@@ -1,28 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <gtest/gtest.h>
-
-#include "../ann_ivf_flat.cuh"
-
-namespace cuvs::neighbors::ivf_flat {
-
-typedef AnnIVFFlatTest<float, int8_t, std::int64_t> AnnIVFFlatTestF_int8;
-TEST_P(AnnIVFFlatTestF_int8, AnnIVFFlat) { this->testIVFFlat(); }
-
-INSTANTIATE_TEST_CASE_P(AnnIVFFlatTest, AnnIVFFlatTestF_int8, ::testing::ValuesIn(inputs));
-
-}  // namespace cuvs::neighbors::ivf_flat
diff --git a/cpp/test/neighbors/ann_ivf_flat/test_uint8_t_int64_t.cu b/cpp/test/neighbors/ann_ivf_flat/test_uint8_t_int64_t.cu
deleted file mode 100644
index a97a831bc..000000000
--- a/cpp/test/neighbors/ann_ivf_flat/test_uint8_t_int64_t.cu
+++ /dev/null
@@ -1,28 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <gtest/gtest.h>
-
-#include "../ann_ivf_flat.cuh"
-
-namespace cuvs::neighbors::ivf_flat {
-
-typedef AnnIVFFlatTest<float, uint8_t, std::int64_t> AnnIVFFlatTestF_uint8;
-TEST_P(AnnIVFFlatTestF_uint8, AnnIVFFlat) { this->testIVFFlat(); }
-
-INSTANTIATE_TEST_CASE_P(AnnIVFFlatTest, AnnIVFFlatTestF_uint8, ::testing::ValuesIn(inputs));
-
-}  // namespace cuvs::neighbors::ivf_flat
diff --git a/cpp/test/neighbors/ann_ivf_pq.cuh b/cpp/test/neighbors/ann_ivf_pq.cuh
deleted file mode 100644
index 51adf4dc9..000000000
--- a/cpp/test/neighbors/ann_ivf_pq.cuh
+++ /dev/null
@@ -1,1095 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#include "../test_utils.cuh"
-#include "ann_utils.cuh"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cuvs_internal/neighbors/naive_knn.cuh>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/ivf_pq.cuh>
-#include <cuvs/neighbors/ivf_pq_helpers.cuh>
-#include <cuvs/neighbors/ivf_pq_serialize.cuh>
-#include <cuvs/neighbors/sample_filter.cuh>
-#include <raft/core/logger.hpp>
-#include <raft/linalg/map.cuh>
-#include <raft/linalg/map_reduce.cuh>
-#include <raft/matrix/gather.cuh>
-#include <raft/random/rng.cuh>
-
-#include <rmm/cuda_stream_view.hpp>
-#include <rmm/device_buffer.hpp>
-#include <rmm/device_vector.hpp>
-#include <rmm/mr/device/managed_memory_resource.hpp>
-
-#include <gtest/gtest.h>
-
-#include <cub/cub.cuh>
-#include <thrust/sequence.h>
-
-#include <algorithm>
-#include <cstddef>
-#include <iostream>
-#include <optional>
-#include <vector>
-
-namespace cuvs::neighbors::ivf_pq {
-
-struct test_ivf_sample_filter {
-  static constexpr unsigned offset = 1500;
-};
-
-struct ivf_pq_inputs {
-  uint32_t num_db_vecs             = 4096;
-  uint32_t num_queries             = 1024;
-  uint32_t dim                     = 64;
-  uint32_t k                       = 32;
-  std::optional<double> min_recall = std::nullopt;
-
-  ivf_pq::index_params index_params;
-  ivf_pq::search_params search_params;
-
-  // Set some default parameters for tests
-  ivf_pq_inputs()
-  {
-    index_params.n_lists                  = max(32u, min(1024u, num_db_vecs / 128u));
-    index_params.kmeans_trainset_fraction = 1.0;
-  }
-};
-
-inline auto operator<<(std::ostream& os, const ivf_pq::codebook_gen& p) -> std::ostream&
-{
-  switch (p) {
-    case ivf_pq::codebook_gen::PER_CLUSTER: os << "codebook_gen::PER_CLUSTER"; break;
-    case ivf_pq::codebook_gen::PER_SUBSPACE: os << "codebook_gen::PER_SUBSPACE"; break;
-    default: RAFT_FAIL("unreachable code");
-  }
-  return os;
-}
-
-inline auto operator<<(std::ostream& os, const ivf_pq_inputs& p) -> std::ostream&
-{
-  ivf_pq_inputs dflt;
-  bool need_comma = false;
-#define PRINT_DIFF_V(spec, val)       \
-  do {                                \
-    if (dflt spec != p spec) {        \
-      if (need_comma) { os << ", "; } \
-      os << #spec << " = " << val;    \
-      need_comma = true;              \
-    }                                 \
-  } while (0)
-#define PRINT_DIFF(spec) PRINT_DIFF_V(spec, p spec)
-
-  os << "ivf_pq_inputs {";
-  PRINT_DIFF(.num_db_vecs);
-  PRINT_DIFF(.num_queries);
-  PRINT_DIFF(.dim);
-  PRINT_DIFF(.k);
-  PRINT_DIFF_V(.min_recall, p.min_recall.value_or(0));
-  PRINT_DIFF_V(.index_params.metric, print_metric{p.index_params.metric});
-  PRINT_DIFF(.index_params.metric_arg);
-  PRINT_DIFF(.index_params.add_data_on_build);
-  PRINT_DIFF(.index_params.n_lists);
-  PRINT_DIFF(.index_params.kmeans_n_iters);
-  PRINT_DIFF(.index_params.kmeans_trainset_fraction);
-  PRINT_DIFF(.index_params.pq_bits);
-  PRINT_DIFF(.index_params.pq_dim);
-  PRINT_DIFF(.index_params.codebook_kind);
-  PRINT_DIFF(.index_params.force_random_rotation);
-  PRINT_DIFF(.search_params.n_probes);
-  PRINT_DIFF_V(.search_params.lut_dtype, print_dtype{p.search_params.lut_dtype});
-  PRINT_DIFF_V(.search_params.internal_distance_dtype,
-               print_dtype{p.search_params.internal_distance_dtype});
-  os << "}";
-  return os;
-}
-
-template <typename T>
-void compare_vectors_l2(
-  const raft::resources& res, T a, T b, uint32_t label, double compression_ratio, double eps)
-{
-  auto n_rows = a.extent(0);
-  auto dim    = a.extent(1);
-  rmm::mr::managed_memory_resource managed_memory;
-  auto dist = make_device_mdarray<double>(res, &managed_memory, make_extents<uint32_t>(n_rows));
-  linalg::map_offset(res, dist.view(), [a, b, dim] __device__(uint32_t i) {
-    spatial::knn::detail::utils::mapping<float> f{};
-    double d = 0.0f;
-    for (uint32_t j = 0; j < dim; j++) {
-      double t = f(a(i, j)) - f(b(i, j));
-      d += t * t;
-    }
-    return sqrt(d / double(dim));
-  });
-  resource::sync_stream(res);
-  for (uint32_t i = 0; i < n_rows; i++) {
-    double d = dist(i);
-    // The theoretical estimate of the error is hard to come up with,
-    // the estimate below is based on experimentation + curse of dimensionality
-    ASSERT_LE(d, 1.2 * eps * std::pow(2.0, compression_ratio))
-      << " (label = " << label << ", ix = " << i << ", eps = " << eps << ")";
-  }
-}
-
-template <typename IdxT>
-auto min_output_size(const raft::resources& handle,
-                     const ivf_pq::index<IdxT>& index,
-                     uint32_t n_probes) -> IdxT
-{
-  auto acc_sizes        = index.accum_sorted_sizes();
-  uint32_t last_nonzero = index.n_lists();
-  while (last_nonzero > 0 && acc_sizes(last_nonzero - 1) == acc_sizes(last_nonzero)) {
-    last_nonzero--;
-  }
-  return acc_sizes(last_nonzero) - acc_sizes(last_nonzero - std::min(last_nonzero, n_probes));
-}
-
-template <typename EvalT, typename DataT, typename IdxT>
-class ivf_pq_test : public ::testing::TestWithParam<ivf_pq_inputs> {
- public:
-  ivf_pq_test()
-    : stream_(resource::get_cuda_stream(handle_)),
-      ps(::testing::TestWithParam<ivf_pq_inputs>::GetParam()),
-      database(0, stream_),
-      search_queries(0, stream_)
-  {
-  }
-
-  void gen_data()
-  {
-    database.resize(size_t{ps.num_db_vecs} * size_t{ps.dim}, stream_);
-    search_queries.resize(size_t{ps.num_queries} * size_t{ps.dim}, stream_);
-
-    raft::random::RngState r(1234ULL);
-    if constexpr (std::is_same<DataT, float>{}) {
-      raft::random::uniform(
-        handle_, r, database.data(), ps.num_db_vecs * ps.dim, DataT(0.1), DataT(2.0));
-      raft::random::uniform(
-        handle_, r, search_queries.data(), ps.num_queries * ps.dim, DataT(0.1), DataT(2.0));
-    } else {
-      raft::random::uniformInt(
-        handle_, r, database.data(), ps.num_db_vecs * ps.dim, DataT(1), DataT(20));
-      raft::random::uniformInt(
-        handle_, r, search_queries.data(), ps.num_queries * ps.dim, DataT(1), DataT(20));
-    }
-    resource::sync_stream(handle_);
-  }
-
-  void calc_ref()
-  {
-    size_t queries_size = size_t{ps.num_queries} * size_t{ps.k};
-    rmm::device_uvector<EvalT> distances_naive_dev(queries_size, stream_);
-    rmm::device_uvector<IdxT> indices_naive_dev(queries_size, stream_);
-    naive_knn<EvalT, DataT, IdxT>(handle_,
-                                  distances_naive_dev.data(),
-                                  indices_naive_dev.data(),
-                                  search_queries.data(),
-                                  database.data(),
-                                  ps.num_queries,
-                                  ps.num_db_vecs,
-                                  ps.dim,
-                                  ps.k,
-                                  ps.index_params.metric);
-    distances_ref.resize(queries_size);
-    update_host(distances_ref.data(), distances_naive_dev.data(), queries_size, stream_);
-    indices_ref.resize(queries_size);
-    update_host(indices_ref.data(), indices_naive_dev.data(), queries_size, stream_);
-    resource::sync_stream(handle_);
-  }
-
-  auto build_only()
-  {
-    auto ipams              = ps.index_params;
-    ipams.add_data_on_build = true;
-
-    auto index_view =
-      raft::make_device_matrix_view<DataT, IdxT>(database.data(), ps.num_db_vecs, ps.dim);
-    return ivf_pq::build<DataT, IdxT>(handle_, ipams, index_view);
-  }
-
-  auto build_2_extends()
-  {
-    auto db_indices = make_device_vector<IdxT>(handle_, ps.num_db_vecs);
-    linalg::map_offset(handle_, db_indices.view(), identity_op{});
-    resource::sync_stream(handle_);
-    auto size_1 = IdxT(ps.num_db_vecs) / 2;
-    auto size_2 = IdxT(ps.num_db_vecs) - size_1;
-    auto vecs_1 = database.data();
-    auto vecs_2 = database.data() + size_t(size_1) * size_t(ps.dim);
-    auto inds_1 = db_indices.data_handle();
-    auto inds_2 = db_indices.data_handle() + size_t(size_1);
-
-    auto ipams              = ps.index_params;
-    ipams.add_data_on_build = false;
-
-    auto database_view =
-      raft::make_device_matrix_view<DataT, IdxT>(database.data(), ps.num_db_vecs, ps.dim);
-    auto idx = ivf_pq::build<DataT, IdxT>(handle_, ipams, database_view);
-
-    auto vecs_2_view = raft::make_device_matrix_view<DataT, IdxT>(vecs_2, size_2, ps.dim);
-    auto inds_2_view = raft::make_device_vector_view<IdxT, IdxT>(inds_2, size_2);
-    ivf_pq::extend<DataT, IdxT>(handle_, vecs_2_view, inds_2_view, &idx);
-
-    auto vecs_1_view =
-      raft::make_device_matrix_view<DataT, IdxT, row_major>(vecs_1, size_1, ps.dim);
-    auto inds_1_view = raft::make_device_vector_view<const IdxT, IdxT>(inds_1, size_1);
-    ivf_pq::extend<DataT, IdxT>(handle_, vecs_1_view, inds_1_view, &idx);
-    return idx;
-  }
-
-  auto build_serialize()
-  {
-    ivf_pq::serialize<IdxT>(handle_, "ivf_pq_index", build_only());
-    return ivf_pq::deserialize<IdxT>(handle_, "ivf_pq_index");
-  }
-
-  void check_reconstruction(const index<IdxT>& index,
-                            double compression_ratio,
-                            uint32_t label,
-                            uint32_t n_take,
-                            uint32_t n_skip)
-  {
-    auto& rec_list = index.lists()[label];
-    auto dim       = index.dim();
-    n_take         = std::min<uint32_t>(n_take, rec_list->size.load());
-    n_skip         = std::min<uint32_t>(n_skip, rec_list->size.load() - n_take);
-
-    if (n_take == 0) { return; }
-
-    auto rec_data  = make_device_matrix<DataT>(handle_, n_take, dim);
-    auto orig_data = make_device_matrix<DataT>(handle_, n_take, dim);
-
-    ivf_pq::helpers::reconstruct_list_data(handle_, index, rec_data.view(), label, n_skip);
-
-    matrix::gather(database.data(),
-                   IdxT{dim},
-                   IdxT{n_take},
-                   rec_list->indices.data_handle() + n_skip,
-                   IdxT{n_take},
-                   orig_data.data_handle(),
-                   stream_);
-
-    compare_vectors_l2(handle_, rec_data.view(), orig_data.view(), label, compression_ratio, 0.06);
-  }
-
-  void check_reconstruct_extend(index<IdxT>* index, double compression_ratio, uint32_t label)
-  {
-    // NB: this is not reference, the list is retained; the index will have to create a new list on
-    // `erase_list` op.
-    auto old_list = index->lists()[label];
-    auto n_rows   = old_list->size.load();
-    if (n_rows == 0) { return; }
-
-    auto vectors_1 = make_device_matrix<EvalT>(handle_, n_rows, index->dim());
-    auto indices   = make_device_vector<IdxT>(handle_, n_rows);
-    copy(indices.data_handle(), old_list->indices.data_handle(), n_rows, stream_);
-
-    ivf_pq::helpers::reconstruct_list_data(handle_, *index, vectors_1.view(), label, 0);
-    ivf_pq::helpers::erase_list(handle_, index, label);
-    // NB: passing the type parameter because const->non-const implicit conversion of the mdspans
-    // breaks type inference
-    ivf_pq::helpers::extend_list<EvalT, IdxT>(
-      handle_, index, vectors_1.view(), indices.view(), label);
-
-    auto& new_list = index->lists()[label];
-    ASSERT_NE(old_list.get(), new_list.get())
-      << "The old list should have been shared and retained after ivf_pq index has erased the "
-         "corresponding cluster.";
-
-    auto vectors_2 = make_device_matrix<EvalT>(handle_, n_rows, index->dim());
-    ivf_pq::helpers::reconstruct_list_data(handle_, *index, vectors_2.view(), label, 0);
-    // The code search is unstable, and there's high chance of repeating values of the lvl-2 codes.
-    // Hence, encoding-decoding chain often leads to altering both the PQ codes and the
-    // reconstructed data.
-    compare_vectors_l2(
-      handle_, vectors_1.view(), vectors_2.view(), label, compression_ratio, 0.04);  // 0.025);
-  }
-
-  void check_packing(index<IdxT>* index, uint32_t label)
-  {
-    auto old_list = index->lists()[label];
-    auto n_rows   = old_list->size.load();
-
-    if (n_rows == 0) { return; }
-
-    auto codes   = make_device_matrix<uint8_t>(handle_, n_rows, index->pq_dim());
-    auto indices = make_device_vector<IdxT>(handle_, n_rows);
-    copy(indices.data_handle(), old_list->indices.data_handle(), n_rows, stream_);
-
-    ivf_pq::helpers::unpack_list_data(handle_, *index, codes.view(), label, 0);
-    ivf_pq::helpers::erase_list(handle_, index, label);
-    ivf_pq::helpers::extend_list_with_codes<IdxT>(
-      handle_, index, codes.view(), indices.view(), label);
-
-    auto& new_list = index->lists()[label];
-    ASSERT_NE(old_list.get(), new_list.get())
-      << "The old list should have been shared and retained after ivf_pq index has erased the "
-         "corresponding cluster.";
-    auto list_data_size = (n_rows / ivf_pq::kIndexGroupSize) * new_list->data.extent(1) *
-                          new_list->data.extent(2) * new_list->data.extent(3);
-
-    ASSERT_TRUE(old_list->data.size() >= list_data_size);
-    ASSERT_TRUE(new_list->data.size() >= list_data_size);
-    ASSERT_TRUE(devArrMatch(old_list->data.data_handle(),
-                            new_list->data.data_handle(),
-                            list_data_size,
-                            Compare<uint8_t>{}));
-
-    // Pack a few vectors back to the list.
-    int row_offset = 9;
-    int n_vec      = 3;
-    ASSERT_TRUE(row_offset + n_vec < n_rows);
-    size_t offset      = row_offset * index->pq_dim();
-    auto codes_to_pack = make_device_matrix_view<const uint8_t, uint32_t>(
-      codes.data_handle() + offset, n_vec, index->pq_dim());
-    ivf_pq::helpers::pack_list_data(handle_, index, codes_to_pack, label, row_offset);
-    ASSERT_TRUE(devArrMatch(old_list->data.data_handle(),
-                            new_list->data.data_handle(),
-                            list_data_size,
-                            Compare<uint8_t>{}));
-
-    // Another test with the API that take list_data directly
-    auto list_data  = index->lists()[label]->data.view();
-    uint32_t n_take = 4;
-    ASSERT_TRUE(row_offset + n_take < n_rows);
-    auto codes2 = raft::make_device_matrix<uint8_t>(handle_, n_take, index->pq_dim());
-    ivf_pq::helpers::codepacker::unpack(
-      handle_, list_data, index->pq_bits(), row_offset, codes2.view());
-
-    // Write it back
-    ivf_pq::helpers::codepacker::pack(
-      handle_, make_const_mdspan(codes2.view()), index->pq_bits(), row_offset, list_data);
-    ASSERT_TRUE(devArrMatch(old_list->data.data_handle(),
-                            new_list->data.data_handle(),
-                            list_data_size,
-                            Compare<uint8_t>{}));
-  }
-  void check_packing_contiguous(index<IdxT>* index, uint32_t label)
-  {
-    auto old_list = index->lists()[label];
-    auto n_rows   = old_list->size.load();
-
-    if (n_rows == 0) { return; }
-
-    auto codes   = make_device_matrix<uint8_t>(handle_, n_rows, index->pq_dim());
-    auto indices = make_device_vector<IdxT>(handle_, n_rows);
-    copy(indices.data_handle(), old_list->indices.data_handle(), n_rows, stream_);
-
-    uint32_t code_size = ceildiv<uint32_t>(index->pq_dim() * index->pq_bits(), 8);
-
-    auto codes_compressed = make_device_matrix<uint8_t>(handle_, n_rows, code_size);
-
-    ivf_pq::helpers::unpack_contiguous_list_data(
-      handle_, *index, codes_compressed.data_handle(), n_rows, label, 0);
-    ivf_pq::helpers::erase_list(handle_, index, label);
-    ivf_pq::detail::extend_list_prepare(handle_, index, make_const_mdspan(indices.view()), label);
-    ivf_pq::helpers::pack_contiguous_list_data<IdxT>(
-      handle_, index, codes_compressed.data_handle(), n_rows, label, 0);
-    ivf_pq::helpers::recompute_internal_state(handle_, index);
-
-    auto& new_list = index->lists()[label];
-    ASSERT_NE(old_list.get(), new_list.get())
-      << "The old list should have been shared and retained after ivf_pq index has erased the "
-         "corresponding cluster.";
-    auto list_data_size = (n_rows / ivf_pq::kIndexGroupSize) * new_list->data.extent(1) *
-                          new_list->data.extent(2) * new_list->data.extent(3);
-
-    ASSERT_TRUE(old_list->data.size() >= list_data_size);
-    ASSERT_TRUE(new_list->data.size() >= list_data_size);
-    ASSERT_TRUE(devArrMatch(old_list->data.data_handle(),
-                            new_list->data.data_handle(),
-                            list_data_size,
-                            Compare<uint8_t>{}));
-
-    // Pack a few vectors back to the list.
-    uint32_t row_offset = 9;
-    uint32_t n_vec      = 3;
-    ASSERT_TRUE(row_offset + n_vec < n_rows);
-    size_t offset      = row_offset * code_size;
-    auto codes_to_pack = make_device_matrix_view<uint8_t, uint32_t>(
-      codes_compressed.data_handle() + offset, n_vec, index->pq_dim());
-    ivf_pq::helpers::pack_contiguous_list_data(
-      handle_, index, codes_to_pack.data_handle(), n_vec, label, row_offset);
-    ASSERT_TRUE(devArrMatch(old_list->data.data_handle(),
-                            new_list->data.data_handle(),
-                            list_data_size,
-                            Compare<uint8_t>{}));
-
-    // // Another test with the API that take list_data directly
-    auto list_data  = index->lists()[label]->data.view();
-    uint32_t n_take = 4;
-    ASSERT_TRUE(row_offset + n_take < n_rows);
-    auto codes2 = raft::make_device_matrix<uint8_t>(handle_, n_take, code_size);
-    ivf_pq::helpers::codepacker::unpack_contiguous(handle_,
-                                                   list_data,
-                                                   index->pq_bits(),
-                                                   row_offset,
-                                                   n_take,
-                                                   index->pq_dim(),
-                                                   codes2.data_handle());
-
-    // Write it back
-    ivf_pq::helpers::codepacker::pack_contiguous(handle_,
-                                                 codes2.data_handle(),
-                                                 n_vec,
-                                                 index->pq_dim(),
-                                                 index->pq_bits(),
-                                                 row_offset,
-                                                 list_data);
-    ASSERT_TRUE(devArrMatch(old_list->data.data_handle(),
-                            new_list->data.data_handle(),
-                            list_data_size,
-                            Compare<uint8_t>{}));
-  }
-
-  template <typename BuildIndex>
-  void run(BuildIndex build_index)
-  {
-    index<IdxT> index = build_index();
-
-    double compression_ratio =
-      static_cast<double>(ps.dim * 8) / static_cast<double>(index.pq_dim() * index.pq_bits());
-
-    for (uint32_t label = 0; label < index.n_lists(); label++) {
-      switch (label % 3) {
-        case 0: {
-          // Reconstruct and re-write vectors for one label
-          check_reconstruct_extend(&index, compression_ratio, label);
-        } break;
-        case 1: {
-          // Dump and re-write codes for one label
-          check_packing(&index, label);
-          check_packing_contiguous(&index, label);
-        } break;
-        default: {
-          // check a small subset of data in a randomly chosen cluster to see if the data
-          // reconstruction works well.
-          check_reconstruction(index, compression_ratio, label, 100, 7);
-        }
-      }
-    }
-
-    size_t queries_size = ps.num_queries * ps.k;
-    std::vector<IdxT> indices_ivf_pq(queries_size);
-    std::vector<EvalT> distances_ivf_pq(queries_size);
-
-    rmm::device_uvector<EvalT> distances_ivf_pq_dev(queries_size, stream_);
-    rmm::device_uvector<IdxT> indices_ivf_pq_dev(queries_size, stream_);
-
-    auto query_view =
-      raft::make_device_matrix_view<DataT, uint32_t>(search_queries.data(), ps.num_queries, ps.dim);
-    auto inds_view = raft::make_device_matrix_view<IdxT, uint32_t>(
-      indices_ivf_pq_dev.data(), ps.num_queries, ps.k);
-    auto dists_view = raft::make_device_matrix_view<EvalT, uint32_t>(
-      distances_ivf_pq_dev.data(), ps.num_queries, ps.k);
-
-    ivf_pq::search<DataT, IdxT>(
-      handle_, ps.search_params, index, query_view, inds_view, dists_view);
-
-    update_host(distances_ivf_pq.data(), distances_ivf_pq_dev.data(), queries_size, stream_);
-    update_host(indices_ivf_pq.data(), indices_ivf_pq_dev.data(), queries_size, stream_);
-    resource::sync_stream(handle_);
-
-    // A very conservative lower bound on recall
-    double min_recall =
-      static_cast<double>(ps.search_params.n_probes) / static_cast<double>(ps.index_params.n_lists);
-    // Using a heuristic to lower the required recall due to code-packing errors
-    min_recall =
-      std::min(std::erfc(0.05 * compression_ratio / std::max(min_recall, 0.5)), min_recall);
-    // Use explicit per-test min recall value if provided.
-    min_recall = ps.min_recall.value_or(min_recall);
-
-    ASSERT_TRUE(eval_neighbours(indices_ref,
-                                indices_ivf_pq,
-                                distances_ref,
-                                distances_ivf_pq,
-                                ps.num_queries,
-                                ps.k,
-                                0.0001 * compression_ratio,
-                                min_recall))
-      << ps;
-
-    // Test a few extra invariants
-    IdxT min_results = min_output_size(handle_, index, ps.search_params.n_probes);
-    IdxT max_oob     = ps.k <= min_results ? 0 : ps.k - min_results;
-    IdxT found_oob   = 0;
-    for (uint32_t query_ix = 0; query_ix < ps.num_queries; query_ix++) {
-      for (uint32_t k = 0; k < ps.k; k++) {
-        auto flat_i   = query_ix * ps.k + k;
-        auto found_ix = indices_ivf_pq[flat_i];
-        if (found_ix == ivf_pq::kOutOfBoundsRecord<IdxT>) {
-          found_oob++;
-          continue;
-        }
-        ASSERT_NE(found_ix, ivf::kInvalidRecord<IdxT>)
-          << "got an invalid record at query_ix = " << query_ix << ", k = " << k
-          << " (distance = " << distances_ivf_pq[flat_i] << ")";
-        ASSERT_LT(found_ix, ps.num_db_vecs)
-          << "got an impossible index = " << found_ix << " at query_ix = " << query_ix
-          << ", k = " << k << " (distance = " << distances_ivf_pq[flat_i] << ")";
-      }
-    }
-    ASSERT_LE(found_oob, max_oob)
-      << "got too many records out-of-bounds (see ivf_pq::kOutOfBoundsRecord<IdxT>).";
-    if (found_oob > 0) {
-      RAFT_LOG_WARN(
-        "Got %zu results out-of-bounds because of large top-k (%zu) and small n_probes (%u) and "
-        "small DB size/n_lists ratio (%zu / %u)",
-        size_t(found_oob),
-        size_t(ps.k),
-        ps.search_params.n_probes,
-        size_t(ps.num_db_vecs),
-        ps.index_params.n_lists);
-    }
-  }
-
-  void SetUp() override  // NOLINT
-  {
-    gen_data();
-    calc_ref();
-  }
-
-  void TearDown() override  // NOLINT
-  {
-    cudaGetLastError();
-    resource::sync_stream(handle_);
-    database.resize(0, stream_);
-    search_queries.resize(0, stream_);
-  }
-
- private:
-  raft::resources handle_;
-  rmm::cuda_stream_view stream_;
-  ivf_pq_inputs ps;                           // NOLINT
-  rmm::device_uvector<DataT> database;        // NOLINT
-  rmm::device_uvector<DataT> search_queries;  // NOLINT
-  std::vector<IdxT> indices_ref;              // NOLINT
-  std::vector<EvalT> distances_ref;           // NOLINT
-};
-
-template <typename EvalT, typename DataT, typename IdxT>
-class ivf_pq_filter_test : public ::testing::TestWithParam<ivf_pq_inputs> {
- public:
-  ivf_pq_filter_test()
-    : stream_(resource::get_cuda_stream(handle_)),
-      ps(::testing::TestWithParam<ivf_pq_inputs>::GetParam()),
-      database(0, stream_),
-      search_queries(0, stream_)
-  {
-  }
-
-  void gen_data()
-  {
-    database.resize(size_t{ps.num_db_vecs} * size_t{ps.dim}, stream_);
-    search_queries.resize(size_t{ps.num_queries} * size_t{ps.dim}, stream_);
-
-    raft::random::RngState r(1234ULL);
-    if constexpr (std::is_same<DataT, float>{}) {
-      raft::random::uniform(
-        handle_, r, database.data(), ps.num_db_vecs * ps.dim, DataT(0.1), DataT(2.0));
-      raft::random::uniform(
-        handle_, r, search_queries.data(), ps.num_queries * ps.dim, DataT(0.1), DataT(2.0));
-    } else {
-      raft::random::uniformInt(
-        handle_, r, database.data(), ps.num_db_vecs * ps.dim, DataT(1), DataT(20));
-      raft::random::uniformInt(
-        handle_, r, search_queries.data(), ps.num_queries * ps.dim, DataT(1), DataT(20));
-    }
-    resource::sync_stream(handle_);
-  }
-
-  void calc_ref()
-  {
-    size_t queries_size = size_t{ps.num_queries} * size_t{ps.k};
-    rmm::device_uvector<EvalT> distances_naive_dev(queries_size, stream_);
-    rmm::device_uvector<IdxT> indices_naive_dev(queries_size, stream_);
-    naive_knn<EvalT, DataT, IdxT>(handle_,
-                                  distances_naive_dev.data(),
-                                  indices_naive_dev.data(),
-                                  search_queries.data(),
-                                  database.data() + test_ivf_sample_filter::offset * ps.dim,
-                                  ps.num_queries,
-                                  ps.num_db_vecs - test_ivf_sample_filter::offset,
-                                  ps.dim,
-                                  ps.k,
-                                  ps.index_params.metric);
-    raft::linalg::addScalar(indices_naive_dev.data(),
-                            indices_naive_dev.data(),
-                            IdxT(test_ivf_sample_filter::offset),
-                            queries_size,
-                            stream_);
-    distances_ref.resize(queries_size);
-    update_host(distances_ref.data(), distances_naive_dev.data(), queries_size, stream_);
-    indices_ref.resize(queries_size);
-    update_host(indices_ref.data(), indices_naive_dev.data(), queries_size, stream_);
-    resource::sync_stream(handle_);
-  }
-
-  auto build_only()
-  {
-    auto ipams              = ps.index_params;
-    ipams.add_data_on_build = true;
-
-    auto index_view =
-      raft::make_device_matrix_view<DataT, IdxT>(database.data(), ps.num_db_vecs, ps.dim);
-    return ivf_pq::build<DataT, IdxT>(handle_, ipams, index_view);
-  }
-
-  template <typename BuildIndex>
-  void run(BuildIndex build_index)
-  {
-    index<IdxT> index = build_index();
-
-    double compression_ratio =
-      static_cast<double>(ps.dim * 8) / static_cast<double>(index.pq_dim() * index.pq_bits());
-    size_t queries_size = ps.num_queries * ps.k;
-    std::vector<IdxT> indices_ivf_pq(queries_size);
-    std::vector<EvalT> distances_ivf_pq(queries_size);
-
-    rmm::device_uvector<EvalT> distances_ivf_pq_dev(queries_size, stream_);
-    rmm::device_uvector<IdxT> indices_ivf_pq_dev(queries_size, stream_);
-
-    auto query_view =
-      raft::make_device_matrix_view<DataT, uint32_t>(search_queries.data(), ps.num_queries, ps.dim);
-    auto inds_view = raft::make_device_matrix_view<IdxT, uint32_t>(
-      indices_ivf_pq_dev.data(), ps.num_queries, ps.k);
-    auto dists_view = raft::make_device_matrix_view<EvalT, uint32_t>(
-      distances_ivf_pq_dev.data(), ps.num_queries, ps.k);
-
-    // Create Bitset filter
-    auto removed_indices =
-      raft::make_device_vector<IdxT, int64_t>(handle_, test_ivf_sample_filter::offset);
-    thrust::sequence(
-      resource::get_thrust_policy(handle_),
-      thrust::device_pointer_cast(removed_indices.data_handle()),
-      thrust::device_pointer_cast(removed_indices.data_handle() + test_ivf_sample_filter::offset));
-    resource::sync_stream(handle_);
-
-    raft::core::bitset<std::uint32_t, IdxT> removed_indices_bitset(
-      handle_, removed_indices.view(), ps.num_db_vecs);
-    ivf_pq::search_with_filtering<DataT, IdxT>(
-      handle_,
-      ps.search_params,
-      index,
-      query_view,
-      inds_view,
-      dists_view,
-      cuvs::neighbors::filtering::bitset_filter(removed_indices_bitset.view()));
-
-    update_host(distances_ivf_pq.data(), distances_ivf_pq_dev.data(), queries_size, stream_);
-    update_host(indices_ivf_pq.data(), indices_ivf_pq_dev.data(), queries_size, stream_);
-    resource::sync_stream(handle_);
-
-    // A very conservative lower bound on recall
-    double min_recall =
-      static_cast<double>(ps.search_params.n_probes) / static_cast<double>(ps.index_params.n_lists);
-    // Using a heuristic to lower the required recall due to code-packing errors
-    min_recall =
-      std::min(std::erfc(0.05 * compression_ratio / std::max(min_recall, 0.5)), min_recall);
-    // Use explicit per-test min recall value if provided.
-    min_recall = ps.min_recall.value_or(min_recall);
-
-    ASSERT_TRUE(eval_neighbours(indices_ref,
-                                indices_ivf_pq,
-                                distances_ref,
-                                distances_ivf_pq,
-                                ps.num_queries,
-                                ps.k,
-                                0.0001 * compression_ratio,
-                                min_recall))
-      << ps;
-  }
-
-  void SetUp() override  // NOLINT
-  {
-    gen_data();
-    calc_ref();
-  }
-
-  void TearDown() override  // NOLINT
-  {
-    cudaGetLastError();
-    resource::sync_stream(handle_);
-    database.resize(0, stream_);
-    search_queries.resize(0, stream_);
-  }
-
- private:
-  raft::resources handle_;
-  rmm::cuda_stream_view stream_;
-  ivf_pq_inputs ps;                           // NOLINT
-  rmm::device_uvector<DataT> database;        // NOLINT
-  rmm::device_uvector<DataT> search_queries;  // NOLINT
-  std::vector<IdxT> indices_ref;              // NOLINT
-  std::vector<EvalT> distances_ref;           // NOLINT
-};
-
-/* Test cases */
-using test_cases_t = std::vector<ivf_pq_inputs>;
-
-// concatenate parameter sets for different type
-template <typename T>
-auto operator+(const std::vector<T>& a, const std::vector<T>& b) -> std::vector<T>
-{
-  std::vector<T> res = a;
-  res.insert(res.end(), b.begin(), b.end());
-  return res;
-}
-
-inline auto defaults() -> test_cases_t { return {ivf_pq_inputs{}}; }
-
-template <typename B, typename A, typename F>
-auto map(const std::vector<A>& xs, F f) -> std::vector<B>
-{
-  std::vector<B> ys(xs.size());
-  std::transform(xs.begin(), xs.end(), ys.begin(), f);
-  return ys;
-}
-
-inline auto with_dims(const std::vector<uint32_t>& dims) -> test_cases_t
-{
-  return map<ivf_pq_inputs>(dims, [](uint32_t d) {
-    ivf_pq_inputs x;
-    x.dim = d;
-    return x;
-  });
-}
-
-/** These will surely trigger the fastest kernel available. */
-inline auto small_dims() -> test_cases_t { return with_dims({1, 2, 3, 4, 5, 8, 15, 16, 17}); }
-
-inline auto small_dims_per_cluster() -> test_cases_t
-{
-  return map<ivf_pq_inputs>(small_dims(), [](const ivf_pq_inputs& x) {
-    ivf_pq_inputs y(x);
-    y.index_params.codebook_kind = ivf_pq::codebook_gen::PER_CLUSTER;
-    return y;
-  });
-}
-
-inline auto big_dims() -> test_cases_t
-{
-  // with_dims({512, 513, 1023, 1024, 1025, 2048, 2049, 2050, 2053, 6144, 8192, 12288, 16384});
-  auto xs = with_dims({512, 513, 1023, 1024, 1025, 2048, 2049, 2050, 2053, 6144});
-  return map<ivf_pq_inputs>(xs, [](const ivf_pq_inputs& x) {
-    ivf_pq_inputs y(x);
-    uint32_t pq_len       = 2;
-    y.index_params.pq_dim = div_rounding_up_safe(x.dim, pq_len);
-    // This comes from pure experimentation, also the recall depens a lot on pq_len.
-    y.min_recall = 0.48 + 0.028 * std::log2(x.dim);
-    return y;
-  });
-}
-
-/** These will surely trigger no-smem-lut kernel.  */
-inline auto big_dims_moderate_lut() -> test_cases_t
-{
-  return map<ivf_pq_inputs>(big_dims(), [](const ivf_pq_inputs& x) {
-    ivf_pq_inputs y(x);
-    uint32_t pq_len           = 2;
-    y.index_params.pq_dim     = round_up_safe(div_rounding_up_safe(x.dim, pq_len), 4u);
-    y.index_params.pq_bits    = 6;
-    y.search_params.lut_dtype = CUDA_R_16F;
-    y.min_recall              = 0.69;
-    return y;
-  });
-}
-
-/** Some of these should trigger no-basediff kernel.  */
-inline auto big_dims_small_lut() -> test_cases_t
-{
-  return map<ivf_pq_inputs>(big_dims(), [](const ivf_pq_inputs& x) {
-    ivf_pq_inputs y(x);
-    uint32_t pq_len           = 8;
-    y.index_params.pq_dim     = round_up_safe(div_rounding_up_safe(x.dim, pq_len), 4u);
-    y.index_params.pq_bits    = 6;
-    y.search_params.lut_dtype = CUDA_R_8U;
-    y.min_recall              = 0.21;
-    return y;
-  });
-}
-
-/**
- * A minimal set of tests to check various enum-like parameters.
- */
-inline auto enum_variety() -> test_cases_t
-{
-  test_cases_t xs;
-#define ADD_CASE(f)                               \
-  do {                                            \
-    xs.push_back({});                             \
-    ([](ivf_pq_inputs & x) f)(xs[xs.size() - 1]); \
-  } while (0);
-
-  ADD_CASE({
-    x.index_params.codebook_kind = ivf_pq::codebook_gen::PER_CLUSTER;
-    x.min_recall                 = 0.86;
-  });
-  ADD_CASE({
-    x.index_params.codebook_kind = ivf_pq::codebook_gen::PER_SUBSPACE;
-    x.min_recall                 = 0.86;
-  });
-  ADD_CASE({
-    x.index_params.codebook_kind = ivf_pq::codebook_gen::PER_CLUSTER;
-    x.index_params.pq_bits       = 4;
-    x.min_recall                 = 0.79;
-  });
-  ADD_CASE({
-    x.index_params.codebook_kind = ivf_pq::codebook_gen::PER_CLUSTER;
-    x.index_params.pq_bits       = 5;
-    x.min_recall                 = 0.83;
-  });
-
-  ADD_CASE({
-    x.index_params.pq_bits = 6;
-    x.min_recall           = 0.84;
-  });
-  ADD_CASE({
-    x.index_params.pq_bits = 7;
-    x.min_recall           = 0.85;
-  });
-  ADD_CASE({
-    x.index_params.pq_bits = 8;
-    x.min_recall           = 0.86;
-  });
-
-  ADD_CASE({
-    x.index_params.force_random_rotation = true;
-    x.min_recall                         = 0.86;
-  });
-  ADD_CASE({
-    x.index_params.force_random_rotation = false;
-    x.min_recall                         = 0.86;
-  });
-
-  ADD_CASE({
-    x.search_params.lut_dtype = CUDA_R_32F;
-    x.min_recall              = 0.86;
-  });
-  ADD_CASE({
-    x.search_params.lut_dtype = CUDA_R_16F;
-    x.min_recall              = 0.86;
-  });
-  ADD_CASE({
-    x.search_params.lut_dtype = CUDA_R_8U;
-    x.min_recall              = 0.84;
-  });
-
-  ADD_CASE({
-    x.search_params.internal_distance_dtype = CUDA_R_32F;
-    x.min_recall                            = 0.86;
-  });
-  ADD_CASE({
-    x.search_params.internal_distance_dtype = CUDA_R_16F;
-    x.search_params.lut_dtype               = CUDA_R_16F;
-    x.min_recall                            = 0.86;
-  });
-
-  return xs;
-}
-
-inline auto enum_variety_l2() -> test_cases_t
-{
-  return map<ivf_pq_inputs>(enum_variety(), [](const ivf_pq_inputs& x) {
-    ivf_pq_inputs y(x);
-    y.index_params.metric = distance::DistanceType::L2Expanded;
-    return y;
-  });
-}
-
-inline auto enum_variety_ip() -> test_cases_t
-{
-  return map<ivf_pq_inputs>(enum_variety(), [](const ivf_pq_inputs& x) {
-    ivf_pq_inputs y(x);
-    if (y.min_recall.has_value()) {
-      if (y.search_params.lut_dtype == CUDA_R_8U) {
-        // InnerProduct score is signed,
-        // thus we're forced to used signed 8-bit representation,
-        // thus we have one bit less precision
-        y.min_recall = y.min_recall.value() * 0.90;
-      } else {
-        // In other cases it seems to perform a little bit better, still worse than L2
-        y.min_recall = y.min_recall.value() * 0.94;
-      }
-    }
-    y.index_params.metric = distance::DistanceType::InnerProduct;
-    return y;
-  });
-}
-
-inline auto enum_variety_l2sqrt() -> test_cases_t
-{
-  return map<ivf_pq_inputs>(enum_variety(), [](const ivf_pq_inputs& x) {
-    ivf_pq_inputs y(x);
-    y.index_params.metric = distance::DistanceType::L2SqrtExpanded;
-    return y;
-  });
-}
-
-/**
- * Try different number of n_probes, some of which may trigger the non-fused version of the search
- * kernel.
- */
-inline auto var_n_probes() -> test_cases_t
-{
-  ivf_pq_inputs dflt;
-  std::vector<uint32_t> xs;
-  for (auto x = dflt.index_params.n_lists; x >= 1; x /= 2) {
-    xs.push_back(x);
-  }
-  return map<ivf_pq_inputs>(xs, [](uint32_t n_probes) {
-    ivf_pq_inputs x;
-    x.search_params.n_probes = n_probes;
-    return x;
-  });
-}
-
-/**
- * Try different number of nearest neighbours.
- * Values smaller than 32 test if the code behaves well when Capacity (== 32) does not change,
- * but `k <= Capacity` changes.
- *
- * Values between `32 and ivf_pq::detail::kMaxCapacity` test various instantiations of the
- * main kernel (Capacity-templated)
- *
- * Values above ivf_pq::detail::kMaxCapacity should trigger the non-fused version of the kernel
- * (manage_local_topk = false).
- *
- * Also we test here various values that are close-but-not-power-of-two to catch any problems
- * related to rounding/alignment.
- *
- * Note, we cannot control explicitly which instance of the search kernel to choose, hence it's
- * important to try a variety of different values of `k` to make sure all paths are triggered.
- *
- * Set the log level to DEBUG (5) or above to inspect the selected kernel instances.
- */
-inline auto var_k() -> test_cases_t
-{
-  return map<ivf_pq_inputs, uint32_t>(
-    {1, 2, 3, 5, 8, 15, 16, 32, 63, 65, 127, 128, 256, 257, 1023, 2048, 2049}, [](uint32_t k) {
-      ivf_pq_inputs x;
-      x.k = k;
-      // when there's not enough data, try more cluster probes
-      x.search_params.n_probes = max(x.search_params.n_probes, min(x.index_params.n_lists, k));
-      return x;
-    });
-}
-
-/**
- * Cases brought up from downstream projects.
- */
-inline auto special_cases() -> test_cases_t
-{
-  test_cases_t xs;
-
-#define ADD_CASE(f)                               \
-  do {                                            \
-    xs.push_back({});                             \
-    ([](ivf_pq_inputs & x) f)(xs[xs.size() - 1]); \
-  } while (0);
-
-  ADD_CASE({
-    x.num_db_vecs                = 1183514;
-    x.dim                        = 100;
-    x.num_queries                = 10000;
-    x.k                          = 10;
-    x.index_params.codebook_kind = ivf_pq::codebook_gen::PER_SUBSPACE;
-    x.index_params.pq_dim        = 10;
-    x.index_params.pq_bits       = 8;
-    x.index_params.n_lists       = 1024;
-    x.search_params.n_probes     = 50;
-  });
-
-  ADD_CASE({
-    x.num_db_vecs                = 10000;
-    x.dim                        = 16;
-    x.num_queries                = 500;
-    x.k                          = 128;
-    x.index_params.metric        = distance::DistanceType::L2Expanded;
-    x.index_params.codebook_kind = ivf_pq::codebook_gen::PER_SUBSPACE;
-    x.index_params.pq_bits       = 8;
-    x.index_params.n_lists       = 100;
-    x.search_params.n_probes     = 100;
-  });
-
-  ADD_CASE({
-    x.num_db_vecs                = 10000;
-    x.dim                        = 16;
-    x.num_queries                = 500;
-    x.k                          = 129;
-    x.index_params.metric        = distance::DistanceType::L2Expanded;
-    x.index_params.codebook_kind = ivf_pq::codebook_gen::PER_SUBSPACE;
-    x.index_params.pq_bits       = 8;
-    x.index_params.n_lists       = 100;
-    x.search_params.n_probes     = 100;
-  });
-
-  ADD_CASE({
-    x.num_db_vecs                = 4335;
-    x.dim                        = 4;
-    x.num_queries                = 100000;
-    x.k                          = 12;
-    x.index_params.metric        = distance::DistanceType::L2Expanded;
-    x.index_params.codebook_kind = ivf_pq::codebook_gen::PER_SUBSPACE;
-    x.index_params.pq_dim        = 2;
-    x.index_params.pq_bits       = 8;
-    x.index_params.n_lists       = 69;
-    x.search_params.n_probes     = 69;
-  });
-
-  ADD_CASE({
-    x.num_db_vecs                = 4335;
-    x.dim                        = 4;
-    x.num_queries                = 100000;
-    x.k                          = 12;
-    x.index_params.metric        = distance::DistanceType::L2Expanded;
-    x.index_params.codebook_kind = ivf_pq::codebook_gen::PER_CLUSTER;
-    x.index_params.pq_dim        = 2;
-    x.index_params.pq_bits       = 8;
-    x.index_params.n_lists       = 69;
-    x.search_params.n_probes     = 69;
-  });
-
-  return xs;
-}
-
-/* Test instantiations */
-
-#define TEST_BUILD_SEARCH(type)                         \
-  TEST_P(type, build_search) /* NOLINT */               \
-  {                                                     \
-    this->run([this]() { return this->build_only(); }); \
-  }
-
-#define TEST_BUILD_EXTEND_SEARCH(type)                       \
-  TEST_P(type, build_extend_search) /* NOLINT */             \
-  {                                                          \
-    this->run([this]() { return this->build_2_extends(); }); \
-  }
-
-#define TEST_BUILD_SERIALIZE_SEARCH(type)                    \
-  TEST_P(type, build_serialize_search) /* NOLINT */          \
-  {                                                          \
-    this->run([this]() { return this->build_serialize(); }); \
-  }
-
-#define INSTANTIATE(type, vals) \
-  INSTANTIATE_TEST_SUITE_P(IvfPq, type, ::testing::ValuesIn(vals)); /* NOLINT */
-
-}  // namespace cuvs::neighbors::ivf_pq
diff --git a/cpp/test/neighbors/ann_ivf_pq/test_filter_float_int64_t.cu b/cpp/test/neighbors/ann_ivf_pq/test_filter_float_int64_t.cu
deleted file mode 100644
index e14993813..000000000
--- a/cpp/test/neighbors/ann_ivf_pq/test_filter_float_int64_t.cu
+++ /dev/null
@@ -1,26 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#undef RAFT_EXPLICIT_INSTANTIATE_ONLY  // Enable instantiation of search with filter
-#include "../ann_ivf_pq.cuh"
-
-namespace cuvs::neighbors::ivf_pq {
-
-using f32_f32_i64_filter = ivf_pq_filter_test<float, float, int64_t>;
-
-TEST_BUILD_SEARCH(f32_f32_i64_filter)
-INSTANTIATE(f32_f32_i64_filter, defaults() + big_dims_moderate_lut());
-}  // namespace cuvs::neighbors::ivf_pq
diff --git a/cpp/test/neighbors/ann_ivf_pq/test_filter_int8_t_int64_t.cu b/cpp/test/neighbors/ann_ivf_pq/test_filter_int8_t_int64_t.cu
deleted file mode 100644
index 2377e729d..000000000
--- a/cpp/test/neighbors/ann_ivf_pq/test_filter_int8_t_int64_t.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#undef RAFT_EXPLICIT_INSTANTIATE_ONLY  // Enable instantiation of search with filter
-#include "../ann_ivf_pq.cuh"
-
-namespace cuvs::neighbors::ivf_pq {
-
-using f32_i08_i64_filter = ivf_pq_filter_test<float, int8_t, int64_t>;
-
-TEST_BUILD_SEARCH(f32_i08_i64_filter)
-INSTANTIATE(f32_i08_i64_filter, big_dims());
-
-}  // namespace cuvs::neighbors::ivf_pq
diff --git a/cpp/test/neighbors/ann_ivf_pq/test_float_int64_t.cu b/cpp/test/neighbors/ann_ivf_pq/test_float_int64_t.cu
deleted file mode 100644
index dc3d3331c..000000000
--- a/cpp/test/neighbors/ann_ivf_pq/test_float_int64_t.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../ann_ivf_pq.cuh"
-
-namespace cuvs::neighbors::ivf_pq {
-
-using f32_f32_i64 = ivf_pq_test<float, float, int64_t>;
-
-TEST_BUILD_EXTEND_SEARCH(f32_f32_i64)
-TEST_BUILD_SERIALIZE_SEARCH(f32_f32_i64)
-INSTANTIATE(f32_f32_i64, defaults() + small_dims() + big_dims_moderate_lut());
-
-}  // namespace cuvs::neighbors::ivf_pq
diff --git a/cpp/test/neighbors/ann_ivf_pq/test_float_uint32_t.cu b/cpp/test/neighbors/ann_ivf_pq/test_float_uint32_t.cu
deleted file mode 100644
index 4d7a540da..000000000
--- a/cpp/test/neighbors/ann_ivf_pq/test_float_uint32_t.cu
+++ /dev/null
@@ -1,37 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-// XXX: the uint32_t instance is not compiled in libraft.so. So we allow
-// instantiating the template here.
-//
-// TODO: consider removing this test or consider adding an instantiation to the
-// library.
-#undef RAFT_EXPLICIT_INSTANTIATE_ONLY
-
-#include "../ann_ivf_pq.cuh"
-
-namespace cuvs::neighbors::ivf_pq {
-
-using f32_f32_u32        = ivf_pq_test<float, float, uint32_t>;
-using f32_f32_u32_filter = ivf_pq_filter_test<float, float, uint32_t>;
-
-TEST_BUILD_SEARCH(f32_f32_u32)
-TEST_BUILD_SERIALIZE_SEARCH(f32_f32_u32)
-INSTANTIATE(f32_f32_u32, defaults() + var_n_probes() + var_k() + special_cases());
-
-TEST_BUILD_SEARCH(f32_f32_u32_filter)
-INSTANTIATE(f32_f32_u32_filter, defaults());
-}  // namespace cuvs::neighbors::ivf_pq
diff --git a/cpp/test/neighbors/ann_ivf_pq/test_int8_t_int64_t.cu b/cpp/test/neighbors/ann_ivf_pq/test_int8_t_int64_t.cu
deleted file mode 100644
index c1c06d0a3..000000000
--- a/cpp/test/neighbors/ann_ivf_pq/test_int8_t_int64_t.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../ann_ivf_pq.cuh"
-
-namespace cuvs::neighbors::ivf_pq {
-
-using f32_i08_i64 = ivf_pq_test<float, int8_t, int64_t>;
-
-TEST_BUILD_SEARCH(f32_i08_i64)
-TEST_BUILD_SERIALIZE_SEARCH(f32_i08_i64)
-INSTANTIATE(f32_i08_i64, defaults() + big_dims() + var_k());
-
-}  // namespace cuvs::neighbors::ivf_pq
diff --git a/cpp/test/neighbors/ann_ivf_pq/test_uint8_t_int64_t.cu b/cpp/test/neighbors/ann_ivf_pq/test_uint8_t_int64_t.cu
deleted file mode 100644
index 6d9eef85c..000000000
--- a/cpp/test/neighbors/ann_ivf_pq/test_uint8_t_int64_t.cu
+++ /dev/null
@@ -1,27 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../ann_ivf_pq.cuh"
-
-namespace cuvs::neighbors::ivf_pq {
-
-using f32_u08_i64 = ivf_pq_test<float, uint8_t, int64_t>;
-
-TEST_BUILD_SEARCH(f32_u08_i64)
-TEST_BUILD_EXTEND_SEARCH(f32_u08_i64)
-INSTANTIATE(f32_u08_i64, small_dims_per_cluster() + enum_variety());
-
-}  // namespace cuvs::neighbors::ivf_pq
diff --git a/cpp/test/neighbors/ann_nn_descent.cuh b/cpp/test/neighbors/ann_nn_descent.cuh
deleted file mode 100644
index e03364198..000000000
--- a/cpp/test/neighbors/ann_nn_descent.cuh
+++ /dev/null
@@ -1,156 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#pragma once
-
-#include "../test_utils.cuh"
-#include "ann_utils.cuh"
-
-#include <cuvs_internal/neighbors/naive_knn.cuh>
-
-#include <cuvs/neighbors/nn_descent.cuh>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/util/itertools.hpp>
-
-#include <gtest/gtest.h>
-
-#include <cstddef>
-#include <iostream>
-#include <string>
-#include <vector>
-
-namespace cuvs::neighbors::experimental::nn_descent {
-
-struct AnnNNDescentInputs {
-  int n_rows;
-  int dim;
-  int graph_degree;
-  cuvs::distance::DistanceType metric;
-  bool host_dataset;
-  double min_recall;
-};
-
-inline ::std::ostream& operator<<(::std::ostream& os, const AnnNNDescentInputs& p)
-{
-  os << "dataset shape=" << p.n_rows << "x" << p.dim << ", graph_degree=" << p.graph_degree
-     << ", metric=" << static_cast<int>(p.metric) << (p.host_dataset ? ", host" : ", device")
-     << std::endl;
-  return os;
-}
-
-template <typename DistanceT, typename DataT, typename IdxT>
-class AnnNNDescentTest : public ::testing::TestWithParam<AnnNNDescentInputs> {
- public:
-  AnnNNDescentTest()
-    : stream_(resource::get_cuda_stream(handle_)),
-      ps(::testing::TestWithParam<AnnNNDescentInputs>::GetParam()),
-      database(0, stream_)
-  {
-  }
-
- protected:
-  void testNNDescent()
-  {
-    size_t queries_size = ps.n_rows * ps.graph_degree;
-    std::vector<IdxT> indices_NNDescent(queries_size);
-    std::vector<IdxT> indices_naive(queries_size);
-
-    {
-      rmm::device_uvector<DistanceT> distances_naive_dev(queries_size, stream_);
-      rmm::device_uvector<IdxT> indices_naive_dev(queries_size, stream_);
-      naive_knn<DistanceT, DataT, IdxT>(handle_,
-                                        distances_naive_dev.data(),
-                                        indices_naive_dev.data(),
-                                        database.data(),
-                                        database.data(),
-                                        ps.n_rows,
-                                        ps.n_rows,
-                                        ps.dim,
-                                        ps.graph_degree,
-                                        ps.metric);
-      update_host(indices_naive.data(), indices_naive_dev.data(), queries_size, stream_);
-      resource::sync_stream(handle_);
-    }
-
-    {
-      {
-        nn_descent::index_params index_params;
-        index_params.metric                    = ps.metric;
-        index_params.graph_degree              = ps.graph_degree;
-        index_params.intermediate_graph_degree = 2 * ps.graph_degree;
-        index_params.max_iterations            = 100;
-
-        auto database_view = raft::make_device_matrix_view<const DataT, int64_t>(
-          (const DataT*)database.data(), ps.n_rows, ps.dim);
-
-        {
-          if (ps.host_dataset) {
-            auto database_host = raft::make_host_matrix<DataT, int64_t>(ps.n_rows, ps.dim);
-            raft::copy(database_host.data_handle(), database.data(), database.size(), stream_);
-            auto database_host_view = raft::make_host_matrix_view<const DataT, int64_t>(
-              (const DataT*)database_host.data_handle(), ps.n_rows, ps.dim);
-            auto index = nn_descent::build<DataT, IdxT>(handle_, index_params, database_host_view);
-            update_host(
-              indices_NNDescent.data(), index.graph().data_handle(), queries_size, stream_);
-          } else {
-            auto index = nn_descent::build<DataT, IdxT>(handle_, index_params, database_view);
-            update_host(
-              indices_NNDescent.data(), index.graph().data_handle(), queries_size, stream_);
-          };
-        }
-        resource::sync_stream(handle_);
-      }
-
-      double min_recall = ps.min_recall;
-      EXPECT_TRUE(eval_recall(
-        indices_naive, indices_NNDescent, ps.n_rows, ps.graph_degree, 0.001, min_recall));
-    }
-  }
-
-  void SetUp() override
-  {
-    database.resize(((size_t)ps.n_rows) * ps.dim, stream_);
-    raft::random::RngState r(1234ULL);
-    if constexpr (std::is_same<DataT, float>{}) {
-      raft::random::normal(handle_, r, database.data(), ps.n_rows * ps.dim, DataT(0.1), DataT(2.0));
-    } else {
-      raft::random::uniformInt(
-        handle_, r, database.data(), ps.n_rows * ps.dim, DataT(1), DataT(20));
-    }
-    resource::sync_stream(handle_);
-  }
-
-  void TearDown() override
-  {
-    resource::sync_stream(handle_);
-    database.resize(0, stream_);
-  }
-
- private:
-  raft::resources handle_;
-  rmm::cuda_stream_view stream_;
-  AnnNNDescentInputs ps;
-  rmm::device_uvector<DataT> database;
-};
-
-const std::vector<AnnNNDescentInputs> inputs = raft::util::itertools::product<AnnNNDescentInputs>(
-  {1000, 2000},                                              // n_rows
-  {3, 5, 7, 8, 17, 64, 128, 137, 192, 256, 512, 619, 1024},  // dim
-  {32, 64},                                                  // graph_degree
-  {cuvs::distance::DistanceType::L2Expanded},
-  {false, true},
-  {0.90});
-
-}  // namespace cuvs::neighbors::experimental::nn_descent
diff --git a/cpp/test/neighbors/ann_nn_descent/test_float_uint32_t.cu b/cpp/test/neighbors/ann_nn_descent/test_float_uint32_t.cu
deleted file mode 100644
index 882ba5f83..000000000
--- a/cpp/test/neighbors/ann_nn_descent/test_float_uint32_t.cu
+++ /dev/null
@@ -1,28 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <gtest/gtest.h>
-
-#include "../ann_nn_descent.cuh"
-
-namespace cuvs::neighbors::experimental::nn_descent {
-
-typedef AnnNNDescentTest<float, float, std::uint32_t> AnnNNDescentTestF_U32;
-TEST_P(AnnNNDescentTestF_U32, AnnNNDescent) { this->testNNDescent(); }
-
-INSTANTIATE_TEST_CASE_P(AnnNNDescentTest, AnnNNDescentTestF_U32, ::testing::ValuesIn(inputs));
-
-}  // namespace cuvs::neighbors::experimental::nn_descent
diff --git a/cpp/test/neighbors/ann_nn_descent/test_int8_t_uint32_t.cu b/cpp/test/neighbors/ann_nn_descent/test_int8_t_uint32_t.cu
deleted file mode 100644
index 6a1fb6c8c..000000000
--- a/cpp/test/neighbors/ann_nn_descent/test_int8_t_uint32_t.cu
+++ /dev/null
@@ -1,28 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <gtest/gtest.h>
-
-#include "../ann_nn_descent.cuh"
-
-namespace cuvs::neighbors::experimental::nn_descent {
-
-typedef AnnNNDescentTest<float, int8_t, std::uint32_t> AnnNNDescentTestI8_U32;
-TEST_P(AnnNNDescentTestI8_U32, AnnNNDescent) { this->testNNDescent(); }
-
-INSTANTIATE_TEST_CASE_P(AnnNNDescentTest, AnnNNDescentTestI8_U32, ::testing::ValuesIn(inputs));
-
-}  // namespace cuvs::neighbors::experimental::nn_descent
diff --git a/cpp/test/neighbors/ann_nn_descent/test_uint8_t_uint32_t.cu b/cpp/test/neighbors/ann_nn_descent/test_uint8_t_uint32_t.cu
deleted file mode 100644
index 088f2c328..000000000
--- a/cpp/test/neighbors/ann_nn_descent/test_uint8_t_uint32_t.cu
+++ /dev/null
@@ -1,28 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <gtest/gtest.h>
-
-#include "../ann_nn_descent.cuh"
-
-namespace cuvs::neighbors::experimental::nn_descent {
-
-typedef AnnNNDescentTest<float, uint8_t, std::uint32_t> AnnNNDescentTestUI8_U32;
-TEST_P(AnnNNDescentTestUI8_U32, AnnNNDescent) { this->testNNDescent(); }
-
-INSTANTIATE_TEST_CASE_P(AnnNNDescentTest, AnnNNDescentTestUI8_U32, ::testing::ValuesIn(inputs));
-
-}  // namespace cuvs::neighbors::experimental::nn_descent
diff --git a/cpp/test/neighbors/ball_cover.cu b/cpp/test/neighbors/ball_cover.cu
deleted file mode 100644
index 1002d61c5..000000000
--- a/cpp/test/neighbors/ball_cover.cu
+++ /dev/null
@@ -1,372 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "spatial_data.h"
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/ball_cover.cuh>
-#include <cuvs/neighbors/brute_force.cuh>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resource/thrust_policy.hpp>
-#include <raft/random/make_blobs.cuh>
-#include <raft/util/cudart_utils.hpp>
-
-#include <rmm/device_uvector.hpp>
-#include <rmm/exec_policy.hpp>
-
-#include <thrust/count.h>
-#include <thrust/fill.h>
-#include <thrust/transform.h>
-
-#include <cstdint>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <vector>
-
-namespace cuvs::neighbors::ball_cover {
-using namespace std;
-
-template <typename value_idx, typename value_t>
-RAFT_KERNEL count_discrepancies_kernel(value_idx* actual_idx,
-                                       value_idx* expected_idx,
-                                       value_t* actual,
-                                       value_t* expected,
-                                       uint32_t m,
-                                       uint32_t n,
-                                       uint32_t* out,
-                                       float thres = 1e-3)
-{
-  uint32_t row = blockDim.x * blockIdx.x + threadIdx.x;
-
-  int n_diffs = 0;
-  if (row < m) {
-    for (uint32_t i = 0; i < n; i++) {
-      value_t d    = actual[row * n + i] - expected[row * n + i];
-      bool matches = (fabsf(d) <= thres) || (actual_idx[row * n + i] == expected_idx[row * n + i] &&
-                                             actual_idx[row * n + i] == row);
-
-      if (!matches) {
-        printf(
-          "row=%ud, n=%ud, actual_dist=%f, actual_ind=%ld, expected_dist=%f, expected_ind=%ld\n",
-          row,
-          i,
-          actual[row * n + i],
-          actual_idx[row * n + i],
-          expected[row * n + i],
-          expected_idx[row * n + i]);
-      }
-      n_diffs += !matches;
-      out[row] = n_diffs;
-    }
-  }
-}
-
-struct is_nonzero {
-  __host__ __device__ bool operator()(uint32_t& i) { return i > 0; }
-};
-
-template <typename value_idx, typename value_t>
-uint32_t count_discrepancies(value_idx* actual_idx,
-                             value_idx* expected_idx,
-                             value_t* actual,
-                             value_t* expected,
-                             uint32_t m,
-                             uint32_t n,
-                             uint32_t* out,
-                             cudaStream_t stream)
-{
-  uint32_t tpb = 256;
-  count_discrepancies_kernel<<<raft::ceildiv(m, tpb), tpb, 0, stream>>>(
-    actual_idx, expected_idx, actual, expected, m, n, out);
-
-  auto exec_policy = rmm::exec_policy(stream);
-
-  uint32_t result = thrust::count_if(exec_policy, out, out + m, is_nonzero());
-  return result;
-}
-
-template <typename value_t>
-void compute_bfknn(const raft::resources& handle,
-                   const value_t* X1,
-                   const value_t* X2,
-                   uint32_t n_rows,
-                   uint32_t n_query_rows,
-                   uint32_t d,
-                   uint32_t k,
-                   const cuvs::distance::DistanceType metric,
-                   value_t* dists,
-                   int64_t* inds)
-{
-  std::vector<raft::device_matrix_view<const value_t, uint32_t>> input_vec = {
-    make_device_matrix_view(X1, n_rows, d)};
-
-  cuvs::neighbors::brute_force::knn(handle,
-                                    input_vec,
-                                    make_device_matrix_view(X2, n_query_rows, d),
-                                    make_device_matrix_view(inds, n_query_rows, k),
-                                    make_device_matrix_view(dists, n_query_rows, k),
-                                    metric);
-}
-
-struct ToRadians {
-  __device__ __host__ float operator()(float a) { return a * (CUDART_PI_F / 180.0); }
-};
-
-template <typename value_int = std::uint32_t>
-struct BallCoverInputs {
-  value_int k;
-  value_int n_rows;
-  value_int n_cols;
-  float weight;
-  value_int n_query;
-  cuvs::distance::DistanceType metric;
-};
-
-template <typename value_idx, typename value_t, typename value_int = std::uint32_t>
-class BallCoverKNNQueryTest : public ::testing::TestWithParam<BallCoverInputs<value_int>> {
- protected:
-  void basicTest()
-  {
-    params = ::testing::TestWithParam<BallCoverInputs<value_int>>::GetParam();
-    raft::resources handle;
-
-    uint32_t k         = params.k;
-    uint32_t n_centers = 25;
-    float weight       = params.weight;
-    auto metric        = params.metric;
-
-    rmm::device_uvector<value_t> X(params.n_rows * params.n_cols,
-                                   resource::get_cuda_stream(handle));
-    rmm::device_uvector<uint32_t> Y(params.n_rows, resource::get_cuda_stream(handle));
-
-    // Make sure the train and query sets are completely disjoint
-    rmm::device_uvector<value_t> X2(params.n_query * params.n_cols,
-                                    resource::get_cuda_stream(handle));
-    rmm::device_uvector<uint32_t> Y2(params.n_query, resource::get_cuda_stream(handle));
-
-    raft::random::make_blobs(X.data(),
-                             Y.data(),
-                             params.n_rows,
-                             params.n_cols,
-                             n_centers,
-                             resource::get_cuda_stream(handle));
-
-    raft::random::make_blobs(X2.data(),
-                             Y2.data(),
-                             params.n_query,
-                             params.n_cols,
-                             n_centers,
-                             resource::get_cuda_stream(handle));
-
-    rmm::device_uvector<value_idx> d_ref_I(params.n_query * k, resource::get_cuda_stream(handle));
-    rmm::device_uvector<value_t> d_ref_D(params.n_query * k, resource::get_cuda_stream(handle));
-
-    if (metric == cuvs::distance::DistanceType::Haversine) {
-      thrust::transform(
-        resource::get_thrust_policy(handle), X.data(), X.data() + X.size(), X.data(), ToRadians());
-      thrust::transform(resource::get_thrust_policy(handle),
-                        X2.data(),
-                        X2.data() + X2.size(),
-                        X2.data(),
-                        ToRadians());
-    }
-
-    compute_bfknn(handle,
-                  X.data(),
-                  X2.data(),
-                  params.n_rows,
-                  params.n_query,
-                  params.n_cols,
-                  k,
-                  metric,
-                  d_ref_D.data(),
-                  d_ref_I.data());
-
-    resource::sync_stream(handle);
-
-    // Allocate predicted arrays
-    rmm::device_uvector<value_idx> d_pred_I(params.n_query * k, resource::get_cuda_stream(handle));
-    rmm::device_uvector<value_t> d_pred_D(params.n_query * k, resource::get_cuda_stream(handle));
-
-    auto X_view =
-      raft::make_device_matrix_view<value_t, value_int>(X.data(), params.n_rows, params.n_cols);
-    auto X2_view = raft::make_device_matrix_view<const value_t, value_int>(
-      (const value_t*)X2.data(), params.n_query, params.n_cols);
-
-    auto d_pred_I_view =
-      raft::make_device_matrix_view<value_idx, value_int>(d_pred_I.data(), params.n_query, k);
-    auto d_pred_D_view =
-      raft::make_device_matrix_view<value_t, value_int>(d_pred_D.data(), params.n_query, k);
-
-    BallCoverIndex<value_idx, value_t, value_int, value_int> index(handle, X_view, metric);
-
-    build_index(handle, index);
-    knn_query(handle, index, X2_view, d_pred_I_view, d_pred_D_view, k, true);
-
-    resource::sync_stream(handle);
-    // What we really want are for the distances to match exactly. The
-    // indices may or may not match exactly, depending upon the ordering which
-    // can be nondeterministic.
-
-    rmm::device_uvector<uint32_t> discrepancies(params.n_query, resource::get_cuda_stream(handle));
-    thrust::fill(resource::get_thrust_policy(handle),
-                 discrepancies.data(),
-                 discrepancies.data() + discrepancies.size(),
-                 0);
-    //
-    int res = count_discrepancies(d_ref_I.data(),
-                                  d_pred_I.data(),
-                                  d_ref_D.data(),
-                                  d_pred_D.data(),
-                                  params.n_query,
-                                  k,
-                                  discrepancies.data(),
-                                  resource::get_cuda_stream(handle));
-
-    ASSERT_TRUE(res == 0);
-  }
-
-  void SetUp() override {}
-
-  void TearDown() override {}
-
- protected:
-  uint32_t d = 2;
-  BallCoverInputs<value_int> params;
-};
-
-template <typename value_idx, typename value_t, typename value_int = std::uint32_t>
-class BallCoverAllKNNTest : public ::testing::TestWithParam<BallCoverInputs<value_int>> {
- protected:
-  void basicTest()
-  {
-    params = ::testing::TestWithParam<BallCoverInputs<value_int>>::GetParam();
-    raft::resources handle;
-
-    uint32_t k         = params.k;
-    uint32_t n_centers = 25;
-    float weight       = params.weight;
-    auto metric        = params.metric;
-
-    rmm::device_uvector<value_t> X(params.n_rows * params.n_cols,
-                                   resource::get_cuda_stream(handle));
-    rmm::device_uvector<uint32_t> Y(params.n_rows, resource::get_cuda_stream(handle));
-
-    raft::random::make_blobs(X.data(),
-                             Y.data(),
-                             params.n_rows,
-                             params.n_cols,
-                             n_centers,
-                             resource::get_cuda_stream(handle));
-
-    rmm::device_uvector<value_idx> d_ref_I(params.n_rows * k, resource::get_cuda_stream(handle));
-    rmm::device_uvector<value_t> d_ref_D(params.n_rows * k, resource::get_cuda_stream(handle));
-
-    auto X_view = raft::make_device_matrix_view<const value_t, value_int>(
-      (const value_t*)X.data(), params.n_rows, params.n_cols);
-
-    if (metric == cuvs::distance::DistanceType::Haversine) {
-      thrust::transform(
-        resource::get_thrust_policy(handle), X.data(), X.data() + X.size(), X.data(), ToRadians());
-    }
-
-    compute_bfknn(handle,
-                  X.data(),
-                  X.data(),
-                  params.n_rows,
-                  params.n_rows,
-                  params.n_cols,
-                  k,
-                  metric,
-                  d_ref_D.data(),
-                  d_ref_I.data());
-
-    resource::sync_stream(handle);
-
-    // Allocate predicted arrays
-    rmm::device_uvector<value_idx> d_pred_I(params.n_rows * k, resource::get_cuda_stream(handle));
-    rmm::device_uvector<value_t> d_pred_D(params.n_rows * k, resource::get_cuda_stream(handle));
-
-    auto d_pred_I_view =
-      raft::make_device_matrix_view<value_idx, value_int>(d_pred_I.data(), params.n_rows, k);
-    auto d_pred_D_view =
-      raft::make_device_matrix_view<value_t, value_int>(d_pred_D.data(), params.n_rows, k);
-
-    BallCoverIndex<value_idx, value_t> index(handle, X_view, metric);
-
-    all_knn_query(handle, index, d_pred_I_view, d_pred_D_view, k, true);
-
-    resource::sync_stream(handle);
-    // What we really want are for the distances to match exactly. The
-    // indices may or may not match exactly, depending upon the ordering which
-    // can be nondeterministic.
-
-    rmm::device_uvector<uint32_t> discrepancies(params.n_rows, resource::get_cuda_stream(handle));
-    thrust::fill(resource::get_thrust_policy(handle),
-                 discrepancies.data(),
-                 discrepancies.data() + discrepancies.size(),
-                 0);
-    //
-    uint32_t res = count_discrepancies(d_ref_I.data(),
-                                       d_pred_I.data(),
-                                       d_ref_D.data(),
-                                       d_pred_D.data(),
-                                       params.n_rows,
-                                       k,
-                                       discrepancies.data(),
-                                       resource::get_cuda_stream(handle));
-
-    // TODO: There seem to be discrepancies here only when
-    // the entire test suite is executed.
-    // Ref: https://github.com/rapidsai/raft/issues/
-    // 1-5 mismatches in 8000 samples is 0.0125% - 0.0625%
-    ASSERT_TRUE(res <= 5);
-  }
-
-  void SetUp() override {}
-
-  void TearDown() override {}
-
- protected:
-  BallCoverInputs<value_int> params;
-};
-
-typedef BallCoverAllKNNTest<int64_t, float> BallCoverAllKNNTestF;
-typedef BallCoverKNNQueryTest<int64_t, float> BallCoverKNNQueryTestF;
-
-const std::vector<BallCoverInputs<std::uint32_t>> ballcover_inputs = {
-  {11, 5000, 2, 1.0, 10000, cuvs::distance::DistanceType::Haversine},
-  {25, 10000, 2, 1.0, 5000, cuvs::distance::DistanceType::Haversine},
-  {2, 10000, 2, 1.0, 5000, cuvs::distance::DistanceType::L2SqrtUnexpanded},
-  {2, 5000, 2, 1.0, 10000, cuvs::distance::DistanceType::Haversine},
-  {11, 10000, 2, 1.0, 5000, cuvs::distance::DistanceType::L2SqrtUnexpanded},
-  {25, 5000, 2, 1.0, 10000, cuvs::distance::DistanceType::L2SqrtUnexpanded},
-  {5, 8000, 3, 1.0, 10000, cuvs::distance::DistanceType::L2SqrtUnexpanded},
-  {11, 6000, 3, 1.0, 10000, cuvs::distance::DistanceType::L2SqrtUnexpanded},
-  {25, 10000, 3, 1.0, 5000, cuvs::distance::DistanceType::L2SqrtUnexpanded}};
-
-INSTANTIATE_TEST_CASE_P(BallCoverAllKNNTest,
-                        BallCoverAllKNNTestF,
-                        ::testing::ValuesIn(ballcover_inputs));
-INSTANTIATE_TEST_CASE_P(BallCoverKNNQueryTest,
-                        BallCoverKNNQueryTestF,
-                        ::testing::ValuesIn(ballcover_inputs));
-
-TEST_P(BallCoverAllKNNTestF, Fit) { basicTest(); }
-TEST_P(BallCoverKNNQueryTestF, Fit) { basicTest(); }
-
-}  // namespace cuvs::neighbors::ball_cover
diff --git a/cpp/test/neighbors/epsilon_neighborhood.cu b/cpp/test/neighbors/epsilon_neighborhood.cu
deleted file mode 100644
index 803a8ed76..000000000
--- a/cpp/test/neighbors/epsilon_neighborhood.cu
+++ /dev/null
@@ -1,121 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <cuvs/spatial/knn/epsilon_neighborhood.cuh>
-#include <gtest/gtest.h>
-#include <memory>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/random/make_blobs.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <rmm/device_uvector.hpp>
-
-namespace raft {
-namespace spatial {
-namespace knn {
-template <typename T, typename IdxT>
-struct EpsInputs {
-  IdxT n_row, n_col, n_centers, n_batches;
-  T eps;
-};
-
-template <typename T, typename IdxT>
-::std::ostream& operator<<(::std::ostream& os, const EpsInputs<T, IdxT>& p)
-{
-  return os;
-}
-
-template <typename T, typename IdxT>
-class EpsNeighTest : public ::testing::TestWithParam<EpsInputs<T, IdxT>> {
- protected:
-  EpsNeighTest()
-    : data(0, resource::get_cuda_stream(handle)),
-      adj(0, resource::get_cuda_stream(handle)),
-      labels(0, resource::get_cuda_stream(handle)),
-      vd(0, resource::get_cuda_stream(handle))
-  {
-  }
-
-  void SetUp() override
-  {
-    auto stream = resource::get_cuda_stream(handle);
-    param       = ::testing::TestWithParam<EpsInputs<T, IdxT>>::GetParam();
-    data.resize(param.n_row * param.n_col, stream);
-    labels.resize(param.n_row, stream);
-    batchSize = param.n_row / param.n_batches;
-    adj.resize(param.n_row * batchSize, stream);
-    vd.resize(batchSize + 1, stream);
-    RAFT_CUDA_TRY(cudaMemsetAsync(vd.data(), 0, vd.size() * sizeof(IdxT), stream));
-    random::make_blobs<T, IdxT>(data.data(),
-                                labels.data(),
-                                param.n_row,
-                                param.n_col,
-                                param.n_centers,
-                                stream,
-                                true,
-                                nullptr,
-                                nullptr,
-                                T(0.01),
-                                false);
-  }
-
-  const raft::resources handle;
-  EpsInputs<T, IdxT> param;
-  cudaStream_t stream = 0;
-  rmm::device_uvector<T> data;
-  rmm::device_uvector<bool> adj;
-  rmm::device_uvector<IdxT> labels, vd;
-  IdxT batchSize;
-};  // class EpsNeighTest
-
-const std::vector<EpsInputs<float, int>> inputsfi = {
-  {15000, 16, 5, 1, 2.f},
-  {14000, 16, 5, 1, 2.f},
-  {15000, 17, 5, 1, 2.f},
-  {14000, 17, 5, 1, 2.f},
-  {15000, 18, 5, 1, 2.f},
-  {14000, 18, 5, 1, 2.f},
-  {15000, 32, 5, 1, 2.f},
-  {14000, 32, 5, 1, 2.f},
-  {20000, 10000, 10, 1, 2.f},
-  {20000, 10000, 10, 2, 2.f},
-};
-typedef EpsNeighTest<float, int> EpsNeighTestFI;
-TEST_P(EpsNeighTestFI, Result)
-{
-  for (int i = 0; i < param.n_batches; ++i) {
-    RAFT_CUDA_TRY(cudaMemsetAsync(adj.data(), 0, sizeof(bool) * param.n_row * batchSize, stream));
-    RAFT_CUDA_TRY(cudaMemsetAsync(vd.data(), 0, sizeof(int) * (batchSize + 1), stream));
-
-    auto adj_view = make_device_matrix_view<bool, int>(adj.data(), param.n_row, batchSize);
-    auto vd_view  = make_device_vector_view<int, int>(vd.data(), batchSize + 1);
-    auto x_view   = make_device_matrix_view<float, int>(data.data(), param.n_row, param.n_col);
-    auto y_view   = make_device_matrix_view<float, int>(
-      data.data() + (i * batchSize * param.n_col), batchSize, param.n_col);
-
-    eps_neighbors_l2sq<float, int, int>(
-      handle, x_view, y_view, adj_view, vd_view, param.eps * param.eps);
-
-    ASSERT_TRUE(raft::devArrMatch(
-      param.n_row / param.n_centers, vd.data(), batchSize, raft::Compare<int>(), stream));
-  }
-}
-INSTANTIATE_TEST_CASE_P(EpsNeighTests, EpsNeighTestFI, ::testing::ValuesIn(inputsfi));
-
-};  // namespace knn
-};  // namespace spatial
-};  // namespace raft
diff --git a/cpp/test/neighbors/fused_l2_knn.cu b/cpp/test/neighbors/fused_l2_knn.cu
deleted file mode 100644
index 770720e92..000000000
--- a/cpp/test/neighbors/fused_l2_knn.cu
+++ /dev/null
@@ -1,175 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "./knn_utils.cuh"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/brute_force.cuh>
-#include <cuvs/spatial/knn/knn.cuh>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/random/rng.cuh>
-
-#include <cuvs/distance/distance.cuh>
-
-#include <rmm/device_buffer.hpp>
-
-#include <gtest/gtest.h>
-
-#include <cstddef>
-#include <iostream>
-#include <vector>
-
-namespace raft {
-namespace spatial {
-namespace knn {
-struct FusedL2KNNInputs {
-  int num_queries;
-  int num_db_vecs;
-  int dim;
-  int k;
-  cuvs::distance::DistanceType metric_;
-};
-
-template <typename T>
-class FusedL2KNNTest : public ::testing::TestWithParam<FusedL2KNNInputs> {
- public:
-  FusedL2KNNTest()
-    : stream_(resource::get_cuda_stream(handle_)),
-      params_(::testing::TestWithParam<FusedL2KNNInputs>::GetParam()),
-      database(params_.num_db_vecs * params_.dim, stream_),
-      search_queries(params_.num_queries * params_.dim, stream_),
-      raft_indices_(params_.num_queries * params_.k, stream_),
-      raft_distances_(params_.num_queries * params_.k, stream_),
-      ref_indices_(params_.num_queries * params_.k, stream_),
-      ref_distances_(params_.num_queries * params_.k, stream_)
-  {
-    RAFT_CUDA_TRY(cudaMemsetAsync(database.data(), 0, database.size() * sizeof(T), stream_));
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(search_queries.data(), 0, search_queries.size() * sizeof(T), stream_));
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(raft_indices_.data(), 0, raft_indices_.size() * sizeof(int64_t), stream_));
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(raft_distances_.data(), 0, raft_distances_.size() * sizeof(T), stream_));
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(ref_indices_.data(), 0, ref_indices_.size() * sizeof(int64_t), stream_));
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(ref_distances_.data(), 0, ref_distances_.size() * sizeof(T), stream_));
-  }
-
- protected:
-  void testBruteForce()
-  {
-    // calculate the naive knn, by calculating the full pairwise distances and doing a k-select
-    rmm::device_uvector<T> temp_distances(num_db_vecs * num_queries, stream_);
-    distance::pairwise_distance(
-      handle_,
-      raft::make_device_matrix_view<T, int32_t>(search_queries.data(), num_queries, dim),
-      raft::make_device_matrix_view<T, int32_t>(database.data(), num_db_vecs, dim),
-      raft::make_device_matrix_view<T, int32_t>(temp_distances.data(), num_queries, num_db_vecs),
-      metric);
-
-    spatial::knn::select_k<int64_t, T>(temp_distances.data(),
-                                       nullptr,
-                                       num_queries,
-                                       num_db_vecs,
-                                       ref_distances_.data(),
-                                       ref_indices_.data(),
-                                       true,
-                                       k_,
-                                       stream_);
-
-    auto index_view =
-      raft::make_device_matrix_view<const T, int64_t>(database.data(), num_db_vecs, dim);
-    auto query_view =
-      raft::make_device_matrix_view<const T, int64_t>(search_queries.data(), num_queries, dim);
-    auto out_indices_view =
-      raft::make_device_matrix_view<int64_t, int64_t>(raft_indices_.data(), num_queries, k_);
-    auto out_dists_view =
-      raft::make_device_matrix_view<T, int64_t>(raft_distances_.data(), num_queries, k_);
-    cuvs::neighbors::brute_force::fused_l2_knn(
-      handle_, index_view, query_view, out_indices_view, out_dists_view, metric);
-
-    // verify.
-    ASSERT_TRUE(devArrMatchKnnPair(ref_indices_.data(),
-                                   raft_indices_.data(),
-                                   ref_distances_.data(),
-                                   raft_distances_.data(),
-                                   num_queries,
-                                   k_,
-                                   float(0.001),
-                                   stream_));
-  }
-
-  void SetUp() override
-  {
-    num_queries = params_.num_queries;
-    num_db_vecs = params_.num_db_vecs;
-    dim         = params_.dim;
-    k_          = params_.k;
-    metric      = params_.metric_;
-
-    unsigned long long int seed = 1234ULL;
-    raft::random::RngState r(seed);
-    uniform(handle_, r, database.data(), num_db_vecs * dim, T(-1.0), T(1.0));
-    uniform(handle_, r, search_queries.data(), num_queries * dim, T(-1.0), T(1.0));
-  }
-
- private:
-  raft::resources handle_;
-  cudaStream_t stream_ = 0;
-  FusedL2KNNInputs params_;
-  int num_queries;
-  int num_db_vecs;
-  int dim;
-  rmm::device_uvector<T> database;
-  rmm::device_uvector<T> search_queries;
-  rmm::device_uvector<int64_t> raft_indices_;
-  rmm::device_uvector<T> raft_distances_;
-  rmm::device_uvector<int64_t> ref_indices_;
-  rmm::device_uvector<T> ref_distances_;
-  int k_;
-  cuvs::distance::DistanceType metric;
-};
-
-const std::vector<FusedL2KNNInputs> inputs = {
-  {100, 1000, 16, 10, cuvs::distance::DistanceType::L2Expanded},
-  {256, 256, 30, 10, cuvs::distance::DistanceType::L2Expanded},
-  {1000, 10000, 16, 10, cuvs::distance::DistanceType::L2Expanded},
-  {100, 1000, 16, 50, cuvs::distance::DistanceType::L2Expanded},
-  {20, 10000, 16, 10, cuvs::distance::DistanceType::L2Expanded},
-  {1000, 10000, 16, 50, cuvs::distance::DistanceType::L2Expanded},
-  {1000, 10000, 32, 50, cuvs::distance::DistanceType::L2Expanded},
-  {10000, 40000, 32, 30, cuvs::distance::DistanceType::L2Expanded},
-  // L2 unexpanded
-  {100, 1000, 16, 10, cuvs::distance::DistanceType::L2Unexpanded},
-  {1000, 10000, 16, 10, cuvs::distance::DistanceType::L2Unexpanded},
-  {100, 1000, 16, 50, cuvs::distance::DistanceType::L2Unexpanded},
-  {20, 10000, 16, 50, cuvs::distance::DistanceType::L2Unexpanded},
-  {1000, 10000, 16, 50, cuvs::distance::DistanceType::L2Unexpanded},
-  {1000, 10000, 32, 50, cuvs::distance::DistanceType::L2Unexpanded},
-  {10000, 40000, 32, 30, cuvs::distance::DistanceType::L2Unexpanded},
-};
-
-typedef FusedL2KNNTest<float> FusedL2KNNTestF;
-TEST_P(FusedL2KNNTestF, FusedBruteForce) { this->testBruteForce(); }
-
-INSTANTIATE_TEST_CASE_P(FusedL2KNNTest, FusedL2KNNTestF, ::testing::ValuesIn(inputs));
-
-}  // namespace knn
-}  // namespace spatial
-}  // namespace raft
diff --git a/cpp/test/neighbors/haversine.cu b/cpp/test/neighbors/haversine.cu
deleted file mode 100644
index 4cf9c169c..000000000
--- a/cpp/test/neighbors/haversine.cu
+++ /dev/null
@@ -1,133 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/spatial/knn/detail/haversine_distance.cuh>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <rmm/device_uvector.hpp>
-#include <vector>
-
-namespace raft {
-namespace spatial {
-namespace knn {
-
-template <typename value_idx, typename value_t>
-class HaversineKNNTest : public ::testing::Test {
- public:
-  HaversineKNNTest()
-    : stream(resource::get_cuda_stream(handle)),
-      d_train_inputs(0, stream),
-      d_ref_I(0, stream),
-      d_ref_D(0, stream),
-      d_pred_I(0, stream),
-      d_pred_D(0, stream)
-  {
-  }
-
- protected:
-  void basicTest()
-  {
-    // Allocate input
-    d_train_inputs.resize(n * d, stream);
-
-    // Allocate reference arrays
-    d_ref_I.resize(n * n, stream);
-    d_ref_D.resize(n * n, stream);
-
-    // Allocate predicted arrays
-    d_pred_I.resize(n * n, stream);
-    d_pred_D.resize(n * n, stream);
-
-    // make testdata on host
-    std::vector<value_t> h_train_inputs = {0.71113885,
-                                           -1.29215058,
-                                           0.59613176,
-                                           -2.08048115,
-                                           0.74932804,
-                                           -1.33634042,
-                                           0.51486728,
-                                           -1.65962873,
-                                           0.53154002,
-                                           -1.47049808,
-                                           0.72891737,
-                                           -1.54095137};
-
-    h_train_inputs.resize(d_train_inputs.size());
-    raft::update_device(
-      d_train_inputs.data(), h_train_inputs.data(), d_train_inputs.size(), stream);
-
-    std::vector<value_t> h_res_D = {0., 0.05041587, 0.18767063, 0.23048252, 0.35749438, 0.62925595,
-                                    0., 0.36575755, 0.44288665, 0.5170737,  0.59501296, 0.62925595,
-                                    0., 0.05041587, 0.152463,   0.2426416,  0.34925285, 0.59501296,
-                                    0., 0.16461092, 0.2345792,  0.34925285, 0.35749438, 0.36575755,
-                                    0., 0.16461092, 0.20535265, 0.23048252, 0.2426416,  0.5170737,
-                                    0., 0.152463,   0.18767063, 0.20535265, 0.2345792,  0.44288665};
-    h_res_D.resize(n * n);
-    raft::update_device(d_ref_D.data(), h_res_D.data(), n * n, stream);
-
-    std::vector<value_idx> h_res_I = {0, 2, 5, 4, 3, 1, 1, 3, 5, 4, 2, 0, 2, 0, 5, 4, 3, 1,
-                                      3, 4, 5, 2, 0, 1, 4, 3, 5, 0, 2, 1, 5, 2, 0, 4, 3, 1};
-    h_res_I.resize(n * n);
-    raft::update_device<value_idx>(d_ref_I.data(), h_res_I.data(), n * n, stream);
-
-    cuvs::spatial::knn::detail::haversine_knn(d_pred_I.data(),
-                                              d_pred_D.data(),
-                                              d_train_inputs.data(),
-                                              d_train_inputs.data(),
-                                              n,
-                                              n,
-                                              k,
-                                              stream);
-
-    resource::sync_stream(handle, stream);
-  }
-
-  void SetUp() override { basicTest(); }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream;
-
-  rmm::device_uvector<value_t> d_train_inputs;
-
-  int n = 6;
-  int d = 2;
-
-  int k = 6;
-
-  rmm::device_uvector<value_idx> d_pred_I;
-  rmm::device_uvector<value_t> d_pred_D;
-
-  rmm::device_uvector<value_idx> d_ref_I;
-  rmm::device_uvector<value_t> d_ref_D;
-};
-
-typedef HaversineKNNTest<int, float> HaversineKNNTestF;
-
-TEST_F(HaversineKNNTestF, Fit)
-{
-  ASSERT_TRUE(raft::devArrMatch(
-    d_ref_D.data(), d_pred_D.data(), n * n, raft::CompareApprox<float>(1e-3), stream));
-  ASSERT_TRUE(
-    raft::devArrMatch(d_ref_I.data(), d_pred_I.data(), n * n, raft::Compare<int>(), stream));
-}
-
-}  // namespace knn
-}  // namespace spatial
-}  // namespace raft
diff --git a/cpp/test/neighbors/knn.cu b/cpp/test/neighbors/knn.cu
deleted file mode 100644
index 907520e42..000000000
--- a/cpp/test/neighbors/knn.cu
+++ /dev/null
@@ -1,197 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/brute_force.cuh>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/logger.hpp>
-
-#include <rmm/device_buffer.hpp>
-
-#include <gtest/gtest.h>
-
-#include <cstddef>
-#include <iostream>
-#include <vector>
-
-namespace cuvs::neighbors::brute_force {
-struct KNNInputs {
-  std::vector<std::vector<float>> input;
-  int k;
-  std::vector<int> labels;
-};
-
-template <typename IdxT>
-RAFT_KERNEL build_actual_output(
-  int* output, int n_rows, int k, const int* idx_labels, const IdxT* indices)
-{
-  int element = threadIdx.x + blockDim.x * blockIdx.x;
-  if (element >= n_rows * k) return;
-
-  output[element] = idx_labels[indices[element]];
-}
-
-RAFT_KERNEL build_expected_output(int* output, int n_rows, int k, const int* labels)
-{
-  int row = threadIdx.x + blockDim.x * blockIdx.x;
-  if (row >= n_rows) return;
-
-  int cur_label = labels[row];
-  for (int i = 0; i < k; i++) {
-    output[row * k + i] = cur_label;
-  }
-}
-
-template <typename T, typename IdxT>
-class KNNTest : public ::testing::TestWithParam<KNNInputs> {
- public:
-  KNNTest()
-    : params_(::testing::TestWithParam<KNNInputs>::GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      actual_labels_(0, stream),
-      expected_labels_(0, stream),
-      input_(0, stream),
-      search_data_(0, stream),
-      indices_(0, stream),
-      distances_(0, stream),
-      search_labels_(0, stream)
-  {
-  }
-
- protected:
-  void testBruteForce()
-  {
-    // #if (RAFT_ACTIVE_LEVEL >= RAFT_LEVEL_DEBUG)
-    raft::print_device_vector("Input array: ", input_.data(), rows_ * cols_, std::cout);
-    std::cout << "K: " << k_ << std::endl;
-    raft::print_device_vector("Labels array: ", search_labels_.data(), rows_, std::cout);
-    // #endif
-
-    std::vector<device_matrix_view<const T, IdxT, row_major>> index = {
-      make_device_matrix_view((const T*)(input_.data()), rows_, cols_)};
-    auto search = raft::make_device_matrix_view<const T, IdxT, row_major>(
-      (const T*)(search_data_.data()), rows_, cols_);
-
-    auto indices = raft::make_device_matrix_view<IdxT, IdxT, row_major>(indices_.data(), rows_, k_);
-    auto distances =
-      raft::make_device_matrix_view<T, IdxT, row_major>(distances_.data(), rows_, k_);
-
-    auto metric = cuvs::distance::DistanceType::L2Unexpanded;
-    knn(handle, index, search, indices, distances, metric, std::make_optional<IdxT>(0));
-
-    build_actual_output<<<raft::ceildiv(rows_ * k_, 32), 32, 0, stream>>>(
-      actual_labels_.data(), rows_, k_, search_labels_.data(), indices_.data());
-
-    build_expected_output<<<raft::ceildiv(rows_ * k_, 32), 32, 0, stream>>>(
-      expected_labels_.data(), rows_, k_, search_labels_.data());
-
-    ASSERT_TRUE(devArrMatch(
-      expected_labels_.data(), actual_labels_.data(), rows_ * k_, raft::Compare<int>(), stream));
-  }
-
-  void SetUp() override
-  {
-    rows_ = params_.input.size();
-    cols_ = params_.input[0].size();
-    k_    = params_.k;
-
-    actual_labels_.resize(rows_ * k_, stream);
-    expected_labels_.resize(rows_ * k_, stream);
-    input_.resize(rows_ * cols_, stream);
-    search_data_.resize(rows_ * cols_, stream);
-    indices_.resize(rows_ * k_, stream);
-    distances_.resize(rows_ * k_, stream);
-    search_labels_.resize(rows_, stream);
-
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(actual_labels_.data(), 0, actual_labels_.size() * sizeof(int), stream));
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(expected_labels_.data(), 0, expected_labels_.size() * sizeof(int), stream));
-    RAFT_CUDA_TRY(cudaMemsetAsync(input_.data(), 0, input_.size() * sizeof(float), stream));
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(search_data_.data(), 0, search_data_.size() * sizeof(float), stream));
-    RAFT_CUDA_TRY(cudaMemsetAsync(indices_.data(), 0, indices_.size() * sizeof(IdxT), stream));
-    RAFT_CUDA_TRY(cudaMemsetAsync(distances_.data(), 0, distances_.size() * sizeof(float), stream));
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(search_labels_.data(), 0, search_labels_.size() * sizeof(int), stream));
-
-    std::vector<float> row_major_input;
-    for (std::size_t i = 0; i < params_.input.size(); ++i) {
-      for (std::size_t j = 0; j < params_.input[i].size(); ++j) {
-        row_major_input.push_back(params_.input[i][j]);
-      }
-    }
-    rmm::device_buffer input_d =
-      rmm::device_buffer(row_major_input.data(), row_major_input.size() * sizeof(float), stream);
-    float* input_ptr = static_cast<float*>(input_d.data());
-
-    rmm::device_buffer labels_d =
-      rmm::device_buffer(params_.labels.data(), params_.labels.size() * sizeof(int), stream);
-    int* labels_ptr = static_cast<int*>(labels_d.data());
-
-    raft::copy(input_.data(), input_ptr, rows_ * cols_, stream);
-    raft::copy(search_data_.data(), input_ptr, rows_ * cols_, stream);
-    raft::copy(search_labels_.data(), labels_ptr, rows_, stream);
-    resource::sync_stream(handle, stream);
-  }
-
- private:
-  raft::resources handle;
-  cudaStream_t stream;
-
-  KNNInputs params_;
-  int rows_;
-  int cols_;
-  rmm::device_uvector<float> input_;
-  rmm::device_uvector<float> search_data_;
-  rmm::device_uvector<IdxT> indices_;
-  rmm::device_uvector<float> distances_;
-  int k_;
-
-  rmm::device_uvector<int> search_labels_;
-  rmm::device_uvector<int> actual_labels_;
-  rmm::device_uvector<int> expected_labels_;
-};
-
-const std::vector<KNNInputs> inputs = {
-  // 2D
-  {{
-     {2.7810836, 2.550537003},
-     {1.465489372, 2.362125076},
-     {3.396561688, 4.400293529},
-     {1.38807019, 1.850220317},
-     {3.06407232, 3.005305973},
-     {7.627531214, 2.759262235},
-     {5.332441248, 2.088626775},
-     {6.922596716, 1.77106367},
-     {8.675418651, -0.242068655},
-     {7.673756466, 3.508563011},
-   },
-   2,
-   {0, 0, 0, 0, 0, 1, 1, 1, 1, 1}}};
-
-typedef KNNTest<float, int> KNNTestFint32_t;
-TEST_P(KNNTestFint32_t, BruteForce) { this->testBruteForce(); }
-typedef KNNTest<float, uint32_t> KNNTestFuint32_t;
-TEST_P(KNNTestFuint32_t, BruteForce) { this->testBruteForce(); }
-
-INSTANTIATE_TEST_CASE_P(KNNTest, KNNTestFint32_t, ::testing::ValuesIn(inputs));
-INSTANTIATE_TEST_CASE_P(KNNTest, KNNTestFuint32_t, ::testing::ValuesIn(inputs));
-
-}  // namespace cuvs::neighbors::brute_force
diff --git a/cpp/test/neighbors/knn_utils.cuh b/cpp/test/neighbors/knn_utils.cuh
deleted file mode 100644
index b9df1c173..000000000
--- a/cpp/test/neighbors/knn_utils.cuh
+++ /dev/null
@@ -1,94 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#pragma once
-
-#include <memory>
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-
-#include <raft/util/cudart_utils.hpp>
-
-namespace cuvs::spatial::knn {
-template <typename IdxT, typename DistT, typename compareDist>
-struct idx_dist_pair {
-  IdxT idx;
-  DistT dist;
-  compareDist eq_compare;
-  bool operator==(const idx_dist_pair<IdxT, DistT, compareDist>& a) const
-  {
-    if (idx == a.idx) return true;
-    if (eq_compare(dist, a.dist)) return true;
-    return false;
-  }
-  idx_dist_pair(IdxT x, DistT y, compareDist op) : idx(x), dist(y), eq_compare(op) {}
-};
-
-template <typename T, typename DistT>
-testing::AssertionResult devArrMatchKnnPair(const T* expected_idx,
-                                            const T* actual_idx,
-                                            const DistT* expected_dist,
-                                            const DistT* actual_dist,
-                                            size_t rows,
-                                            size_t cols,
-                                            const DistT eps,
-                                            cudaStream_t stream = 0,
-                                            bool sort_inputs    = false)
-{
-  size_t size = rows * cols;
-  std::unique_ptr<T[]> exp_idx_h(new T[size]);
-  std::unique_ptr<T[]> act_idx_h(new T[size]);
-  std::unique_ptr<DistT[]> exp_dist_h(new DistT[size]);
-  std::unique_ptr<DistT[]> act_dist_h(new DistT[size]);
-  raft::update_host<T>(exp_idx_h.get(), expected_idx, size, stream);
-  raft::update_host<T>(act_idx_h.get(), actual_idx, size, stream);
-  raft::update_host<DistT>(exp_dist_h.get(), expected_dist, size, stream);
-  raft::update_host<DistT>(act_dist_h.get(), actual_dist, size, stream);
-
-  RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
-  for (size_t i(0); i < rows; ++i) {
-    std::vector<std::pair<DistT, T>> actual;
-    std::vector<std::pair<DistT, T>> expected;
-    for (size_t j(0); j < cols; ++j) {
-      auto idx      = i * cols + j;  // row major assumption!
-      auto exp_idx  = exp_idx_h.get()[idx];
-      auto act_idx  = act_idx_h.get()[idx];
-      auto exp_dist = exp_dist_h.get()[idx];
-      auto act_dist = act_dist_h.get()[idx];
-      actual.push_back(std::make_pair(act_dist, act_idx));
-      expected.push_back(std::make_pair(exp_dist, exp_idx));
-    }
-    if (sort_inputs) {
-      // inputs could be unsorted here, sort for comparison
-      std::sort(actual.begin(), actual.end());
-      std::sort(expected.begin(), expected.end());
-    }
-    for (size_t j(0); j < cols; ++j) {
-      auto act = actual[j];
-      auto exp = expected[j];
-      idx_dist_pair exp_kvp(exp.second, exp.first, raft::CompareApprox<DistT>(eps));
-      idx_dist_pair act_kvp(act.second, act.first, raft::CompareApprox<DistT>(eps));
-      if (!(exp_kvp == act_kvp)) {
-        return testing::AssertionFailure()
-               << "actual=" << act_kvp.idx << "," << act_kvp.dist << "!="
-               << "expected" << exp_kvp.idx << "," << exp_kvp.dist << " @" << i << "," << j;
-      }
-    }
-  }
-  return testing::AssertionSuccess();
-}
-}  // namespace cuvs::spatial::knn
diff --git a/cpp/test/neighbors/refine.cu b/cpp/test/neighbors/refine.cu
deleted file mode 100644
index 953770ebf..000000000
--- a/cpp/test/neighbors/refine.cu
+++ /dev/null
@@ -1,129 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "ann_utils.cuh"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cuvs_internal/neighbors/refine_helper.cuh>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/detail/refine.cuh>
-#include <cuvs/neighbors/refine.cuh>
-#include <cuvs/spatial/knn/ann.cuh>
-#include <raft/core/logger.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/util/itertools.hpp>
-
-#include <rmm/cuda_stream_view.hpp>
-
-#include <gtest/gtest.h>
-
-#include <vector>
-
-namespace cuvs::neighbors {
-
-template <typename DataT, typename DistanceT, typename IdxT>
-class RefineTest : public ::testing::TestWithParam<RefineInputs<IdxT>> {
- public:
-  RefineTest()
-    : stream_(resource::get_cuda_stream(handle_)),
-      data(handle_, ::testing::TestWithParam<RefineInputs<IdxT>>::GetParam())
-  {
-  }
-
- protected:
- public:  // tamas remove
-  void testRefine()
-  {
-    std::vector<IdxT> indices(data.p.n_queries * data.p.k);
-    std::vector<DistanceT> distances(data.p.n_queries * data.p.k);
-
-    if (data.p.host_data) {
-      cuvs::neighbors::refine<IdxT, DataT, DistanceT, IdxT>(handle_,
-                                                            data.dataset_host.view(),
-                                                            data.queries_host.view(),
-                                                            data.candidates_host.view(),
-                                                            data.refined_indices_host.view(),
-                                                            data.refined_distances_host.view(),
-                                                            data.p.metric);
-      raft::copy(indices.data(),
-                 data.refined_indices_host.data_handle(),
-                 data.refined_indices_host.size(),
-                 stream_);
-      raft::copy(distances.data(),
-                 data.refined_distances_host.data_handle(),
-                 data.refined_distances_host.size(),
-                 stream_);
-
-    } else {
-      cuvs::neighbors::refine<IdxT, DataT, DistanceT, IdxT>(handle_,
-                                                            data.dataset.view(),
-                                                            data.queries.view(),
-                                                            data.candidates.view(),
-                                                            data.refined_indices.view(),
-                                                            data.refined_distances.view(),
-                                                            data.p.metric);
-      update_host(distances.data(),
-                  data.refined_distances.data_handle(),
-                  data.refined_distances.size(),
-                  stream_);
-      update_host(
-        indices.data(), data.refined_indices.data_handle(), data.refined_indices.size(), stream_);
-    }
-    resource::sync_stream(handle_);
-
-    double min_recall = 1;
-
-    ASSERT_TRUE(cuvs::neighbors::eval_neighbours(data.true_refined_indices_host,
-                                                 indices,
-                                                 data.true_refined_distances_host,
-                                                 distances,
-                                                 data.p.n_queries,
-                                                 data.p.k,
-                                                 0.001,
-                                                 min_recall));
-  }
-
- public:
-  raft::resources handle_;
-  rmm::cuda_stream_view stream_;
-  RefineHelper<DataT, DistanceT, IdxT> data;
-};
-
-const std::vector<RefineInputs<int64_t>> inputs =
-  raft::util::itertools::product<RefineInputs<int64_t>>(
-    {static_cast<int64_t>(137)},
-    {static_cast<int64_t>(1000)},
-    {static_cast<int64_t>(16)},
-    {static_cast<int64_t>(1), static_cast<int64_t>(10), static_cast<int64_t>(33)},
-    {static_cast<int64_t>(33)},
-    {cuvs::distance::DistanceType::L2Expanded, cuvs::distance::DistanceType::InnerProduct},
-    {false, true});
-
-typedef RefineTest<float, float, std::int64_t> RefineTestF;
-TEST_P(RefineTestF, AnnRefine) { this->testRefine(); }
-
-INSTANTIATE_TEST_CASE_P(RefineTest, RefineTestF, ::testing::ValuesIn(inputs));
-
-typedef RefineTest<uint8_t, float, std::int64_t> RefineTestF_uint8;
-TEST_P(RefineTestF_uint8, AnnRefine) { this->testRefine(); }
-INSTANTIATE_TEST_CASE_P(RefineTest, RefineTestF_uint8, ::testing::ValuesIn(inputs));
-
-typedef RefineTest<int8_t, float, std::int64_t> RefineTestF_int8;
-TEST_P(RefineTestF_int8, AnnRefine) { this->testRefine(); }
-INSTANTIATE_TEST_CASE_P(RefineTest, RefineTestF_int8, ::testing::ValuesIn(inputs));
-}  // namespace cuvs::neighbors
diff --git a/cpp/test/neighbors/selection.cu b/cpp/test/neighbors/selection.cu
deleted file mode 100644
index 2a6e77ce0..000000000
--- a/cpp/test/neighbors/selection.cu
+++ /dev/null
@@ -1,499 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <algorithm>
-#include <cuvs/neighbors/detail/selection_faiss.cuh>
-#include <cuvs/neighbors/detail/selection_faiss_helpers.cuh>  // kFaissMax
-#include <gtest/gtest.h>
-#include <numeric>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/random/rng.cuh>
-#include <raft/util/cudart_utils.hpp>
-
-#include "../test_utils.cuh"
-
-#include <cuvs/spatial/knn/knn.cuh>
-#include <raft/sparse/detail/utils.h>
-
-namespace cuvs::spatial::selection {
-
-using namespace raft;
-using namespace raft::sparse;
-
-struct SelectTestSpec {
-  int n_inputs;
-  int input_len;
-  int k;
-  int select_min;
-  bool use_index_input = true;
-};
-
-std::ostream& operator<<(std::ostream& os, const SelectTestSpec& ss)
-{
-  os << "spec{size: " << ss.input_len << "*" << ss.n_inputs << ", k: " << ss.k;
-  os << (ss.select_min ? "; min}" : "; max}");
-  return os;
-}
-
-template <typename IdxT>
-auto gen_simple_ids(int n_inputs, int input_len, const raft::resources& handle) -> std::vector<IdxT>
-{
-  std::vector<IdxT> out(n_inputs * input_len);
-  auto s = resource::get_cuda_stream(handle);
-  rmm::device_uvector<IdxT> out_d(out.size(), s);
-  iota_fill(out_d.data(), IdxT(n_inputs), IdxT(input_len), s);
-  update_host(out.data(), out_d.data(), out.size(), s);
-  s.synchronize();
-  return out;
-}
-
-template <typename KeyT, typename IdxT>
-struct SelectInOutSimple {
- public:
-  bool not_supported = false;
-
-  SelectInOutSimple(std::shared_ptr<raft::resources> handle,
-                    const SelectTestSpec& spec,
-                    const std::vector<KeyT>& in_dists,
-                    const std::vector<KeyT>& out_dists,
-                    const std::vector<IdxT>& out_ids)
-    : in_dists_(in_dists),
-      in_ids_(gen_simple_ids<IdxT>(spec.n_inputs, spec.input_len, *handle.get())),
-      out_dists_(out_dists),
-      out_ids_(out_ids),
-      handle_(handle)
-  {
-  }
-
-  auto get_in_dists() -> std::vector<KeyT>& { return in_dists_; }
-  auto get_in_ids() -> std::vector<IdxT>& { return in_ids_; }
-  auto get_out_dists() -> std::vector<KeyT>& { return out_dists_; }
-  auto get_out_ids() -> std::vector<IdxT>& { return out_ids_; }
-
- private:
-  std::shared_ptr<raft::resources> handle_;
-  std::vector<KeyT> in_dists_;
-  std::vector<IdxT> in_ids_;
-  std::vector<KeyT> out_dists_;
-  std::vector<IdxT> out_ids_;
-};
-
-template <typename KeyT, typename IdxT>
-struct SelectInOutComputed {
- public:
-  bool not_supported = false;
-
-  SelectInOutComputed(std::shared_ptr<raft::resources> handle,
-                      const SelectTestSpec& spec,
-                      knn::SelectKAlgo algo,
-                      const std::vector<KeyT>& in_dists,
-                      const std::optional<std::vector<IdxT>>& in_ids = std::nullopt)
-    : handle_(handle),
-      in_dists_(in_dists),
-      in_ids_(in_ids.value_or(gen_simple_ids<IdxT>(spec.n_inputs, spec.input_len, *handle.get()))),
-      out_dists_(spec.n_inputs * spec.k),
-      out_ids_(spec.n_inputs * spec.k)
-
-  {
-    // check if the size is supported by the algorithm
-    switch (algo) {
-      case knn::SelectKAlgo::WARP_SORT:
-        if (spec.k > raft::matrix::detail::select::warpsort::kMaxCapacity) {
-          not_supported = true;
-          return;
-        }
-        break;
-      case knn::SelectKAlgo::FAISS:
-        if (spec.k > cuvs::neighbors::detail::kFaissMaxK<IdxT, KeyT>()) {
-          not_supported = true;
-          return;
-        }
-        break;
-      default: break;
-    }
-
-    auto stream = resource::get_cuda_stream(*handle_);
-
-    rmm::device_uvector<KeyT> in_dists_d(in_dists_.size(), stream);
-    rmm::device_uvector<IdxT> in_ids_d(in_ids_.size(), stream);
-    rmm::device_uvector<KeyT> out_dists_d(out_dists_.size(), stream);
-    rmm::device_uvector<IdxT> out_ids_d(out_ids_.size(), stream);
-
-    update_device(in_dists_d.data(), in_dists_.data(), in_dists_.size(), stream);
-    update_device(in_ids_d.data(), in_ids_.data(), in_ids_.size(), stream);
-
-    cuvs::spatial::knn::select_k<IdxT, KeyT>(in_dists_d.data(),
-                                             spec.use_index_input ? in_ids_d.data() : nullptr,
-                                             spec.n_inputs,
-                                             spec.input_len,
-                                             out_dists_d.data(),
-                                             out_ids_d.data(),
-                                             spec.select_min,
-                                             spec.k,
-                                             stream,
-                                             algo);
-
-    update_host(out_dists_.data(), out_dists_d.data(), out_dists_.size(), stream);
-    update_host(out_ids_.data(), out_ids_d.data(), out_ids_.size(), stream);
-
-    interruptible::synchronize(stream);
-
-    auto p = topk_sort_permutation(out_dists_, out_ids_, spec.k, spec.select_min);
-    apply_permutation(out_dists_, p);
-    apply_permutation(out_ids_, p);
-  }
-
-  auto get_in_dists() -> std::vector<KeyT>& { return in_dists_; }
-  auto get_in_ids() -> std::vector<IdxT>& { return in_ids_; }
-  auto get_out_dists() -> std::vector<KeyT>& { return out_dists_; }
-  auto get_out_ids() -> std::vector<IdxT>& { return out_ids_; }
-
- private:
-  std::shared_ptr<raft::resources> handle_;
-  std::vector<KeyT> in_dists_;
-  std::vector<IdxT> in_ids_;
-  std::vector<KeyT> out_dists_;
-  std::vector<IdxT> out_ids_;
-
-  auto topk_sort_permutation(const std::vector<KeyT>& vec,
-                             const std::vector<IdxT>& inds,
-                             int k,
-                             bool select_min) -> std::vector<IdxT>
-  {
-    std::vector<IdxT> p(vec.size());
-    std::iota(p.begin(), p.end(), 0);
-    if (select_min) {
-      std::sort(p.begin(), p.end(), [&vec, &inds, k](IdxT i, IdxT j) {
-        const IdxT ik = i / k;
-        const IdxT jk = j / k;
-        if (ik == jk) {
-          if (vec[i] == vec[j]) { return inds[i] < inds[j]; }
-          return vec[i] < vec[j];
-        }
-        return ik < jk;
-      });
-    } else {
-      std::sort(p.begin(), p.end(), [&vec, &inds, k](IdxT i, IdxT j) {
-        const IdxT ik = i / k;
-        const IdxT jk = j / k;
-        if (ik == jk) {
-          if (vec[i] == vec[j]) { return inds[i] < inds[j]; }
-          return vec[i] > vec[j];
-        }
-        return ik < jk;
-      });
-    }
-    return p;
-  }
-
-  template <typename T>
-  void apply_permutation(std::vector<T>& vec, const std::vector<IdxT>& p)
-  {
-    for (auto i = IdxT(vec.size()) - 1; i > 0; i--) {
-      auto j = p[i];
-      while (j > i)
-        j = p[j];
-      std::swap(vec[j], vec[i]);
-    }
-  }
-};
-
-template <typename InOut>
-using Params =
-  std::tuple<SelectTestSpec, knn::SelectKAlgo, InOut, std::shared_ptr<raft::resources>>;
-
-template <typename KeyT, typename IdxT, template <typename, typename> typename ParamsReader>
-class SelectionTest : public testing::TestWithParam<typename ParamsReader<KeyT, IdxT>::ParamsIn> {
- protected:
-  std::shared_ptr<raft::resources> handle_;
-  const SelectTestSpec spec;
-  const knn::SelectKAlgo algo;
-
-  typename ParamsReader<KeyT, IdxT>::InOut ref;
-  SelectInOutComputed<KeyT, IdxT> res;
-
- public:
-  explicit SelectionTest(Params<typename ParamsReader<KeyT, IdxT>::InOut> ps)
-    : handle_(std::get<3>(ps)),
-      spec(std::get<0>(ps)),
-      algo(std::get<1>(ps)),
-      ref(std::get<2>(ps)),
-      res(handle_, spec, algo, ref.get_in_dists(), ref.get_in_ids())
-  {
-  }
-
-  explicit SelectionTest(typename ParamsReader<KeyT, IdxT>::ParamsIn ps)
-    : SelectionTest(ParamsReader<KeyT, IdxT>::read(ps))
-  {
-  }
-
-  SelectionTest()
-    : SelectionTest(testing::TestWithParam<typename ParamsReader<KeyT, IdxT>::ParamsIn>::GetParam())
-  {
-  }
-
-  void run()
-  {
-    if (ref.not_supported || res.not_supported) { GTEST_SKIP(); }
-
-    ASSERT_TRUE(hostVecMatch(ref.get_out_dists(), res.get_out_dists(), Compare<KeyT>()));
-    // If the dists (keys) are the same, different corresponding ids may end up in the selection due
-    // to non-deterministic nature of some implementations.
-    auto& in_ids   = ref.get_in_ids();
-    auto& in_dists = ref.get_in_dists();
-
-    auto compare_ids = [&in_ids, &in_dists](const IdxT& i, const IdxT& j) {
-      if (i == j) return true;
-      auto ix_i = size_t(std::find(in_ids.begin(), in_ids.end(), i) - in_ids.begin());
-      auto ix_j = size_t(std::find(in_ids.begin(), in_ids.end(), j) - in_ids.begin());
-      if (ix_i >= in_ids.size() || ix_j >= in_ids.size()) return false;
-      auto dist_i = in_dists[ix_i];
-      auto dist_j = in_dists[ix_j];
-      if (dist_i == dist_j) return true;
-      std::cout << "ERROR: ref[" << ix_i << "] = " << dist_i << " != "
-                << "res[" << ix_j << "] = " << dist_j << std::endl;
-      return false;
-    };
-    ASSERT_TRUE(hostVecMatch(ref.get_out_ids(), res.get_out_ids(), compare_ids));
-  }
-};
-
-template <typename KeyT, typename IdxT>
-struct params_simple {
-  using InOut = SelectInOutSimple<KeyT, IdxT>;
-  using Inputs =
-    std::tuple<SelectTestSpec, std::vector<KeyT>, std::vector<KeyT>, std::vector<IdxT>>;
-  using Handle   = std::shared_ptr<raft::resources>;
-  using ParamsIn = std::tuple<Inputs, knn::SelectKAlgo, Handle>;
-
-  static auto read(ParamsIn ps) -> Params<InOut>
-  {
-    auto ins    = std::get<0>(ps);
-    auto algo   = std::get<1>(ps);
-    auto handle = std::get<2>(ps);
-    return std::make_tuple(
-      std::get<0>(ins),
-      algo,
-      SelectInOutSimple<KeyT, IdxT>(
-        handle, std::get<0>(ins), std::get<1>(ins), std::get<2>(ins), std::get<3>(ins)),
-      handle);
-  }
-};
-
-auto inputs_simple_f = testing::Values(
-  params_simple<float, int>::Inputs(
-    {5, 5, 5, true, true},
-    {5.0, 4.0, 3.0, 2.0, 1.0, 1.0, 2.0, 3.0, 4.0, 5.0, 2.0, 3.0, 5.0,
-     1.0, 4.0, 5.0, 3.0, 2.0, 4.0, 1.0, 1.0, 3.0, 2.0, 5.0, 4.0},
-    {1.0, 2.0, 3.0, 4.0, 5.0, 1.0, 2.0, 3.0, 4.0, 5.0, 1.0, 2.0, 3.0,
-     4.0, 5.0, 1.0, 2.0, 3.0, 4.0, 5.0, 1.0, 2.0, 3.0, 4.0, 5.0},
-    {4, 3, 2, 1, 0, 0, 1, 2, 3, 4, 3, 0, 1, 4, 2, 4, 2, 1, 3, 0, 0, 2, 1, 4, 3}),
-  params_simple<float, int>::Inputs(
-    {5, 5, 3, true, true},
-    {5.0, 4.0, 3.0, 2.0, 1.0, 1.0, 2.0, 3.0, 4.0, 5.0, 2.0, 3.0, 5.0,
-     1.0, 4.0, 5.0, 3.0, 2.0, 4.0, 1.0, 1.0, 3.0, 2.0, 5.0, 4.0},
-    {1.0, 2.0, 3.0, 1.0, 2.0, 3.0, 1.0, 2.0, 3.0, 1.0, 2.0, 3.0, 1.0, 2.0, 3.0},
-    {4, 3, 2, 0, 1, 2, 3, 0, 1, 4, 2, 1, 0, 2, 1}),
-  params_simple<float, int>::Inputs(
-    {5, 5, 5, true, false},
-    {5.0, 4.0, 3.0, 2.0, 1.0, 1.0, 2.0, 3.0, 4.0, 5.0, 2.0, 3.0, 5.0,
-     1.0, 4.0, 5.0, 3.0, 2.0, 4.0, 1.0, 1.0, 3.0, 2.0, 5.0, 4.0},
-    {1.0, 2.0, 3.0, 4.0, 5.0, 1.0, 2.0, 3.0, 4.0, 5.0, 1.0, 2.0, 3.0,
-     4.0, 5.0, 1.0, 2.0, 3.0, 4.0, 5.0, 1.0, 2.0, 3.0, 4.0, 5.0},
-    {4, 3, 2, 1, 0, 0, 1, 2, 3, 4, 3, 0, 1, 4, 2, 4, 2, 1, 3, 0, 0, 2, 1, 4, 3}),
-  params_simple<float, int>::Inputs(
-    {5, 5, 3, true, false},
-    {5.0, 4.0, 3.0, 2.0, 1.0, 1.0, 2.0, 3.0, 4.0, 5.0, 2.0, 3.0, 5.0,
-     1.0, 4.0, 5.0, 3.0, 2.0, 4.0, 1.0, 1.0, 3.0, 2.0, 5.0, 4.0},
-    {1.0, 2.0, 3.0, 1.0, 2.0, 3.0, 1.0, 2.0, 3.0, 1.0, 2.0, 3.0, 1.0, 2.0, 3.0},
-    {4, 3, 2, 0, 1, 2, 3, 0, 1, 4, 2, 1, 0, 2, 1}),
-  params_simple<float, int>::Inputs(
-    {5, 7, 3, true, true},
-    {5.0, 4.0, 3.0, 2.0, 1.3, 7.5, 19.0, 9.0, 2.0, 3.0, 3.0, 5.0, 6.0, 4.0, 2.0, 3.0, 5.0, 1.0,
-     4.0, 1.0, 1.0, 5.0, 7.0, 2.5, 4.0,  7.0, 8.0, 8.0, 1.0, 3.0, 2.0, 5.0, 4.0, 1.1, 1.2},
-    {1.3, 2.0, 3.0, 2.0, 3.0, 3.0, 1.0, 1.0, 1.0, 2.5, 4.0, 5.0, 1.0, 1.1, 1.2},
-    {4, 3, 2, 1, 2, 3, 3, 5, 6, 2, 3, 0, 0, 5, 6}),
-  params_simple<float, int>::Inputs(
-    {1, 7, 3, true, true}, {2.0, 3.0, 5.0, 1.0, 4.0, 1.0, 1.0}, {1.0, 1.0, 1.0}, {3, 5, 6}),
-  params_simple<float, int>::Inputs(
-    {1, 7, 3, false, false}, {2.0, 3.0, 5.0, 1.0, 4.0, 1.0, 1.0}, {5.0, 4.0, 3.0}, {2, 4, 1}),
-  params_simple<float, int>::Inputs(
-    {1, 7, 3, false, true}, {2.0, 3.0, 5.0, 9.0, 4.0, 9.0, 9.0}, {9.0, 9.0, 9.0}, {3, 5, 6}),
-  params_simple<float, int>::Inputs(
-    {1, 130, 5, false, true},
-    {19, 1, 0, 1, 0, 1,  0,  1,  0,  1,  0,  1,  0,  1,  0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1,
-     0,  1, 0, 1, 0, 1,  0,  1,  0,  1,  0,  1,  0,  1,  0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1,
-     0,  1, 0, 1, 0, 1,  0,  1,  1,  2,  1,  2,  1,  2,  1, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1, 2,
-     1,  2, 1, 2, 1, 2,  1,  2,  1,  2,  1,  2,  1,  2,  1, 2, 1, 2, 1, 2, 1, 2, 1, 2, 3, 4,
-     5,  6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 4, 4, 2, 3, 2, 3, 2, 3, 2, 3, 2, 20},
-    {20, 19, 18, 17, 16},
-    {129, 0, 117, 116, 115}),
-  params_simple<float, int>::Inputs(
-    {1, 130, 15, false, true},
-    {19, 1, 0, 1, 0, 1,  0,  1,  0,  1,  0,  1,  0,  1,  0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1,
-     0,  1, 0, 1, 0, 1,  0,  1,  0,  1,  0,  1,  0,  1,  0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1,
-     0,  1, 0, 1, 0, 1,  0,  1,  1,  2,  1,  2,  1,  2,  1, 2, 1, 2, 1, 2, 1, 2, 1, 2, 1, 2,
-     1,  2, 1, 2, 1, 2,  1,  2,  1,  2,  1,  2,  1,  2,  1, 2, 1, 2, 1, 2, 1, 2, 1, 2, 3, 4,
-     5,  6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 4, 4, 2, 3, 2, 3, 2, 3, 2, 3, 2, 20},
-    {20, 19, 18, 17, 16, 15, 14, 13, 12, 11, 10, 9, 8, 7, 6},
-    {129, 0, 117, 116, 115, 114, 113, 112, 111, 110, 109, 108, 107, 106, 105}));
-
-typedef SelectionTest<float, int, params_simple> SimpleFloatInt;
-TEST_P(SimpleFloatInt, Run) { run(); }
-INSTANTIATE_TEST_CASE_P(SelectionTest,
-                        SimpleFloatInt,
-                        testing::Combine(inputs_simple_f,
-                                         testing::Values(knn::SelectKAlgo::FAISS,
-                                                         knn::SelectKAlgo::RADIX_8_BITS,
-                                                         knn::SelectKAlgo::RADIX_11_BITS,
-                                                         knn::SelectKAlgo::WARP_SORT),
-                                         testing::Values(std::make_shared<raft::resources>())));
-
-template <knn::SelectKAlgo RefAlgo>
-struct with_ref {
-  template <typename KeyT, typename IdxT>
-  struct params_random {
-    using InOut    = SelectInOutComputed<KeyT, IdxT>;
-    using Handle   = std::shared_ptr<raft::resources>;
-    using ParamsIn = std::tuple<SelectTestSpec, knn::SelectKAlgo, Handle>;
-
-    static auto read(ParamsIn ps) -> Params<InOut>
-    {
-      auto spec   = std::get<0>(ps);
-      auto algo   = std::get<1>(ps);
-      auto handle = std::get<2>(ps);
-
-      std::vector<KeyT> dists(spec.input_len * spec.n_inputs);
-
-      {
-        auto s = resource::get_cuda_stream(*handle);
-        rmm::device_uvector<KeyT> dists_d(spec.input_len * spec.n_inputs, s);
-        raft::random::RngState r(42);
-        normal(*(handle.get()), r, dists_d.data(), dists_d.size(), KeyT(10.0), KeyT(100.0));
-        update_host(dists.data(), dists_d.data(), dists_d.size(), s);
-        s.synchronize();
-      }
-
-      return std::make_tuple(
-        spec, algo, SelectInOutComputed<KeyT, IdxT>(handle, spec, RefAlgo, dists), handle);
-    }
-  };
-};
-
-auto inputs_random_longlist = testing::Values(SelectTestSpec{1, 130, 15, false},
-                                              SelectTestSpec{1, 128, 15, false},
-                                              SelectTestSpec{20, 700, 1, true},
-                                              SelectTestSpec{20, 700, 2, true},
-                                              SelectTestSpec{20, 700, 3, true},
-                                              SelectTestSpec{20, 700, 4, true},
-                                              SelectTestSpec{20, 700, 5, true},
-                                              SelectTestSpec{20, 700, 6, true},
-                                              SelectTestSpec{20, 700, 7, true},
-                                              SelectTestSpec{20, 700, 8, true},
-                                              SelectTestSpec{20, 700, 9, true},
-                                              SelectTestSpec{20, 700, 10, true, false},
-                                              SelectTestSpec{20, 700, 11, true},
-                                              SelectTestSpec{20, 700, 12, true},
-                                              SelectTestSpec{20, 700, 16, true},
-                                              SelectTestSpec{100, 1700, 17, true},
-                                              SelectTestSpec{100, 1700, 31, true, false},
-                                              SelectTestSpec{100, 1700, 32, false},
-                                              SelectTestSpec{100, 1700, 33, false},
-                                              SelectTestSpec{100, 1700, 63, false},
-                                              SelectTestSpec{100, 1700, 64, false, false},
-                                              SelectTestSpec{100, 1700, 65, false},
-                                              SelectTestSpec{100, 1700, 255, true},
-                                              SelectTestSpec{100, 1700, 256, true},
-                                              SelectTestSpec{100, 1700, 511, false},
-                                              SelectTestSpec{100, 1700, 512, true},
-                                              SelectTestSpec{100, 1700, 1023, false, false},
-                                              SelectTestSpec{100, 1700, 1024, true},
-                                              SelectTestSpec{100, 1700, 1700, true});
-
-auto inputs_random_largesize = testing::Values(SelectTestSpec{100, 100000, 1, true},
-                                               SelectTestSpec{100, 100000, 2, true},
-                                               SelectTestSpec{100, 100000, 3, true, false},
-                                               SelectTestSpec{100, 100000, 7, true},
-                                               SelectTestSpec{100, 100000, 16, true},
-                                               SelectTestSpec{100, 100000, 31, true},
-                                               SelectTestSpec{100, 100000, 32, true, false},
-                                               SelectTestSpec{100, 100000, 60, true},
-                                               SelectTestSpec{100, 100000, 100, true, false},
-                                               SelectTestSpec{100, 100000, 200, true},
-                                               SelectTestSpec{100000, 100, 100, false},
-                                               SelectTestSpec{1, 100000000, 1, true},
-                                               SelectTestSpec{1, 100000000, 16, false, false},
-                                               SelectTestSpec{1, 100000000, 64, false},
-                                               SelectTestSpec{1, 100000000, 128, true, false},
-                                               SelectTestSpec{1, 100000000, 256, false, false});
-
-auto inputs_random_largek = testing::Values(SelectTestSpec{100, 100000, 1000, true},
-                                            SelectTestSpec{100, 100000, 2000, false},
-                                            SelectTestSpec{100, 100000, 100000, true, false},
-                                            SelectTestSpec{100, 100000, 2048, false},
-                                            SelectTestSpec{100, 100000, 1237, true});
-
-typedef SelectionTest<float, int, with_ref<knn::SelectKAlgo::FAISS>::params_random>
-  ReferencedRandomFloatInt;
-TEST_P(ReferencedRandomFloatInt, Run) { run(); }
-INSTANTIATE_TEST_CASE_P(SelectionTest,
-                        ReferencedRandomFloatInt,
-                        testing::Combine(inputs_random_longlist,
-                                         testing::Values(knn::SelectKAlgo::RADIX_8_BITS,
-                                                         knn::SelectKAlgo::RADIX_11_BITS,
-                                                         knn::SelectKAlgo::WARP_SORT),
-                                         testing::Values(std::make_shared<raft::resources>())));
-
-typedef SelectionTest<double, size_t, with_ref<knn::SelectKAlgo::FAISS>::params_random>
-  ReferencedRandomDoubleSizeT;
-TEST_P(ReferencedRandomDoubleSizeT, Run) { run(); }
-INSTANTIATE_TEST_CASE_P(SelectionTest,
-                        ReferencedRandomDoubleSizeT,
-                        testing::Combine(inputs_random_longlist,
-                                         testing::Values(knn::SelectKAlgo::RADIX_8_BITS,
-                                                         knn::SelectKAlgo::RADIX_11_BITS,
-                                                         knn::SelectKAlgo::WARP_SORT),
-                                         testing::Values(std::make_shared<raft::resources>())));
-
-typedef SelectionTest<double, int, with_ref<knn::SelectKAlgo::FAISS>::params_random>
-  ReferencedRandomDoubleInt;
-TEST_P(ReferencedRandomDoubleInt, LargeSize) { run(); }
-INSTANTIATE_TEST_CASE_P(SelectionTest,
-                        ReferencedRandomDoubleInt,
-                        testing::Combine(inputs_random_largesize,
-                                         testing::Values(knn::SelectKAlgo::WARP_SORT),
-                                         testing::Values(std::make_shared<raft::resources>())));
-
-/** TODO: Fix test failure in RAFT CI
- *
- *  SelectionTest/ReferencedRandomFloatSizeT.LargeK/0
- *  Indicices do not match! ref[91628] = 131.359 != res[36504] = 158.438
- *  Actual: false (actual=36504 != expected=91628 @38999;
- *
- *  SelectionTest/ReferencedRandomFloatSizeT.LargeK/1
- *  ERROR: ref[57977] = 58.9079 != res[21973] = 54.9354
- *  Actual: false (actual=21973 != expected=57977 @107999;
- *
- */
-typedef SelectionTest<float, size_t, with_ref<knn::SelectKAlgo::RADIX_11_BITS>::params_random>
-  ReferencedRandomFloatSizeT;
-TEST_P(ReferencedRandomFloatSizeT, LargeK) { run(); }
-INSTANTIATE_TEST_CASE_P(SelectionTest,
-                        ReferencedRandomFloatSizeT,
-                        testing::Combine(inputs_random_largek,
-                                         testing::Values(knn::SelectKAlgo::FAISS),
-                                         testing::Values(std::make_shared<raft::resources>())));
-}  // namespace cuvs::spatial::selection
diff --git a/cpp/test/neighbors/spatial_data.h b/cpp/test/neighbors/spatial_data.h
deleted file mode 100644
index 32cbcc71f..000000000
--- a/cpp/test/neighbors/spatial_data.h
+++ /dev/null
@@ -1,38 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <vector>
-
-namespace raft {
-namespace spatial {
-
-// Latitude and longitude coordinates of 51 US states / territories
-std::vector<float> spatial_data = {
-  63.588753, -154.493062, 32.318231, -86.902298,  35.20105,  -91.831833,  34.048928, -111.093731,
-  36.778261, -119.417932, 39.550051, -105.782067, 41.603221, -73.087749,  38.905985, -77.033418,
-  38.910832, -75.52767,   27.664827, -81.515754,  32.157435, -82.907123,  19.898682, -155.665857,
-  41.878003, -93.097702,  44.068202, -114.742041, 40.633125, -89.398528,  40.551217, -85.602364,
-  39.011902, -98.484246,  37.839333, -84.270018,  31.244823, -92.145024,  42.407211, -71.382437,
-  39.045755, -76.641271,  45.253783, -69.445469,  44.314844, -85.602364,  46.729553, -94.6859,
-  37.964253, -91.831833,  32.354668, -89.398528,  46.879682, -110.362566, 35.759573, -79.0193,
-  47.551493, -101.002012, 41.492537, -99.901813,  43.193852, -71.572395,  40.058324, -74.405661,
-  34.97273,  -105.032363, 38.80261,  -116.419389, 43.299428, -74.217933,  40.417287, -82.907123,
-  35.007752, -97.092877,  43.804133, -120.554201, 41.203322, -77.194525,  18.220833, -66.590149,
-  41.580095, -71.477429,  33.836081, -81.163725,  43.969515, -99.901813,  35.517491, -86.580447,
-  31.968599, -99.901813,  39.32098,  -111.093731, 37.431573, -78.656894,  44.558803, -72.577841,
-  47.751074, -120.740139, 43.78444,  -88.787868,  38.597626, -80.454903,  43.075968, -107.290284};
-};  // namespace spatial
-};  // namespace raft
\ No newline at end of file
diff --git a/cpp/test/neighbors/tiled_knn.cu b/cpp/test/neighbors/tiled_knn.cu
deleted file mode 100644
index 41d03ad26..000000000
--- a/cpp/test/neighbors/tiled_knn.cu
+++ /dev/null
@@ -1,354 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include "./ann_utils.cuh"
-#include "./knn_utils.cuh"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cuvs/distance/distance.cuh>  // cuvs::distance::pairwise_distance
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/neighbors/brute_force.cuh>
-#include <cuvs/neighbors/detail/knn_brute_force.cuh>  // cuvs::neighbors::detail::brute_force_knn_impl
-#include <cuvs/neighbors/detail/selection_faiss.cuh>  // cuvs::neighbors::detail::select_k
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/logger.hpp>
-#include <raft/linalg/transpose.cuh>
-#include <raft/matrix/init.cuh>
-
-#include <rmm/device_buffer.hpp>
-
-#include <gtest/gtest.h>
-
-#include <cstddef>
-#include <iostream>
-#include <vector>
-
-namespace cuvs::neighbors::brute_force {
-
-struct TiledKNNInputs {
-  int num_queries;
-  int num_db_vecs;
-  int dim;
-  int k;
-  int row_tiles;
-  int col_tiles;
-  cuvs::distance::DistanceType metric;
-  bool row_major;
-};
-
-std::ostream& operator<<(std::ostream& os, const TiledKNNInputs& input)
-{
-  return os << "num_queries:" << input.num_queries << " num_vecs:" << input.num_db_vecs
-            << " dim:" << input.dim << " k:" << input.k << " row_tiles:" << input.row_tiles
-            << " col_tiles:" << input.col_tiles << " metric:" << print_metric{input.metric}
-            << " row_major:" << input.row_major;
-}
-
-template <typename T>
-class TiledKNNTest : public ::testing::TestWithParam<TiledKNNInputs> {
- public:
-  TiledKNNTest()
-    : stream_(resource::get_cuda_stream(handle_)),
-      params_(::testing::TestWithParam<TiledKNNInputs>::GetParam()),
-      database(params_.num_db_vecs * params_.dim, stream_),
-      search_queries(params_.num_queries * params_.dim, stream_),
-      raft_indices_(params_.num_queries * params_.k, stream_),
-      raft_distances_(params_.num_queries * params_.k, stream_),
-      ref_indices_(params_.num_queries * params_.k, stream_),
-      ref_distances_(params_.num_queries * params_.k, stream_)
-  {
-    raft::matrix::fill(
-      handle_,
-      raft::make_device_matrix_view(database.data(), params_.num_db_vecs, params_.dim),
-      T{0.0});
-    raft::matrix::fill(
-      handle_,
-      raft::make_device_matrix_view(search_queries.data(), params_.num_queries, params_.dim),
-      T{0.0});
-    raft::matrix::fill(
-      handle_,
-      raft::make_device_matrix_view(raft_indices_.data(), params_.num_queries, params_.k),
-      0);
-    raft::matrix::fill(
-      handle_,
-      raft::make_device_matrix_view(raft_distances_.data(), params_.num_queries, params_.k),
-      T{0.0});
-    raft::matrix::fill(
-      handle_,
-      raft::make_device_matrix_view(ref_indices_.data(), params_.num_queries, params_.k),
-      0);
-    raft::matrix::fill(
-      handle_,
-      raft::make_device_matrix_view(ref_distances_.data(), params_.num_queries, params_.k),
-      T{0.0});
-  }
-
- protected:
-  void testBruteForce()
-  {
-    float metric_arg = 3.0;
-
-    // calculate the naive knn, by calculating the full pairwise distances and doing a k-select
-    rmm::device_uvector<T> temp_distances(num_db_vecs * num_queries, stream_);
-    rmm::device_uvector<char> workspace(0, stream_);
-    distance::pairwise_distance(handle_,
-                                search_queries.data(),
-                                database.data(),
-                                temp_distances.data(),
-                                num_queries,
-                                num_db_vecs,
-                                dim,
-                                workspace,
-                                metric,
-                                params_.row_major,
-                                metric_arg);
-
-    // setting the 'isRowMajor' flag in the pairwise distances api, not only sets
-    // the inputs as colmajor - but also the output. this means we have to transpose in this
-    // case
-    auto temp_dist = temp_distances.data();
-    rmm::device_uvector<T> temp_row_major_dist(num_db_vecs * num_queries, stream_);
-    if (!params_.row_major) {
-      raft::linalg::transpose(
-        handle_, temp_dist, temp_row_major_dist.data(), num_queries, num_db_vecs, stream_);
-      temp_dist = temp_row_major_dist.data();
-    }
-
-    cuvs::neighbors::detail::select_k<int, T>(temp_dist,
-                                              nullptr,
-                                              num_queries,
-                                              num_db_vecs,
-                                              ref_distances_.data(),
-                                              ref_indices_.data(),
-                                              cuvs::distance::is_min_close(metric),
-                                              k_,
-                                              stream_);
-
-    if ((params_.row_tiles == 0) && (params_.col_tiles == 0)) {
-      std::vector<T*> input{database.data()};
-      std::vector<size_t> sizes{static_cast<size_t>(num_db_vecs)};
-      neighbors::detail::brute_force_knn_impl<size_t, int, T>(handle_,
-                                                              input,
-                                                              sizes,
-                                                              dim,
-                                                              const_cast<T*>(search_queries.data()),
-                                                              num_queries,
-                                                              raft_indices_.data(),
-                                                              raft_distances_.data(),
-                                                              k_,
-                                                              params_.row_major,
-                                                              params_.row_major,
-                                                              nullptr,
-                                                              metric,
-                                                              metric_arg);
-    } else {
-      neighbors::detail::tiled_brute_force_knn(handle_,
-                                               search_queries.data(),
-                                               database.data(),
-                                               num_queries,
-                                               num_db_vecs,
-                                               dim,
-                                               k_,
-                                               raft_distances_.data(),
-                                               raft_indices_.data(),
-                                               metric,
-                                               metric_arg,
-                                               params_.row_tiles,
-                                               params_.col_tiles);
-    }
-
-    // verify.
-    ASSERT_TRUE(cuvs::spatial::knn::devArrMatchKnnPair(ref_indices_.data(),
-                                                       raft_indices_.data(),
-                                                       ref_distances_.data(),
-                                                       raft_distances_.data(),
-                                                       num_queries,
-                                                       k_,
-                                                       float(0.001),
-                                                       stream_,
-                                                       true));
-
-    // Also test out the 'index' api - where we can use precomputed norms
-    if (params_.row_major) {
-      auto idx =
-        cuvs::neighbors::brute_force::build<T>(handle_,
-                                               raft::make_device_matrix_view<const T, int64_t>(
-                                                 database.data(), params_.num_db_vecs, params_.dim),
-                                               metric,
-                                               metric_arg);
-
-      auto query_view = raft::make_device_matrix_view<const T, int64_t>(
-        search_queries.data(), params_.num_queries, params_.dim);
-
-      cuvs::neighbors::brute_force::search<T, int>(
-        handle_,
-        idx,
-        query_view,
-        raft::make_device_matrix_view<int, int64_t>(
-          raft_indices_.data(), params_.num_queries, params_.k),
-        raft::make_device_matrix_view<T, int64_t>(
-          raft_distances_.data(), params_.num_queries, params_.k));
-
-      ASSERT_TRUE(cuvs::spatial::knn::devArrMatchKnnPair(ref_indices_.data(),
-                                                         raft_indices_.data(),
-                                                         ref_distances_.data(),
-                                                         raft_distances_.data(),
-                                                         num_queries,
-                                                         k_,
-                                                         float(0.001),
-                                                         stream_,
-                                                         true));
-      // also test out the batch api. First get new reference results (all k, up to a certain
-      // max size)
-      auto all_size      = std::min(params_.num_db_vecs, 1024);
-      auto all_indices   = raft::make_device_matrix<int, int64_t>(handle_, num_queries, all_size);
-      auto all_distances = raft::make_device_matrix<T, int64_t>(handle_, num_queries, all_size);
-      cuvs::neighbors::brute_force::search<T, int>(
-        handle_, idx, query_view, all_indices.view(), all_distances.view());
-
-      int64_t offset = 0;
-      auto query     = make_batch_k_query<T, int>(handle_, idx, query_view, k_);
-      for (auto batch : *query) {
-        auto batch_size = batch.batch_size();
-        auto indices    = raft::make_device_matrix<int, int64_t>(handle_, num_queries, batch_size);
-        auto distances  = raft::make_device_matrix<T, int64_t>(handle_, num_queries, batch_size);
-
-        matrix::slice_coordinates<int64_t> coords{0, offset, num_queries, offset + batch_size};
-
-        matrix::slice(handle_, raft::make_const_mdspan(all_indices.view()), indices.view(), coords);
-        matrix::slice(
-          handle_, raft::make_const_mdspan(all_distances.view()), distances.view(), coords);
-
-        ASSERT_TRUE(cuvs::spatial::knn::devArrMatchKnnPair(indices.data_handle(),
-                                                           batch.indices().data_handle(),
-                                                           distances.data_handle(),
-                                                           batch.distances().data_handle(),
-                                                           num_queries,
-                                                           batch_size,
-                                                           float(0.001),
-                                                           stream_,
-                                                           true));
-
-        offset += batch_size;
-        if (offset + batch_size > all_size) break;
-      }
-
-      // also test out with variable batch sizes
-      offset             = 0;
-      int64_t batch_size = k_;
-      query              = make_batch_k_query<T, int>(handle_, idx, query_view, batch_size);
-      for (auto it = query->begin(); it != query->end(); it.advance(batch_size)) {
-        // batch_size could be less than requested (in the case of final batch). handle.
-        ASSERT_TRUE(it->indices().extent(1) <= batch_size);
-        batch_size = it->indices().extent(1);
-
-        auto indices   = raft::make_device_matrix<int, int64_t>(handle_, num_queries, batch_size);
-        auto distances = raft::make_device_matrix<T, int64_t>(handle_, num_queries, batch_size);
-
-        matrix::slice_coordinates<int64_t> coords{0, offset, num_queries, offset + batch_size};
-        matrix::slice(handle_, raft::make_const_mdspan(all_indices.view()), indices.view(), coords);
-        matrix::slice(
-          handle_, raft::make_const_mdspan(all_distances.view()), distances.view(), coords);
-
-        ASSERT_TRUE(cuvs::spatial::knn::devArrMatchKnnPair(indices.data_handle(),
-                                                           it->indices().data_handle(),
-                                                           distances.data_handle(),
-                                                           it->distances().data_handle(),
-                                                           num_queries,
-                                                           batch_size,
-                                                           float(0.001),
-                                                           stream_,
-                                                           true));
-
-        offset += batch_size;
-        if (offset + batch_size > all_size) break;
-
-        batch_size += 2;
-      }
-    }
-  }
-
-  void SetUp() override
-  {
-    num_queries = params_.num_queries;
-    num_db_vecs = params_.num_db_vecs;
-    dim         = params_.dim;
-    k_          = params_.k;
-    metric      = params_.metric;
-
-    unsigned long long int seed = 1234ULL;
-    raft::random::RngState r(seed);
-
-    // JensenShannon distance requires positive values
-    T min_val = metric == cuvs::distance::DistanceType::JensenShannon ? T(0.0) : T(-1.0);
-    uniform(handle_, r, database.data(), num_db_vecs * dim, min_val, T(1.0));
-    uniform(handle_, r, search_queries.data(), num_queries * dim, min_val, T(1.0));
-  }
-
- private:
-  raft::resources handle_;
-  cudaStream_t stream_ = 0;
-  TiledKNNInputs params_;
-  int num_queries;
-  int num_db_vecs;
-  int dim;
-  rmm::device_uvector<T> database;
-  rmm::device_uvector<T> search_queries;
-  rmm::device_uvector<int> raft_indices_;
-  rmm::device_uvector<T> raft_distances_;
-  rmm::device_uvector<int> ref_indices_;
-  rmm::device_uvector<T> ref_distances_;
-  int k_;
-  cuvs::distance::DistanceType metric;
-};
-
-const std::vector<TiledKNNInputs> random_inputs = {
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::L2Expanded, true},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::L2Unexpanded, true},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::L2SqrtExpanded, true},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::L2SqrtUnexpanded, true},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::L1, true},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::Linf, true},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::InnerProduct, true},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::CorrelationExpanded, true},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::CosineExpanded, true},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::LpUnexpanded, true},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::JensenShannon, true},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::L2SqrtExpanded, true},
-  // BrayCurtis isn't currently supported by pairwise_distance api
-  // {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::BrayCurtis},
-  {256, 512, 16, 8, 16, 8, cuvs::distance::DistanceType::Canberra, true},
-  {10000, 40000, 32, 30, 512, 1024, cuvs::distance::DistanceType::L2Expanded, true},
-  {345, 1023, 16, 128, 512, 1024, cuvs::distance::DistanceType::CosineExpanded, true},
-  {789, 20516, 64, 256, 512, 4096, cuvs::distance::DistanceType::L2SqrtExpanded, true},
-  // Test where the final column tile has < K items:
-  {4, 12, 32, 6, 4, 8, cuvs::distance::DistanceType::L2Expanded, true},
-  // Test where passing column_tiles < K
-  {1, 40, 32, 30, 1, 8, cuvs::distance::DistanceType::L2Expanded, true},
-  // Passing tile sizes of 0 means to use brute_force_knn_impl (instead of the
-  // tiled_brute_force_knn api).
-  {1000, 500000, 128, 128, 0, 0, cuvs::distance::DistanceType::L2Expanded, true},
-  {1000, 500000, 128, 128, 0, 0, cuvs::distance::DistanceType::L2Expanded, false},
-  {1000, 5000, 128, 128, 0, 0, cuvs::distance::DistanceType::LpUnexpanded, true},
-  {1000, 5000, 128, 128, 0, 0, cuvs::distance::DistanceType::L2SqrtExpanded, false},
-  {1000, 5000, 128, 128, 0, 0, cuvs::distance::DistanceType::InnerProduct, false}};
-
-typedef TiledKNNTest<float> TiledKNNTestF;
-TEST_P(TiledKNNTestF, BruteForce) { this->testBruteForce(); }
-
-INSTANTIATE_TEST_CASE_P(TiledKNNTest, TiledKNNTestF, ::testing::ValuesIn(random_inputs));
-}  // namespace cuvs::neighbors::brute_force
diff --git a/cpp/test/sparse/dist_coo_spmv.cu b/cpp/test/sparse/dist_coo_spmv.cu
deleted file mode 100644
index 4e2a83f83..000000000
--- a/cpp/test/sparse/dist_coo_spmv.cu
+++ /dev/null
@@ -1,697 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <gtest/gtest.h>
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/core/operators.cuh>
-#include <raft/core/operators.hpp>
-#include <raft/linalg/unary_op.cuh>
-#include <raft/sparse/detail/cusparse_wrappers.h>
-#include <raft/util/cudart_utils.hpp>
-#include <rmm/device_uvector.hpp>
-
-#include <raft/sparse/convert/coo.cuh>
-#include <raft/sparse/distance/detail/coo_spmv.cuh>
-
-#include "../test_utils.cuh"
-
-#include <type_traits>
-
-namespace raft {
-namespace sparse {
-namespace distance {
-
-using namespace raft;
-using namespace raft::sparse;
-
-template <typename value_idx, typename value_t>
-struct InputConfiguration {
-  value_idx n_cols;
-
-  std::vector<value_idx> indptr_h;
-  std::vector<value_idx> indices_h;
-  std::vector<value_t> data_h;
-
-  std::vector<value_t> out_dists_ref_h;
-
-  cuvs::distance::DistanceType metric;
-
-  float metric_arg = 0.0;
-};
-
-using dense_smem_strategy_t = detail::dense_smem_strategy<int, float, 1024>;
-using hash_strategy_t       = detail::hash_strategy<int, float, 1024>;
-
-template <typename value_idx, typename value_t, typename strategy_t>
-struct SparseDistanceCOOSPMVInputs {
-  InputConfiguration<value_idx, value_t> input_configuration;
-
-  float capacity_threshold = 0.5;
-  int map_size             = detail::hash_strategy<value_idx, value_t, 1024>::get_map_size();
-};
-
-template <typename value_idx, typename value_t, typename strategy_t>
-::std::ostream& operator<<(::std::ostream& os,
-                           const SparseDistanceCOOSPMVInputs<value_idx, value_t, strategy_t>& dims)
-{
-  return os;
-}
-
-template <typename value_idx, typename value_t, typename strategy_t>
-class SparseDistanceCOOSPMVTest
-  : public ::testing::TestWithParam<SparseDistanceCOOSPMVInputs<value_idx, value_t, strategy_t>> {
- public:
-  SparseDistanceCOOSPMVTest()
-    : dist_config(handle),
-      indptr(0, resource::get_cuda_stream(handle)),
-      indices(0, resource::get_cuda_stream(handle)),
-      data(0, resource::get_cuda_stream(handle)),
-      out_dists(0, resource::get_cuda_stream(handle)),
-      out_dists_ref(0, resource::get_cuda_stream(handle))
-  {
-  }
-
-  template <typename U, std::enable_if_t<std::is_same_v<U, hash_strategy_t>>* = nullptr>
-  U make_strategy()
-  {
-    return strategy_t(dist_config, params.capacity_threshold, params.map_size);
-  }
-
-  template <typename U, std::enable_if_t<std::is_same_v<U, dense_smem_strategy_t>>* = nullptr>
-  U make_strategy()
-  {
-    return strategy_t(dist_config);
-  }
-
-  template <typename reduce_f, typename accum_f, typename write_f>
-  void compute_dist(reduce_f reduce_func, accum_f accum_func, write_f write_func, bool rev = true)
-  {
-    rmm::device_uvector<value_idx> coo_rows(max(dist_config.b_nnz, dist_config.a_nnz),
-                                            resource::get_cuda_stream(dist_config.handle));
-
-    raft::sparse::convert::csr_to_coo(dist_config.b_indptr,
-                                      dist_config.b_nrows,
-                                      coo_rows.data(),
-                                      dist_config.b_nnz,
-                                      resource::get_cuda_stream(dist_config.handle));
-
-    strategy_t selected_strategy = make_strategy<strategy_t>();
-    detail::balanced_coo_pairwise_generalized_spmv<value_idx, value_t>(out_dists.data(),
-                                                                       dist_config,
-                                                                       coo_rows.data(),
-                                                                       reduce_func,
-                                                                       accum_func,
-                                                                       write_func,
-                                                                       selected_strategy);
-
-    if (rev) {
-      raft::sparse::convert::csr_to_coo(dist_config.a_indptr,
-                                        dist_config.a_nrows,
-                                        coo_rows.data(),
-                                        dist_config.a_nnz,
-                                        resource::get_cuda_stream(dist_config.handle));
-
-      detail::balanced_coo_pairwise_generalized_spmv_rev<value_idx, value_t>(out_dists.data(),
-                                                                             dist_config,
-                                                                             coo_rows.data(),
-                                                                             reduce_func,
-                                                                             accum_func,
-                                                                             write_func,
-                                                                             selected_strategy);
-    }
-  }
-
-  void run_spmv()
-  {
-    switch (params.input_configuration.metric) {
-      case cuvs::distance::DistanceType::InnerProduct:
-        compute_dist(raft::mul_op(), raft::add_op(), raft::atomic_add_op(), true);
-        break;
-      case cuvs::distance::DistanceType::L2Unexpanded:
-        compute_dist(raft::sqdiff_op(), raft::add_op(), raft::atomic_add_op());
-        break;
-      case cuvs::distance::DistanceType::Canberra:
-        compute_dist(
-          [] __device__(value_t a, value_t b) { return fabsf(a - b) / (fabsf(a) + fabsf(b)); },
-          raft::add_op(),
-          raft::atomic_add_op());
-        break;
-      case cuvs::distance::DistanceType::L1:
-        compute_dist(absdiff_op(), raft::add_op(), raft::atomic_add_op());
-        break;
-      case cuvs::distance::DistanceType::Linf:
-        compute_dist(absdiff_op(), raft::max_op(), raft::atomic_max_op());
-        break;
-      case cuvs::distance::DistanceType::LpUnexpanded: {
-        compute_dist(
-          raft::compose_op(raft::pow_const_op<value_t>(params.input_configuration.metric_arg),
-                           raft::sub_op()),
-          raft::add_op(),
-          raft::atomic_add_op());
-        value_t p = value_t{1} / params.input_configuration.metric_arg;
-        raft::linalg::unaryOp<value_t>(out_dists.data(),
-                                       out_dists.data(),
-                                       dist_config.a_nrows * dist_config.b_nrows,
-                                       raft::pow_const_op<value_t>{p},
-                                       resource::get_cuda_stream(dist_config.handle));
-
-      } break;
-      default: throw raft::exception("Unknown distance");
-    }
-  }
-
- protected:
-  void make_data()
-  {
-    std::vector<value_idx> indptr_h  = params.input_configuration.indptr_h;
-    std::vector<value_idx> indices_h = params.input_configuration.indices_h;
-    std::vector<value_t> data_h      = params.input_configuration.data_h;
-
-    auto stream = resource::get_cuda_stream(handle);
-    indptr.resize(indptr_h.size(), stream);
-    indices.resize(indices_h.size(), stream);
-    data.resize(data_h.size(), stream);
-
-    update_device(indptr.data(), indptr_h.data(), indptr_h.size(), stream);
-    update_device(indices.data(), indices_h.data(), indices_h.size(), stream);
-    update_device(data.data(), data_h.data(), data_h.size(), stream);
-
-    std::vector<value_t> out_dists_ref_h = params.input_configuration.out_dists_ref_h;
-
-    out_dists_ref.resize((indptr_h.size() - 1) * (indptr_h.size() - 1), stream);
-
-    update_device(out_dists_ref.data(), out_dists_ref_h.data(), out_dists_ref_h.size(), stream);
-  }
-
-  void SetUp() override
-  {
-    params = ::testing::TestWithParam<
-      SparseDistanceCOOSPMVInputs<value_idx, value_t, strategy_t>>::GetParam();
-
-    make_data();
-
-    dist_config.b_nrows   = params.input_configuration.indptr_h.size() - 1;
-    dist_config.b_ncols   = params.input_configuration.n_cols;
-    dist_config.b_nnz     = params.input_configuration.indices_h.size();
-    dist_config.b_indptr  = indptr.data();
-    dist_config.b_indices = indices.data();
-    dist_config.b_data    = data.data();
-    dist_config.a_nrows   = params.input_configuration.indptr_h.size() - 1;
-    dist_config.a_ncols   = params.input_configuration.n_cols;
-    dist_config.a_nnz     = params.input_configuration.indices_h.size();
-    dist_config.a_indptr  = indptr.data();
-    dist_config.a_indices = indices.data();
-    dist_config.a_data    = data.data();
-
-    int out_size = dist_config.a_nrows * dist_config.b_nrows;
-
-    out_dists.resize(out_size, resource::get_cuda_stream(handle));
-
-    run_spmv();
-
-    RAFT_CUDA_TRY(cudaStreamSynchronize(resource::get_cuda_stream(handle)));
-  }
-
-  void compare()
-  {
-    ASSERT_TRUE(devArrMatch(out_dists_ref.data(),
-                            out_dists.data(),
-                            params.input_configuration.out_dists_ref_h.size(),
-                            CompareApprox<value_t>(1e-3)));
-  }
-
- protected:
-  raft::resources handle;
-
-  // input data
-  rmm::device_uvector<value_idx> indptr, indices;
-  rmm::device_uvector<value_t> data;
-
-  // output data
-  rmm::device_uvector<value_t> out_dists, out_dists_ref;
-
-  raft::sparse::distance::detail::distances_config_t<value_idx, value_t> dist_config;
-
-  SparseDistanceCOOSPMVInputs<value_idx, value_t, strategy_t> params;
-};
-
-const InputConfiguration<int, float> input_inner_product = {
-  2,
-  {0, 2, 4, 6, 8},
-  {0, 1, 0, 1, 0, 1, 0, 1},
-  {1.0f, 2.0f, 1.0f, 2.0f, 1.0f, 2.0f, 1.0f, 2.0f},
-  {5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0},
-  cuvs::distance::DistanceType::InnerProduct,
-  0.0};
-
-const InputConfiguration<int, float> input_l2_unexpanded = {
-  2,
-  {0, 2, 4, 6, 8},
-  {0, 1, 0, 1, 0, 1, 0, 1},  // indices
-  {1.0f, 3.0f, 1.0f, 5.0f, 50.0f, 28.0f, 16.0f, 2.0f},
-  {
-    // dense output
-    0.0,
-    4.0,
-    3026.0,
-    226.0,
-    4.0,
-    0.0,
-    2930.0,
-    234.0,
-    3026.0,
-    2930.0,
-    0.0,
-    1832.0,
-    226.0,
-    234.0,
-    1832.0,
-    0.0,
-  },
-  cuvs::distance::DistanceType::L2Unexpanded,
-  0.0};
-
-const InputConfiguration<int, float> input_canberra = {
-  10,
-  {0, 5, 11, 15, 20, 27, 32, 36, 43, 47, 50},
-  {0, 1, 3, 6, 8, 0, 1, 2, 3, 5, 6, 1, 2, 4, 8, 0, 2, 3, 4, 7, 0, 1, 2, 3, 4,
-   6, 8, 0, 1, 2, 5, 7, 1, 5, 8, 9, 0, 1, 2, 5, 6, 8, 9, 2, 4, 5, 7, 0, 3, 9},  // indices
-  {0.5438, 0.2695, 0.4377, 0.7174, 0.9251, 0.7648, 0.3322, 0.7279, 0.4131, 0.5167,
-   0.8655, 0.0730, 0.0291, 0.9036, 0.7988, 0.5019, 0.7663, 0.2190, 0.8206, 0.3625,
-   0.0411, 0.3995, 0.5688, 0.7028, 0.8706, 0.3199, 0.4431, 0.0535, 0.2225, 0.8853,
-   0.1932, 0.3761, 0.3379, 0.1771, 0.2107, 0.228,  0.5279, 0.4885, 0.3495, 0.5079,
-   0.2325, 0.2331, 0.3018, 0.6231, 0.2645, 0.8429, 0.6625, 0.0797, 0.2724, 0.4218},
-  {0.0,
-   3.3954660629919076,
-   5.6469232737388815,
-   6.373112846266441,
-   4.0212880272531715,
-   6.916281504639404,
-   5.741508386786526,
-   5.411470999663036,
-   9.0,
-   4.977014354725805,
-   3.3954660629919076,
-   0.0,
-   7.56256082439209,
-   5.540261147481582,
-   4.832322929216881,
-   4.62003193872216,
-   6.498056792320361,
-   4.309846252268695,
-   6.317531174829905,
-   6.016362684141827,
-   5.6469232737388815,
-   7.56256082439209,
-   0.0,
-   5.974878731322299,
-   4.898357301336036,
-   6.442097410320605,
-   5.227077347287883,
-   7.134101195584642,
-   5.457753923371659,
-   7.0,
-   6.373112846266441,
-   5.540261147481582,
-   5.974878731322299,
-   0.0,
-   5.5507273748583,
-   4.897749658726415,
-   9.0,
-   8.398776718824767,
-   3.908281400328807,
-   4.83431066343688,
-   4.0212880272531715,
-   4.832322929216881,
-   4.898357301336036,
-   5.5507273748583,
-   0.0,
-   6.632989819428174,
-   7.438852294822894,
-   5.6631570310967465,
-   7.579428202635459,
-   6.760811985364303,
-   6.916281504639404,
-   4.62003193872216,
-   6.442097410320605,
-   4.897749658726415,
-   6.632989819428174,
-   0.0,
-   5.249404187382862,
-   6.072559523278559,
-   4.07661278488929,
-   6.19678948003145,
-   5.741508386786526,
-   6.498056792320361,
-   5.227077347287883,
-   9.0,
-   7.438852294822894,
-   5.249404187382862,
-   0.0,
-   3.854811639654704,
-   6.652724827169063,
-   5.298236851430971,
-   5.411470999663036,
-   4.309846252268695,
-   7.134101195584642,
-   8.398776718824767,
-   5.6631570310967465,
-   6.072559523278559,
-   3.854811639654704,
-   0.0,
-   7.529184598969917,
-   6.903282911791188,
-   9.0,
-   6.317531174829905,
-   5.457753923371659,
-   3.908281400328807,
-   7.579428202635459,
-   4.07661278488929,
-   6.652724827169063,
-   7.529184598969917,
-   0.0,
-   7.0,
-   4.977014354725805,
-   6.016362684141827,
-   7.0,
-   4.83431066343688,
-   6.760811985364303,
-   6.19678948003145,
-   5.298236851430971,
-   6.903282911791188,
-   7.0,
-   0.0},
-  cuvs::distance::DistanceType::Canberra,
-  0.0};
-
-const InputConfiguration<int, float> input_lp_unexpanded = {
-  10,
-  {0, 5, 11, 15, 20, 27, 32, 36, 43, 47, 50},
-  {0, 1, 3, 6, 8, 0, 1, 2, 3, 5, 6, 1, 2, 4, 8, 0, 2, 3, 4, 7, 0, 1, 2, 3, 4,
-   6, 8, 0, 1, 2, 5, 7, 1, 5, 8, 9, 0, 1, 2, 5, 6, 8, 9, 2, 4, 5, 7, 0, 3, 9},  // indices
-  {0.5438, 0.2695, 0.4377, 0.7174, 0.9251, 0.7648, 0.3322, 0.7279, 0.4131, 0.5167,
-   0.8655, 0.0730, 0.0291, 0.9036, 0.7988, 0.5019, 0.7663, 0.2190, 0.8206, 0.3625,
-   0.0411, 0.3995, 0.5688, 0.7028, 0.8706, 0.3199, 0.4431, 0.0535, 0.2225, 0.8853,
-   0.1932, 0.3761, 0.3379, 0.1771, 0.2107, 0.228,  0.5279, 0.4885, 0.3495, 0.5079,
-   0.2325, 0.2331, 0.3018, 0.6231, 0.2645, 0.8429, 0.6625, 0.0797, 0.2724, 0.4218},
-  {0.0,
-   1.31462855332296,
-   1.3690307816129905,
-   1.698603990921237,
-   1.3460470789553531,
-   1.6636670712582544,
-   1.2651744044972217,
-   1.1938329352055201,
-   1.8811409082590185,
-   1.3653115050624267,
-   1.31462855332296,
-   0.0,
-   1.9447722703291133,
-   1.42818777206562,
-   1.4685491458946494,
-   1.3071999866010466,
-   1.4988622861692171,
-   0.9698559287406783,
-   1.4972023224597841,
-   1.5243383567266802,
-   1.3690307816129905,
-   1.9447722703291133,
-   0.0,
-   1.2748400840107568,
-   1.0599569946448246,
-   1.546591282841402,
-   1.147526531928459,
-   1.447002179128145,
-   1.5982242387673176,
-   1.3112533607072414,
-   1.698603990921237,
-   1.42818777206562,
-   1.2748400840107568,
-   0.0,
-   1.038121552545461,
-   1.011788365364402,
-   1.3907391109256988,
-   1.3128200942311496,
-   1.19595706584447,
-   1.3233328139624725,
-   1.3460470789553531,
-   1.4685491458946494,
-   1.0599569946448246,
-   1.038121552545461,
-   0.0,
-   1.3642741698145529,
-   1.3493868683808095,
-   1.394942694628328,
-   1.572881849642552,
-   1.380122665319464,
-   1.6636670712582544,
-   1.3071999866010466,
-   1.546591282841402,
-   1.011788365364402,
-   1.3642741698145529,
-   0.0,
-   1.018961640373018,
-   1.0114394258945634,
-   0.8338711034820684,
-   1.1247823842299223,
-   1.2651744044972217,
-   1.4988622861692171,
-   1.147526531928459,
-   1.3907391109256988,
-   1.3493868683808095,
-   1.018961640373018,
-   0.0,
-   0.7701238110357329,
-   1.245486437864406,
-   0.5551259549534626,
-   1.1938329352055201,
-   0.9698559287406783,
-   1.447002179128145,
-   1.3128200942311496,
-   1.394942694628328,
-   1.0114394258945634,
-   0.7701238110357329,
-   0.0,
-   1.1886800117391216,
-   1.0083692448135637,
-   1.8811409082590185,
-   1.4972023224597841,
-   1.5982242387673176,
-   1.19595706584447,
-   1.572881849642552,
-   0.8338711034820684,
-   1.245486437864406,
-   1.1886800117391216,
-   0.0,
-   1.3661374102525012,
-   1.3653115050624267,
-   1.5243383567266802,
-   1.3112533607072414,
-   1.3233328139624725,
-   1.380122665319464,
-   1.1247823842299223,
-   0.5551259549534626,
-   1.0083692448135637,
-   1.3661374102525012,
-   0.0},
-  cuvs::distance::DistanceType::LpUnexpanded,
-  2.0};
-
-const InputConfiguration<int, float> input_linf = {
-  10,
-  {0, 5, 11, 15, 20, 27, 32, 36, 43, 47, 50},
-  {0, 1, 3, 6, 8, 0, 1, 2, 3, 5, 6, 1, 2, 4, 8, 0, 2, 3, 4, 7, 0, 1, 2, 3, 4,
-   6, 8, 0, 1, 2, 5, 7, 1, 5, 8, 9, 0, 1, 2, 5, 6, 8, 9, 2, 4, 5, 7, 0, 3, 9},  // indices
-  {0.5438, 0.2695, 0.4377, 0.7174, 0.9251, 0.7648, 0.3322, 0.7279, 0.4131, 0.5167,
-   0.8655, 0.0730, 0.0291, 0.9036, 0.7988, 0.5019, 0.7663, 0.2190, 0.8206, 0.3625,
-   0.0411, 0.3995, 0.5688, 0.7028, 0.8706, 0.3199, 0.4431, 0.0535, 0.2225, 0.8853,
-   0.1932, 0.3761, 0.3379, 0.1771, 0.2107, 0.228,  0.5279, 0.4885, 0.3495, 0.5079,
-   0.2325, 0.2331, 0.3018, 0.6231, 0.2645, 0.8429, 0.6625, 0.0797, 0.2724, 0.4218},
-  {0.0,
-   0.9251771844789913,
-   0.9036452083899731,
-   0.9251771844789913,
-   0.8706483735804971,
-   0.9251771844789913,
-   0.717493881903289,
-   0.6920214832303888,
-   0.9251771844789913,
-   0.9251771844789913,
-   0.9251771844789913,
-   0.0,
-   0.9036452083899731,
-   0.8655339692155823,
-   0.8706483735804971,
-   0.8655339692155823,
-   0.8655339692155823,
-   0.6329837991017668,
-   0.8655339692155823,
-   0.8655339692155823,
-   0.9036452083899731,
-   0.9036452083899731,
-   0.0,
-   0.7988276152181608,
-   0.7028075145996631,
-   0.9036452083899731,
-   0.9036452083899731,
-   0.9036452083899731,
-   0.8429599432532096,
-   0.9036452083899731,
-   0.9251771844789913,
-   0.8655339692155823,
-   0.7988276152181608,
-   0.0,
-   0.48376552205293305,
-   0.8206394616536681,
-   0.8206394616536681,
-   0.8206394616536681,
-   0.8429599432532096,
-   0.8206394616536681,
-   0.8706483735804971,
-   0.8706483735804971,
-   0.7028075145996631,
-   0.48376552205293305,
-   0.0,
-   0.8706483735804971,
-   0.8706483735804971,
-   0.8706483735804971,
-   0.8429599432532096,
-   0.8706483735804971,
-   0.9251771844789913,
-   0.8655339692155823,
-   0.9036452083899731,
-   0.8206394616536681,
-   0.8706483735804971,
-   0.0,
-   0.8853924473642432,
-   0.535821510936138,
-   0.6497196601457607,
-   0.8853924473642432,
-   0.717493881903289,
-   0.8655339692155823,
-   0.9036452083899731,
-   0.8206394616536681,
-   0.8706483735804971,
-   0.8853924473642432,
-   0.0,
-   0.5279604218147174,
-   0.6658348373853169,
-   0.33799874888632914,
-   0.6920214832303888,
-   0.6329837991017668,
-   0.9036452083899731,
-   0.8206394616536681,
-   0.8706483735804971,
-   0.535821510936138,
-   0.5279604218147174,
-   0.0,
-   0.662579808115858,
-   0.5079750812968089,
-   0.9251771844789913,
-   0.8655339692155823,
-   0.8429599432532096,
-   0.8429599432532096,
-   0.8429599432532096,
-   0.6497196601457607,
-   0.6658348373853169,
-   0.662579808115858,
-   0.0,
-   0.8429599432532096,
-   0.9251771844789913,
-   0.8655339692155823,
-   0.9036452083899731,
-   0.8206394616536681,
-   0.8706483735804971,
-   0.8853924473642432,
-   0.33799874888632914,
-   0.5079750812968089,
-   0.8429599432532096,
-   0.0},
-  cuvs::distance::DistanceType::Linf,
-  0.0};
-
-const InputConfiguration<int, float> input_l1 = {4,
-                                                 {0, 1, 1, 2, 4},
-                                                 {3, 2, 0, 1},  // indices
-                                                 {0.99296, 0.42180, 0.11687, 0.305869},
-                                                 {
-                                                   // dense output
-                                                   0.0,
-                                                   0.99296,
-                                                   1.41476,
-                                                   1.415707,
-                                                   0.99296,
-                                                   0.0,
-                                                   0.42180,
-                                                   0.42274,
-                                                   1.41476,
-                                                   0.42180,
-                                                   0.0,
-                                                   0.84454,
-                                                   1.41570,
-                                                   0.42274,
-                                                   0.84454,
-                                                   0.0,
-                                                 },
-                                                 cuvs::distance::DistanceType::L1,
-                                                 0.0};
-
-// test dense smem strategy
-const std::vector<SparseDistanceCOOSPMVInputs<int, float, dense_smem_strategy_t>>
-  inputs_dense_strategy = {{input_inner_product},
-                           {input_l2_unexpanded},
-                           {input_canberra},
-                           {input_lp_unexpanded},
-                           {input_linf},
-                           {input_l1}};
-
-typedef SparseDistanceCOOSPMVTest<int, float, dense_smem_strategy_t>
-  SparseDistanceCOOSPMVTestDenseStrategyF;
-TEST_P(SparseDistanceCOOSPMVTestDenseStrategyF, Result) { compare(); }
-INSTANTIATE_TEST_CASE_P(SparseDistanceCOOSPMVTests,
-                        SparseDistanceCOOSPMVTestDenseStrategyF,
-                        ::testing::ValuesIn(inputs_dense_strategy));
-
-// test hash and chunk strategy
-const std::vector<SparseDistanceCOOSPMVInputs<int, float, hash_strategy_t>> inputs_hash_strategy = {
-  {input_inner_product},
-  {input_inner_product, 0.5, 2},
-  {input_l2_unexpanded},
-  {input_l2_unexpanded, 0.5, 2},
-  {input_canberra},
-  {input_canberra, 0.5, 2},
-  {input_canberra, 0.5, 6},
-  {input_lp_unexpanded},
-  {input_lp_unexpanded, 0.5, 2},
-  {input_lp_unexpanded, 0.5, 6},
-  {input_linf},
-  {input_linf, 0.5, 2},
-  {input_linf, 0.5, 6},
-  {input_l1},
-  {input_l1, 0.5, 2}};
-
-typedef SparseDistanceCOOSPMVTest<int, float, hash_strategy_t>
-  SparseDistanceCOOSPMVTestHashStrategyF;
-TEST_P(SparseDistanceCOOSPMVTestHashStrategyF, Result) { compare(); }
-INSTANTIATE_TEST_CASE_P(SparseDistanceCOOSPMVTests,
-                        SparseDistanceCOOSPMVTestHashStrategyF,
-                        ::testing::ValuesIn(inputs_hash_strategy));
-
-};  // namespace distance
-};  // end namespace sparse
-};  // end namespace raft
diff --git a/cpp/test/sparse/distance.cu b/cpp/test/sparse/distance.cu
deleted file mode 100644
index e2b3bc45f..000000000
--- a/cpp/test/sparse/distance.cu
+++ /dev/null
@@ -1,853 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <gtest/gtest.h>
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cusparse_v2.h>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/sparse/detail/cusparse_wrappers.h>
-#include <raft/util/cudart_utils.hpp>
-
-#include <raft/sparse/distance/distance.cuh>
-
-#include "../test_utils.cuh"
-
-namespace raft {
-namespace sparse {
-namespace distance {
-
-using namespace raft;
-using namespace raft::sparse;
-
-template <typename value_idx, typename value_t>
-struct SparseDistanceInputs {
-  value_idx n_cols;
-
-  std::vector<value_idx> indptr_h;
-  std::vector<value_idx> indices_h;
-  std::vector<value_t> data_h;
-
-  std::vector<value_t> out_dists_ref_h;
-
-  cuvs::distance::DistanceType metric;
-
-  float metric_arg = 0.0;
-};
-
-template <typename value_idx, typename value_t>
-::std::ostream& operator<<(::std::ostream& os, const SparseDistanceInputs<value_idx, value_t>& dims)
-{
-  return os;
-}
-
-template <typename value_idx, typename value_t>
-class SparseDistanceTest
-  : public ::testing::TestWithParam<SparseDistanceInputs<value_idx, value_t>> {
- public:
-  SparseDistanceTest()
-    : params(::testing::TestWithParam<SparseDistanceInputs<value_idx, value_t>>::GetParam()),
-      indptr(0, resource::get_cuda_stream(handle)),
-      indices(0, resource::get_cuda_stream(handle)),
-      data(0, resource::get_cuda_stream(handle)),
-      out_dists(0, resource::get_cuda_stream(handle)),
-      out_dists_ref(0, resource::get_cuda_stream(handle))
-  {
-  }
-
-  void SetUp() override
-  {
-    make_data();
-
-    int out_size = static_cast<value_idx>(params.indptr_h.size() - 1) *
-                   static_cast<value_idx>(params.indptr_h.size() - 1);
-
-    out_dists.resize(out_size, resource::get_cuda_stream(handle));
-
-    auto out = raft::make_device_matrix_view<value_t, value_idx>(
-      out_dists.data(),
-      static_cast<value_idx>(params.indptr_h.size() - 1),
-      static_cast<value_idx>(params.indptr_h.size() - 1));
-
-    auto x_structure = raft::make_device_compressed_structure_view<value_idx, value_idx, value_idx>(
-      indptr.data(),
-      indices.data(),
-      static_cast<value_idx>(params.indptr_h.size() - 1),
-      params.n_cols,
-      static_cast<value_idx>(params.indices_h.size()));
-    auto x = raft::make_device_csr_matrix_view<const value_t>(data.data(), x_structure);
-
-    pairwise_distance(handle, x, x, out, params.metric, params.metric_arg);
-
-    RAFT_CUDA_TRY(cudaStreamSynchronize(resource::get_cuda_stream(handle)));
-  }
-
-  void compare()
-  {
-    ASSERT_TRUE(devArrMatch(out_dists_ref.data(),
-                            out_dists.data(),
-                            params.out_dists_ref_h.size(),
-                            CompareApprox<value_t>(1e-3)));
-  }
-
- protected:
-  void make_data()
-  {
-    std::vector<value_idx> indptr_h  = params.indptr_h;
-    std::vector<value_idx> indices_h = params.indices_h;
-    std::vector<value_t> data_h      = params.data_h;
-
-    auto stream = resource::get_cuda_stream(handle);
-    indptr.resize(indptr_h.size(), stream);
-    indices.resize(indices_h.size(), stream);
-    data.resize(data_h.size(), stream);
-
-    update_device(indptr.data(), indptr_h.data(), indptr_h.size(), stream);
-    update_device(indices.data(), indices_h.data(), indices_h.size(), stream);
-    update_device(data.data(), data_h.data(), data_h.size(), stream);
-
-    std::vector<value_t> out_dists_ref_h = params.out_dists_ref_h;
-
-    out_dists_ref.resize((indptr_h.size() - 1) * (indptr_h.size() - 1), stream);
-
-    update_device(out_dists_ref.data(),
-                  out_dists_ref_h.data(),
-                  out_dists_ref_h.size(),
-                  resource::get_cuda_stream(handle));
-  }
-
-  raft::resources handle;
-
-  // input data
-  rmm::device_uvector<value_idx> indptr, indices;
-  rmm::device_uvector<value_t> data;
-
-  // output data
-  rmm::device_uvector<value_t> out_dists, out_dists_ref;
-
-  SparseDistanceInputs<value_idx, value_t> params;
-};
-
-const std::vector<SparseDistanceInputs<int, float>> inputs_i32_f = {
-  {5,
-   {0, 0, 1, 2},
-
-   {1, 2},
-   {0.5, 0.5},
-   {0, 1, 1, 1, 0, 1, 1, 1, 0},
-   cuvs::distance::DistanceType::CosineExpanded,
-   0.0},
-  {5,
-   {0, 0, 1, 2},
-
-   {1, 2},
-   {1.0, 1.0},
-   {0, 1, 1, 1, 0, 1, 1, 1, 0},
-   cuvs::distance::DistanceType::JaccardExpanded,
-   0.0},
-  {2,
-   {0, 2, 4, 6, 8},
-   {0, 1, 0, 1, 0, 1, 0, 1},  // indices
-   {1.0f, 3.0f, 1.0f, 5.0f, 50.0f, 28.0f, 16.0f, 2.0f},
-   {
-     // dense output
-     0.0,
-     4.0,
-     3026.0,
-     226.0,
-     4.0,
-     0.0,
-     2930.0,
-     234.0,
-     3026.0,
-     2930.0,
-     0.0,
-     1832.0,
-     226.0,
-     234.0,
-     1832.0,
-     0.0,
-   },
-   cuvs::distance::DistanceType::L2Expanded,
-   0.0},
-  {2,
-   {0, 2, 4, 6, 8},
-   {0, 1, 0, 1, 0, 1, 0, 1},
-   {1.0f, 2.0f, 1.0f, 2.0f, 1.0f, 2.0f, 1.0f, 2.0f},
-   {5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0, 5.0},
-   cuvs::distance::DistanceType::InnerProduct,
-   0.0},
-  {2,
-   {0, 2, 4, 6, 8},
-   {0, 1, 0, 1, 0, 1, 0, 1},  // indices
-   {1.0f, 3.0f, 1.0f, 5.0f, 50.0f, 28.0f, 16.0f, 2.0f},
-   {
-     // dense output
-     0.0,
-     4.0,
-     3026.0,
-     226.0,
-     4.0,
-     0.0,
-     2930.0,
-     234.0,
-     3026.0,
-     2930.0,
-     0.0,
-     1832.0,
-     226.0,
-     234.0,
-     1832.0,
-     0.0,
-   },
-   cuvs::distance::DistanceType::L2Unexpanded,
-   0.0},
-
-  {10,
-   {0, 5, 11, 15, 20, 27, 32, 36, 43, 47, 50},
-   {0, 1, 3, 6, 8, 0, 1, 2, 3, 5, 6, 1, 2, 4, 8, 0, 2, 3, 4, 7, 0, 1, 2, 3, 4,
-    6, 8, 0, 1, 2, 5, 7, 1, 5, 8, 9, 0, 1, 2, 5, 6, 8, 9, 2, 4, 5, 7, 0, 3, 9},  // indices
-   {0.5438, 0.2695, 0.4377, 0.7174, 0.9251, 0.7648, 0.3322, 0.7279, 0.4131, 0.5167,
-    0.8655, 0.0730, 0.0291, 0.9036, 0.7988, 0.5019, 0.7663, 0.2190, 0.8206, 0.3625,
-    0.0411, 0.3995, 0.5688, 0.7028, 0.8706, 0.3199, 0.4431, 0.0535, 0.2225, 0.8853,
-    0.1932, 0.3761, 0.3379, 0.1771, 0.2107, 0.228,  0.5279, 0.4885, 0.3495, 0.5079,
-    0.2325, 0.2331, 0.3018, 0.6231, 0.2645, 0.8429, 0.6625, 0.0797, 0.2724, 0.4218},
-   {0.,         0.39419924, 0.54823225, 0.79593037, 0.45658883, 0.93634219, 0.58146987, 0.44940102,
-    1.,         0.76978799, 0.39419924, 0.,         0.97577154, 0.48904013, 0.48300801, 0.45087445,
-    0.73323749, 0.21050481, 0.54847744, 0.78021386, 0.54823225, 0.97577154, 0.,         0.51413997,
-    0.31195441, 0.96546343, 0.67534399, 0.81665436, 0.8321819,  1.,         0.79593037, 0.48904013,
-    0.51413997, 0.,         0.28605559, 0.35772784, 1.,         0.60889396, 0.43324829, 0.84923694,
-    0.45658883, 0.48300801, 0.31195441, 0.28605559, 0.,         0.58623212, 0.6745457,  0.60287165,
-    0.67676228, 0.73155632, 0.93634219, 0.45087445, 0.96546343, 0.35772784, 0.58623212, 0.,
-    0.77917274, 0.48390993, 0.24558392, 0.99166225, 0.58146987, 0.73323749, 0.67534399, 1.,
-    0.6745457,  0.77917274, 0.,         0.27605686, 0.76064776, 0.61547536, 0.44940102, 0.21050481,
-    0.81665436, 0.60889396, 0.60287165, 0.48390993, 0.27605686, 0.,         0.51360432, 0.68185144,
-    1.,         0.54847744, 0.8321819,  0.43324829, 0.67676228, 0.24558392, 0.76064776, 0.51360432,
-    0.,         1.,         0.76978799, 0.78021386, 1.,         0.84923694, 0.73155632, 0.99166225,
-    0.61547536, 0.68185144, 1.,         0.},
-   cuvs::distance::DistanceType::CosineExpanded,
-   0.0},
-
-  {10,
-   {0, 5, 11, 15, 20, 27, 32, 36, 43, 47, 50},
-   {0, 1, 3, 6, 8, 0, 1, 2, 3, 5, 6, 1, 2, 4, 8, 0, 2, 3, 4, 7, 0, 1, 2, 3, 4,
-    6, 8, 0, 1, 2, 5, 7, 1, 5, 8, 9, 0, 1, 2, 5, 6, 8, 9, 2, 4, 5, 7, 0, 3, 9},  // indices
-   {1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
-    1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
-    1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.},
-   {0.0,
-    0.42857142857142855,
-    0.7142857142857143,
-    0.75,
-    0.2857142857142857,
-    0.75,
-    0.7142857142857143,
-    0.5,
-    1.0,
-    0.6666666666666666,
-    0.42857142857142855,
-    0.0,
-    0.75,
-    0.625,
-    0.375,
-    0.42857142857142855,
-    0.75,
-    0.375,
-    0.75,
-    0.7142857142857143,
-    0.7142857142857143,
-    0.75,
-    0.0,
-    0.7142857142857143,
-    0.42857142857142855,
-    0.7142857142857143,
-    0.6666666666666666,
-    0.625,
-    0.6666666666666666,
-    1.0,
-    0.75,
-    0.625,
-    0.7142857142857143,
-    0.0,
-    0.5,
-    0.5714285714285714,
-    1.0,
-    0.8,
-    0.5,
-    0.6666666666666666,
-    0.2857142857142857,
-    0.375,
-    0.42857142857142855,
-    0.5,
-    0.0,
-    0.6666666666666666,
-    0.7777777777777778,
-    0.4444444444444444,
-    0.7777777777777778,
-    0.75,
-    0.75,
-    0.42857142857142855,
-    0.7142857142857143,
-    0.5714285714285714,
-    0.6666666666666666,
-    0.0,
-    0.7142857142857143,
-    0.5,
-    0.5,
-    0.8571428571428571,
-    0.7142857142857143,
-    0.75,
-    0.6666666666666666,
-    1.0,
-    0.7777777777777778,
-    0.7142857142857143,
-    0.0,
-    0.42857142857142855,
-    0.8571428571428571,
-    0.8333333333333334,
-    0.5,
-    0.375,
-    0.625,
-    0.8,
-    0.4444444444444444,
-    0.5,
-    0.42857142857142855,
-    0.0,
-    0.7777777777777778,
-    0.75,
-    1.0,
-    0.75,
-    0.6666666666666666,
-    0.5,
-    0.7777777777777778,
-    0.5,
-    0.8571428571428571,
-    0.7777777777777778,
-    0.0,
-    1.0,
-    0.6666666666666666,
-    0.7142857142857143,
-    1.0,
-    0.6666666666666666,
-    0.75,
-    0.8571428571428571,
-    0.8333333333333334,
-    0.75,
-    1.0,
-    0.0},
-   cuvs::distance::DistanceType::JaccardExpanded,
-   0.0},
-
-  {10,
-   {0, 5, 11, 15, 20, 27, 32, 36, 43, 47, 50},
-   {0, 1, 3, 6, 8, 0, 1, 2, 3, 5, 6, 1, 2, 4, 8, 0, 2, 3, 4, 7, 0, 1, 2, 3, 4,
-    6, 8, 0, 1, 2, 5, 7, 1, 5, 8, 9, 0, 1, 2, 5, 6, 8, 9, 2, 4, 5, 7, 0, 3, 9},  // indices
-   {0.5438, 0.2695, 0.4377, 0.7174, 0.9251, 0.7648, 0.3322, 0.7279, 0.4131, 0.5167,
-    0.8655, 0.0730, 0.0291, 0.9036, 0.7988, 0.5019, 0.7663, 0.2190, 0.8206, 0.3625,
-    0.0411, 0.3995, 0.5688, 0.7028, 0.8706, 0.3199, 0.4431, 0.0535, 0.2225, 0.8853,
-    0.1932, 0.3761, 0.3379, 0.1771, 0.2107, 0.228,  0.5279, 0.4885, 0.3495, 0.5079,
-    0.2325, 0.2331, 0.3018, 0.6231, 0.2645, 0.8429, 0.6625, 0.0797, 0.2724, 0.4218},
-   {0.0,
-    3.3954660629919076,
-    5.6469232737388815,
-    6.373112846266441,
-    4.0212880272531715,
-    6.916281504639404,
-    5.741508386786526,
-    5.411470999663036,
-    9.0,
-    4.977014354725805,
-    3.3954660629919076,
-    0.0,
-    7.56256082439209,
-    5.540261147481582,
-    4.832322929216881,
-    4.62003193872216,
-    6.498056792320361,
-    4.309846252268695,
-    6.317531174829905,
-    6.016362684141827,
-    5.6469232737388815,
-    7.56256082439209,
-    0.0,
-    5.974878731322299,
-    4.898357301336036,
-    6.442097410320605,
-    5.227077347287883,
-    7.134101195584642,
-    5.457753923371659,
-    7.0,
-    6.373112846266441,
-    5.540261147481582,
-    5.974878731322299,
-    0.0,
-    5.5507273748583,
-    4.897749658726415,
-    9.0,
-    8.398776718824767,
-    3.908281400328807,
-    4.83431066343688,
-    4.0212880272531715,
-    4.832322929216881,
-    4.898357301336036,
-    5.5507273748583,
-    0.0,
-    6.632989819428174,
-    7.438852294822894,
-    5.6631570310967465,
-    7.579428202635459,
-    6.760811985364303,
-    6.916281504639404,
-    4.62003193872216,
-    6.442097410320605,
-    4.897749658726415,
-    6.632989819428174,
-    0.0,
-    5.249404187382862,
-    6.072559523278559,
-    4.07661278488929,
-    6.19678948003145,
-    5.741508386786526,
-    6.498056792320361,
-    5.227077347287883,
-    9.0,
-    7.438852294822894,
-    5.249404187382862,
-    0.0,
-    3.854811639654704,
-    6.652724827169063,
-    5.298236851430971,
-    5.411470999663036,
-    4.309846252268695,
-    7.134101195584642,
-    8.398776718824767,
-    5.6631570310967465,
-    6.072559523278559,
-    3.854811639654704,
-    0.0,
-    7.529184598969917,
-    6.903282911791188,
-    9.0,
-    6.317531174829905,
-    5.457753923371659,
-    3.908281400328807,
-    7.579428202635459,
-    4.07661278488929,
-    6.652724827169063,
-    7.529184598969917,
-    0.0,
-    7.0,
-    4.977014354725805,
-    6.016362684141827,
-    7.0,
-    4.83431066343688,
-    6.760811985364303,
-    6.19678948003145,
-    5.298236851430971,
-    6.903282911791188,
-    7.0,
-    0.0},
-   cuvs::distance::DistanceType::Canberra,
-   0.0},
-
-  {10,
-   {0, 5, 11, 15, 20, 27, 32, 36, 43, 47, 50},
-   {0, 1, 3, 6, 8, 0, 1, 2, 3, 5, 6, 1, 2, 4, 8, 0, 2, 3, 4, 7, 0, 1, 2, 3, 4,
-    6, 8, 0, 1, 2, 5, 7, 1, 5, 8, 9, 0, 1, 2, 5, 6, 8, 9, 2, 4, 5, 7, 0, 3, 9},  // indices
-   {0.5438, 0.2695, 0.4377, 0.7174, 0.9251, 0.7648, 0.3322, 0.7279, 0.4131, 0.5167,
-    0.8655, 0.0730, 0.0291, 0.9036, 0.7988, 0.5019, 0.7663, 0.2190, 0.8206, 0.3625,
-    0.0411, 0.3995, 0.5688, 0.7028, 0.8706, 0.3199, 0.4431, 0.0535, 0.2225, 0.8853,
-    0.1932, 0.3761, 0.3379, 0.1771, 0.2107, 0.228,  0.5279, 0.4885, 0.3495, 0.5079,
-    0.2325, 0.2331, 0.3018, 0.6231, 0.2645, 0.8429, 0.6625, 0.0797, 0.2724, 0.4218},
-   {0.0,
-    1.31462855332296,
-    1.3690307816129905,
-    1.698603990921237,
-    1.3460470789553531,
-    1.6636670712582544,
-    1.2651744044972217,
-    1.1938329352055201,
-    1.8811409082590185,
-    1.3653115050624267,
-    1.31462855332296,
-    0.0,
-    1.9447722703291133,
-    1.42818777206562,
-    1.4685491458946494,
-    1.3071999866010466,
-    1.4988622861692171,
-    0.9698559287406783,
-    1.4972023224597841,
-    1.5243383567266802,
-    1.3690307816129905,
-    1.9447722703291133,
-    0.0,
-    1.2748400840107568,
-    1.0599569946448246,
-    1.546591282841402,
-    1.147526531928459,
-    1.447002179128145,
-    1.5982242387673176,
-    1.3112533607072414,
-    1.698603990921237,
-    1.42818777206562,
-    1.2748400840107568,
-    0.0,
-    1.038121552545461,
-    1.011788365364402,
-    1.3907391109256988,
-    1.3128200942311496,
-    1.19595706584447,
-    1.3233328139624725,
-    1.3460470789553531,
-    1.4685491458946494,
-    1.0599569946448246,
-    1.038121552545461,
-    0.0,
-    1.3642741698145529,
-    1.3493868683808095,
-    1.394942694628328,
-    1.572881849642552,
-    1.380122665319464,
-    1.6636670712582544,
-    1.3071999866010466,
-    1.546591282841402,
-    1.011788365364402,
-    1.3642741698145529,
-    0.0,
-    1.018961640373018,
-    1.0114394258945634,
-    0.8338711034820684,
-    1.1247823842299223,
-    1.2651744044972217,
-    1.4988622861692171,
-    1.147526531928459,
-    1.3907391109256988,
-    1.3493868683808095,
-    1.018961640373018,
-    0.0,
-    0.7701238110357329,
-    1.245486437864406,
-    0.5551259549534626,
-    1.1938329352055201,
-    0.9698559287406783,
-    1.447002179128145,
-    1.3128200942311496,
-    1.394942694628328,
-    1.0114394258945634,
-    0.7701238110357329,
-    0.0,
-    1.1886800117391216,
-    1.0083692448135637,
-    1.8811409082590185,
-    1.4972023224597841,
-    1.5982242387673176,
-    1.19595706584447,
-    1.572881849642552,
-    0.8338711034820684,
-    1.245486437864406,
-    1.1886800117391216,
-    0.0,
-    1.3661374102525012,
-    1.3653115050624267,
-    1.5243383567266802,
-    1.3112533607072414,
-    1.3233328139624725,
-    1.380122665319464,
-    1.1247823842299223,
-    0.5551259549534626,
-    1.0083692448135637,
-    1.3661374102525012,
-    0.0},
-   cuvs::distance::DistanceType::LpUnexpanded,
-   2.0},
-
-  {10,
-   {0, 5, 11, 15, 20, 27, 32, 36, 43, 47, 50},
-   {0, 1, 3, 6, 8, 0, 1, 2, 3, 5, 6, 1, 2, 4, 8, 0, 2, 3, 4, 7, 0, 1, 2, 3, 4,
-    6, 8, 0, 1, 2, 5, 7, 1, 5, 8, 9, 0, 1, 2, 5, 6, 8, 9, 2, 4, 5, 7, 0, 3, 9},  // indices
-   {0.5438, 0.2695, 0.4377, 0.7174, 0.9251, 0.7648, 0.3322, 0.7279, 0.4131, 0.5167,
-    0.8655, 0.0730, 0.0291, 0.9036, 0.7988, 0.5019, 0.7663, 0.2190, 0.8206, 0.3625,
-    0.0411, 0.3995, 0.5688, 0.7028, 0.8706, 0.3199, 0.4431, 0.0535, 0.2225, 0.8853,
-    0.1932, 0.3761, 0.3379, 0.1771, 0.2107, 0.228,  0.5279, 0.4885, 0.3495, 0.5079,
-    0.2325, 0.2331, 0.3018, 0.6231, 0.2645, 0.8429, 0.6625, 0.0797, 0.2724, 0.4218},
-   {0.0,
-    0.9251771844789913,
-    0.9036452083899731,
-    0.9251771844789913,
-    0.8706483735804971,
-    0.9251771844789913,
-    0.717493881903289,
-    0.6920214832303888,
-    0.9251771844789913,
-    0.9251771844789913,
-    0.9251771844789913,
-    0.0,
-    0.9036452083899731,
-    0.8655339692155823,
-    0.8706483735804971,
-    0.8655339692155823,
-    0.8655339692155823,
-    0.6329837991017668,
-    0.8655339692155823,
-    0.8655339692155823,
-    0.9036452083899731,
-    0.9036452083899731,
-    0.0,
-    0.7988276152181608,
-    0.7028075145996631,
-    0.9036452083899731,
-    0.9036452083899731,
-    0.9036452083899731,
-    0.8429599432532096,
-    0.9036452083899731,
-    0.9251771844789913,
-    0.8655339692155823,
-    0.7988276152181608,
-    0.0,
-    0.48376552205293305,
-    0.8206394616536681,
-    0.8206394616536681,
-    0.8206394616536681,
-    0.8429599432532096,
-    0.8206394616536681,
-    0.8706483735804971,
-    0.8706483735804971,
-    0.7028075145996631,
-    0.48376552205293305,
-    0.0,
-    0.8706483735804971,
-    0.8706483735804971,
-    0.8706483735804971,
-    0.8429599432532096,
-    0.8706483735804971,
-    0.9251771844789913,
-    0.8655339692155823,
-    0.9036452083899731,
-    0.8206394616536681,
-    0.8706483735804971,
-    0.0,
-    0.8853924473642432,
-    0.535821510936138,
-    0.6497196601457607,
-    0.8853924473642432,
-    0.717493881903289,
-    0.8655339692155823,
-    0.9036452083899731,
-    0.8206394616536681,
-    0.8706483735804971,
-    0.8853924473642432,
-    0.0,
-    0.5279604218147174,
-    0.6658348373853169,
-    0.33799874888632914,
-    0.6920214832303888,
-    0.6329837991017668,
-    0.9036452083899731,
-    0.8206394616536681,
-    0.8706483735804971,
-    0.535821510936138,
-    0.5279604218147174,
-    0.0,
-    0.662579808115858,
-    0.5079750812968089,
-    0.9251771844789913,
-    0.8655339692155823,
-    0.8429599432532096,
-    0.8429599432532096,
-    0.8429599432532096,
-    0.6497196601457607,
-    0.6658348373853169,
-    0.662579808115858,
-    0.0,
-    0.8429599432532096,
-    0.9251771844789913,
-    0.8655339692155823,
-    0.9036452083899731,
-    0.8206394616536681,
-    0.8706483735804971,
-    0.8853924473642432,
-    0.33799874888632914,
-    0.5079750812968089,
-    0.8429599432532096,
-    0.0},
-   cuvs::distance::DistanceType::Linf,
-   0.0},
-
-  {15,
-   {0, 5, 8, 9, 15, 20, 26, 31, 34, 38, 45},
-   {0, 1, 5,  6, 9, 1,  4,  14, 7, 3,  4,  7, 9, 11, 14, 0, 3, 7, 8, 12, 0,  2, 5,
-    7, 8, 14, 4, 9, 10, 11, 13, 4, 10, 14, 5, 6, 8,  9,  0, 2, 3, 4, 6,  10, 11},
-   {0.13537497, 0.51440163, 0.17231936, 0.02417618, 0.15372786, 0.17760507, 0.73789274, 0.08450219,
-    1.,         0.20184723, 0.18036963, 0.12581403, 0.13867603, 0.24040536, 0.11288773, 0.00290246,
-    0.09120187, 0.31190555, 0.43245423, 0.16153588, 0.3233026,  0.05279589, 0.1387149,  0.05962761,
-    0.41751856, 0.00804045, 0.03262381, 0.27507131, 0.37245804, 0.16378881, 0.15605804, 0.3867739,
-    0.24908977, 0.36413632, 0.37643732, 0.28910679, 0.0198409,  0.31461499, 0.24412279, 0.08327667,
-    0.04444576, 0.05047969, 0.26190054, 0.2077349,  0.10803964},
-   {1.05367121e-08, 8.35309089e-01, 1.00000000e+00, 9.24116813e-01,
-    9.90039274e-01, 7.97613546e-01, 8.91271059e-01, 1.00000000e+00,
-    6.64669302e-01, 8.59439512e-01, 8.35309089e-01, 1.05367121e-08,
-    1.00000000e+00, 7.33151506e-01, 1.00000000e+00, 9.86880955e-01,
-    9.19154851e-01, 5.38849774e-01, 1.00000000e+00, 8.98332369e-01,
-    1.00000000e+00, 1.00000000e+00, 0.00000000e+00, 8.03303970e-01,
-    6.64465915e-01, 8.69374690e-01, 1.00000000e+00, 1.00000000e+00,
-    1.00000000e+00, 1.00000000e+00, 9.24116813e-01, 7.33151506e-01,
-    8.03303970e-01, 0.00000000e+00, 8.16225843e-01, 9.39818306e-01,
-    7.27700415e-01, 7.30155528e-01, 8.89451011e-01, 8.05419635e-01,
-    9.90039274e-01, 1.00000000e+00, 6.64465915e-01, 8.16225843e-01,
-    0.00000000e+00, 6.38804490e-01, 1.00000000e+00, 1.00000000e+00,
-    9.52559809e-01, 9.53789212e-01, 7.97613546e-01, 9.86880955e-01,
-    8.69374690e-01, 9.39818306e-01, 6.38804490e-01, 0.0,
-    1.00000000e+00, 9.72569112e-01, 8.24907516e-01, 8.07933016e-01,
-    8.91271059e-01, 9.19154851e-01, 1.00000000e+00, 7.27700415e-01,
-    1.00000000e+00, 1.00000000e+00, 0.00000000e+00, 7.63596268e-01,
-    8.40131263e-01, 7.40428532e-01, 1.00000000e+00, 5.38849774e-01,
-    1.00000000e+00, 7.30155528e-01, 1.00000000e+00, 9.72569112e-01,
-    7.63596268e-01, 0.00000000e+00, 1.00000000e+00, 7.95485011e-01,
-    6.64669302e-01, 1.00000000e+00, 1.00000000e+00, 8.89451011e-01,
-    9.52559809e-01, 8.24907516e-01, 8.40131263e-01, 1.00000000e+00,
-    0.00000000e+00, 8.51370877e-01, 8.59439512e-01, 8.98332369e-01,
-    1.00000000e+00, 8.05419635e-01, 9.53789212e-01, 8.07933016e-01,
-    7.40428532e-01, 7.95485011e-01, 8.51370877e-01, 1.49011612e-08},
-   // Dataset is L1 normalized into pdfs
-   cuvs::distance::DistanceType::HellingerExpanded,
-   0.0},
-
-  {4,
-   {0, 1, 1, 2, 4},
-   {3, 2, 0, 1},  // indices
-   {0.99296, 0.42180, 0.11687, 0.305869},
-   {
-     // dense output
-     0.0,
-     0.99296,
-     1.41476,
-     1.415707,
-     0.99296,
-     0.0,
-     0.42180,
-     0.42274,
-     1.41476,
-     0.42180,
-     0.0,
-     0.84454,
-     1.41570,
-     0.42274,
-     0.84454,
-     0.0,
-   },
-   cuvs::distance::DistanceType::L1,
-   0.0},
-  {5,
-   {0, 3, 8, 12, 16, 20, 25, 30, 35, 40, 45},
-   {0, 3, 4, 0, 1, 2, 3, 4, 1, 2, 3, 4, 0, 2, 3, 4, 0, 1, 3, 4, 0, 1, 2,
-    3, 4, 0, 1, 2, 3, 4, 0, 1, 2, 3, 4, 0, 1, 2, 3, 4, 0, 1, 2, 3, 4},
-   {0.70862347, 0.8232774,  0.12108795, 0.84527547, 0.94937088, 0.03258545, 0.99584118, 0.76835667,
-    0.34426657, 0.2357925,  0.01274851, 0.11422017, 0.3437756,  0.31967718, 0.5956055,  0.31610373,
-    0.04147273, 0.03724415, 0.21515727, 0.04751052, 0.50283183, 0.99957274, 0.01395933, 0.96032529,
-    0.88438711, 0.46095378, 0.27432481, 0.54294211, 0.54280225, 0.59503329, 0.61364678, 0.22837736,
-    0.56609561, 0.29809423, 0.76736686, 0.56460608, 0.98165371, 0.02140123, 0.19881268, 0.26057815,
-    0.31648823, 0.89874295, 0.27366735, 0.5119944,  0.11416134},
-   {// dense output
-    0.,         0.48769777, 1.88014197, 0.26127048, 0.26657011, 0.7874794,  0.76962708, 1.122858,
-    1.1232498,  1.08166081, 0.48769777, 0.,         1.31332116, 0.98318907, 0.42661815, 0.09279052,
-    1.35187836, 1.38429055, 0.40658897, 0.56136388, 1.88014197, 1.31332116, 0.,         1.82943642,
-    1.54826077, 1.05918884, 1.59360067, 1.34698954, 0.60215168, 0.46993848, 0.26127048, 0.98318907,
-    1.82943642, 0.,         0.29945563, 1.08494093, 0.22934281, 0.82801925, 1.74288748, 1.50610116,
-    0.26657011, 0.42661815, 1.54826077, 0.29945563, 0.,         0.45060069, 0.77814948, 1.45245711,
-    1.18328348, 0.82486987, 0.7874794,  0.09279052, 1.05918884, 1.08494093, 0.45060069, 0.,
-    1.29899154, 1.40683824, 0.48505269, 0.53862363, 0.76962708, 1.35187836, 1.59360067, 0.22934281,
-    0.77814948, 1.29899154, 0.,         0.33202426, 1.92108999, 1.88812175, 1.122858,   1.38429055,
-    1.34698954, 0.82801925, 1.45245711, 1.40683824, 0.33202426, 0.,         1.47318624, 1.92660889,
-    1.1232498,  0.40658897, 0.60215168, 1.74288748, 1.18328348, 0.48505269, 1.92108999, 1.47318624,
-    0.,         0.24992619, 1.08166081, 0.56136388, 0.46993848, 1.50610116, 0.82486987, 0.53862363,
-    1.88812175, 1.92660889, 0.24992619, 0.},
-   cuvs::distance::DistanceType::CorrelationExpanded,
-   0.0},
-  {5,
-   {0, 1, 2, 4, 4, 5, 6, 7, 9, 9, 10},
-   {1, 4, 0, 4, 1, 3, 0, 1, 3, 0},
-   {1., 1., 1., 1., 1., 1., 1., 1., 1., 1.},
-   {// dense output
-    0.,  1.,  1.,  1., 0.8, 1., 1.,  0.8, 1., 1.,  1.,  0., 0.8, 1., 1.,  1.,  1.,  1.,  1., 1.,
-    1.,  0.8, 0.,  1., 1.,  1., 0.8, 1.,  1., 0.8, 1.,  1., 1.,  0., 1.,  1.,  1.,  1.,  1., 1.,
-    0.8, 1.,  1.,  1., 0.,  1., 1.,  0.8, 1., 1.,  1.,  1., 1.,  1., 1.,  0.,  1.,  0.8, 1., 1.,
-    1.,  1.,  0.8, 1., 1.,  1., 0.,  1.,  1., 0.8, 0.8, 1., 1.,  1., 0.8, 0.8, 1.,  0.,  1., 1.,
-    1.,  1.,  1.,  1., 1.,  1., 1.,  1.,  0., 1.,  1.,  1., 0.8, 1., 1.,  1.,  0.8, 1.,  1., 0.},
-   cuvs::distance::DistanceType::RusselRaoExpanded,
-   0.0},
-  {5,
-   {0, 1, 1, 3, 3, 4, 4, 6, 9, 10, 10},
-   {0, 3, 4, 4, 2, 3, 0, 2, 3, 2},
-   {1., 1., 1., 1., 1., 1., 1., 1., 1., 1.},
-   {// dense output
-    0.,  0.2, 0.6, 0.2, 0.4, 0.2, 0.6, 0.4, 0.4, 0.2, 0.2, 0.,  0.4, 0.,  0.2, 0.,  0.4,
-    0.6, 0.2, 0.,  0.6, 0.4, 0.,  0.4, 0.2, 0.4, 0.4, 0.6, 0.6, 0.4, 0.2, 0.,  0.4, 0.,
-    0.2, 0.,  0.4, 0.6, 0.2, 0.,  0.4, 0.2, 0.2, 0.2, 0.,  0.2, 0.6, 0.8, 0.4, 0.2, 0.2,
-    0.,  0.4, 0.,  0.2, 0.,  0.4, 0.6, 0.2, 0.,  0.6, 0.4, 0.4, 0.4, 0.6, 0.4, 0.,  0.2,
-    0.2, 0.4, 0.4, 0.6, 0.6, 0.6, 0.8, 0.6, 0.2, 0.,  0.4, 0.6, 0.4, 0.2, 0.6, 0.2, 0.4,
-    0.2, 0.2, 0.4, 0.,  0.2, 0.2, 0.,  0.4, 0.,  0.2, 0.,  0.4, 0.6, 0.2, 0.},
-   cuvs::distance::DistanceType::HammingUnexpanded,
-   0.0},
-  {3,
-   {0, 1, 2},
-   {0, 1},
-   {1.0, 1.0},
-   {0.0, 0.83255, 0.83255, 0.0},
-   cuvs::distance::DistanceType::JensenShannon,
-   0.0},
-  {2,
-   {0, 1, 3},
-   {0, 0, 1},
-   {1.0, 0.5, 0.5},
-   {0, 0.4645014, 0.4645014, 0},
-   cuvs::distance::DistanceType::JensenShannon,
-   0.0},
-  {3,
-   {0, 1, 2},
-   {0, 0},
-   {1.0, 1.0},
-   {0.0, 0.0, 0.0, 0.0},
-   cuvs::distance::DistanceType::JensenShannon,
-   0.0},
-
-  {3,
-   {0, 1, 2},
-   {0, 1},
-   {1.0, 1.0},
-   {0.0, 1.0, 1.0, 0.0},
-   cuvs::distance::DistanceType::DiceExpanded,
-   0.0},
-  {3,
-   {0, 1, 3},
-   {0, 0, 1},
-   {1.0, 1.0, 1.0},
-   {0, 0.333333, 0.333333, 0},
-   cuvs::distance::DistanceType::DiceExpanded,
-   0.0},
-
-};
-
-typedef SparseDistanceTest<int, float> SparseDistanceTestF;
-TEST_P(SparseDistanceTestF, Result) { compare(); }
-INSTANTIATE_TEST_CASE_P(SparseDistanceTests,
-                        SparseDistanceTestF,
-                        ::testing::ValuesIn(inputs_i32_f));
-
-};  // namespace distance
-};  // end namespace sparse
-};  // end namespace raft
diff --git a/cpp/test/sparse/gram.cu b/cpp/test/sparse/gram.cu
deleted file mode 100644
index fb7a9c61e..000000000
--- a/cpp/test/sparse/gram.cu
+++ /dev/null
@@ -1,328 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#if defined RAFT_DISTANCE_COMPILED
-#include <cuvs/distance/specializations.cuh>
-#include <raft/core/resource/cuda_stream.hpp>
-#endif
-
-#include "../distance/gram_base.cuh"
-#include "../test_utils.cuh"
-#include <cuvs/distance/distance_types.hpp>
-#include <cuvs/distance/kernels.cuh>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <memory>
-#include <raft/random/rng.cuh>
-#include <raft/sparse/convert/dense.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <raft/util/itertools.hpp>
-#include <rmm/device_uvector.hpp>
-
-namespace cuvs::distance::kernels {
-
-/**
- * Structure to describe structure of the input matrices:
- *  - DENSE: dense, dense
- *  - MIX: CSR, dense
- *  - CSR: CSR, CSR
- */
-enum SparseType { DENSE, MIX, CSR };
-
-struct GramMatrixInputs {
-  int n1;      // feature vectors in matrix 1
-  int n2;      // featuer vectors in matrix 2
-  int n_cols;  // number of elements in a feature vector
-  bool is_row_major;
-  SparseType sparse_input;
-  KernelParams kernel;
-  int ld1;
-  int ld2;
-  int ld_out;
-  // We will generate random input using the dimensions given here.
-  // The reference output is calculated by a custom kernel.
-};
-
-std::ostream& operator<<(std::ostream& os, const GramMatrixInputs& p)
-{
-  std::vector<std::string> kernel_names{"linear", "poly", "rbf", "tanh"};
-  os << "/" << p.n1 << "x" << p.n2 << "x" << p.n_cols << "/"
-     << (p.is_row_major ? "RowMajor/" : "ColMajor/")
-     << (p.sparse_input == SparseType::DENSE
-           ? "DenseDense/"
-           : (p.sparse_input == SparseType::MIX ? "CsrDense/" : "CsrCsr/"))
-     << kernel_names[p.kernel.kernel] << "/ld_" << p.ld1 << "x" << p.ld2 << "x" << p.ld_out;
-  return os;
-}
-
-/*struct KernelParams {
-  // Kernel function parameters
-  KernelType kernel;  //!< Type of the kernel function
-  int degree;         //!< Degree of polynomial kernel (ignored by others)
-  double gamma;       //!< multiplier in the
-  double coef0;       //!< additive constant in poly and tanh kernels
-};*/
-
-// const KernelParams linear_kernel_params{.kernel=KernelType::LINEAR};
-
-// {KernelType::POLYNOMIAL, 2, 0.5, 2.4}, {KernelType::TANH, 0, 0.5, 2.4}, {KernelType::RBF, 0, 0.5}
-const std::vector<GramMatrixInputs> inputs = raft::util::itertools::product<GramMatrixInputs>(
-  {42},
-  {137},
-  {2},
-  {true, false},
-  {SparseType::DENSE, SparseType::MIX, SparseType::CSR},
-  {KernelParams{KernelType::LINEAR},
-   KernelParams{KernelType::POLYNOMIAL, 2, 0.5, 2.4},
-   KernelParams{KernelType::TANH, 0, 0.5, 2.4},
-   KernelParams{KernelType::RBF, 0, 0.5}});
-
-// (ld_1, ld_2, ld_out) not supported by RBF and CSR
-const std::vector<GramMatrixInputs> inputs_ld = raft::util::itertools::product<GramMatrixInputs>(
-  {137},
-  {42},
-  {2},
-  {true, false},
-  {SparseType::DENSE, SparseType::MIX},
-  {KernelParams{KernelType::LINEAR},
-   KernelParams{KernelType::POLYNOMIAL, 2, 0.5, 2.4},
-   KernelParams{KernelType::TANH, 0, 0.5, 2.4}},
-  {159},
-  {73},
-  {144});
-
-// (ld_1, ld_2) are supported by CSR
-const std::vector<GramMatrixInputs> inputs_ld_csr =
-  raft::util::itertools::product<GramMatrixInputs>(
-    {42},
-    {137},
-    {2},
-    {true, false},
-    {SparseType::CSR, SparseType::MIX},
-    {KernelParams{KernelType::LINEAR},
-     KernelParams{KernelType::POLYNOMIAL, 2, 0.5, 2.4},
-     KernelParams{KernelType::TANH, 0, 0.5, 2.4}},
-    {64},
-    {155},
-    {0});
-
-template <typename math_t>
-class GramMatrixTest : public ::testing::TestWithParam<GramMatrixInputs> {
- protected:
-  GramMatrixTest()
-    : params(GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      x1(0, stream),
-      x2(0, stream),
-      x1_csr_indptr(0, stream),
-      x1_csr_indices(0, stream),
-      x1_csr_data(0, stream),
-      x2_csr_indptr(0, stream),
-      x2_csr_indices(0, stream),
-      x2_csr_data(0, stream),
-      gram(0, stream),
-      gram_host(0)
-  {
-    if (params.ld1 == 0) { params.ld1 = params.is_row_major ? params.n_cols : params.n1; }
-    if (params.ld2 == 0) { params.ld2 = params.is_row_major ? params.n_cols : params.n2; }
-    if (params.ld_out == 0) { params.ld_out = params.is_row_major ? params.n2 : params.n1; }
-    // Derive the size of the output from the offset of the last element.
-    size_t size = get_offset(params.n1 - 1, params.n_cols - 1, params.ld1, params.is_row_major) + 1;
-    x1.resize(size, stream);
-    size = get_offset(params.n2 - 1, params.n_cols - 1, params.ld2, params.is_row_major) + 1;
-    x2.resize(size, stream);
-    size = get_offset(params.n1 - 1, params.n2 - 1, params.ld_out, params.is_row_major) + 1;
-
-    gram.resize(size, stream);
-    RAFT_CUDA_TRY(cudaMemsetAsync(gram.data(), 0, gram.size() * sizeof(math_t), stream));
-    gram_host.resize(gram.size());
-    std::fill(gram_host.begin(), gram_host.end(), 0);
-
-    raft::random::RngState r(42137ULL);
-    raft::random::uniform(handle, r, x1.data(), x1.size(), math_t(0), math_t(1));
-    raft::random::uniform(handle, r, x2.data(), x2.size(), math_t(0), math_t(1));
-
-    RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
-  }
-
-  ~GramMatrixTest() override {}
-
-  int prepareCsr(math_t* dense, int n_rows, int ld, int* indptr, int* indices, math_t* data)
-  {
-    int nnz           = 0;
-    double eps        = 1e-6;
-    int n_cols        = params.n_cols;
-    bool is_row_major = params.is_row_major;
-    size_t dense_size = get_offset(n_rows - 1, n_cols - 1, ld, is_row_major) + 1;
-
-    std::vector<math_t> dense_host(dense_size);
-    raft::update_host(dense_host.data(), dense, dense_size, stream);
-    resource::sync_stream(handle, stream);
-
-    std::vector<int> indptr_host(n_rows + 1);
-    std::vector<int> indices_host(n_rows * n_cols);
-    std::vector<math_t> data_host(n_rows * n_cols);
-
-    // create csr matrix from dense (with threshold)
-    for (int i = 0; i < n_rows; ++i) {
-      indptr_host[i] = nnz;
-      for (int j = 0; j < n_cols; ++j) {
-        math_t value = dense_host[get_offset(i, j, ld, is_row_major)];
-        if (value > eps) {
-          indices_host[nnz] = j;
-          data_host[nnz]    = value;
-          nnz++;
-        }
-      }
-    }
-    indptr_host[n_rows] = nnz;
-
-    // fill back dense matrix from CSR
-    std::fill(dense_host.data(), dense_host.data() + dense_size, 0);
-    for (int i = 0; i < n_rows; ++i) {
-      for (int idx = indptr_host[i]; idx < indptr_host[i + 1]; ++idx) {
-        dense_host[get_offset(i, indices_host[idx], ld, is_row_major)] = data_host[idx];
-      }
-    }
-
-    raft::update_device(dense, dense_host.data(), dense_size, stream);
-    raft::update_device(indptr, indptr_host.data(), n_rows + 1, stream);
-    raft::update_device(indices, indices_host.data(), nnz, stream);
-    raft::update_device(data, data_host.data(), nnz, stream);
-    resource::sync_stream(handle, stream);
-    return nnz;
-  }
-
-  void runTest()
-  {
-    std::unique_ptr<GramMatrixBase<math_t>> kernel =
-      std::unique_ptr<GramMatrixBase<math_t>>(KernelFactory<math_t>::create(params.kernel));
-
-    auto x1_span =
-      params.is_row_major
-        ? raft::make_device_strided_matrix_view<const math_t, int, raft::layout_c_contiguous>(
-            x1.data(), params.n1, params.n_cols, params.ld1)
-        : raft::make_device_strided_matrix_view<const math_t, int, raft::layout_f_contiguous>(
-            x1.data(), params.n1, params.n_cols, params.ld1);
-    auto x2_span =
-      params.is_row_major
-        ? raft::make_device_strided_matrix_view<const math_t, int, raft::layout_c_contiguous>(
-            x2.data(), params.n2, params.n_cols, params.ld2)
-        : raft::make_device_strided_matrix_view<const math_t, int, raft::layout_f_contiguous>(
-            x2.data(), params.n2, params.n_cols, params.ld2);
-    auto out_span =
-      params.is_row_major
-        ? raft::make_device_strided_matrix_view<math_t, int, raft::layout_c_contiguous>(
-            gram.data(), params.n1, params.n2, params.ld_out)
-        : raft::make_device_strided_matrix_view<math_t, int, raft::layout_f_contiguous>(
-            gram.data(), params.n1, params.n2, params.ld_out);
-
-    if (params.sparse_input == SparseType::DENSE) {
-      (*kernel)(handle, x1_span, x2_span, out_span);
-    } else {
-      x1_csr_indptr.reserve(params.n1 + 1, stream);
-      x1_csr_indices.reserve(params.n1 * params.n_cols, stream);
-      x1_csr_data.reserve(params.n1 * params.n_cols, stream);
-      int x1_nnz = prepareCsr(x1.data(),
-                              params.n1,
-                              params.ld1,
-                              x1_csr_indptr.data(),
-                              x1_csr_indices.data(),
-                              x1_csr_data.data());
-
-      auto x1_csr_structure = raft::make_device_compressed_structure_view<int, int, int>(
-        x1_csr_indptr.data(), x1_csr_indices.data(), params.n1, params.n_cols, x1_nnz);
-      auto x1_csr = raft::device_csr_matrix_view<const math_t, int, int, int>(
-        raft::device_span<const math_t>(x1_csr_data.data(), x1_csr_structure.get_nnz()),
-        x1_csr_structure);
-
-      if (params.sparse_input == SparseType::MIX) {
-        (*kernel)(handle, x1_csr, x2_span, out_span);
-      } else {
-        x2_csr_indptr.reserve(params.n2 + 1, stream);
-        x2_csr_indices.reserve(params.n2 * params.n_cols, stream);
-        x2_csr_data.reserve(params.n2 * params.n_cols, stream);
-        int x2_nnz = prepareCsr(x2.data(),
-                                params.n2,
-                                params.ld2,
-                                x2_csr_indptr.data(),
-                                x2_csr_indices.data(),
-                                x2_csr_data.data());
-
-        auto x2_csr_structure = raft::make_device_compressed_structure_view<int, int, int>(
-          x2_csr_indptr.data(), x2_csr_indices.data(), params.n2, params.n_cols, x2_nnz);
-        auto x2_csr = raft::device_csr_matrix_view<const math_t, int, int, int>(
-          raft::device_span<const math_t>(x2_csr_data.data(), x2_csr_structure.get_nnz()),
-          x2_csr_structure);
-
-        (*kernel)(handle, x1_csr, x2_csr, out_span);
-      }
-    }
-    // Something in gram is executing not on the 'stream' and therefore
-    // a full device sync is required
-    RAFT_CUDA_TRY(cudaDeviceSynchronize());
-    naiveGramMatrixKernel(params.n1,
-                          params.n2,
-                          params.n_cols,
-                          x1,
-                          x2,
-                          gram_host.data(),
-                          params.ld1,
-                          params.ld2,
-                          params.ld_out,
-                          params.is_row_major,
-                          params.kernel,
-                          stream,
-                          handle);
-    resource::sync_stream(handle, stream);
-
-    ASSERT_TRUE(raft::devArrMatchHost(
-      gram_host.data(), gram.data(), gram.size(), raft::CompareApprox<math_t>(1e-6f), stream));
-  }
-
-  raft::resources handle;
-  cudaStream_t stream = 0;
-  GramMatrixInputs params;
-
-  rmm::device_uvector<math_t> x1;
-  rmm::device_uvector<math_t> x2;
-
-  rmm::device_uvector<int> x1_csr_indptr;
-  rmm::device_uvector<int> x1_csr_indices;
-  rmm::device_uvector<math_t> x1_csr_data;
-  rmm::device_uvector<int> x2_csr_indptr;
-  rmm::device_uvector<int> x2_csr_indices;
-  rmm::device_uvector<math_t> x2_csr_data;
-
-  rmm::device_uvector<math_t> gram;
-  std::vector<math_t> gram_host;
-};
-
-typedef GramMatrixTest<float> GramMatrixTestFloatStandard;
-typedef GramMatrixTest<float> GramMatrixTestFloatLd;
-typedef GramMatrixTest<float> GramMatrixTestFloatLdCsr;
-
-TEST_P(GramMatrixTestFloatStandard, Gram) { runTest(); }
-TEST_P(GramMatrixTestFloatLd, Gram) { runTest(); }
-TEST_P(GramMatrixTestFloatLdCsr, Gram) { runTest(); }
-
-INSTANTIATE_TEST_SUITE_P(GramMatrixTests, GramMatrixTestFloatStandard, ::testing::ValuesIn(inputs));
-INSTANTIATE_TEST_SUITE_P(GramMatrixTests, GramMatrixTestFloatLd, ::testing::ValuesIn(inputs_ld));
-INSTANTIATE_TEST_SUITE_P(GramMatrixTests,
-                         GramMatrixTestFloatLdCsr,
-                         ::testing::ValuesIn(inputs_ld_csr));
-};  // end namespace cuvs::distance::kernels
diff --git a/cpp/test/sparse/neighbors/brute_force.cu b/cpp/test/sparse/neighbors/brute_force.cu
deleted file mode 100644
index 34507ec74..000000000
--- a/cpp/test/sparse/neighbors/brute_force.cu
+++ /dev/null
@@ -1,179 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <cusparse_v2.h>
-#include <gtest/gtest.h>
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include "../../test_utils.cuh"
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/sparse/neighbors/knn.cuh>
-
-#include <raft/util/cudart_utils.hpp>
-
-namespace raft {
-namespace sparse {
-namespace selection {
-
-using namespace raft;
-using namespace raft::sparse;
-
-template <typename value_idx, typename value_t>
-struct SparseKNNInputs {
-  value_idx n_cols;
-
-  std::vector<value_idx> indptr_h;
-  std::vector<value_idx> indices_h;
-  std::vector<value_t> data_h;
-
-  std::vector<value_t> out_dists_ref_h;
-  std::vector<value_idx> out_indices_ref_h;
-
-  int k;
-
-  int batch_size_index = 2;
-  int batch_size_query = 2;
-
-  cuvs::distance::DistanceType metric = cuvs::distance::DistanceType::L2SqrtExpanded;
-};
-
-template <typename value_idx, typename value_t>
-::std::ostream& operator<<(::std::ostream& os, const SparseKNNInputs<value_idx, value_t>& dims)
-{
-  return os;
-}
-
-template <typename value_idx, typename value_t>
-class SparseKNNTest : public ::testing::TestWithParam<SparseKNNInputs<value_idx, value_t>> {
- public:
-  SparseKNNTest()
-    : params(::testing::TestWithParam<SparseKNNInputs<value_idx, value_t>>::GetParam()),
-      indptr(0, resource::get_cuda_stream(handle)),
-      indices(0, resource::get_cuda_stream(handle)),
-      data(0, resource::get_cuda_stream(handle)),
-      out_indices(0, resource::get_cuda_stream(handle)),
-      out_dists(0, resource::get_cuda_stream(handle)),
-      out_indices_ref(0, resource::get_cuda_stream(handle)),
-      out_dists_ref(0, resource::get_cuda_stream(handle))
-  {
-  }
-
- protected:
-  void SetUp() override
-  {
-    n_rows = params.indptr_h.size() - 1;
-    nnz    = params.indices_h.size();
-    k      = params.k;
-
-    make_data();
-
-    raft::sparse::neighbors::brute_force_knn<value_idx, value_t>(indptr.data(),
-                                                                 indices.data(),
-                                                                 data.data(),
-                                                                 nnz,
-                                                                 n_rows,
-                                                                 params.n_cols,
-                                                                 indptr.data(),
-                                                                 indices.data(),
-                                                                 data.data(),
-                                                                 nnz,
-                                                                 n_rows,
-                                                                 params.n_cols,
-                                                                 out_indices.data(),
-                                                                 out_dists.data(),
-                                                                 k,
-                                                                 handle,
-                                                                 params.batch_size_index,
-                                                                 params.batch_size_query,
-                                                                 params.metric);
-
-    RAFT_CUDA_TRY(cudaStreamSynchronize(resource::get_cuda_stream(handle)));
-  }
-
-  void compare()
-  {
-    ASSERT_TRUE(devArrMatch(
-      out_dists_ref.data(), out_dists.data(), n_rows * k, CompareApprox<value_t>(1e-4)));
-    ASSERT_TRUE(
-      devArrMatch(out_indices_ref.data(), out_indices.data(), n_rows * k, Compare<value_idx>()));
-  }
-
- protected:
-  void make_data()
-  {
-    std::vector<value_idx> indptr_h  = params.indptr_h;
-    std::vector<value_idx> indices_h = params.indices_h;
-    std::vector<value_t> data_h      = params.data_h;
-
-    auto stream = resource::get_cuda_stream(handle);
-    indptr.resize(indptr_h.size(), stream);
-    indices.resize(indices_h.size(), stream);
-    data.resize(data_h.size(), stream);
-
-    update_device(indptr.data(), indptr_h.data(), indptr_h.size(), stream);
-    update_device(indices.data(), indices_h.data(), indices_h.size(), stream);
-    update_device(data.data(), data_h.data(), data_h.size(), stream);
-
-    std::vector<value_t> out_dists_ref_h     = params.out_dists_ref_h;
-    std::vector<value_idx> out_indices_ref_h = params.out_indices_ref_h;
-
-    out_indices_ref.resize(out_indices_ref_h.size(), stream);
-    out_dists_ref.resize(out_dists_ref_h.size(), stream);
-
-    update_device(
-      out_indices_ref.data(), out_indices_ref_h.data(), out_indices_ref_h.size(), stream);
-    update_device(out_dists_ref.data(), out_dists_ref_h.data(), out_dists_ref_h.size(), stream);
-
-    out_dists.resize(n_rows * k, stream);
-    out_indices.resize(n_rows * k, stream);
-  }
-
-  raft::resources handle;
-
-  int n_rows, nnz, k;
-
-  // input data
-  rmm::device_uvector<value_idx> indptr, indices;
-  rmm::device_uvector<value_t> data;
-
-  // output data
-  rmm::device_uvector<value_idx> out_indices;
-  rmm::device_uvector<value_t> out_dists;
-
-  rmm::device_uvector<value_idx> out_indices_ref;
-  rmm::device_uvector<value_t> out_dists_ref;
-
-  SparseKNNInputs<value_idx, value_t> params;
-};
-
-const std::vector<SparseKNNInputs<int, float>> inputs_i32_f = {
-  {9,                                                 // ncols
-   {0, 2, 4, 6, 8},                                   // indptr
-   {0, 4, 0, 3, 0, 2, 0, 8},                          // indices
-   {0.0f, 1.0f, 5.0f, 6.0f, 5.0f, 6.0f, 0.0f, 1.0f},  // data
-   {0, 1.41421, 0, 7.87401, 0, 7.87401, 0, 1.41421},  // dists
-   {0, 3, 1, 0, 2, 0, 3, 0},                          // inds
-   2,
-   2,
-   2,
-   cuvs::distance::DistanceType::L2SqrtExpanded}};
-typedef SparseKNNTest<int, float> SparseKNNTestF;
-TEST_P(SparseKNNTestF, Result) { compare(); }
-INSTANTIATE_TEST_CASE_P(SparseKNNTest, SparseKNNTestF, ::testing::ValuesIn(inputs_i32_f));
-
-};  // end namespace selection
-};  // end namespace sparse
-};  // end namespace raft
diff --git a/cpp/test/sparse/neighbors/cross_component_nn.cu b/cpp/test/sparse/neighbors/cross_component_nn.cu
deleted file mode 100644
index d32c832d9..000000000
--- a/cpp/test/sparse/neighbors/cross_component_nn.cu
+++ /dev/null
@@ -1,1036 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-// XXX: We allow the instantiation of masked_l2_nn here:
-// raft::linkage::FixConnectivitiesRedOp<value_idx, value_t> red_op(params.n_row);
-// raft::linkage::cross_component_nn<value_idx, value_t>(
-//   handle, out_edges, data.data(), colors.data(), params.n_row, params.n_col, red_op);
-//
-// TODO: consider adding this to libraft.so or creating an instance in a
-// separate translation unit for this test.
-//
-// TODO: edge case testing. Reference: https://github.com/rapidsai/raft/issues/1669
-
-#include <gtest/gtest.h>
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <cub/cub.cuh>
-
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <vector>
-
-#include <raft/sparse/linalg/symmetrize.cuh>
-#include <raft/sparse/mst/mst.cuh>
-#include <raft/sparse/neighbors/knn_graph.cuh>
-#include <raft/sparse/selection/cross_component_nn.cuh>
-
-#include <cuvs/distance/distance_types.hpp>
-#include <raft/linalg/transpose.cuh>
-#include <raft/sparse/convert/csr.cuh>
-#include <raft/sparse/coo.hpp>
-#include <raft/sparse/hierarchy/single_linkage.cuh>
-#include <rmm/device_uvector.hpp>
-
-#include "../../test_utils.cuh"
-
-namespace raft {
-namespace sparse {
-
-using namespace std;
-
-template <typename value_t, typename value_idx>
-struct ConnectComponentsInputs {
-  value_idx n_row;
-  value_idx n_col;
-  std::vector<value_t> data;
-
-  int c;
-};
-
-template <typename value_idx, typename value_t>
-class ConnectComponentsTest
-  : public ::testing::TestWithParam<ConnectComponentsInputs<value_t, value_idx>> {
- protected:
-  void basicTest()
-  {
-    raft::resources handle;
-
-    auto stream = resource::get_cuda_stream(handle);
-
-    params = ::testing::TestWithParam<ConnectComponentsInputs<value_t, value_idx>>::GetParam();
-
-    raft::sparse::COO<value_t, value_idx> out_edges(resource::get_cuda_stream(handle));
-    raft::sparse::COO<value_t, value_idx> out_edges_batched(resource::get_cuda_stream(handle));
-
-    rmm::device_uvector<value_t> data(params.n_row * params.n_col,
-                                      resource::get_cuda_stream(handle));
-
-    raft::copy(data.data(), params.data.data(), data.size(), resource::get_cuda_stream(handle));
-
-    rmm::device_uvector<value_idx> indptr(params.n_row + 1, stream);
-
-    /**
-     * 1. Construct knn graph
-     */
-    raft::sparse::COO<value_t, value_idx> knn_graph_coo(stream);
-
-    raft::sparse::neighbors::knn_graph(handle,
-                                       data.data(),
-                                       params.n_row,
-                                       params.n_col,
-                                       cuvs::distance::DistanceType::L2SqrtExpanded,
-                                       knn_graph_coo,
-                                       params.c);
-
-    raft::sparse::convert::sorted_coo_to_csr(
-      knn_graph_coo.rows(), knn_graph_coo.nnz, indptr.data(), params.n_row + 1, stream);
-
-    /**
-     * 2. Construct MST, sorted by weights
-     */
-    rmm::device_uvector<value_idx> colors(params.n_row, stream);
-
-    auto mst_coo = raft::mst::mst<value_idx, value_idx, value_t, double>(handle,
-                                                                         indptr.data(),
-                                                                         knn_graph_coo.cols(),
-                                                                         knn_graph_coo.vals(),
-                                                                         params.n_row,
-                                                                         knn_graph_coo.nnz,
-                                                                         colors.data(),
-                                                                         stream,
-                                                                         false,
-                                                                         true);
-
-    /**
-     * 3. cross_component_nn to fix connectivities
-     */
-    raft::linkage::FixConnectivitiesRedOp<value_idx, value_t> red_op(params.n_row);
-    raft::linkage::cross_component_nn<value_idx, value_t>(handle,
-                                                          out_edges,
-                                                          data.data(),
-                                                          colors.data(),
-                                                          params.n_row,
-                                                          params.n_col,
-                                                          red_op,
-                                                          params.n_row,
-                                                          params.n_col);
-
-    raft::linkage::cross_component_nn<value_idx, value_t>(handle,
-                                                          out_edges_batched,
-                                                          data.data(),
-                                                          colors.data(),
-                                                          params.n_row,
-                                                          params.n_col,
-                                                          red_op,
-                                                          params.n_row / 2,
-                                                          params.n_col / 2);
-
-    ASSERT_TRUE(out_edges.nnz == out_edges_batched.nnz);
-
-    ASSERT_TRUE(
-      devArrMatch(out_edges.rows(), out_edges_batched.rows(), out_edges.nnz, Compare<int>()));
-
-    ASSERT_TRUE(
-      devArrMatch(out_edges.cols(), out_edges_batched.cols(), out_edges.nnz, Compare<int>()));
-
-    ASSERT_TRUE(devArrMatch(
-      out_edges.vals(), out_edges_batched.vals(), out_edges.nnz, CompareApprox<float>(1e-4)));
-
-    /**
-     * Construct final edge list
-     */
-    rmm::device_uvector<value_idx> indptr2(params.n_row + 1, stream);
-
-    raft::sparse::convert::sorted_coo_to_csr(
-      out_edges.rows(), out_edges.nnz, indptr2.data(), params.n_row + 1, stream);
-
-    auto output_mst = raft::mst::mst<value_idx, value_idx, value_t>(handle,
-                                                                    indptr2.data(),
-                                                                    out_edges.cols(),
-                                                                    out_edges.vals(),
-                                                                    params.n_row,
-                                                                    out_edges.nnz,
-                                                                    colors.data(),
-                                                                    stream,
-                                                                    false,
-                                                                    false);
-
-    resource::sync_stream(handle, stream);
-
-    // The sum of edges for both MST runs should be n_rows - 1
-    final_edges = output_mst.n_edges + mst_coo.n_edges;
-  }
-
-  void SetUp() override { basicTest(); }
-
-  void TearDown() override {}
-
- protected:
-  ConnectComponentsInputs<value_t, value_idx> params;
-
-  value_idx final_edges;
-};
-
-const std::vector<ConnectComponentsInputs<float, int>> fix_conn_inputsf2 = {
-  // Test n_clusters == n_points
-  {10,
-   5,
-   {0.21390334, 0.50261639, 0.91036676, 0.59166485, 0.71162682, 0.10248392, 0.77782677, 0.43772379,
-    0.4035871,  0.3282796,  0.47544681, 0.59862974, 0.12319357, 0.06239463, 0.28200272, 0.1345717,
-    0.50498218, 0.5113505,  0.16233086, 0.62165332, 0.42281548, 0.933117,   0.41386077, 0.23264562,
-    0.73325968, 0.37537541, 0.70719873, 0.14522645, 0.73279625, 0.9126674,  0.84854131, 0.28890216,
-    0.85267903, 0.74703138, 0.83842071, 0.34942792, 0.27864171, 0.70911132, 0.21338564, 0.32035554,
-    0.73788331, 0.46926692, 0.57570162, 0.42559178, 0.87120209, 0.22734951, 0.01847905, 0.75549396,
-    0.76166195, 0.66613745},
-   -1},
-  // Test n_points == 100
-  {100,
-   10,
-   {6.26168372e-01, 9.30437651e-01, 6.02450208e-01, 2.73025296e-01, 9.53050619e-01, 3.32164396e-01,
-    6.88942598e-01, 5.79163537e-01, 6.70341547e-01, 2.70140602e-02, 9.30429671e-01, 7.17721157e-01,
-    9.89948537e-01, 7.75253347e-01, 1.34491522e-02, 2.48522428e-02, 3.51413378e-01, 7.64405834e-01,
-    7.86373507e-01, 7.18748577e-01, 8.66998621e-01, 6.80316582e-01, 2.51288712e-01, 4.91078420e-01,
-    3.76246281e-01, 4.86828710e-01, 5.67464772e-01, 5.30734742e-01, 8.99478296e-01, 7.66699088e-01,
-    9.49339111e-01, 3.55248484e-01, 9.06046929e-01, 4.48407772e-01, 6.96395305e-01, 2.44277335e-01,
-    7.74840000e-01, 5.21046603e-01, 4.66423971e-02, 5.12019638e-02, 8.95019614e-01, 5.28956953e-01,
-    4.31536306e-01, 5.83857744e-01, 4.41787364e-01, 4.68656523e-01, 5.73971433e-01, 6.79989654e-01,
-    3.19650588e-01, 6.12579596e-01, 6.49126442e-02, 8.39131142e-01, 2.85252117e-01, 5.84848929e-01,
-    9.46507115e-01, 8.58440748e-01, 3.61528940e-01, 2.44215959e-01, 3.80101125e-01, 4.57128957e-02,
-    8.82216988e-01, 8.31498633e-01, 7.23474381e-01, 7.75788607e-01, 1.40864146e-01, 6.62092382e-01,
-    5.13985168e-01, 3.00686418e-01, 8.70109949e-01, 2.43187753e-01, 2.89391938e-01, 2.84214238e-01,
-    8.70985521e-01, 8.77491176e-01, 6.72537226e-01, 3.30929686e-01, 1.85934324e-01, 9.16222614e-01,
-    6.18239142e-01, 2.64768597e-01, 5.76145451e-01, 8.62961369e-01, 6.84757925e-01, 7.60549082e-01,
-    1.27645356e-01, 4.51004673e-01, 3.92292980e-01, 4.63170803e-01, 4.35449330e-02, 2.17583404e-01,
-    5.71832605e-02, 2.06763039e-01, 3.70116249e-01, 2.09750028e-01, 6.17283019e-01, 8.62549231e-01,
-    9.84156240e-02, 2.66249156e-01, 3.87635103e-01, 2.85591012e-02, 4.24826068e-01, 4.45795088e-01,
-    6.86227676e-01, 1.08848960e-01, 5.96731841e-02, 3.71770228e-01, 1.91548833e-01, 6.95136078e-01,
-    9.00700636e-01, 8.76363105e-01, 2.67334632e-01, 1.80619709e-01, 7.94060419e-01, 1.42854171e-02,
-    1.09372387e-01, 8.74028108e-01, 6.46403232e-01, 4.86588834e-01, 5.93446175e-02, 6.11886291e-01,
-    8.83865057e-01, 3.15879821e-01, 2.27043992e-01, 9.76764951e-01, 6.15620336e-01, 9.76199360e-01,
-    2.40548962e-01, 3.21795663e-01, 8.75087904e-02, 8.11234663e-01, 6.96070480e-01, 8.12062321e-01,
-    1.21958818e-01, 3.44348628e-02, 8.72630414e-01, 3.06162776e-01, 1.76043529e-02, 9.45894971e-01,
-    5.33896401e-01, 6.21642973e-01, 4.93062535e-01, 4.48984262e-01, 2.24560379e-01, 4.24052195e-02,
-    4.43447610e-01, 8.95646149e-01, 6.05220676e-01, 1.81840491e-01, 9.70831206e-01, 2.12563586e-02,
-    6.92582693e-01, 7.55946922e-01, 7.95086143e-01, 6.05328941e-01, 3.99350764e-01, 4.32846636e-01,
-    9.81114529e-01, 4.98266428e-01, 6.37127930e-03, 1.59085889e-01, 6.34682067e-05, 5.59429440e-01,
-    7.38827633e-01, 8.93214770e-01, 2.16494306e-01, 9.35430573e-02, 4.75665868e-02, 7.80503518e-01,
-    7.86240041e-01, 7.06854594e-01, 2.13725879e-02, 7.68246091e-01, 4.50234808e-01, 5.21231104e-01,
-    5.01989826e-03, 4.22081572e-02, 1.65337732e-01, 8.54134740e-01, 4.99430262e-01, 8.94525601e-01,
-    1.14028379e-01, 3.69739861e-01, 1.32955599e-01, 2.65563824e-01, 2.52811151e-01, 1.44792843e-01,
-    6.88449594e-01, 4.44921417e-01, 8.23296587e-01, 1.93266317e-01, 1.19033309e-01, 1.36368966e-01,
-    3.42600285e-01, 5.64505195e-01, 5.57594559e-01, 7.44257892e-01, 8.38231569e-02, 4.11548847e-01,
-    3.21010077e-01, 8.55081359e-01, 4.30105779e-01, 1.16229135e-01, 9.87731964e-02, 3.14712335e-01,
-    4.50880592e-01, 2.72289598e-01, 6.31615256e-01, 8.97432958e-01, 4.44764250e-01, 8.03776440e-01,
-    2.68767748e-02, 2.43374608e-01, 4.02141103e-01, 4.98881209e-01, 5.33173003e-01, 8.82890436e-01,
-    7.16149148e-01, 4.19664401e-01, 2.29335357e-01, 2.88637806e-01, 3.44696803e-01, 6.78171906e-01,
-    5.69849716e-01, 5.86454477e-01, 3.54474989e-01, 9.03876540e-01, 6.45980000e-01, 6.34887593e-01,
-    7.88039746e-02, 2.04814126e-01, 7.82251754e-01, 2.43147074e-01, 7.50951808e-01, 1.72799092e-02,
-    2.95349590e-01, 6.57991826e-01, 8.81214312e-01, 5.73970708e-01, 2.77610881e-01, 1.82155097e-01,
-    7.69797417e-02, 6.44792402e-01, 9.46950998e-01, 7.73064845e-01, 6.04733624e-01, 5.80094567e-01,
-    1.67498426e-01, 2.66514296e-01, 6.50140368e-01, 1.91170299e-01, 2.08752199e-01, 3.01664091e-01,
-    9.85033484e-01, 2.92909152e-01, 8.65816607e-01, 1.85222119e-01, 2.28814559e-01, 1.34286382e-02,
-    2.89234322e-01, 8.18668708e-01, 4.71706924e-01, 9.23199803e-01, 2.80879188e-01, 1.47319284e-01,
-    4.13915748e-01, 9.31274932e-02, 6.66322195e-01, 9.66953974e-01, 3.19405786e-01, 6.69486551e-01,
-    5.03096313e-02, 6.95225201e-01, 5.78469859e-01, 6.29481655e-01, 1.39252534e-01, 1.22564968e-01,
-    6.80663678e-01, 6.34607157e-01, 6.42765834e-01, 1.57127410e-02, 2.92132086e-01, 5.24423878e-01,
-    4.68676824e-01, 2.86003928e-01, 7.18608322e-01, 8.95617933e-01, 5.48844309e-01, 1.74517278e-01,
-    5.24379196e-01, 2.13526524e-01, 5.88375435e-01, 9.88560185e-01, 4.17435771e-01, 6.14438688e-01,
-    9.53760881e-01, 5.27151288e-01, 7.03017278e-01, 3.44448559e-01, 4.47059676e-01, 2.83414901e-01,
-    1.98979011e-01, 4.24917361e-01, 5.73172761e-01, 2.32398853e-02, 1.65887230e-01, 4.05552785e-01,
-    9.29665524e-01, 2.26135696e-01, 9.20563384e-01, 7.65259963e-01, 4.54820075e-01, 8.97710267e-01,
-    3.78559302e-03, 9.15219382e-01, 3.55705698e-01, 6.94905124e-01, 8.58540202e-01, 3.89790666e-01,
-    2.49478206e-01, 7.93679304e-01, 4.75830027e-01, 4.40425353e-01, 3.70579459e-01, 1.40578049e-01,
-    1.70386675e-01, 7.04056121e-01, 4.85963102e-01, 9.68450060e-01, 6.77178001e-01, 2.65934654e-01,
-    2.58915007e-01, 6.70052890e-01, 2.61945109e-01, 8.46207759e-01, 1.01928951e-01, 2.85611334e-01,
-    2.45776933e-01, 2.66658783e-01, 3.71724077e-01, 4.34319025e-01, 4.24407347e-01, 7.15417683e-01,
-    8.07997684e-01, 1.64296275e-01, 6.01638065e-01, 8.60606804e-02, 2.68719187e-01, 5.11764101e-01,
-    9.75844338e-01, 7.81226782e-01, 2.20925515e-01, 7.18135040e-01, 9.82395577e-01, 8.39160243e-01,
-    9.08058083e-01, 6.88010677e-01, 8.14271847e-01, 5.12460821e-01, 1.17311345e-01, 5.96075228e-01,
-    9.17455497e-01, 2.12052706e-01, 7.04074603e-01, 8.72872565e-02, 8.76047818e-01, 6.96235046e-01,
-    8.54801557e-01, 2.49729159e-01, 9.76594604e-01, 2.87386363e-01, 2.36461559e-02, 9.94075254e-01,
-    4.25193986e-01, 7.61869994e-01, 5.13334255e-01, 6.44711165e-02, 8.92156689e-01, 3.55235167e-01,
-    1.08154647e-01, 8.78446825e-01, 2.43833016e-01, 9.23071293e-01, 2.72724115e-01, 9.46631338e-01,
-    3.74510294e-01, 4.08451278e-02, 9.78392777e-01, 3.65079221e-01, 6.37199516e-01, 5.51144906e-01,
-    5.25978080e-01, 1.42803678e-01, 4.05451674e-01, 7.79788219e-01, 6.26009784e-01, 3.35249497e-01,
-    1.43159543e-02, 1.80363779e-01, 5.05096904e-01, 2.82619947e-01, 5.83561392e-01, 3.10951324e-01,
-    8.73223968e-01, 4.38545619e-01, 4.81348800e-01, 6.68497085e-01, 3.79345401e-01, 9.58832501e-01,
-    1.89869550e-01, 2.34083070e-01, 2.94066207e-01, 5.74892667e-02, 6.92106828e-02, 9.61127686e-02,
-    6.72650672e-02, 8.47345378e-01, 2.80916761e-01, 7.32177357e-03, 9.80785961e-01, 5.73192225e-02,
-    8.48781331e-01, 8.83225408e-01, 7.34398275e-01, 7.70381941e-01, 6.20778343e-01, 8.96822048e-01,
-    5.40732486e-01, 3.69704071e-01, 5.77305837e-01, 2.08221827e-01, 7.34275341e-01, 1.06110900e-01,
-    3.49496706e-01, 8.34948910e-01, 1.56403291e-02, 6.78576376e-01, 8.96141268e-01, 5.94835119e-01,
-    1.43943153e-01, 3.49618530e-01, 2.10440392e-01, 3.46585620e-01, 1.05153093e-01, 3.45446174e-01,
-    2.72177079e-01, 7.07946300e-01, 4.33717726e-02, 3.31232203e-01, 3.91874320e-01, 4.76338141e-01,
-    6.22777789e-01, 2.95989228e-02, 4.32855769e-01, 7.61049310e-01, 3.63279149e-01, 9.47210350e-01,
-    6.43721247e-01, 6.58025802e-01, 1.05247633e-02, 5.29974442e-01, 7.30675767e-01, 4.30041079e-01,
-    6.62634841e-01, 8.25936616e-01, 9.91253704e-01, 6.79399281e-01, 5.44177006e-01, 7.52876048e-01,
-    3.32139049e-01, 7.98732398e-01, 7.38865223e-01, 9.16055132e-01, 6.11736493e-01, 9.63672879e-01,
-    1.83778839e-01, 7.27558919e-02, 5.91602822e-01, 3.25235484e-01, 2.34741217e-01, 9.52346277e-01,
-    9.18556407e-01, 9.35373324e-01, 6.89209070e-01, 2.56049054e-01, 6.17975395e-01, 7.82285691e-01,
-    9.84983432e-01, 6.62322741e-01, 2.04144457e-01, 3.98446577e-01, 1.38918297e-01, 3.05919921e-01,
-    3.14043787e-01, 5.91072666e-01, 7.44703771e-01, 8.92272567e-01, 9.78017873e-01, 9.01203161e-01,
-    1.41526372e-01, 4.14878484e-01, 6.80683651e-01, 5.01733152e-02, 8.14635389e-01, 2.27926375e-01,
-    9.03269815e-01, 8.68443745e-01, 9.86939190e-01, 7.40779486e-01, 2.61005311e-01, 3.19276232e-01,
-    9.69509248e-01, 1.11908818e-01, 4.49198556e-01, 1.27056715e-01, 3.84064823e-01, 5.14591811e-01,
-    2.10747488e-01, 9.53884090e-01, 8.43167950e-01, 4.51187972e-01, 3.75331782e-01, 6.23566461e-01,
-    3.55290379e-01, 2.95705968e-01, 1.69622690e-01, 1.42981830e-01, 2.72180991e-01, 9.46468040e-01,
-    3.70932500e-01, 9.94292830e-01, 4.62587505e-01, 7.14817405e-01, 2.45370540e-02, 3.00906377e-01,
-    5.75768304e-01, 9.71448393e-01, 6.95574827e-02, 3.93693854e-01, 5.29306116e-01, 5.04694554e-01,
-    6.73797120e-02, 6.76596969e-01, 5.50948898e-01, 3.24909641e-01, 7.70337719e-01, 6.51842631e-03,
-    3.03264879e-01, 7.61037886e-03, 2.72289601e-01, 1.50502041e-01, 6.71103888e-02, 7.41503703e-01,
-    1.92088941e-01, 2.19043977e-01, 9.09320161e-01, 2.37993569e-01, 6.18107973e-02, 8.31447852e-01,
-    2.23355609e-01, 1.84789435e-01, 4.16104518e-01, 4.21573859e-01, 8.72446305e-02, 2.97294197e-01,
-    4.50328256e-01, 8.72199917e-01, 2.51279916e-01, 4.86219272e-01, 7.57071329e-01, 4.85655942e-01,
-    1.06187277e-01, 4.92341327e-01, 1.46017513e-01, 5.25421017e-01, 4.22637906e-01, 2.24685018e-01,
-    8.72648431e-01, 5.54051490e-01, 1.80745062e-01, 2.12756336e-01, 5.20883169e-01, 7.60363654e-01,
-    8.30254678e-01, 5.00003328e-01, 4.69017439e-01, 6.38105527e-01, 3.50638261e-02, 5.22217353e-02,
-    9.06516882e-02, 8.52975842e-01, 1.19985883e-01, 3.74926753e-01, 6.50302066e-01, 1.98875727e-01,
-    6.28362507e-02, 4.32693501e-01, 3.10500685e-01, 6.20732833e-01, 4.58503272e-01, 3.20790034e-01,
-    7.91284868e-01, 7.93054570e-01, 2.93406765e-01, 8.95399023e-01, 1.06441034e-01, 7.53085241e-02,
-    8.67523104e-01, 1.47963482e-01, 1.25584706e-01, 3.81545040e-02, 6.34338619e-01, 1.76368938e-02,
-    5.75553531e-02, 5.31607516e-01, 2.63869588e-01, 9.41945823e-01, 9.24028838e-02, 5.21496463e-01,
-    7.74866558e-01, 5.65210610e-01, 7.28015327e-02, 6.51963790e-01, 8.94727453e-01, 4.49571590e-01,
-    1.29932405e-01, 8.64026259e-01, 9.92599934e-01, 7.43721560e-01, 8.87300215e-01, 1.06369925e-01,
-    8.11335531e-01, 7.87734900e-01, 9.87344678e-01, 5.32502820e-01, 4.42612382e-01, 9.64041183e-01,
-    1.66085871e-01, 1.12937664e-01, 5.24423470e-01, 6.54689333e-01, 4.59119726e-01, 5.22774091e-01,
-    3.08722276e-02, 6.26979315e-01, 4.49754105e-01, 8.07495757e-01, 2.34199499e-01, 1.67765675e-01,
-    9.22168418e-01, 3.73210378e-01, 8.04432575e-01, 5.61890354e-01, 4.47025593e-01, 6.43155678e-01,
-    2.40407640e-01, 5.91631279e-01, 1.59369206e-01, 7.75799090e-01, 8.32067212e-01, 5.59791576e-02,
-    6.39105224e-01, 4.85274738e-01, 2.12630838e-01, 2.81431312e-02, 7.16205363e-01, 6.83885011e-01,
-    5.23869697e-01, 9.99418314e-01, 8.35331599e-01, 4.69877463e-02, 6.74712562e-01, 7.99273684e-01,
-    2.77001890e-02, 5.75809742e-01, 2.78513031e-01, 8.36209905e-01, 7.25472379e-01, 4.87173943e-01,
-    7.88311357e-01, 9.64676177e-01, 1.75752651e-01, 4.98112580e-01, 8.08850418e-02, 6.40981131e-01,
-    4.06647450e-01, 8.46539387e-01, 2.12620694e-01, 9.11012851e-01, 8.25041445e-01, 8.90065575e-01,
-    9.63626055e-01, 5.96689242e-01, 1.63372670e-01, 4.51640148e-01, 3.43026542e-01, 5.80658851e-01,
-    2.82327625e-01, 4.75535418e-01, 6.27760926e-01, 8.46314115e-01, 9.61961932e-01, 3.19806094e-01,
-    5.05508062e-01, 5.28102944e-01, 6.13045057e-01, 7.44714938e-01, 1.50586073e-01, 7.91878033e-01,
-    4.89839179e-01, 3.10496849e-01, 8.82309038e-01, 2.86922314e-01, 4.84687559e-01, 5.20838630e-01,
-    4.62955493e-01, 2.38185305e-01, 5.47259907e-02, 7.10916137e-01, 7.31887202e-01, 6.25602317e-01,
-    8.77741168e-01, 4.19881322e-01, 4.81222328e-01, 1.28224501e-01, 2.46034010e-01, 3.34971854e-01,
-    7.37216484e-01, 5.62134821e-02, 7.14089724e-01, 9.85549393e-01, 4.66295827e-01, 3.08722434e-03,
-    4.70237690e-01, 2.66524167e-01, 7.93875484e-01, 4.54795911e-02, 8.09702944e-01, 1.47709735e-02,
-    1.70082405e-01, 6.35905179e-01, 3.75379109e-01, 4.30315011e-01, 3.15788760e-01, 5.58065230e-01,
-    2.24643800e-01, 2.42142981e-01, 6.57283636e-01, 3.34921891e-01, 1.26588975e-01, 7.68064155e-01,
-    9.43856291e-01, 4.47518596e-01, 5.44453573e-01, 9.95764932e-01, 7.16444391e-01, 8.51019765e-01,
-    1.01179183e-01, 4.45473958e-01, 4.60327322e-01, 4.96895844e-02, 4.72907738e-01, 5.58987444e-01,
-    3.41027487e-01, 1.56175026e-01, 7.58283148e-01, 6.83600909e-01, 2.14623396e-01, 3.27348880e-01,
-    3.92517893e-01, 6.70418431e-01, 5.16440832e-01, 8.63140348e-01, 5.73277464e-01, 3.46608058e-01,
-    7.39396341e-01, 7.20852434e-01, 2.35653246e-02, 3.89935659e-01, 7.53783745e-01, 6.34563528e-01,
-    8.79339335e-01, 7.41599159e-02, 5.62433904e-01, 6.15553852e-01, 4.56956324e-01, 5.20047447e-01,
-    5.26845015e-02, 5.58471266e-01, 1.63632233e-01, 5.38936665e-02, 6.49593683e-01, 2.56838748e-01,
-    8.99035326e-01, 7.20847756e-01, 5.68954684e-01, 7.43684755e-01, 5.70924238e-01, 3.82318724e-01,
-    4.89328290e-01, 5.62208561e-01, 4.97540804e-02, 4.18011085e-01, 6.88041565e-01, 2.16234653e-01,
-    7.89548214e-01, 8.46136387e-01, 8.46816189e-01, 1.73842353e-01, 6.11627842e-02, 8.44440559e-01,
-    4.50646654e-01, 3.74785037e-01, 4.87196697e-01, 4.56276448e-01, 9.13284391e-01, 4.15715464e-01,
-    7.13597697e-01, 1.23641270e-02, 5.10031271e-01, 4.74601930e-02, 2.55731159e-01, 3.22090006e-01,
-    1.91165703e-01, 4.51170940e-01, 7.50843157e-01, 4.42420576e-01, 4.25380660e-01, 4.50667257e-01,
-    6.55689206e-01, 9.68257670e-02, 1.96528793e-01, 8.97343028e-01, 4.99940904e-01, 6.65504083e-01,
-    9.41828079e-01, 4.54397338e-01, 5.61893331e-01, 5.09839880e-01, 4.53117514e-01, 8.96804127e-02,
-    1.74888861e-01, 6.65641378e-01, 2.81668336e-01, 1.89532742e-01, 5.61668382e-01, 8.68330157e-02,
-    8.25092797e-01, 5.18106324e-01, 1.71904024e-01, 3.68385523e-01, 1.62005436e-01, 7.48507399e-01,
-    9.30274827e-01, 2.38198517e-01, 9.52222901e-01, 5.23587800e-01, 6.94384557e-01, 1.09338652e-01,
-    4.83356794e-01, 2.73050402e-01, 3.68027050e-01, 5.92366466e-01, 1.83192289e-01, 8.60376029e-01,
-    7.13926203e-01, 8.16750052e-01, 1.57890291e-01, 6.25691951e-01, 5.24831646e-01, 1.73873797e-01,
-    1.02429784e-01, 9.17488471e-01, 4.03584434e-01, 9.31170884e-01, 2.79386137e-01, 8.77745206e-01,
-    2.45200576e-01, 1.28896951e-01, 3.15713052e-01, 5.27874291e-01, 2.16444335e-01, 7.03883817e-01,
-    7.74738919e-02, 8.42422142e-01, 3.75598924e-01, 3.51002411e-01, 6.22752776e-01, 4.82407943e-01,
-    7.43107867e-01, 9.46182666e-01, 9.44344819e-01, 3.28124763e-01, 1.06147431e-01, 1.65102684e-01,
-    3.84060507e-01, 2.91057722e-01, 7.68173662e-02, 1.03543651e-01, 6.76698940e-01, 1.43141994e-01,
-    7.21342202e-01, 6.69471294e-03, 9.07298311e-01, 5.57080171e-01, 8.10954489e-01, 4.11120526e-01,
-    2.06407453e-01, 2.59590556e-01, 7.58512718e-01, 5.79873897e-01, 2.92875650e-01, 2.83686529e-01,
-    2.42829343e-01, 9.19323719e-01, 3.46832864e-01, 3.58238858e-01, 7.42827585e-01, 2.05760059e-01,
-    9.58438860e-01, 5.66326411e-01, 6.60292846e-01, 5.61095078e-02, 6.79465531e-01, 7.05118513e-01,
-    4.44713264e-01, 2.09732933e-01, 5.22732436e-01, 1.74396512e-01, 5.29356748e-01, 4.38475687e-01,
-    4.94036404e-01, 4.09785794e-01, 6.40025507e-01, 5.79371821e-01, 1.57726118e-01, 6.04572263e-01,
-    5.41072639e-01, 5.18847173e-01, 1.97093284e-01, 8.91767002e-01, 4.29050835e-01, 8.25490570e-01,
-    3.87699807e-01, 4.50705808e-01, 2.49371643e-01, 3.36074898e-01, 9.29925118e-01, 6.65393649e-01,
-    9.07275994e-01, 3.73075859e-01, 4.14044139e-03, 2.37463702e-01, 2.25893784e-01, 2.46900245e-01,
-    4.50350196e-01, 3.48618117e-01, 5.07193932e-01, 5.23435142e-01, 8.13611417e-01, 8.92715622e-01,
-    1.02623450e-01, 3.06088345e-01, 7.80461650e-01, 2.21453645e-01, 2.01419652e-01, 2.84254457e-01,
-    3.68286735e-01, 7.39358243e-01, 8.97879394e-01, 9.81599566e-01, 7.56526442e-01, 7.37645545e-01,
-    4.23976657e-02, 8.25922012e-01, 2.60956996e-01, 2.90702065e-01, 8.98388344e-01, 3.03733299e-01,
-    8.49071471e-01, 3.45835425e-01, 7.65458276e-01, 5.68094872e-01, 8.93770930e-01, 9.93161641e-01,
-    5.63368667e-02, 4.26548945e-01, 5.46745780e-01, 5.75674571e-01, 7.94599487e-01, 7.18935553e-02,
-    4.46492976e-01, 6.40240123e-01, 2.73246969e-01, 2.00465968e-01, 1.30718835e-01, 1.92492005e-01,
-    1.96617189e-01, 6.61271644e-01, 8.12687657e-01, 8.66342445e-01
-
-   },
-   -4}};
-
-typedef ConnectComponentsTest<int, float> ConnectComponentsTestF_Int;
-TEST_P(ConnectComponentsTestF_Int, Result)
-{
-  /**
-   * Verify the src & dst vertices on each edge have different colors
-   */
-  EXPECT_TRUE(final_edges == params.n_row - 1);
-}
-
-INSTANTIATE_TEST_CASE_P(ConnectComponentsTest,
-                        ConnectComponentsTestF_Int,
-                        ::testing::ValuesIn(fix_conn_inputsf2));
-
-template <typename value_idx, typename value_t>
-struct MutualReachabilityFixConnectivitiesRedOp {
-  value_t* core_dists;
-  value_idx m;
-
-  DI MutualReachabilityFixConnectivitiesRedOp() : m(0) {}
-
-  MutualReachabilityFixConnectivitiesRedOp(value_t* core_dists_, value_idx m_)
-    : core_dists(core_dists_), m(m_){};
-
-  typedef typename raft::KeyValuePair<value_idx, value_t> KVP;
-  DI void operator()(value_idx rit, KVP* out, const KVP& other) const
-  {
-    if (rit < m && other.value < std::numeric_limits<value_t>::max()) {
-      value_t core_dist_rit   = core_dists[rit];
-      value_t core_dist_other = max(core_dist_rit, max(core_dists[other.key], other.value));
-
-      value_t core_dist_out;
-      if (out->key > -1) {
-        core_dist_out = max(core_dist_rit, max(core_dists[out->key], out->value));
-      } else {
-        core_dist_out = out->value;
-      }
-
-      bool smaller = core_dist_other < core_dist_out;
-      out->key     = smaller ? other.key : out->key;
-      out->value   = smaller ? core_dist_other : core_dist_out;
-    }
-  }
-
-  DI KVP operator()(value_idx rit, const KVP& a, const KVP& b) const
-  {
-    if (rit < m && a.key > -1) {
-      value_t core_dist_rit = core_dists[rit];
-      value_t core_dist_a   = max(core_dist_rit, max(core_dists[a.key], a.value));
-
-      value_t core_dist_b;
-      if (b.key > -1) {
-        core_dist_b = max(core_dist_rit, max(core_dists[b.key], b.value));
-      } else {
-        core_dist_b = b.value;
-      }
-
-      return core_dist_a < core_dist_b ? KVP(a.key, core_dist_a) : KVP(b.key, core_dist_b);
-    }
-
-    return b;
-  }
-
-  DI void init(value_t* out, value_t maxVal) const { *out = maxVal; }
-  DI void init(KVP* out, value_t maxVal) const
-  {
-    out->key   = -1;
-    out->value = maxVal;
-  }
-
-  DI void init_key(value_t& out, value_idx idx) const { return; }
-  DI void init_key(KVP& out, value_idx idx) const { out.key = idx; }
-
-  DI value_t get_value(KVP& out) const { return out.value; }
-  DI value_t get_value(value_t& out) const { return out; }
-
-  void gather(const raft::resources& handle, value_idx* map)
-  {
-    auto tmp_core_dists = raft::make_device_vector<value_t>(handle, m);
-    thrust::gather(raft::resource::get_thrust_policy(handle),
-                   map,
-                   map + m,
-                   core_dists,
-                   tmp_core_dists.data_handle());
-    raft::copy_async(
-      core_dists, tmp_core_dists.data_handle(), m, raft::resource::get_cuda_stream(handle));
-  }
-
-  void scatter(const raft::resources& handle, value_idx* map)
-  {
-    auto tmp_core_dists = raft::make_device_vector<value_t>(handle, m);
-    thrust::scatter(raft::resource::get_thrust_policy(handle),
-                    core_dists,
-                    core_dists + m,
-                    map,
-                    tmp_core_dists.data_handle());
-    raft::copy_async(
-      core_dists, tmp_core_dists.data_handle(), m, raft::resource::get_cuda_stream(handle));
-  }
-};
-
-template <typename value_t, typename value_idx>
-struct ConnectComponentsMutualReachabilityInputs {
-  value_idx n_row;
-  value_idx n_col;
-  std::vector<value_t> data;
-  std::vector<value_t> core_dists;
-  std::vector<value_idx> colors;
-  std::vector<value_idx> expected_rows;
-  std::vector<value_idx> expected_cols;
-  std::vector<value_t> expected_vals;
-};
-
-template <typename value_idx, typename value_t>
-class ConnectComponentsEdgesTest
-  : public ::testing::TestWithParam<ConnectComponentsMutualReachabilityInputs<value_t, value_idx>> {
- protected:
-  void basicTest()
-  {
-    raft::resources handle;
-
-    auto stream = resource::get_cuda_stream(handle);
-
-    params = ::testing::TestWithParam<
-      ConnectComponentsMutualReachabilityInputs<value_t, value_idx>>::GetParam();
-
-    raft::sparse::COO<value_t, value_idx> out_edges_unbatched(resource::get_cuda_stream(handle));
-    raft::sparse::COO<value_t, value_idx> out_edges_batched(resource::get_cuda_stream(handle));
-
-    rmm::device_uvector<value_t> data(params.n_row * params.n_col,
-                                      resource::get_cuda_stream(handle));
-    rmm::device_uvector<value_t> core_dists(params.n_row, resource::get_cuda_stream(handle));
-    rmm::device_uvector<value_idx> colors(params.n_row, resource::get_cuda_stream(handle));
-
-    raft::copy(data.data(), params.data.data(), data.size(), resource::get_cuda_stream(handle));
-    raft::copy(core_dists.data(),
-               params.core_dists.data(),
-               core_dists.size(),
-               resource::get_cuda_stream(handle));
-    raft::copy(
-      colors.data(), params.colors.data(), colors.size(), resource::get_cuda_stream(handle));
-
-    /**
-     * 3. cross_component_nn to fix connectivities
-     */
-    MutualReachabilityFixConnectivitiesRedOp<value_idx, value_t> red_op(core_dists.data(),
-                                                                        params.n_row);
-
-    raft::linkage::cross_component_nn<value_idx, value_t>(handle,
-                                                          out_edges_unbatched,
-                                                          data.data(),
-                                                          colors.data(),
-                                                          params.n_row,
-                                                          params.n_col,
-                                                          red_op,
-                                                          params.n_row,
-                                                          params.n_col);
-
-    raft::linkage::cross_component_nn<value_idx, value_t>(handle,
-                                                          out_edges_batched,
-                                                          data.data(),
-                                                          colors.data(),
-                                                          params.n_row,
-                                                          params.n_col,
-                                                          red_op,
-                                                          11,
-                                                          1);
-
-    ASSERT_TRUE(out_edges_unbatched.nnz == out_edges_batched.nnz &&
-                out_edges_unbatched.nnz == params.expected_rows.size());
-
-    ASSERT_TRUE(devArrMatch(out_edges_unbatched.rows(),
-                            params.expected_rows.data(),
-                            out_edges_unbatched.nnz,
-                            Compare<int>()));
-
-    ASSERT_TRUE(devArrMatch(out_edges_unbatched.cols(),
-                            params.expected_cols.data(),
-                            out_edges_unbatched.nnz,
-                            Compare<int>()));
-
-    ASSERT_TRUE(devArrMatch(out_edges_unbatched.vals(),
-                            params.expected_vals.data(),
-                            out_edges_unbatched.nnz,
-                            CompareApprox<float>(1e-4)));
-
-    ASSERT_TRUE(devArrMatch(out_edges_batched.rows(),
-                            params.expected_rows.data(),
-                            out_edges_batched.nnz,
-                            Compare<int>()));
-
-    ASSERT_TRUE(devArrMatch(out_edges_batched.cols(),
-                            params.expected_cols.data(),
-                            out_edges_batched.nnz,
-                            Compare<int>()));
-
-    ASSERT_TRUE(devArrMatch(out_edges_batched.vals(),
-                            params.expected_vals.data(),
-                            out_edges_batched.nnz,
-                            CompareApprox<float>(1e-4)));
-  }
-
-  void SetUp() override { basicTest(); }
-
-  void TearDown() override {}
-
- protected:
-  ConnectComponentsMutualReachabilityInputs<value_t, value_idx> params;
-};
-
-const std::vector<ConnectComponentsMutualReachabilityInputs<float, int>> mr_fix_conn_inputsf2 = {
-  {100,
-   2,
-   {-7.72642, -8.39496, 5.4534,   0.742305, -2.97867, 9.55685,   6.04267,  0.571319,   -6.52184,
-    -6.31932, 3.64934,  1.40687,  -2.17793, 9.98983,  4.42021,   2.33028,  4.73696,    2.94181,
-    -3.66019, 9.38998,  -3.05358, 9.12521,  -6.65217, -5.57297,  -6.35769, -6.58313,   -3.61553,
-    7.81808,  -1.77073, 9.18565,  -7.95052, -6.39764, -6.60294,  -6.05293, -2.58121,   10.0178,
-    -7.76348, -6.72638, -6.40639, -6.95294, -2.97262, 8.54856,   -6.95673, -6.53896,   -7.32614,
-    -6.02371, -2.1478,  10.5523,  -2.54502, 10.5789,  -2.96984,  10.0714,  3.22451,    1.55252,
-    -6.25396, -7.73727, -7.85431, -6.09303, -8.11658, -8.20057,  -7.55965, -6.64786,   4.936,
-    2.23423,  4.44752,  2.27472,  -5.72103, -7.70079, -0.929985, 9.78172,  -3.10984,   8.72259,
-    -2.44167, 7.58954,  -2.18511, 8.6292,   5.55528,  2.30192,   4.73164,  -0.0143992, -8.2573,
-    -7.81793, -2.98837, 8.82863,  4.60517,  0.804492, -3.83738,  9.21115,  -2.62485,   8.71318,
-    3.57758,  2.44676,  -8.48711, -6.69548, -6.70645, -6.49479,  -6.86663, -5.42658,   3.83139,
-    1.47141,  2.02013,  2.79507,  4.64499,  1.73858,  -1.69667,  10.3705,  -6.61974,   -6.09829,
-    -6.05757, -4.98332, -7.10309, -6.16611, -3.52203, 9.32853,   -2.26724, 7.10101,    6.11777,
-    1.4549,   -4.23412, 8.452,    -6.58655, -7.59446, 3.93783,   1.64551,  -7.12502,   -7.63385,
-    2.72111,  1.94666,  -7.14428, -4.15994, -6.66553, -8.12585,  4.70011,  4.43641,    -7.76914,
-    -7.69592, 4.11012,  2.48644,  4.89743,  1.89872,  4.29716,   1.17089,  -6.62913,   -6.53366,
-    -8.07093, -6.22356, -2.16558, 7.25125,  4.73953,  1.46969,   -5.91625, -6.46733,   5.43091,
-    1.06378,  -6.82142, -8.02308, 6.52606,  2.14775,  3.08922,   2.04173,  -2.14756,   8.36917,
-    3.85663,  1.65111,  -1.68665, 7.79344,  -5.01385, -6.40628,  -2.52269, 7.95658,    -2.30033,
-    7.05462,  -1.04355, 8.78851,  3.72045,  3.5231,   -3.98772,  8.29444,  4.24777,    0.509655,
-    4.72693,  1.67416,  5.7827,   2.7251,   -3.41722, 7.60198,   5.22674,  4.16363,    -3.1109,
-    10.8666,  -3.18612, 9.62596,  -1.4782,  9.94557,  4.47859,   2.37722,  -5.79658,   -5.82631,
-    -3.34842, 8.70507},
-   {0.978428, 1.01917,  0.608673, 1.45629,  0.310713, 0.689461, 0.701126, 0.63296,  0.774788,
-    0.701648, 0.513282, 0.757651, 0.45638,  0.973111, 0.901396, 0.613692, 0.482497, 0.688143,
-    0.72428,  0.666345, 0.58232,  0.554756, 0.710315, 0.903611, 0.694115, 0.796099, 0.639759,
-    0.798998, 0.639839, 1.30727,  0.663729, 0.57476,  0.571348, 1.14662,  1.26518,  0.485068,
-    0.78207,  0.791621, 1.01678,  1.28509,  1.14715,  0.381395, 0.850507, 0.788511, 0.588341,
-    0.878516, 0.928669, 0.405874, 0.776421, 0.612274, 1.84963,  0.57476,  0.95226,  0.488078,
-    1.24868,  0.515136, 0.589378, 0.903632, 1.01678,  1.09964,  0.666345, 0.713265, 0.877168,
-    1.10053,  1.96887,  1.03574,  2.03728,  0.969553, 0.774788, 0.586338, 0.65168,  0.435472,
-    0.664396, 0.790584, 0.678637, 0.715964, 0.865494, 0.978428, 1.59242,  0.861109, 0.833259,
-    0.65168,  0.903632, 1.49599,  0.76347,  0.960453, 1.1848,   1.37398,  0.928957, 1.07848,
-    0.661798, 1.21104,  1.04579,  1.89047,  1.24288,  0.529553, 0.903611, 0.620897, 0.882467,
-    0.647189},
-   {0, 1, 2, 1, 0, 1, 2, 1, 1, 2, 2, 0, 0, 2, 2, 0, 0, 2, 0, 0, 2, 0, 0, 2, 2,
-    2, 1, 0, 0, 0, 0, 1, 1, 0, 2, 2, 2, 2, 1, 1, 0, 2, 1, 2, 2, 1, 0, 0, 0, 1,
-    1, 1, 2, 0, 0, 0, 2, 2, 1, 2, 0, 1, 0, 1, 0, 0, 1, 0, 1, 1, 1, 0, 0, 2, 1,
-    0, 1, 0, 1, 1, 2, 1, 2, 0, 2, 2, 2, 1, 2, 1, 1, 1, 2, 1, 2, 2, 2, 1, 0, 2},
-   {50, 54, 57, 63, 82, 87},
-   {57, 63, 50, 54, 87, 82},
-   {6.0764, 11.1843, 6.0764, 11.1843, 6.89004, 6.89004}},
-  {1000,
-   2,
-   {-6.59634,  -7.13901,  -6.13753,  -6.58082, 5.19821,    2.04918,     -2.96856,  8.16444,
-    -2.76879,  7.51114,   -6.82261,  -6.61152, 5.02008,    2.58376,     5.55621,   2.31966,
-    4.86379,   3.33731,   5.84639,   1.15623,  -2.17159,   8.60241,     -4.97844,  -6.94077,
-    -2.31014,  8.41407,   5.5582,    0.402669, 5.25265,    0.919754,    5.85298,   2.11489,
-    -3.29245,  8.69222,   -1.9621,   8.81209,  -1.53408,   8.86723,     -2.18227,  8.79519,
-    4.60519,   2.20738,   -6.4759,   -6.9043,  -7.18766,   -6.10045,    -9.00148,  -7.48793,
-    4.01674,   1.41769,   -2.45347,  10.1085,  -3.20892,   9.22827,     -3.18612,  9.62596,
-    4.81977,   3.36517,   4.90693,   2.8628,   -6.44269,   -5.68946,    -8.30144,  -5.37878,
-    4.61485,   2.79094,   -1.98726,  9.31127,  -3.66019,   9.38998,     -6.58607,  -8.23669,
-    -7.46015,  -6.29153,  4.08468,   3.85433,  -6.36842,   -5.50645,    -6.83602,  -5.18506,
-    -0.627173, 10.3597,   3.98846,   1.48928,  -2.9968,    8.58173,     -7.2144,   -7.28376,
-    -0.660242, 10.1409,   -4.23528,  -8.38308, -3.15984,   8.52716,     -2.40987,  9.76567,
-    -8.7548,   -6.76508,  4.56971,   0.312209, -7.5487,    -5.8402,     -1.6096,   9.32159,
-    5.04813,   0.270586,  -7.6525,   -6.47306, -1.79758,   7.88964,     -9.0153,   -3.74236,
-    -3.5715,   9.48788,   -1.65154,  8.85435,  -3.47412,   9.70034,     6.31245,   2.39219,
-    4.03851,   2.29295,   -3.17098,  9.86672,  -6.90693,   -7.81338,    -6.22373,  -6.68537,
-    -3.22204,  9.12072,   -0.365254, 9.6482,   -7.76712,   -7.31757,    4.15669,   3.54716,
-    4.1937,    0.083629,  -3.03896,  9.52755,  -6.29293,   -7.35501,    -2.95926,  9.63714,
-    4.02709,   1.58547,   4.56828,   1.93595,  5.6242,     1.75918,     -7.36237,  -7.83344,
-    5.32177,   3.81988,   -2.43183,  8.153,    -1.97939,   10.4559,     -3.49492,  9.51833,
-    3.39602,   1.28026,   -2.42215,  8.71528,  -3.57682,   8.87191,     -2.77385,  11.7345,
-    5.71351,   0.946654,  -6.50253,  -6.90937, 4.08239,    0.603367,    -5.64134,  -6.85884,
-    -2.76177,  7.7665,    -2.25165,  8.93984,  -3.49071,   9.47639,     -1.06792,  7.57842,
-    5.15754,   1.24743,   3.63574,   1.20537,  -6.07969,   -8.49642,    4.12227,   2.19696,
-    -7.17144,  -8.4433,   -1.92234,  11.2047,  3.23237,    1.19535,     3.85389,   0.641937,
-    4.82665,   1.21779,   -7.68923,  -6.45605, -7.00816,   -8.76196,    -5.12894,  9.83619,
-    -5.66247,  -5.35879,  3.05598,   2.73358,  6.06038,    1.40242,     -1.69568,  7.78342,
-    5.13391,   2.23384,   -2.96984,  10.0714,  -5.36618,   -6.2493,     5.55896,   1.6829,
-    3.55882,   2.58911,   5.36155,   0.844118, -0.0634456, 9.14351,     4.88368,   1.40909,
-    -7.04675,  -6.59753,  -7.78333,  -6.55575, 5.39881,    2.25436,     -2.85189,  8.64285,
-    -2.22821,  8.39159,   3.88591,   1.69249,  -7.55481,   -7.02463,    4.60032,   2.65467,
-    -6.90615,  -7.76198,  -6.76005,  -7.85318, 4.15044,    3.01733,     -7.18884,  -7.63227,
-    4.68874,   2.01376,   3.51716,   2.35558,  -3.81367,   9.68396,     4.42644,   3.4639,
-    4.81758,   0.637825,  -6.20705,  -4.98023, -1.68603,   9.0876,      -4.99504,  -5.33687,
-    -1.77073,  9.18565,   4.86433,   3.02027,  4.20538,    1.664,       4.59042,   2.64799,
-    -3.09856,  9.86389,   -3.02306,  7.95507,  -6.32402,   -6.79053,    -7.67205,  -7.18807,
-    -8.10918,  -6.38341,  -1.67979,  6.80315,  4.00249,    3.16219,     -2.54391,  7.84561,
-    -3.22764,  8.80084,   -2.63712,  8.05875,  -2.41744,   7.02672,     -6.71117,  -5.56251,
-    5.18348,   1.60256,   -7.40824,  -6.29375, -4.22233,   10.3682,     4.8509,    1.87646,
-    -2.99456,  9.09616,   5.1332,    2.15801,  -2.27358,   9.78515,     -6.73874,  -8.64855,
-    4.96124,   2.39509,   -3.70949,  8.67978,  -4.13674,   9.06237,     2.80367,   2.48116,
-    -0.876786, 7.58414,   -3.7005,   9.67084,  6.48652,    0.903085,    6.28189,   2.98299,
-    -6.07922,  -6.12582,  -5.67921,  -7.537,   4.55014,    3.41329,     -1.63688,  9.19763,
-    -4.02439,  10.3812,   5.23053,   3.08187,  -2.2951,    7.76855,     -6.24491,  -5.77041,
-    6.02415,   2.53708,   -6.91286,  -7.08823, 4.83193,    1.66405,     -7.07454,  -5.74634,
-    -2.09576,  10.8911,   3.29543,   1.05452,  -3.49973,   8.44799,     5.2922,    0.396778,
-    -2.54502,  10.5789,   -6.38865,  -6.14523, -1.75221,   8.09212,     -9.30387,  -5.99606,
-    -2.98113,  10.1032,   -6.2017,   -7.36802, 4.63628,    0.814805,    -1.81905,  8.61307,
-    4.88926,   3.55062,   3.08325,   2.57918,  -2.51717,   10.4942,     -5.75358,  -6.9315,
-    6.36742,   2.40949,   5.74806,   0.933264, 4.74408,    1.91058,     -7.41496,  -6.97064,
-    -2.98414,  8.36096,   6.72825,   1.83358,  -2.95349,   9.39159,     -3.35599,  7.49944,
-    6.18738,   3.76905,   -3.17182,  9.58488,  5.17863,    1.0525,      -3.0397,   8.43847,
-    -2.23874,  8.96405,   3.04689,   2.41364,  6.14064,    2.82339,     -6.33334,  -6.87369,
-    -7.92444,  -8.84647,  3.65129,   0.86958,  5.29842,    3.98337,     -2.06538,  9.78892,
-    -6.89494,  -6.30082,  -2.52144,  8.11703,  -8.11398,   -7.47257,    5.3381,    2.36666,
-    -6.93452,  -6.59456,  -7.50634,  -6.01772, 6.23438,    1.12621,     -2.15218,  8.32138,
-    -7.04777,  -7.3522,   -2.52771,  8.72563,  -2.77907,   8.03552,     4.29123,   1.62391,
-    -8.07551,  -6.43551,  -3.28202,  8.77747,  -2.21308,   9.27534,     -8.25153,  -8.49367,
-    -3.54644,  8.82395,   -8.05867,  -5.69243, 4.46681,    1.98875,     3.8362,    3.61229,
-    -6.96231,  -7.00186,  5.18993,   1.00483,  -5.35116,   -6.37227,    5.23298,   1.66362,
-    -5.68306,  -7.03864,  -9.03144,  -7.59926, -6.10127,   -7.4313,     4.83572,   0.994797,
-    -7.32695,  -5.59909,  0.569683,  10.1339,  3.35957,    2.84563,     -2.4122,   9.60944,
-    5.00855,   1.57983,   -2.57528,  7.80327,  3.96349,    3.77411,     4.59429,   2.21651,
-    -6.54765,  -6.68961,  4.76798,   1.29212,  -1.67351,   7.88458,     5.63615,   1.47941,
-    -2.5301,   9.13161,   4.26075,   1.76959,  4.67788,    2.0932,      4.39955,   1.59835,
-    3.91274,   1.72565,   -4.1786,   9.55765,  -7.34566,   -8.47481,    4.8364,    2.68217,
-    -7.36848,  -7.99973,  -5.84708,  -5.7534,  5.37252,    1.89245,     -2.1707,   8.599,
-    -1.3299,   9.0818,    -6.79122,  -5.40258, 5.56391,    1.78827,     -0.194539, 7.14702,
-    4.60489,   3.74397,   5.50995,   2.46885,  -3.98772,   8.29444,     -5.21837,  -7.33721,
-    -1.63959,  10.3699,   -5.92932,  -5.1695,  -5.88358,   -7.6369,     4.11716,   3.02218,
-    -6.54114,  -7.17551,  3.97179,   2.96521,  -6.75325,   -4.94118,    5.26169,   0.402945,
-    3.25031,   0.327771,  -0.44845,  10.7696,  -2.15141,   9.57507,     7.04329,   1.91555,
-    -3.74615,  7.69383,   -7.52318,  -5.85015, -6.80419,   -8.48208,    -4.57664,  8.92517,
-    4.57574,   2.30193,   4.84098,   3.02382,  -9.43355,   -5.94579,    -3.52203,  9.32853,
-    3.43018,   2.5731,    -6.15725,  -7.25294, -6.69861,   -8.17694,    -2.40955,  8.51081,
-    -4.82342,  -7.98332,  -7.10611,  -6.51274, 5.86755,    0.763529,    -6.56045,  -5.53966,
-    -3.61553,  7.81808,   4.3825,    0.304586, -6.52818,   -5.80996,    4.59972,   0.542395,
-    -6.90603,  -6.59995,  -6.3585,   -6.23489, -6.01915,   -7.46319,    -5.38694,  -7.15123,
-    -7.83475,  -6.45651,  5.89564,   1.07856,  -5.15266,   -7.27975,    -6.97978,  -7.08378,
-    5.83493,   0.449983,  -2.62374,  10.2521,  -7.34494,   -6.98606,    -6.79719,  -8.33766,
-    3.54757,   1.65676,   -8.40528,  -5.61753, -5.85556,   -6.28758,    4.66862,   3.25162,
-    -6.26047,  -4.82261,  4.61552,   4.11544,  -1.36637,   9.76622,     4.2517,    2.14359,
-    -2.45099,  7.87132,   -0.376164, 7.0622,   4.34493,    3.22091,     6.95921,   2.36649,
-    -6.70319,  -7.24714,  -5.56932,  -5.48443, -7.43149,   -4.32191,    -3.23956,  9.23074,
-    -5.77255,  -7.00049,  4.96601,   0.722056, -7.88617,   -5.74023,    4.18757,   -0.45071,
-    -7.12569,  -7.72336,  5.27366,   2.38697,  3.93487,    1.9174,      3.19186,   -0.225636,
-    -3.41722,  7.60198,   -3.08286,  8.46743,  -5.87905,   -7.55073,    -5.26425,  -7.20243,
-    -2.97867,  9.55685,   -1.23153,  8.42272,  -2.33602,   9.3996,      -3.33819,  8.45411,
-    -3.58009,  9.49676,   3.78152,   2.67348,  -1.54582,   9.42707,     -4.04331,  10.292,
-    3.3452,    3.134,     -2.75494,  8.74156,  -3.26555,   7.59203,     -7.27139,  -7.80252,
-    3.5293,    3.72544,   6.11642,   3.35326,  4.01611,    3.8872,      4.89591,   2.95586,
-    -7.06677,  -5.89438,  4.19438,   3.42655,  -6.11355,   -5.65318,    -7.59645,  -8.74665,
-    -5.80362,  -6.8588,   3.80453,   4.11832,  5.70655,    3.14247,     -4.98084,  8.21739,
-    -1.87642,  11.285,    4.39864,   2.32523,  -3.48388,   9.80137,     4.02836,   0.566509,
-    -2.41212,  9.98293,   -5.40846,  -7.08943, 4.01506,    1.99926,     -3.43613,  8.95476,
-    -7.24458,  -7.71932,  6.02204,   2.62188,  -6.29999,   -6.55431,    6.19038,   0.974816,
-    3.55882,   3.02632,   -7.06011,  -3.687,   -1.55877,   8.43738,     -5.14711,  -4.64881,
-    4.7167,    0.690177,  -7.90381,  -5.02602, 4.17218,    2.31967,     -0.643423, 9.48812,
-    -7.95237,  -6.64086,  -4.05986,  9.08285,  -6.24158,   -6.37927,    -6.6105,   -7.2233,
-    -6.21675,  -5.70664,  -3.29967,  9.48575,  3.41775,    2.68617,     -2.24948,  8.10997,
-    -2.24931,  9.79611,   -9.0523,   -6.03269, -2.2587,    9.36073,     5.20965,   2.42088,
-    -3.10159,  8.1503,    -6.67906,  -5.73147, 4.0687,     2.54575,     -1.24229,  8.30662,
-    -2.09627,  8.45056,   -7.87801,  -6.57832, 4.72216,    3.03865,     -0.929985, 9.78172,
-    -8.56307,  -7.68598,  -7.05257,  -5.1684,  -7.09076,   -7.86729,    4.61432,   3.1459,
-    -6.34133,  -5.8076,   -3.82943,  10.8457,  -8.46082,   -5.98507,    5.34763,   1.4107,
-    -1.68714,  10.9111,   -1.67886,  8.1582,   -0.623012,  9.18886,     -4.21258,  8.95874,
-    -2.16744,  10.8905,   -6.57158,  -7.27176, 2.14047,    4.26411,     -8.44217,  -7.40916,
-    5.29008,   1.87399,   4.31824,   4.04992,  -3.77008,   9.93215,     -2.72688,  10.1131,
-    -6.14278,  -7.16144,  -3.92457,  8.59364,  -5.92649,   -6.59299,    4.68369,   1.82617,
-    -6.89905,  -7.18329,  3.95173,   4.22561,  -7.66453,   -6.23183,    -2.44167,  7.58954,
-    -6.36603,  -7.41281,  -6.45081,  -6.187,   -6.6125,    -6.37138,    5.46036,   2.48044,
-    -2.14756,  8.36917,   -2.3889,   9.52872,  3.80752,    2.44459,     -3.98778,  10.158,
-    -6.63887,  -4.27843,  -8.65266,  -5.61819, -7.97003,   -5.46918,    -5.9604,   -7.54825,
-    -0.916011, 8.50307,   -3.69246,  6.97505,  -7.98533,   -7.09503,    -2.30033,  7.05462,
-    4.76218,   2.51647,   -7.04981,  -7.33334, 3.66401,    3.02681,     -2.50408,  8.7797,
-    7.19996,   1.87711,   4.01291,   3.78562,  -0.356015,  8.24694,     -0.958046, 9.12996,
-    4.60675,   3.76773,   6.21945,   1.45031,  4.27744,    0.8535,      -4.72232,  -7.48582,
-    6.03923,   2.8978,    -3.26833,  9.16468,  -7.97059,   -7.29092,    -2.3998,   9.74005,
-    -2.66721,  8.58741,   -7.36269,  -6.73332, -7.87893,   -7.38488,    4.65023,   0.661333,
-    -4.8171,   -7.94764,  -4.11564,  9.21775,  4.80633,    2.46562,     -2.72887,  9.3714,
-    -5.26735,  -5.5652,   4.9826,    2.42992,  -6.17018,   -7.3156,     4.38084,   1.77682,
-    5.35084,   2.41743,   -2.61796,  9.416,    5.27229,    2.94572,     -7.52315,  -5.95227,
-    -1.45077,  7.25555,   -3.79916,  7.71921,  -2.23251,   9.84147,     3.70054,   1.82908,
-    -1.93831,  10.1499,   -6.18324,  -5.9248,  -3.33142,   9.25797,     -6.08536,  -8.1344,
-    5.95727,   2.17077,   4.87366,   0.417274, -6.529,     -6.39092,    -9.24256,  -7.88984,
-    -6.36652,  -7.13966,  -3.90777,  9.57726,  -7.06252,   -5.50523,    -2.26423,  8.50734,
-    -2.84498,  10.6833,   5.0391,    2.62037,  -2.74815,   8.10672,     3.35945,   3.72796,
-    -4.11668,  9.19892,   5.66903,   2.44577,  -1.63807,   8.68826,     -7.42587,  -6.48831,
-    6.17063,   3.19193,   -2.28511,  9.02688,  -7.10088,   -7.15692,    4.46293,   1.17487,
-    -5.91017,  -6.45292,  -2.26724,  7.10101,  -2.43339,   8.33712,     -4.63309,  8.48853,
-    -3.31769,  8.51253,   -2.49078,  10.6907,  -1.30798,   8.60621,     6.30535,   2.98754,
-    -5.79384,  -6.78213,  -1.93213,  8.81124,  4.55773,    3.09047,     6.37584,   2.17108,
-    4.3927,    1.29119,   -3.2245,   9.69388,  -1.69634,   9.64392,     2.799,     0.693593,
-    -2.1426,   8.07441,   -8.4505,   -8.00688, 4.736,      1.51089,     -2.5863,   9.35544,
-    -2.94924,  9.14503,   6.2054,    1.90742,  5.67172,    0.487609,    -5.69071,  -6.17181,
-    -8.24651,  -7.10488,  -7.34424,  -6.67895, -6.71977,   -7.90778,    -1.82294,  7.40157,
-    -9.40991,  -7.16611,  -4.37999,  8.66277,  -1.42615,   10.0681,     -2.00828,  8.03673,
-    -7.50228,  -6.6855,   -5.65859,  -6.29801, -8.02335,   -6.77155,    -3.40761,  9.50621,
-    -2.82447,  9.77326,   -1.5938,   9.34304,  -3.5213,    7.35943,     -3.36961,  8.62973,
-    -7.01708,  -5.92724,  5.20886,   3.60157,  -1.71817,   8.1049,      -2.46363,  8.36269,
-    -2.77809,  7.90776,   -2.75459,  8.26055,  -2.03596,   8.94146,     -4.53434,  9.20074,
-    -7.44387,  -6.69556,  -6.90099,  -7.62732, 3.29169,    2.71643,     6.08686,   2.16972,
-    -2.31111,  8.86993,   -5.75046,  7.9899,   4.69951,    1.32623,     4.71851,   -0.025031,
-    -6.42374,  -4.71511,  -8.04974,  -8.68209, -3.16103,   9.06168,     -6.18267,  -7.21393,
-    -7.94202,  -6.4518,   -7.07697,  -7.03138, 3.93554,    0.564708,    -1.20372,  9.03529,
-    -7.10611,  -7.83955,  -7.47529,  -5.50567, -6.15453,   -6.36393,    -2.98024,  9.24634,
-    -7.75761,  -7.70699,  -3.08597,  9.76968,  -8.04954,   -9.75237,    5.2534,    0.950377,
-    5.63789,   -0.923086, -5.7065,   -6.51047, -8.02132,   -7.07377,    -8.28594,  -6.96322,
-    -7.70722,  -6.79397,  -2.4962,   10.4678,  5.02846,    4.46617,     4.02648,   1.6707,
-    -0.319395, 8.20599,   4.74525,   0.639144, -1.0313,    8.49602,     4.08766,   2.6061,
-    3.63826,   1.69207,   2.55795,   3.66963,  5.2826,     3.30232,     -1.04355,  8.78851,
-    -6.84762,  -7.63353,  -4.70868,  -7.056,   3.53651,    -0.179721,   -3.38482,  7.63149,
-    -5.9265,   -6.36702,  -0.986074, 9.5532,   -2.42261,   8.85861,     -7.42835,  -6.78726,
-    -4.02857,  8.53005,   -8.22675,  -7.85172, -5.57529,   -8.5426,     6.03009,   2.53098,
-    -7.10448,  -7.53011,  -3.4988,   8.8885,   -2.62485,   8.71318,     -6.39489,  -7.72647,
-    3.93789,   1.31027,   4.27627,   1.91622,  -0.923181,  7.77647,     -5.16017,  10.1058,
-    -6.44307,  -5.97617,  -7.24495,  -6.69543, 6.27331,    0.826824,    -6.55655,  -7.13246,
-    5.66245,   4.41292,   -2.13805,  8.4103,   5.23463,    2.82659,     -4.86624,  -6.74357,
-    -6.14082,  -6.26474,  -2.67048,  9.41834,  -1.26311,   6.9409,      -7.20231,  -7.13094,
-    -1.35109,  9.80595,   3.9906,    0.749229, -6.75696,   -5.25543,    4.84826,   -0.0685652,
-    -7.4914,   -6.91715,  4.46725,   2.85683,  -2.95571,   9.87068,     6.32381,   1.51429,
-    -6.81177,  -6.02734,  -2.57188,  9.96943,  -4.28792,   10.5103,     3.65025,   2.91394,
-    -7.11856,  -7.24693,  -6.98693,  -6.43239, 4.7651,     1.54376,     4.00092,   0.65008,
-    -7.14816,  -7.7713,   -7.58803,  -8.39382, 4.3321,     2.19232,     -7.89545,  -6.81843,
-    -2.11475,  8.5933,    -0.743743, 9.41927,  3.64849,    -0.18022,    -1.68665,  7.79344,
-    4.00214,   1.44217,   -6.96799,  -7.25012, -1.58302,   10.9237,     -6.68524,  -7.23328,
-    4.65831,   2.32075,   4.62024,   2.52566,  -4.23412,   8.452,       -0.822056, 9.89593,
-    -7.19868,  -7.67614,  -3.32742,  11.1067,  5.27861,    0.830165,    4.48982,   2.09875,
-    -6.58087,  -7.6319,   -0.880582, 7.63418,  -7.01088,   -6.80326,    -7.31601,  -6.98972,
-    -6.85883,  -7.60811,  6.14328,   2.85053,  -7.49206,   -6.51861,    -2.28174,  10.3214,
-    4.81074,   1.78919,   -5.58987,  -6.20693, 4.08096,    2.35038,     -1.5029,   8.43739,
-    4.11536,   2.46254,   -3.28299,  7.76963,  4.31953,    2.39734,     4.91146,   0.696421,
-    -1.4782,   9.94557,   -3.34842,  8.70507,  -6.97822,   -6.86126,    4.10012,   1.19486,
-    -2.50395,  9.06127,   4.41891,   2.00006,  -2.73266,   9.72829,     3.5436,    0.533119,
-    5.78864,   0.233456,  -6.62589,  -6.41242, -2.21942,   11.0897,     -6.76636,  -8.31839,
-    -2.71732,  8.52129,   -5.20972,  -6.48544, 3.26056,    1.24224,     3.45228,   2.28299,
-    4.72171,   1.87428,   -7.52585,  -5.1048,  5.0695,     2.18086,     -6.55646,  -7.02771,
-    3.23727,   3.72275,   3.41411,   0.508795, -7.80698,   -6.64174,    -5.90443,  -6.37902,
-    -0.387041, 10.0468,   -1.3506,   8.1936,   -6.08614,   -8.62864,    -5.91478,  -5.26453,
-    -2.61623,  7.97904,   4.45459,   1.84335,  -6.66643,   -7.63208,    3.6729,    1.92546,
-    -1.32976,  8.54511,   6.31758,   1.41958,  4.63381,    2.81166,     -7.01394,  -6.0693,
-    -2.7786,   9.73183,   -2.90131,  7.55077,  -7.13842,   -5.28146,    6.71514,   1.28398,
-    -6.98408,  -7.04893,  -3.03946,  8.22141,  -2.76417,   10.5183,     -7.35347,  -6.89456,
-    4.19345,   2.16726,   -2.02819,  9.23817,  4.97076,    2.8067,      -0.544473, 9.04955,
-    4.90727,   2.29487,   -6.31871,  -7.17559, 3.71665,    0.621485,    4.7903,    2.33813,
-    -6.47994,  -7.53147,  -6.80958,  -5.71823, -8.07326,   -5.96096,    4.77342,   1.8207,
-    5.71856,   1.93466,   -2.70156,  9.31583,  -2.1478,    10.5523,     4.78855,   1.63608,
-    5.53507,   2.60834,   -7.00058,  -6.46058, 5.4738,     2.43235,     -1.34603,  9.02452,
-    -7.5337,   -8.71074,  -7.30893,  -7.57253, -5.33752,   -4.87402,    -7.01364,  -6.86542,
-    -7.93331,  -7.94791,  -5.69392,  -6.16116, -7.32291,   -7.76491,    -6.41965,  -7.55783,
-    -7.87996,  -7.55785,  -6.69005,  -5.87906, 3.92147,    2.86809,     -1.5552,   9.66568,
-    5.07989,   1.47112,   -7.48524,  -5.0541,  -1.82724,   8.70402,     -2.00421,  9.88004,
-    -2.62153,  8.79332,   -7.52111,  -6.44819, 4.06424,    2.09518,     -6.65494,  -5.94752,
-    6.93878,   1.61033,   -3.95728,  7.60682,  5.67016,    2.21196,     -7.81507,  -5.79413,
-    -2.41152,  8.24128,   -3.83738,  9.21115,  4.5516,     4.55288,     -5.75551,  -5.93258,
-    4.56545,   2.59384,   -7.45614,  -9.47115, -2.39568,   9.67642,     5.57816,   1.45712,
-    -7.48184,  -6.41134,  -1.99415,  12.867,   -8.35854,   -6.69675,    -7.52559,  -7.6793,
-    5.7454,    3.1602,    2.94692,   1.87483,  -8.77324,   -6.66682,    -3.21125,  8.68662,
-    -6.25806,  -7.24972,  5.17639,   1.0747,   -2.44897,   11.4775,     -3.30172,  8.89955,
-    -2.85191,  8.21201,   -8.85893,  -6.1322,  4.08957,    1.30155,     -5.88132,  -7.31173,
-    -7.10309,  -7.22943,  -2.46068,  8.18334,  -7.01226,   -7.85464,    4.75411,   2.12347,
-    -3.42862,  10.5642,   7.16681,   1.4423,   5.42568,    2.39863,     -6.00833,  -8.22609,
-    -1.7619,   9.62466,   -2.49527,  8.99016,  -2.98837,   8.82863,     -2.97262,  8.54856,
-    -1.34142,  9.26871,   -5.99652,  -6.95795, -1.87061,   7.35277,     -8.68277,  -8.46425,
-    -7.01808,  -8.10441,  -7.04269,  -7.62501, -7.69783,   -6.88348,    -2.19829,  10.4896,
-    4.67396,   1.2032,    -5.58263,  -6.90298, -5.69224,   -4.29055,    4.77285,   1.27305,
-    -3.33469,  8.6929,    -2.54195,  8.47086,  4.46492,    1.21742,     5.41158,   -0.875373,
-    -8.68069,  -7.42278,  -3.88687,  8.07646,  4.6682,     2.00293,     -8.29799,  -8.64092,
-    -1.86382,  10.3829,   -6.51234,  -5.04193, 4.54458,    2.25219,     -1.93264,  9.32554,
-    -3.06285,  7.81641,   -6.90714,  -5.10786, 4.69653,    2.50286,     6.43757,   2.61401,
-    -1.85483,  8.9587,    4.60224,   3.07647,  4.4492,     2.1906,      5.02181,   2.40321,
-    -2.22923,  7.8888,    5.68943,   1.43793,  -6.71097,   -6.43817,    -5.00633,  -5.80006,
-    -2.43763,  8.53663,   5.72577,   2.44787,  -6.57079,   -5.17789,    -5.77867,  -4.92176,
-    -6.57222,  -6.06437,  3.96639,   2.25216,  -7.95177,   -9.80146,    4.92574,   2.30763,
-    -7.6221,   -8.20013,  -6.4132,   -6.91575, 4.01432,    2.36897,     3.0833,    1.54505,
-    -1.99416,  9.52807,   -7.85128,  -8.25973, -0.86423,   8.76525,     -6.31412,  -8.64087,
-    -8.07355,  -6.73717,  -2.52821,  8.01176,  -5.82357,   -6.65687,    -7.08865,  -7.73063,
-    -5.56251,  -6.99818,  -2.12513,  8.98159,  -6.89834,   -7.26863,    -7.92654,  -6.34346,
-    4.86201,   1.49442,   4.92905,   4.42847,  -5.57789,   -5.3186,     4.34232,   3.34888,
-    2.64614,   2.34723,   -4.10363,  8.41491,  -2.18648,   8.18706,     -3.39871,  8.19848,
-    -2.66098,  9.6026,    -6.95927,  -6.42774, -5.61392,   -7.74628,    5.60376,   4.18369,
-    5.28536,   4.13642,   4.8428,    0.457426, -6.33816,   -6.12095,    -2.4394,   8.62897,
-    4.56938,   2.45967,   4.0582,    0.958413, 5.62164,    1.64834,     5.73119,   2.58231,
-    4.66806,   1.96405,   -6.71905,  -6.87706, -2.18503,   8.88414,     -6.03901,  -6.33338,
-    -8.38435,  -6.12005,  0.0641622, 9.0735,   5.19967,    3.05395,     -5.48716,  -7.13016,
-    -6.85541,  -5.46789,  -1.88353,  8.15713,  4.27891,    3.1325,      -2.75816,  9.98586,
-    -2.03022,  9.34795,   -7.66741,  -7.50096, -3.39305,   9.16801,     -8.49476,  -5.71537,
-    -1.68378,  9.8278,    -7.41559,  -6.07205, -3.15577,   7.93274,     5.22381,   1.61388,
-    3.65739,   1.74854,   4.94251,   1.21889,  -7.12832,   -5.27276,    -9.58286,  -6.20223,
-    -2.21613,  8.29993,   5.34799,   2.92987,  4.09496,    2.37231,     -7.25183,  -5.79136,
-    -6.46981,  -7.12137,  -6.28607,  -9.8205,  4.52865,    1.06926,     -3.10984,  8.72259,
-    3.61865,   2.68153,   -5.96604,  -7.68329, 3.11435,    1.28126,     -1.1064,   7.61243,
-    -2.17688,  8.2658,    -3.27246,  7.2094,   -5.55143,   -6.32388,    -1.69667,  10.3705,
-    -2.16558,  7.25125,   -6.36572,  -6.70053, 4.12259,    3.38252,     -4.80554,  -7.79949,
-    -5.23966,  -6.13798,  4.21969,   1.69139,  -1.98985,   10.547,      -2.52269,  7.95658,
-    -6.75642,  -6.32862,  -3.51521,  7.8001,   4.70435,    -0.00229688, 6.25359,   2.4267,
-    5.82935,   0.745562,  5.24778,   2.15978,  5.48052,    1.32055,     -3.05358,  9.12521,
-    -3.18922,  9.24654,   4.47276,   2.11988,  5.36751,    2.02512,     -2.18511,  8.6292,
-    -2.48469,  9.51228,   5.57556,   3.24472,  -2.58121,   10.0178,     -6.12629,  -6.49895,
-    -4.54732,  8.0062,    -4.20166,  10.5438,  -7.61422,   -7.69036,    -4.42797,  8.98777,
-    4.45301,   1.53344,   4.59296,   2.45021,  -6.81264,   -6.36417,    4.62346,   3.16156,
-    -5.93007,  -8.36501,  -2.78425,  6.71237,  -6.17141,   -6.64689,    -5.20608,  8.95999,
-    -7.30598,  -5.73166,  4.39572,   2.93726,  -1.89503,   9.77179,     -5.683,    -7.48989,
-    4.80924,   0.559455,  -2.17793,  9.98983,  5.23728,    2.67434,     -7.03976,  -6.20877,
-    3.90435,   3.20926,   -7.78536,  -7.53388, -1.00684,   9.08838,     -5.26741,  -5.98327,
-    3.28002,   2.71942,   -1.47166,  8.50427,  -2.32733,   9.26251,     5.16271,   1.39947,
-    -6.59093,  -6.61979,  -2.44492,  7.93654,  -1.05805,   9.97356,     -3.1109,   10.8666,
-    3.38834,   3.41693,   4.83098,   2.01961,  -2.74013,   9.71049,     -3.34892,  8.41489,
-    4.94768,   0.263001,  3.57477,   1.66795,  5.78915,    1.26999,     -4.81812,  -5.67174,
-    -1.88508,  9.64263,   3.69048,   4.60555,  4.03037,    1.7862,      -7.4418,   -7.08933},
-   {0.127717,  0.211407, 0.195547, 0.21633,   0.39671,  0.229008,  0.20839,  0.169236,  0.314314,
-    0.322473,  0.169506, 0.45499,  0.147819,  0.296502, 0.15198,   0.356444, 0.0992833, 0.220833,
-    0.296206,  0.178067, 0.135359, 0.189725,  0.243099, 0.519986,  0.168105, 0.273465,  0.126033,
-    0.18045,   0.282832, 0.193901, 0.213704,  0.425046, 0.203191,  0.228674, 0.209267,  0.355039,
-    0.212918,  0.315495, 0.294112, 0.257576,  0.5786,   0.186019,  0.171919, 0.171919,  0.449151,
-    1.34947,   0.171919, 0.16341,  0.641387,  0.342115, 0.267343,  0.246125, 0.277612,  0.181462,
-    0.22944,   1.95598,  0.164897, 0.235803,  0.228273, 0.314629,  0.127403, 0.241241,  0.189362,
-    0.151691,  0.130085, 0.526707, 0.217069,  0.282306, 0.531523,  0.177035, 0.169776,  0.20395,
-    0.177165,  0.146628, 0.280013, 0.223033,  0.50947,  0.184133,  0.295329, 0.183219,  0.28166,
-    0.179348,  0.276462, 1.00283,  0.248147,  0.214453, 0.231732,  0.170672, 0.256893,  0.133271,
-    0.151137,  0.500823, 0.23678,  0.376983,  0.362061, 0.140013,  0.388863, 0.398552,  0.38015,
-    0.190081,  0.167115, 0.206884, 0.473849,  1.05117,  0.435665,  0.323618, 0.326201,  0.32226,
-    0.201787,  0.246496, 0.28325,  0.226596,  0.238153, 0.277268,  0.674629, 0.179433,  0.175651,
-    0.154778,  0.178195, 0.192796, 0.103571,  0.227621, 0.201124,  0.160525, 0.160964,  0.240099,
-    0.258027,  0.134127, 0.127717, 0.341378,  0.311595, 0.282306,  0.168988, 0.40775,   0.246125,
-    0.583131,  0.236804, 0.238633, 0.194824,  0.169315, 0.244227,  0.249511, 0.189725,  0.305662,
-    0.301415,  0.658641, 0.250944, 0.151792,  0.141383, 0.143843,  0.563347, 0.184216,  0.204155,
-    0.221764,  0.314908, 0.144518, 0.228808,  0.255785, 0.163457,  0.424705, 0.170202,  0.312598,
-    0.300629,  0.532614, 0.661392, 0.228273,  0.543432, 0.257175,  0.258994, 0.281413,  0.273897,
-    0.246837,  0.293489, 0.25533,  0.260492,  0.213704, 0.3091,    0.17103,  0.172285,  0.241399,
-    0.35999,   0.372243, 0.269191, 0.390239,  0.31761,  0.200593,  0.22197,  0.752914,  0.266571,
-    0.13102,   0.268659, 0.293723, 0.356294,  0.296258, 0.264531,  0.15468,  0.358535,  0.243711,
-    0.112147,  0.121659, 0.197101, 0.515292,  0.245628, 0.279863,  0.789807, 0.195156,  0.196073,
-    0.149564,  0.118675, 0.389373, 0.233821,  0.176128, 0.481088,  0.360027, 0.553152,  0.208207,
-    0.171608,  0.160489, 0.334298, 0.139426,  0.168603, 0.266199,  0.326458, 0.103571,  0.171208,
-    0.130961,  0.190887, 0.177229, 0.241651,  0.115152, 0.196753,  0.481088, 0.230965,  0.354631,
-    0.14591,   0.328543, 0.141544, 0.195888,  0.290379, 0.245954,  0.184547, 0.575214,  0.186929,
-    0.28527,   0.292213, 1.20033,  0.281528,  0.15625,  0.211524,  0.186398, 0.298061,  0.147393,
-    0.245349,  0.164527, 0.224771, 0.222382,  0.251643, 0.148835,  0.135359, 0.204967,  0.193024,
-    0.486309,  0.389686, 0.211921, 0.307405,  0.38666,  0.26802,   0.16605,  0.323134,  0.268397,
-    0.217894,  0.974118, 0.371618, 0.156201,  0.305787, 0.339305,  0.371032, 0.381765,  0.22747,
-    0.24906,   0.100884, 0.253192, 0.314253,  0.388289, 0.580947,  1.00267,  0.241998,  0.489101,
-    0.341501,  0.247423, 0.328311, 0.440281,  0.14927,  0.244469,  0.846828, 0.191725,  0.217429,
-    0.123403,  0.322875, 0.145373, 0.757259,  0.190086, 0.316286,  0.268397, 0.296721,  0.440472,
-    0.186848,  0.232134, 0.180239, 0.219724,  0.205886, 0.250975,  0.145636, 0.312476,  0.366418,
-    0.128135,  0.315235, 0.264531, 0.161815,  0.31631,  0.296489,  0.37171,  0.197217,  0.195625,
-    0.479579,  0.443037, 0.323347, 0.193616,  0.160251, 0.8952,    0.256291, 0.593345,  0.177165,
-    0.409514,  0.847863, 0.111448, 0.210031,  0.251347, 0.351953,  0.705204, 0.117901,  0.182343,
-    0.230179,  0.83632,  0.22104,  0.145163,  0.200326, 0.23431,   0.21868,  0.253575,  0.186562,
-    0.192757,  0.172716, 0.27396,  0.258581,  0.327892, 0.376138,  0.223477, 0.302375,  0.145845,
-    0.436902,  0.421794, 0.328543, 0.19246,   0.238889, 0.254866,  0.284674, 0.457849,  0.202937,
-    0.392568,  0.453083, 0.782713, 0.465401,  0.178623, 0.304863,  0.190081, 0.228641,  0.255135,
-    0.245037,  0.217526, 0.109584, 0.276462,  0.182301, 0.38582,   0.349942, 1.3889,    0.30235,
-    0.796353,  0.160168, 0.643204, 0.153752,  0.410268, 0.186439,  0.256834, 0.185783,  0.0957629,
-    0.226596,  0.197951, 0.17123,  0.192836,  0.18405,  0.575784,  0.228874, 0.201787,  0.241209,
-    0.217386,  0.195751, 0.291585, 0.144531,  0.14176,  0.157635,  0.410268, 0.476338,  0.308148,
-    0.148077,  0.152093, 0.196791, 0.568087,  0.414026, 0.250587,  0.473463, 0.293645,  0.396768,
-    0.2766,    0.38664,  0.135034, 1.50827,   0.472527, 0.268418,  0.40383,  0.375914,  0.246496,
-    0.176474,  0.340405, 0.220833, 0.138782,  0.159009, 0.444219,  0.259582, 0.33638,   0.195586,
-    0.210974,  0.200288, 0.148129, 0.0974216, 0.211588, 0.280081,  0.44113,  0.773921,  0.553848,
-    0.448079,  0.183136, 0.380854, 0.685021,  0.308767, 0.553276,  0.181578, 0.164759,  0.313889,
-    0.137886,  0.545387, 0.278449, 0.736895,  0.360054, 0.358929,  0.457315, 0.343278,  0.507662,
-    0.280829,  0.113886, 0.23146,  0.160584,  0.192796, 0.147561,  0.241272, 0.168988,  0.730511,
-    0.27836,   0.179847, 0.22555,  0.418069,  0.158348, 0.128965,  0.179454, 0.126366,  0.164434,
-    0.273633,  0.309556, 0.500823, 0.367852,  0.192875, 0.230262,  0.32724,  0.249969,  0.142618,
-    0.494229,  0.36108,  0.227931, 0.23113,   0.742825, 0.190126,  0.33741,  0.280598,  0.145268,
-    0.378423,  0.211921, 0.183594, 0.59201,   0.279563, 0.195683,  0.248101, 0.199754,  0.342494,
-    0.174343,  0.14149,  0.28085,  0.175781,  0.518738, 0.17223,   0.489904, 0.181167,  0.354286,
-    0.297824,  0.280829, 0.219412, 0.22814,   0.195625, 0.313949,  0.294708, 0.211551,  0.236255,
-    0.666933,  0.204808, 0.52591,  0.180725,  0.186889, 0.246589,  0.410575, 0.338348,  0.206219,
-    0.361766,  0.158143, 0.280816, 0.4149,    0.773082, 0.340046,  0.369672, 0.256923,  0.167195,
-    0.197217,  0.252339, 0.172716, 0.191526,  0.263085, 0.345698,  0.168286, 0.243099,  0.434631,
-    0.22944,   0.161862, 0.206589, 0.23457,   0.181924, 0.419063,  0.183427, 0.186152,  0.236352,
-    0.306336,  0.149002, 1.50086,  0.188231,  0.442757, 0.485602,  0.466662, 0.17329,   0.141329,
-    0.180619,  0.160061, 0.192569, 0.270999,  0.117901, 0.362693,  0.217561, 0.208975,  0.233658,
-    0.175173,  1.10307,  0.14625,  1.31124,   0.237608, 0.286784,  0.325112, 0.2485,    0.259641,
-    0.553152,  0.179039, 0.780781, 0.174758,  0.297824, 0.2558,    0.235949, 0.952186,  0.356744,
-    0.312646,  0.189362, 0.574524, 0.705204,  0.213168, 0.225956,  0.424165, 0.169506,  0.137109,
-    0.352451,  0.454554, 0.653302, 0.31261,   0.194412, 0.23719,   0.137886, 0.31498,   0.199085,
-    0.203875,  0.597248, 1.10036,  0.196869,  0.22104,  0.451345,  0.105613, 0.683928,  0.135204,
-    0.25533,   0.607871, 0.219724, 0.184464,  0.725001, 0.160061,  0.333407, 0.192569,  0.234147,
-    0.47178,   0.161815, 0.242455, 0.215305,  0.410575, 0.242376,  0.211335, 0.462804,  0.275065,
-    0.126878,  0.170404, 0.179433, 0.147244,  0.109584, 0.352905,  0.158215, 0.197604,  0.172407,
-    0.407506,  0.645446, 0.313061, 0.165602,  0.136663, 0.55444,   0.15527,  0.133128,  0.125912,
-    0.340405,  0.44521,  0.122783, 0.814526,  0.243773, 0.15743,   0.266743, 0.684458,  0.22221,
-    0.181294,  0.193901, 0.258802, 0.167195,  0.292056, 0.132309,  0.227671, 0.117334,  0.271758,
-    0.146185,  0.225042, 0.225964, 0.194863,  0.290274, 0.138438,  0.196714, 0.266012,  0.267771,
-    0.162544,  0.244258, 0.358038, 0.522617,  0.192875, 0.45066,   0.330396, 0.223477,  0.42967,
-    0.350884,  0.404655, 0.123155, 0.431583,  0.191675, 0.147354,  0.609034, 0.459487,  0.187337,
-    0.215128,  0.604169, 0.330165, 0.494229,  0.40775,  0.167377,  0.192648, 0.234635,  0.275578,
-    0.253094,  0.420063, 0.228299, 0.206478,  0.20395,  0.377656,  0.317393, 0.478623,  0.159009,
-    0.217034,  0.300933, 0.139754, 0.153901,  0.261077, 0.22834,   0.449609, 0.157672,  0.176474,
-    0.285704,  0.180186, 0.212738, 0.266428,  0.388313, 0.0954637, 0.298093, 0.251643,  0.330696,
-    0.159572,  0.210666, 0.149411, 0.139618,  0.338472, 0.450304,  0.208793, 0.583609,  0.185865,
-    0.400576,  0.21626,  0.174867, 0.239144,  0.249113, 0.200402,  0.275065, 0.238793,  0.205784,
-    0.4475,    0.231262, 0.259082, 0.20934,   0.16806,  0.193616,  0.213811, 0.395632,  0.482465,
-    0.274649,  0.307405, 0.165866, 0.334275,  0.683337, 0.368825,  0.14625,  0.780742,  0.163457,
-    0.226596,  0.138713, 1.79155,  0.400443,  0.233658, 0.426399,  0.623024, 0.670955,  0.123588,
-    0.110899,  0.173751, 0.651068, 0.199983,  0.190887, 0.541435,  0.21324,  0.266571,  0.134638,
-    0.179348,  0.145636, 0.170929, 0.623252,  0.587738, 0.109688,  0.515314, 0.217666,  0.213311,
-    0.249144,  0.187947, 0.270999, 0.268311,  0.469782, 0.763609,  0.32124,  0.146315,  0.265223,
-    0.298694,  0.197623, 0.21349,  0.845778,  0.175466, 0.123588,  0.17223,  0.258603,  1.17119,
-    0.538142,  0.407675, 0.120288, 0.587238,  0.244664, 0.333956,  0.132812, 0.21399,   0.302375,
-    0.275882,  0.134284, 0.377555, 0.228541,  0.187307, 0.143804,  0.180545, 0.222451,  0.239638,
-    0.188028,  0.46334,  0.175868, 0.242392,  0.314762, 0.44473,   0.21962,  0.175966,  1.12364,
-    0.138837,  0.400576, 0.18184,  0.137706,  0.409763, 0.216894,  0.466662, 0.376604,  0.487155,
-    0.283143,  0.118547, 0.221591, 0.122783,  0.179007, 0.16628,   0.180999, 0.239845,  0.169607,
-    0.578402,  0.396537, 0.222288, 0.563237,  0.371238, 0.138658,  0.324336, 0.191526,  0.168603,
-    0.357715,  0.640905, 0.460706, 0.220902,  0.240797, 0.164062,  0.157853, 0.34457,   0.196092,
-    0.289353,  0.104597, 0.259641, 0.126878,  0.175781, 0.441458,  0.820108, 0.261864,  0.23431,
-    0.254506,  0.271955, 0.227529, 0.22834,   0.196753, 0.224906,  0.193783, 0.419481,  0.236933,
-    0.229706,  0.29785,  0.222947, 0.177606,  0.216911, 0.305188,  0.933438, 0.116666,  0.278483,
-    0.0973824, 0.271224, 0.127717, 1.28139,   0.276283, 0.180704,  0.234554, 0.285984,  0.290172,
-    0.49594,   0.135879, 0.436784, 0.206219,  0.342215, 0.374165,  0.182217, 0.274864,  0.625,
-    0.356925,  0.194324, 0.342215, 0.113012,  0.155123, 0.254207,  0.438919, 0.262548,  0.302299,
-    0.179528,  0.312744, 0.168513, 0.142618,  0.150543, 0.231361,  0.166004, 0.186725,  0.38848,
-    0.179857,  0.182301, 0.629476, 0.44113,   0.289669, 0.328543,  0.279938, 0.14625,   0.187174,
-    0.157635,  0.396749, 0.798931, 0.201541,  0.778619, 0.265883,  0.258027, 0.218576,  0.266571,
-    0.160168,  0.230303, 0.273633, 0.233298,  0.30175,  0.217069,  0.345145, 0.397901,  0.224499,
-    0.248101,  0.241335, 0.222947, 0.237094,  0.176518, 0.380032,  0.634775, 0.426193,  0.16362,
-    0.231097,  0.219898, 0.343789, 0.275578,  0.282022, 0.628542,  0.232184, 0.848367,  0.200754,
-    0.179177},
-   {0, 0, 2, 3, 3, 0, 2, 2, 2, 2, 3, 0, 3, 2, 2, 2, 3, 3, 3, 3, 2, 0, 0, 0, 2, 3, 3, 3, 2, 2, 0, 0,
-    2, 3, 3, 0, 0, 2, 0, 0, 3, 2, 3, 0, 3, 0, 3, 3, 0, 2, 0, 3, 2, 0, 3, 0, 3, 3, 3, 2, 2, 3, 0, 0,
-    3, 3, 0, 2, 2, 3, 0, 3, 2, 2, 2, 0, 2, 3, 3, 3, 2, 3, 3, 3, 2, 0, 2, 0, 3, 3, 3, 3, 2, 2, 0, 2,
-    0, 3, 2, 2, 2, 0, 0, 3, 0, 2, 2, 3, 2, 3, 0, 2, 2, 2, 3, 2, 0, 0, 2, 3, 3, 2, 0, 2, 0, 0, 2, 0,
-    2, 2, 3, 2, 2, 0, 3, 0, 3, 2, 2, 2, 3, 3, 0, 0, 0, 3, 2, 3, 3, 3, 3, 0, 2, 0, 3, 2, 3, 2, 3, 0,
-    2, 3, 3, 2, 3, 3, 2, 2, 0, 0, 2, 3, 3, 2, 3, 0, 2, 0, 2, 0, 3, 2, 3, 2, 3, 0, 3, 0, 3, 0, 2, 3,
-    2, 2, 3, 0, 2, 2, 2, 0, 3, 2, 3, 3, 2, 3, 2, 3, 3, 2, 2, 0, 0, 2, 2, 3, 0, 3, 0, 2, 0, 0, 2, 3,
-    0, 3, 3, 2, 0, 3, 3, 0, 3, 0, 2, 2, 0, 2, 0, 2, 0, 0, 0, 2, 0, 3, 2, 3, 2, 3, 2, 2, 0, 2, 3, 2,
-    3, 2, 2, 2, 2, 3, 0, 2, 0, 0, 2, 3, 3, 0, 2, 3, 2, 2, 3, 0, 3, 0, 0, 2, 0, 2, 0, 2, 2, 3, 3, 2,
-    3, 0, 0, 3, 2, 2, 0, 3, 2, 0, 0, 3, 0, 0, 2, 0, 3, 2, 0, 2, 0, 0, 0, 0, 0, 2, 0, 0, 2, 3, 0, 0,
-    2, 0, 0, 2, 0, 2, 3, 2, 3, 3, 2, 2, 0, 0, 0, 3, 0, 2, 0, 2, 0, 2, 2, 2, 3, 3, 0, 0, 3, 3, 3, 3,
-    3, 2, 3, 3, 2, 3, 3, 0, 2, 2, 2, 2, 0, 2, 0, 0, 0, 2, 2, 3, 3, 2, 3, 2, 3, 0, 2, 3, 0, 2, 0, 2,
-    2, 0, 3, 0, 2, 0, 2, 3, 0, 3, 0, 0, 0, 3, 2, 3, 3, 0, 3, 2, 3, 0, 2, 3, 3, 0, 2, 3, 0, 0, 0, 2,
-    0, 3, 0, 2, 3, 3, 3, 3, 3, 0, 2, 0, 2, 2, 3, 3, 0, 3, 0, 2, 0, 2, 0, 3, 0, 0, 0, 2, 3, 3, 2, 3,
-    0, 0, 0, 0, 3, 3, 0, 3, 2, 0, 2, 3, 2, 2, 3, 3, 2, 2, 2, 0, 2, 3, 0, 3, 3, 0, 0, 2, 0, 3, 2, 3,
-    0, 2, 0, 2, 2, 3, 2, 0, 3, 3, 3, 2, 3, 0, 3, 0, 2, 2, 0, 0, 0, 3, 0, 3, 3, 2, 3, 2, 3, 2, 3, 0,
-    2, 3, 0, 2, 0, 3, 3, 3, 3, 3, 3, 2, 0, 3, 2, 2, 2, 3, 3, 2, 3, 0, 2, 3, 3, 2, 2, 0, 0, 0, 0, 3,
-    0, 3, 3, 3, 0, 0, 0, 3, 3, 3, 3, 3, 0, 2, 3, 3, 3, 3, 3, 3, 0, 0, 2, 2, 3, 3, 2, 2, 0, 0, 3, 0,
-    0, 0, 2, 3, 0, 0, 0, 3, 0, 3, 0, 2, 2, 0, 0, 0, 0, 3, 2, 2, 3, 2, 3, 2, 2, 2, 2, 3, 0, 0, 2, 3,
-    0, 3, 3, 0, 3, 0, 0, 2, 0, 3, 3, 0, 2, 2, 3, 3, 0, 0, 2, 0, 2, 3, 2, 0, 0, 3, 3, 0, 3, 2, 0, 2,
-    0, 2, 3, 2, 0, 3, 3, 2, 0, 0, 2, 2, 0, 0, 2, 0, 3, 3, 2, 3, 2, 0, 3, 0, 2, 2, 3, 3, 0, 3, 2, 2,
-    0, 3, 0, 0, 0, 2, 0, 3, 2, 0, 2, 3, 2, 3, 2, 2, 3, 3, 0, 2, 3, 2, 3, 2, 2, 0, 3, 0, 3, 0, 2, 2,
-    2, 0, 2, 0, 2, 2, 0, 0, 3, 3, 0, 0, 3, 2, 0, 2, 3, 2, 2, 0, 3, 3, 0, 2, 0, 3, 3, 0, 2, 3, 2, 3,
-    2, 0, 2, 2, 0, 0, 0, 2, 2, 3, 3, 2, 2, 0, 2, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 3, 2, 0, 3, 3,
-    3, 0, 2, 0, 2, 3, 2, 0, 3, 3, 2, 0, 2, 0, 3, 2, 0, 3, 0, 0, 2, 2, 0, 3, 0, 2, 3, 3, 3, 0, 2, 0,
-    0, 3, 0, 2, 3, 2, 2, 0, 3, 3, 3, 3, 3, 0, 3, 0, 0, 0, 0, 3, 2, 0, 0, 2, 3, 3, 2, 2, 0, 3, 2, 0,
-    3, 0, 2, 3, 3, 0, 2, 2, 3, 2, 2, 2, 3, 2, 0, 0, 3, 2, 0, 0, 0, 2, 0, 2, 0, 0, 2, 2, 3, 0, 3, 0,
-    0, 3, 0, 0, 0, 3, 0, 0, 2, 2, 0, 2, 2, 3, 3, 3, 3, 0, 0, 2, 2, 2, 0, 3, 2, 2, 2, 2, 2, 0, 3, 0,
-    0, 3, 2, 0, 0, 3, 2, 3, 3, 0, 3, 0, 3, 0, 3, 2, 2, 2, 0, 0, 3, 2, 2, 0, 0, 0, 2, 3, 2, 0, 2, 3,
-    3, 3, 0, 3, 3, 0, 2, 0, 0, 2, 3, 3, 0, 3, 2, 2, 2, 2, 2, 3, 3, 2, 2, 3, 3, 2, 3, 0, 3, 3, 0, 3,
-    2, 2, 0, 2, 0, 3, 0, 3, 0, 2, 3, 0, 2, 3, 2, 0, 2, 0, 3, 0, 2, 3, 3, 2, 0, 3, 3, 3, 2, 2, 3, 3,
-    2, 2, 2, 0, 3, 2, 2, 0},
-   {271, 271, 329, 343, 387, 426, 426, 601},
-   {426, 601, 426, 387, 343, 271, 329, 271},
-   {3.70991, 4.43491, 3.76334, 9.43944, 9.43944, 3.70991, 3.76334, 4.43491}}};
-
-typedef ConnectComponentsEdgesTest<int, float> ConnectComponentsEdgesTestF_Int;
-TEST_P(ConnectComponentsEdgesTestF_Int, Result) { EXPECT_TRUE(true); }
-
-INSTANTIATE_TEST_CASE_P(ConnectComponentsEdgesTest,
-                        ConnectComponentsEdgesTestF_Int,
-                        ::testing::ValuesIn(mr_fix_conn_inputsf2));
-
-};  // namespace sparse
-};  // end namespace raft
diff --git a/cpp/test/sparse/neighbors/knn_graph.cu b/cpp/test/sparse/neighbors/knn_graph.cu
deleted file mode 100644
index 41e6ec0df..000000000
--- a/cpp/test/sparse/neighbors/knn_graph.cu
+++ /dev/null
@@ -1,127 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/util/cudart_utils.hpp>
-#include <rmm/device_scalar.hpp>
-#include <rmm/device_uvector.hpp>
-
-#include <raft/sparse/coo.hpp>
-#include <raft/sparse/neighbors/knn_graph.cuh>
-
-#include <iostream>
-
-namespace raft {
-namespace sparse {
-
-template <typename value_idx, typename value_t>
-RAFT_KERNEL assert_symmetry(
-  value_idx* rows, value_idx* cols, value_t* vals, value_idx nnz, value_idx* sum)
-{
-  int tid = blockDim.x * blockIdx.x + threadIdx.x;
-
-  if (tid >= nnz) return;
-
-  atomicAdd(sum, rows[tid]);
-  atomicAdd(sum, -1 * cols[tid]);
-}
-
-template <typename value_idx, typename value_t>
-struct KNNGraphInputs {
-  value_idx m;
-  value_idx n;
-
-  std::vector<value_t> X;
-
-  int k = 2;
-};
-
-template <typename value_idx, typename value_t>
-::std::ostream& operator<<(::std::ostream& os, const KNNGraphInputs<value_idx, value_t>& dims)
-{
-  return os;
-}
-
-template <typename value_idx, typename value_t>
-class KNNGraphTest : public ::testing::TestWithParam<KNNGraphInputs<value_idx, value_t>> {
- public:
-  KNNGraphTest()
-    : params(::testing::TestWithParam<KNNGraphInputs<value_idx, value_t>>::GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      X(0, stream)
-  {
-    X.resize(params.X.size(), stream);
-  }
-
- protected:
-  void SetUp() override
-  {
-    out = new raft::sparse::COO<value_t, value_idx>(stream);
-
-    update_device(X.data(), params.X.data(), params.X.size(), stream);
-
-    raft::sparse::neighbors::knn_graph(
-      handle, X.data(), params.m, params.n, cuvs::distance::DistanceType::L2Unexpanded, *out);
-
-    rmm::device_scalar<value_idx> sum(stream);
-    sum.set_value_to_zero_async(stream);
-
-    /**
-     * Assert the knn graph is symmetric
-     */
-    assert_symmetry<<<raft::ceildiv(out->nnz, 256), 256, 0, stream>>>(
-      out->rows(), out->cols(), out->vals(), out->nnz, sum.data());
-
-    sum_h = sum.value(stream);
-    resource::sync_stream(handle, stream);
-  }
-
-  void TearDown() override { delete out; }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream;
-
-  // input data
-  raft::sparse::COO<value_t, value_idx>* out;
-
-  rmm::device_uvector<value_t> X;
-
-  value_idx sum_h;
-
-  KNNGraphInputs<value_idx, value_t> params;
-};
-
-const std::vector<KNNGraphInputs<int, float>> knn_graph_inputs_fint = {
-  // Test n_clusters == n_points
-  {4, 2, {0, 100, 0.01, 0.02, 5000, 10000, -5, -2}, 2}};
-
-typedef KNNGraphTest<int, float> KNNGraphTestF_int;
-TEST_P(KNNGraphTestF_int, Result)
-{
-  // nnz should not be larger than twice m * k
-  ASSERT_TRUE(out->nnz <= (params.m * params.k * 2));
-  ASSERT_TRUE(sum_h == 0);
-}
-
-INSTANTIATE_TEST_CASE_P(KNNGraphTest,
-                        KNNGraphTestF_int,
-                        ::testing::ValuesIn(knn_graph_inputs_fint));
-
-}  // namespace sparse
-}  // namespace raft
diff --git a/cpp/test/sparse/spectral_matrix.cu b/cpp/test/sparse/spectral_matrix.cu
deleted file mode 100644
index 0bed73a72..000000000
--- a/cpp/test/sparse/spectral_matrix.cu
+++ /dev/null
@@ -1,83 +0,0 @@
-/*
- * Copyright (c) 2020-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include <gtest/gtest.h>
-#include <iostream>
-#include <memory>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resource/device_id.hpp>
-#include <raft/core/resources.hpp>
-
-#include <raft/spectral/matrix_wrappers.hpp>
-
-namespace raft {
-namespace spectral {
-namespace matrix {
-namespace {
-template <typename index_type, typename value_type>
-struct csr_view_t {
-  index_type* offsets;
-  index_type* indices;
-  value_type* edge_data;
-  index_type number_of_vertices;
-  index_type number_of_edges;
-};
-}  // namespace
-TEST(Raft, SpectralMatrices)
-{
-  using index_type = int;
-  using value_type = double;
-
-  raft::resources h;
-  ASSERT_EQ(0, raft::resource::get_device_id(h));
-
-  csr_view_t<index_type, value_type> csr_v{nullptr, nullptr, nullptr, 0, 0};
-
-  int const sz = 10;
-  vector_t<index_type> d_v{h, sz};
-
-  index_type* ro{nullptr};
-  index_type* ci{nullptr};
-  value_type* vs{nullptr};
-  index_type nnz   = 0;
-  index_type nrows = 0;
-  sparse_matrix_t<index_type, value_type> sm1{h, ro, ci, vs, nrows, nnz};
-  sparse_matrix_t<index_type, value_type> sm2{h, csr_v};
-  ASSERT_EQ(nullptr, sm1.row_offsets_);
-  ASSERT_EQ(nullptr, sm2.row_offsets_);
-
-  auto stream = resource::get_cuda_stream(h);
-
-  auto cnstr_lm1 = [&h, ro, ci, vs, nrows, nnz](void) {
-    laplacian_matrix_t<index_type, value_type> lm1{h, ro, ci, vs, nrows, nnz};
-  };
-  EXPECT_ANY_THROW(cnstr_lm1());  // because of nullptr ptr args
-
-  auto cnstr_lm2 = [&h, &sm2](void) { laplacian_matrix_t<index_type, value_type> lm2{h, sm2}; };
-  EXPECT_ANY_THROW(cnstr_lm2());  // because of nullptr ptr args
-
-  auto cnstr_mm1 = [&h, ro, ci, vs, nrows, nnz](void) {
-    modularity_matrix_t<index_type, value_type> mm1{h, ro, ci, vs, nrows, nnz};
-  };
-  EXPECT_ANY_THROW(cnstr_mm1());  // because of nullptr ptr args
-
-  auto cnstr_mm2 = [&h, &sm2](void) { modularity_matrix_t<index_type, value_type> mm2{h, sm2}; };
-  EXPECT_ANY_THROW(cnstr_mm2());  // because of nullptr ptr args
-}
-
-}  // namespace matrix
-}  // namespace spectral
-}  // namespace raft
diff --git a/cpp/test/stats/accuracy.cu b/cpp/test/stats/accuracy.cu
deleted file mode 100644
index 5bc0506e7..000000000
--- a/cpp/test/stats/accuracy.cu
+++ /dev/null
@@ -1,106 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <optional>
-#include <raft/core/interruptible.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/random/rng.cuh>
-#include <raft/stats/accuracy.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <rmm/device_uvector.hpp>
-#include <stdio.h>
-#include <stdlib.h>
-#include <vector>
-
-namespace raft {
-namespace stats {
-
-template <typename T>
-struct AccuracyInputs {
-  T tolerance;
-  int nrows;
-  unsigned long long int seed;
-};
-
-template <typename T>
-::std::ostream& operator<<(::std::ostream& os, const AccuracyInputs<T>& dims)
-{
-  return os;
-}
-
-template <typename T>
-class AccuracyTest : public ::testing::TestWithParam<AccuracyInputs<T>> {
- protected:
-  AccuracyTest() : stream(resource::get_cuda_stream(handle)) {}
-
-  void SetUp() override
-  {
-    params = ::testing::TestWithParam<AccuracyInputs<T>>::GetParam();
-    raft::random::RngState r(params.seed);
-    rmm::device_uvector<int> predictions(params.nrows, stream);
-    rmm::device_uvector<int> ref_predictions(params.nrows, stream);
-    uniformInt(handle, r, predictions.data(), params.nrows, 0, 10);
-    uniformInt(handle, r, ref_predictions.data(), params.nrows, 0, 10);
-
-    actualVal =
-      accuracy(handle,
-               raft::make_device_vector_view<const int>(predictions.data(), params.nrows),
-               raft::make_device_vector_view<const int>(ref_predictions.data(), params.nrows));
-    expectedVal = T(0);
-    std::vector<int> h_predictions(params.nrows, 0);
-    std::vector<int> h_ref_predictions(params.nrows, 0);
-    raft::update_host(h_predictions.data(), predictions.data(), params.nrows, stream);
-    raft::update_host(h_ref_predictions.data(), ref_predictions.data(), params.nrows, stream);
-
-    unsigned long long correctly_predicted = 0ULL;
-    for (int i = 0; i < params.nrows; ++i) {
-      correctly_predicted += (h_predictions[i] - h_ref_predictions[i]) == 0;
-    }
-    expectedVal = correctly_predicted * 1.0f / params.nrows;
-    raft::interruptible::synchronize(stream);
-  }
-
- protected:
-  AccuracyInputs<T> params;
-  raft::resources handle;
-  cudaStream_t stream = 0;
-  T expectedVal, actualVal;
-};
-
-const std::vector<AccuracyInputs<float>> inputsf = {
-  {0.001f, 30, 1234ULL}, {0.001f, 100, 1234ULL}, {0.001f, 1000, 1234ULL}};
-typedef AccuracyTest<float> AccuracyTestF;
-TEST_P(AccuracyTestF, Result)
-{
-  auto eq = raft::CompareApprox<float>(params.tolerance);
-  ASSERT_TRUE(match(expectedVal, actualVal, eq));
-}
-INSTANTIATE_TEST_CASE_P(AccuracyTests, AccuracyTestF, ::testing::ValuesIn(inputsf));
-
-const std::vector<AccuracyInputs<double>> inputsd = {
-  {0.001, 30, 1234ULL}, {0.001, 100, 1234ULL}, {0.001, 1000, 1234ULL}};
-typedef AccuracyTest<double> AccuracyTestD;
-TEST_P(AccuracyTestD, Result)
-{
-  auto eq = raft::CompareApprox<double>(params.tolerance);
-  ASSERT_TRUE(match(expectedVal, actualVal, eq));
-}
-INSTANTIATE_TEST_CASE_P(AccuracyTests, AccuracyTestD, ::testing::ValuesIn(inputsd));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/adjusted_rand_index.cu b/cpp/test/stats/adjusted_rand_index.cu
deleted file mode 100644
index fb7b3825f..000000000
--- a/cpp/test/stats/adjusted_rand_index.cu
+++ /dev/null
@@ -1,207 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <algorithm>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/stats/adjusted_rand_index.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <random>
-
-namespace raft {
-namespace stats {
-
-struct adjustedRandIndexParam {
-  int nElements;
-  int lowerLabelRange;
-  int upperLabelRange;
-  bool sameArrays;
-  double tolerance;
-  // if this is true, then it is assumed that `sameArrays` is also true
-  // further it also assumes `lowerLabelRange` and `upperLabelRange` are 0
-  bool testZeroArray;
-};
-
-template <typename T, typename MathT = int>
-class adjustedRandIndexTest : public ::testing::TestWithParam<adjustedRandIndexParam> {
- protected:
-  adjustedRandIndexTest()
-    : stream(resource::get_cuda_stream(handle)),
-      firstClusterArray(0, stream),
-      secondClusterArray(0, stream)
-  {
-  }
-
-  void SetUp() override
-  {
-    params    = ::testing::TestWithParam<adjustedRandIndexParam>::GetParam();
-    nElements = params.nElements;
-
-    firstClusterArray.resize(nElements, stream);
-    secondClusterArray.resize(nElements, stream);
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(firstClusterArray.data(), 0, firstClusterArray.size() * sizeof(T), stream));
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(secondClusterArray.data(), 0, secondClusterArray.size() * sizeof(T), stream));
-
-    if (!params.testZeroArray) {
-      SetUpDifferentArrays();
-    } else {
-      SetupZeroArray();
-    }
-    // allocating and initializing memory to the GPU
-    computed_adjusted_rand_index = adjusted_rand_index<T, MathT>(
-      handle,
-      raft::make_device_vector_view<const T>(firstClusterArray.data(), nElements),
-      raft::make_device_vector_view<const T>(secondClusterArray.data(), nElements));
-  }
-
-  void SetUpDifferentArrays()
-  {
-    lowerLabelRange = params.lowerLabelRange;
-    upperLabelRange = params.upperLabelRange;
-    std::vector<int> arr1(nElements, 0);
-    std::vector<int> arr2(nElements, 0);
-    std::random_device rd;
-    std::default_random_engine dre(rd());
-    std::uniform_int_distribution<int> intGenerator(lowerLabelRange, upperLabelRange);
-    std::generate(arr1.begin(), arr1.end(), [&]() { return intGenerator(dre); });
-    if (params.sameArrays) {
-      arr2 = arr1;
-    } else {
-      std::generate(arr2.begin(), arr2.end(), [&]() { return intGenerator(dre); });
-    }
-    // calculating golden output
-    int numUniqueClasses = upperLabelRange - lowerLabelRange + 1;
-    size_t sizeOfMat     = numUniqueClasses * numUniqueClasses * sizeof(int);
-    int* hGoldenOutput   = (int*)malloc(sizeOfMat);
-    memset(hGoldenOutput, 0, sizeOfMat);
-    for (int i = 0; i < nElements; i++) {
-      int row    = arr1[i] - lowerLabelRange;
-      int column = arr2[i] - lowerLabelRange;
-      hGoldenOutput[row * numUniqueClasses + column] += 1;
-    }
-    int sumOfNijCTwo = 0;
-    int* a           = (int*)malloc(numUniqueClasses * sizeof(int));
-    int* b           = (int*)malloc(numUniqueClasses * sizeof(int));
-    memset(a, 0, numUniqueClasses * sizeof(int));
-    memset(b, 0, numUniqueClasses * sizeof(int));
-    int sumOfAiCTwo = 0;
-    int sumOfBiCTwo = 0;
-    // calculating the sum of number of pairwise points in each index
-    // and also the reducing contingency matrix along row and column
-    for (int i = 0; i < numUniqueClasses; ++i) {
-      for (int j = 0; j < numUniqueClasses; ++j) {
-        int Nij = hGoldenOutput[i * numUniqueClasses + j];
-        sumOfNijCTwo += ((Nij) * (Nij - 1)) / 2;
-        a[i] += hGoldenOutput[i * numUniqueClasses + j];
-        b[i] += hGoldenOutput[j * numUniqueClasses + i];
-      }
-    }
-    // claculating the sum of number pairwise points in ever column sum
-    // claculating the sum of number pairwise points in ever row sum
-    for (int i = 0; i < numUniqueClasses; ++i) {
-      sumOfAiCTwo += ((a[i]) * (a[i] - 1)) / 2;
-      sumOfBiCTwo += ((b[i]) * (b[i] - 1)) / 2;
-    }
-    // calculating the ARI
-    double nCTwo         = double(nElements) * double(nElements - 1) / 2.0;
-    double expectedIndex = (double(sumOfBiCTwo) * double(sumOfAiCTwo)) / double(nCTwo);
-    double maxIndex      = (double(sumOfAiCTwo) + double(sumOfBiCTwo)) / 2.0;
-    double index         = (double)sumOfNijCTwo;
-    if (maxIndex - expectedIndex)
-      truth_adjusted_rand_index = (index - expectedIndex) / (maxIndex - expectedIndex);
-    else
-      truth_adjusted_rand_index = 0;
-    raft::update_device(firstClusterArray.data(), &arr1[0], nElements, stream);
-    raft::update_device(secondClusterArray.data(), &arr2[0], nElements, stream);
-  }
-
-  void SetupZeroArray()
-  {
-    lowerLabelRange           = 0;
-    upperLabelRange           = 0;
-    truth_adjusted_rand_index = 1.0;
-  }
-
-  raft::resources handle;
-  cudaStream_t stream = 0;
-  adjustedRandIndexParam params;
-  T lowerLabelRange, upperLabelRange;
-  rmm::device_uvector<T> firstClusterArray;
-  rmm::device_uvector<T> secondClusterArray;
-  int nElements                       = 0;
-  double truth_adjusted_rand_index    = 0;
-  double computed_adjusted_rand_index = 0;
-};
-
-const std::vector<adjustedRandIndexParam> inputs = {
-  {199, 1, 10, false, 0.000001, false},
-  {200, 15, 100, false, 0.000001, false},
-  {100, 1, 20, false, 0.000001, false},
-  {10, 1, 10, false, 0.000001, false},
-  {198, 1, 100, false, 0.000001, false},
-  {300, 3, 99, false, 0.000001, false},
-  {199, 1, 10, true, 0.000001, false},
-  {200, 15, 100, true, 0.000001, false},
-  {100, 1, 20, true, 0.000001, false},
-  // FIXME: disabled temporarily due to flaky test
-  // {10, 1, 10, true, 0.000001, false},
-  {198, 1, 100, true, 0.000001, false},
-  {300, 3, 99, true, 0.000001, false},
-
-  {199, 0, 0, false, 0.000001, true},
-  {200, 0, 0, false, 0.000001, true},
-  {100, 0, 0, false, 0.000001, true},
-  {10, 0, 0, false, 0.000001, true},
-  {198, 0, 0, false, 0.000001, true},
-  {300, 0, 0, false, 0.000001, true},
-  {199, 0, 0, true, 0.000001, true},
-  {200, 0, 0, true, 0.000001, true},
-  {100, 0, 0, true, 0.000001, true},
-  {10, 0, 0, true, 0.000001, true},
-  {198, 0, 0, true, 0.000001, true},
-  {300, 0, 0, true, 0.000001, true},
-};
-
-const std::vector<adjustedRandIndexParam> large_inputs = {
-  {2000000, 1, 1000, false, 0.000001, false},
-  {2000000, 1, 1000, true, 0.000001, false},
-
-  {2000000, 0, 0, false, 0.000001, true},
-  {2000000, 0, 0, true, 0.000001, true},
-};
-
-typedef adjustedRandIndexTest<int, int> ARI_ii;
-TEST_P(ARI_ii, Result)
-{
-  ASSERT_NEAR(computed_adjusted_rand_index, truth_adjusted_rand_index, params.tolerance);
-}
-INSTANTIATE_TEST_CASE_P(adjusted_rand_index, ARI_ii, ::testing::ValuesIn(inputs));
-
-typedef adjustedRandIndexTest<int, unsigned long long> ARI_il;
-TEST_P(ARI_il, Result)
-{
-  ASSERT_NEAR(computed_adjusted_rand_index, truth_adjusted_rand_index, params.tolerance);
-}
-INSTANTIATE_TEST_CASE_P(adjusted_rand_index, ARI_il, ::testing::ValuesIn(inputs));
-INSTANTIATE_TEST_CASE_P(adjusted_rand_index_large, ARI_il, ::testing::ValuesIn(large_inputs));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/completeness_score.cu b/cpp/test/stats/completeness_score.cu
deleted file mode 100644
index c5c134418..000000000
--- a/cpp/test/stats/completeness_score.cu
+++ /dev/null
@@ -1,136 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "../test_utils.cuh"
-#include <algorithm>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/stats/completeness_score.cuh>
-#include <raft/stats/entropy.cuh>
-#include <raft/stats/mutual_info_score.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <random>
-
-namespace raft {
-namespace stats {
-
-// parameter structure definition
-struct completenessParam {
-  int nElements;
-  int lowerLabelRange;
-  int upperLabelRange;
-  bool sameArrays;
-  double tolerance;
-};
-
-// test fixture class
-template <typename T>
-class completenessTest : public ::testing::TestWithParam<completenessParam> {
- protected:
-  // the constructor
-  completenessTest() : stream(resource::get_cuda_stream(handle)) {}
-
-  void SetUp() override
-  {
-    // getting the parameters
-    params = ::testing::TestWithParam<completenessParam>::GetParam();
-
-    nElements       = params.nElements;
-    lowerLabelRange = params.lowerLabelRange;
-    upperLabelRange = params.upperLabelRange;
-
-    // generating random value test input
-    std::vector<int> arr1(nElements, 0);
-    std::vector<int> arr2(nElements, 0);
-    std::random_device rd;
-    std::default_random_engine dre(rd());
-    std::uniform_int_distribution<int> intGenerator(lowerLabelRange, upperLabelRange);
-
-    std::generate(arr1.begin(), arr1.end(), [&]() { return intGenerator(dre); });
-    if (params.sameArrays) {
-      arr2 = arr1;
-    } else {
-      std::generate(arr2.begin(), arr2.end(), [&]() { return intGenerator(dre); });
-    }
-
-    // allocating and initializing memory to the GPU
-    rmm::device_uvector<T> truthClusterArray(nElements, stream);
-    rmm::device_uvector<T> predClusterArray(nElements, stream);
-    raft::update_device(truthClusterArray.data(), arr1.data(), (int)nElements, stream);
-    raft::update_device(predClusterArray.data(), arr2.data(), (int)nElements, stream);
-
-    // calculating the golden output
-    double truthMI, truthEntropy;
-
-    truthMI      = raft::stats::mutual_info_score(truthClusterArray.data(),
-                                             predClusterArray.data(),
-                                             nElements,
-                                             lowerLabelRange,
-                                             upperLabelRange,
-                                             stream);
-    truthEntropy = raft::stats::entropy(
-      predClusterArray.data(), nElements, lowerLabelRange, upperLabelRange, stream);
-
-    if (truthEntropy) {
-      truthCompleteness = truthMI / truthEntropy;
-    } else
-      truthCompleteness = 1.0;
-
-    if (nElements == 0) truthCompleteness = 1.0;
-
-    // calling the completeness CUDA implementation
-    computedCompleteness = raft::stats::completeness_score(
-      handle,
-      raft::make_device_vector_view<const T>(truthClusterArray.data(), nElements),
-      raft::make_device_vector_view<const T>(predClusterArray.data(), nElements),
-      lowerLabelRange,
-      upperLabelRange);
-  }
-
-  // declaring the data values
-  raft::resources handle;
-  completenessParam params;
-  T lowerLabelRange, upperLabelRange;
-  int nElements               = 0;
-  double truthCompleteness    = 0;
-  double computedCompleteness = 0;
-  cudaStream_t stream         = 0;
-};
-
-// setting test parameter values
-const std::vector<completenessParam> inputs = {{199, 1, 10, false, 0.000001},
-                                               {200, 15, 100, false, 0.000001},
-                                               {100, 1, 20, false, 0.000001},
-                                               {10, 1, 10, false, 0.000001},
-                                               {198, 1, 100, false, 0.000001},
-                                               {300, 3, 99, false, 0.000001},
-                                               {199, 1, 10, true, 0.000001},
-                                               {200, 15, 100, true, 0.000001},
-                                               {100, 1, 20, true, 0.000001},
-                                               {10, 1, 10, true, 0.000001},
-                                               {198, 1, 100, true, 0.000001},
-                                               {300, 3, 99, true, 0.000001}};
-
-// writing the test suite
-typedef completenessTest<int> completenessTestClass;
-TEST_P(completenessTestClass, Result)
-{
-  ASSERT_NEAR(computedCompleteness, truthCompleteness, params.tolerance);
-}
-INSTANTIATE_TEST_CASE_P(completeness, completenessTestClass, ::testing::ValuesIn(inputs));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/contingencyMatrix.cu b/cpp/test/stats/contingencyMatrix.cu
deleted file mode 100644
index acfd1aecf..000000000
--- a/cpp/test/stats/contingencyMatrix.cu
+++ /dev/null
@@ -1,167 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <algorithm>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/interruptible.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/stats/contingency_matrix.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <random>
-#include <rmm/device_uvector.hpp>
-
-namespace raft {
-namespace stats {
-
-struct ContingencyMatrixParam {
-  int nElements;
-  int minClass;
-  int maxClass;
-  bool calcCardinality;
-  bool skipLabels;
-  float tolerance;
-};
-
-template <typename T>
-class ContingencyMatrixTest : public ::testing::TestWithParam<ContingencyMatrixParam> {
- protected:
-  ContingencyMatrixTest()
-    : stream(resource::get_cuda_stream(handle)),
-      dY(0, stream),
-      dYHat(0, stream),
-      dComputedOutput(0, stream),
-      dGoldenOutput(0, stream)
-  {
-  }
-
-  void SetUp() override
-  {
-    params = ::testing::TestWithParam<ContingencyMatrixParam>::GetParam();
-
-    int numElements     = params.nElements;
-    int lowerLabelRange = params.minClass;
-    int upperLabelRange = params.maxClass;
-
-    std::vector<int> y(numElements, 0);
-    std::vector<int> y_hat(numElements, 0);
-    std::random_device rd;
-    std::default_random_engine dre(rd());
-    std::uniform_int_distribution<int> intGenerator(lowerLabelRange, upperLabelRange);
-
-    std::generate(y.begin(), y.end(), [&]() { return intGenerator(dre); });
-    std::generate(y_hat.begin(), y_hat.end(), [&]() { return intGenerator(dre); });
-
-    if (params.skipLabels) {
-      // remove two label value from input arrays
-      int y1 = (upperLabelRange - lowerLabelRange) / 2;
-      int y2 = y1 + (upperLabelRange - lowerLabelRange) / 4;
-
-      // replacement values
-      int y1_R = y1 + 1;
-      int y2_R = y2 + 1;
-
-      std::replace(y.begin(), y.end(), y1, y1_R);
-      std::replace(y.begin(), y.end(), y2, y2_R);
-      std::replace(y_hat.begin(), y_hat.end(), y1, y1_R);
-      std::replace(y_hat.begin(), y_hat.end(), y2, y2_R);
-    }
-
-    dY.resize(numElements, stream);
-    dYHat.resize(numElements, stream);
-
-    raft::update_device(dYHat.data(), &y_hat[0], numElements, stream);
-    raft::update_device(dY.data(), &y[0], numElements, stream);
-
-    if (params.calcCardinality) {
-      raft::stats::get_input_class_cardinality(
-        handle,
-        raft::make_device_vector_view<const T>(dY.data(), numElements),
-        raft::make_host_scalar_view(&minLabel),
-        raft::make_host_scalar_view(&maxLabel));
-    } else {
-      minLabel = lowerLabelRange;
-      maxLabel = upperLabelRange;
-    }
-
-    numUniqueClasses = maxLabel - minLabel + 1;
-
-    dComputedOutput.resize(numUniqueClasses * numUniqueClasses, stream);
-    dGoldenOutput.resize(numUniqueClasses * numUniqueClasses, stream);
-
-    // generate golden output on CPU
-    size_t sizeOfMat = numUniqueClasses * numUniqueClasses * sizeof(int);
-    std::vector<int> hGoldenOutput(sizeOfMat, 0);
-
-    for (int i = 0; i < numElements; i++) {
-      auto row    = y[i] - minLabel;
-      auto column = y_hat[i] - minLabel;
-      hGoldenOutput[row * numUniqueClasses + column] += 1;
-    }
-
-    raft::update_device(
-      dGoldenOutput.data(), hGoldenOutput.data(), numUniqueClasses * numUniqueClasses, stream);
-    raft::interruptible::synchronize(stream);
-  }
-
-  void RunTest()
-  {
-    int numElements = params.nElements;
-    raft::stats::contingency_matrix(
-      handle,
-      raft::make_device_vector_view<const T>(dY.data(), numElements),
-      raft::make_device_vector_view<const T>(dYHat.data(), numElements),
-      raft::make_device_matrix_view(dComputedOutput.data(), numUniqueClasses, numUniqueClasses),
-      std::make_optional(minLabel),
-      std::make_optional(maxLabel));
-
-    raft::interruptible::synchronize(stream);
-    ASSERT_TRUE(raft::devArrMatch(dComputedOutput.data(),
-                                  dGoldenOutput.data(),
-                                  numUniqueClasses * numUniqueClasses,
-                                  raft::Compare<T>()));
-  }
-
-  raft::resources handle;
-  ContingencyMatrixParam params;
-  int numUniqueClasses = -1;
-  T minLabel, maxLabel;
-  cudaStream_t stream = 0;
-  rmm::device_uvector<T> dY, dYHat;
-  rmm::device_uvector<int> dComputedOutput, dGoldenOutput;
-};
-
-const std::vector<ContingencyMatrixParam> inputs = {
-  {10000, 1, 10, true, false, 0.000001},
-  {10000, 1, 5000, true, false, 0.000001},
-  {10000, 1, 10000, true, false, 0.000001},
-  {10000, 1, 20000, true, false, 0.000001},
-  {10000, 1, 10, false, false, 0.000001},
-  {10000, 1, 5000, false, false, 0.000001},
-  {10000, 1, 10000, false, false, 0.000001},
-  {10000, 1, 20000, false, false, 0.000001},
-  {100000, 1, 100, false, false, 0.000001},
-  {1000000, 1, 1200, true, false, 0.000001},
-  {1000000, 1, 10000, false, false, 0.000001},
-  {100000, 1, 100, false, true, 0.000001},
-};
-
-typedef ContingencyMatrixTest<int> ContingencyMatrixTestS;
-TEST_P(ContingencyMatrixTestS, Result) { RunTest(); }
-INSTANTIATE_TEST_CASE_P(ContingencyMatrix, ContingencyMatrixTestS, ::testing::ValuesIn(inputs));
-}  // namespace stats
-}  // namespace raft
diff --git a/cpp/test/stats/cov.cu b/cpp/test/stats/cov.cu
deleted file mode 100644
index ca9d43771..000000000
--- a/cpp/test/stats/cov.cu
+++ /dev/null
@@ -1,193 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/random/rng.cuh>
-#include <raft/stats/cov.cuh>
-#include <raft/stats/mean.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <rmm/device_uvector.hpp>
-
-namespace raft {
-namespace stats {
-
-template <typename T>
-struct CovInputs {
-  T tolerance, mean, var;
-  int rows, cols;
-  bool sample, rowMajor, stable;
-  unsigned long long int seed;
-};
-
-template <typename T>
-::std::ostream& operator<<(::std::ostream& os, const CovInputs<T>& dims)
-{
-  return os;
-}
-
-template <typename T>
-class CovTest : public ::testing::TestWithParam<CovInputs<T>> {
- protected:
-  CovTest()
-    : data(0, stream),
-      mean_act(0, stream),
-      cov_act(0, stream),
-      cov_cm(0, stream),
-      cov_cm_ref(0, stream)
-  {
-  }
-
-  void SetUp() override
-  {
-    raft::resources handle;
-    cudaStream_t stream = resource::get_cuda_stream(handle);
-
-    params = ::testing::TestWithParam<CovInputs<T>>::GetParam();
-    params.tolerance *= 2;
-    raft::random::RngState r(params.seed);
-    int rows = params.rows, cols = params.cols;
-    auto len = rows * cols;
-    T var    = params.var;
-    data.resize(len, stream);
-    mean_act.resize(cols, stream);
-    cov_act.resize(cols * cols, stream);
-
-    normal(handle, r, data.data(), len, params.mean, var);
-    raft::stats::mean(
-      mean_act.data(), data.data(), cols, rows, params.sample, params.rowMajor, stream);
-    if (params.rowMajor) {
-      using layout = raft::row_major;
-      cov(handle,
-          raft::make_device_matrix_view<T, std::uint32_t, layout>(data.data(), rows, cols),
-          raft::make_device_vector_view<const T, std::uint32_t>(mean_act.data(), cols),
-          raft::make_device_matrix_view<T, std::uint32_t, layout>(cov_act.data(), cols, cols),
-          params.sample,
-          params.stable);
-    } else {
-      using layout = raft::col_major;
-      cov(handle,
-          raft::make_device_matrix_view<T, std::uint32_t, layout>(data.data(), rows, cols),
-          raft::make_device_vector_view<const T, std::uint32_t>(mean_act.data(), cols),
-          raft::make_device_matrix_view<T, std::uint32_t, layout>(cov_act.data(), cols, cols),
-          params.sample,
-          params.stable);
-    }
-
-    T data_h[6]       = {1.0, 2.0, 5.0, 4.0, 2.0, 1.0};
-    T cov_cm_ref_h[4] = {4.3333, -2.8333, -2.8333, 2.333};
-
-    cov_cm.resize(4, stream);
-    cov_cm_ref.resize(4, stream);
-    rmm::device_uvector<T> data_cm(6, stream);
-    rmm::device_uvector<T> mean_cm(2, stream);
-
-    raft::update_device(data_cm.data(), data_h, 6, stream);
-    raft::update_device(cov_cm_ref.data(), cov_cm_ref_h, 4, stream);
-
-    raft::stats::mean(mean_cm.data(), data_cm.data(), 2, 3, true, false, stream);
-    cov(handle, cov_cm.data(), data_cm.data(), mean_cm.data(), 2, 3, true, false, true, stream);
-  }
-
- protected:
-  cublasHandle_t handle;
-  cudaStream_t stream = 0;
-  CovInputs<T> params;
-  rmm::device_uvector<T> data, mean_act, cov_act, cov_cm, cov_cm_ref;
-};
-
-///@todo: add stable=false after it has been implemented
-const std::vector<CovInputs<float>> inputsf = {
-  {0.03f, 1.f, 2.f, 32 * 1024, 32, true, false, true, 1234ULL},
-  {0.03f, 1.f, 2.f, 32 * 1024, 64, true, false, true, 1234ULL},
-  {0.03f, 1.f, 2.f, 32 * 1024, 128, true, false, true, 1234ULL},
-  {0.03f, 1.f, 2.f, 32 * 1024, 256, true, false, true, 1234ULL},
-  {0.03f, -1.f, 2.f, 32 * 1024, 32, false, false, true, 1234ULL},
-  {0.03f, -1.f, 2.f, 32 * 1024, 64, false, false, true, 1234ULL},
-  {0.03f, -1.f, 2.f, 32 * 1024, 128, false, false, true, 1234ULL},
-  {0.03f, -1.f, 2.f, 32 * 1024, 256, false, false, true, 1234ULL},
-  {0.03f, 1.f, 2.f, 32 * 1024, 32, true, true, true, 1234ULL},
-  {0.03f, 1.f, 2.f, 32 * 1024, 64, true, true, true, 1234ULL},
-  {0.03f, 1.f, 2.f, 32 * 1024, 128, true, true, true, 1234ULL},
-  {0.03f, 1.f, 2.f, 32 * 1024, 256, true, true, true, 1234ULL},
-  {0.03f, -1.f, 2.f, 32 * 1024, 32, false, true, true, 1234ULL},
-  {0.03f, -1.f, 2.f, 32 * 1024, 64, false, true, true, 1234ULL},
-  {0.03f, -1.f, 2.f, 32 * 1024, 128, false, true, true, 1234ULL},
-  {0.03f, -1.f, 2.f, 32 * 1024, 256, false, true, true, 1234ULL}};
-
-const std::vector<CovInputs<double>> inputsd = {
-  {0.03, 1.0, 2.0, 32 * 1024, 32, true, false, true, 1234ULL},
-  {0.03, 1.0, 2.0, 32 * 1024, 64, true, false, true, 1234ULL},
-  {0.03, 1.0, 2.0, 32 * 1024, 128, true, false, true, 1234ULL},
-  {0.03, 1.0, 2.0, 32 * 1024, 256, true, false, true, 1234ULL},
-  {0.03, -1.0, 2.0, 32 * 1024, 32, false, false, true, 1234ULL},
-  {0.03, -1.0, 2.0, 32 * 1024, 64, false, false, true, 1234ULL},
-  {0.03, -1.0, 2.0, 32 * 1024, 128, false, false, true, 1234ULL},
-  {0.03, -1.0, 2.0, 32 * 1024, 256, false, false, true, 1234ULL},
-  {0.03, 1.0, 2.0, 32 * 1024, 32, true, true, true, 1234ULL},
-  {0.03, 1.0, 2.0, 32 * 1024, 64, true, true, true, 1234ULL},
-  {0.03, 1.0, 2.0, 32 * 1024, 128, true, true, true, 1234ULL},
-  {0.03, 1.0, 2.0, 32 * 1024, 256, true, true, true, 1234ULL},
-  {0.03, -1.0, 2.0, 32 * 1024, 32, false, true, true, 1234ULL},
-  {0.03, -1.0, 2.0, 32 * 1024, 64, false, true, true, 1234ULL},
-  {0.03, -1.0, 2.0, 32 * 1024, 128, false, true, true, 1234ULL},
-  {0.03, -1.0, 2.0, 32 * 1024, 256, false, true, true, 1234ULL}};
-
-typedef CovTest<float> CovTestF;
-TEST_P(CovTestF, Result)
-{
-  ASSERT_TRUE(raft::diagonalMatch(params.var * params.var,
-                                  cov_act.data(),
-                                  params.cols,
-                                  params.cols,
-                                  raft::CompareApprox<float>(params.tolerance)));
-}
-
-typedef CovTest<double> CovTestD;
-TEST_P(CovTestD, Result)
-{
-  ASSERT_TRUE(raft::diagonalMatch(params.var * params.var,
-                                  cov_act.data(),
-                                  params.cols,
-                                  params.cols,
-                                  raft::CompareApprox<double>(params.tolerance)));
-}
-
-typedef CovTest<float> CovTestSmallF;
-TEST_P(CovTestSmallF, Result)
-{
-  ASSERT_TRUE(raft::devArrMatch(
-    cov_cm_ref.data(), cov_cm.data(), 2, 2, raft::CompareApprox<float>(params.tolerance)));
-}
-
-typedef CovTest<double> CovTestSmallD;
-TEST_P(CovTestSmallD, Result)
-{
-  ASSERT_TRUE(raft::devArrMatch(
-    cov_cm_ref.data(), cov_cm.data(), 2, 2, raft::CompareApprox<double>(params.tolerance)));
-}
-
-INSTANTIATE_TEST_CASE_P(CovTests, CovTestF, ::testing::ValuesIn(inputsf));
-
-INSTANTIATE_TEST_CASE_P(CovTests, CovTestD, ::testing::ValuesIn(inputsd));
-
-INSTANTIATE_TEST_CASE_P(CovTests, CovTestSmallF, ::testing::ValuesIn(inputsf));
-
-INSTANTIATE_TEST_CASE_P(CovTests, CovTestSmallD, ::testing::ValuesIn(inputsd));
-
-}  // namespace stats
-}  // namespace raft
diff --git a/cpp/test/stats/dispersion.cu b/cpp/test/stats/dispersion.cu
deleted file mode 100644
index 9ef678050..000000000
--- a/cpp/test/stats/dispersion.cu
+++ /dev/null
@@ -1,132 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <optional>
-#include <raft/core/interruptible.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/random/rng.cuh>
-#include <raft/stats/dispersion.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <rmm/device_uvector.hpp>
-#include <stdio.h>
-#include <stdlib.h>
-#include <vector>
-
-namespace raft {
-namespace stats {
-
-template <typename T>
-struct DispersionInputs {
-  T tolerance;
-  int dim, clusters;
-  unsigned long long int seed;
-};
-
-template <typename T>
-::std::ostream& operator<<(::std::ostream& os, const DispersionInputs<T>& dims)
-{
-  return os;
-}
-
-template <typename T>
-class DispersionTest : public ::testing::TestWithParam<DispersionInputs<T>> {
- protected:
-  DispersionTest()
-    : stream(resource::get_cuda_stream(handle)), exp_mean(0, stream), act_mean(0, stream)
-  {
-  }
-
-  void SetUp() override
-  {
-    params = ::testing::TestWithParam<DispersionInputs<T>>::GetParam();
-    raft::random::RngState r(params.seed);
-    int len = params.clusters * params.dim;
-    rmm::device_uvector<T> data(len, stream);
-    rmm::device_uvector<int> counts(params.clusters, stream);
-    exp_mean.resize(params.dim, stream);
-    act_mean.resize(params.dim, stream);
-    uniform(handle, r, data.data(), len, (T)-1.0, (T)1.0);
-    uniformInt(handle, r, counts.data(), params.clusters, 1, 100);
-    std::vector<int> h_counts(params.clusters, 0);
-    raft::update_host(&(h_counts[0]), counts.data(), params.clusters, stream);
-    npoints = 0;
-    for (const auto& val : h_counts) {
-      npoints += val;
-    }
-    actualVal = cluster_dispersion(
-      handle,
-      raft::make_device_matrix_view<const T, int>(data.data(), params.clusters, params.dim),
-      raft::make_device_vector_view<const int, int>(counts.data(), params.clusters),
-      std::make_optional(raft::make_device_vector_view<T, int>(act_mean.data(), params.dim)),
-      npoints);
-    expectedVal = T(0);
-    std::vector<T> h_data(len, T(0));
-    raft::update_host(&(h_data[0]), data.data(), len, stream);
-    std::vector<T> mean(params.dim, T(0));
-    for (int i = 0; i < params.clusters; ++i) {
-      for (int j = 0; j < params.dim; ++j) {
-        mean[j] += h_data[i * params.dim + j] * T(h_counts[i]);
-      }
-    }
-    for (int i = 0; i < params.dim; ++i) {
-      mean[i] /= T(npoints);
-    }
-    raft::update_device(exp_mean.data(), &(mean[0]), params.dim, stream);
-    for (int i = 0; i < params.clusters; ++i) {
-      for (int j = 0; j < params.dim; ++j) {
-        auto diff = h_data[i * params.dim + j] - mean[j];
-        expectedVal += diff * diff * T(h_counts[i]);
-      }
-    }
-    expectedVal = sqrt(expectedVal);
-    raft::interruptible::synchronize(stream);
-  }
-
- protected:
-  DispersionInputs<T> params;
-  raft::resources handle;
-  rmm::device_uvector<T> exp_mean, act_mean;
-  cudaStream_t stream = 0;
-  int npoints;
-  T expectedVal, actualVal;
-};
-
-const std::vector<DispersionInputs<float>> inputsf = {
-  {0.001f, 10, 1000, 1234ULL}, {0.001f, 100, 100, 1234ULL}, {0.001f, 1000, 1000, 1234ULL}};
-typedef DispersionTest<float> DispersionTestF;
-TEST_P(DispersionTestF, Result)
-{
-  auto eq = raft::CompareApprox<float>(params.tolerance);
-  ASSERT_TRUE(devArrMatch(exp_mean.data(), act_mean.data(), params.dim, eq));
-  ASSERT_TRUE(match(expectedVal, actualVal, eq));
-}
-INSTANTIATE_TEST_CASE_P(DispersionTests, DispersionTestF, ::testing::ValuesIn(inputsf));
-
-const std::vector<DispersionInputs<double>> inputsd = {
-  {0.001, 10, 1000, 1234ULL}, {0.001, 100, 100, 1234ULL}, {0.001, 1000, 1000, 1234ULL}};
-typedef DispersionTest<double> DispersionTestD;
-TEST_P(DispersionTestD, Result)
-{
-  auto eq = raft::CompareApprox<double>(params.tolerance);
-  ASSERT_TRUE(devArrMatch(exp_mean.data(), act_mean.data(), params.dim, eq));
-  ASSERT_TRUE(match(expectedVal, actualVal, eq));
-}
-INSTANTIATE_TEST_CASE_P(DispersionTests, DispersionTestD, ::testing::ValuesIn(inputsd));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/entropy.cu b/cpp/test/stats/entropy.cu
deleted file mode 100644
index dea8828b2..000000000
--- a/cpp/test/stats/entropy.cu
+++ /dev/null
@@ -1,123 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "../test_utils.cuh"
-#include <algorithm>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/interruptible.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/stats/entropy.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <random>
-#include <rmm/device_uvector.hpp>
-
-namespace raft {
-namespace stats {
-
-struct entropyParam {
-  int nElements;
-  int lowerLabelRange;
-  int upperLabelRange;
-  double tolerance;
-};
-
-// test fixture class
-template <typename T>
-class entropyTest : public ::testing::TestWithParam<entropyParam> {
- protected:
-  // the constructor
-  entropyTest() : stream(resource::get_cuda_stream(handle)) {}
-
-  void SetUp() override
-  {
-    // getting the parameters
-    params = ::testing::TestWithParam<entropyParam>::GetParam();
-
-    nElements       = params.nElements;
-    lowerLabelRange = params.lowerLabelRange;
-    upperLabelRange = params.upperLabelRange;
-
-    // generating random value test input
-    std::vector<int> arr1(nElements, 0);
-    std::random_device rd;
-    std::default_random_engine dre(rd());
-    std::uniform_int_distribution<int> intGenerator(lowerLabelRange, upperLabelRange);
-
-    std::generate(arr1.begin(), arr1.end(), [&]() { return intGenerator(dre); });
-
-    // generating the golden output
-    int numUniqueClasses = upperLabelRange - lowerLabelRange + 1;
-
-    int* p = (int*)malloc(numUniqueClasses * sizeof(int));
-    memset(p, 0, numUniqueClasses * sizeof(int));
-
-    // calculating the bincount array
-    for (int i = 0; i < nElements; ++i) {
-      ++p[arr1[i] - lowerLabelRange];
-    }
-
-    // calculating the aggregate entropy
-    for (int i = 0; i < numUniqueClasses; ++i) {
-      if (p[i])
-        truthEntropy +=
-          -1 * (double(p[i]) / double(nElements)) * (log(double(p[i])) - log(double(nElements)));
-    }
-
-    // allocating and initializing memory to the GPU
-    rmm::device_uvector<T> clusterArray(nElements, stream);
-    raft::update_device(clusterArray.data(), &arr1[0], (int)nElements, stream);
-
-    raft::interruptible::synchronize(stream);
-    // calling the entropy CUDA implementation
-    computedEntropy =
-      raft::stats::entropy(handle,
-                           raft::make_device_vector_view<const T>(clusterArray.data(), nElements),
-                           lowerLabelRange,
-                           upperLabelRange);
-  }
-
-  raft::resources handle;
-  // declaring the data values
-  entropyParam params;
-  T lowerLabelRange, upperLabelRange;
-
-  int nElements          = 0;
-  double truthEntropy    = 0;
-  double computedEntropy = 0;
-  cudaStream_t stream    = 0;
-};
-
-// setting test parameter values
-const std::vector<entropyParam> inputs = {{199, 1, 10, 0.000001},
-                                          {200, 15, 100, 0.000001},
-                                          {100, 1, 20, 0.000001},
-                                          {10, 1, 10, 0.000001},
-                                          {198, 1, 100, 0.000001},
-                                          {300, 3, 99, 0.000001},
-                                          {199, 1, 10, 0.000001},
-                                          {200, 15, 100, 0.000001},
-                                          {100, 1, 20, 0.000001},
-                                          {10, 1, 10, 0.000001},
-                                          {198, 1, 100, 0.000001},
-                                          {300, 3, 99, 0.000001}};
-
-// writing the test suite
-typedef entropyTest<int> entropyTestClass;
-TEST_P(entropyTestClass, Result) { ASSERT_NEAR(computedEntropy, truthEntropy, params.tolerance); }
-INSTANTIATE_TEST_CASE_P(entropy, entropyTestClass, ::testing::ValuesIn(inputs));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/histogram.cu b/cpp/test/stats/histogram.cu
deleted file mode 100644
index 86f708db6..000000000
--- a/cpp/test/stats/histogram.cu
+++ /dev/null
@@ -1,318 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <raft/core/interruptible.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/random/rng.cuh>
-#include <raft/stats/histogram.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-
-namespace raft {
-namespace stats {
-
-// Note: this kernel also updates the input vector to take care of OOB bins!
-RAFT_KERNEL naiveHistKernel(int* bins, int nbins, int* in, int nrows)
-{
-  int tid        = threadIdx.x + blockIdx.x * blockDim.x;
-  int stride     = blockDim.x * gridDim.x;
-  auto offset    = blockIdx.y * nrows;
-  auto binOffset = blockIdx.y * nbins;
-  for (; tid < nrows; tid += stride) {
-    int id = in[offset + tid];
-    if (id < 0)
-      id = 0;
-    else if (id >= nbins)
-      id = nbins - 1;
-    in[offset + tid] = id;
-    raft::myAtomicAdd(bins + binOffset + id, 1);
-  }
-}
-
-void naiveHist(int* bins, int nbins, int* in, int nrows, int ncols, cudaStream_t stream)
-{
-  const int TPB = 128;
-  int nblksx    = raft::ceildiv(nrows, TPB);
-  dim3 blks(nblksx, ncols);
-  naiveHistKernel<<<blks, TPB, 0, stream>>>(bins, nbins, in, nrows);
-  RAFT_CUDA_TRY(cudaGetLastError());
-}
-
-struct HistInputs {
-  int nrows, ncols, nbins;
-  bool isNormal;
-  HistType type;
-  int start, end;
-  unsigned long long int seed;
-};
-
-class HistTest : public ::testing::TestWithParam<HistInputs> {
- protected:
-  HistTest()
-    : in(0, resource::get_cuda_stream(handle)),
-      bins(0, resource::get_cuda_stream(handle)),
-      ref_bins(0, resource::get_cuda_stream(handle))
-  {
-  }
-
-  void SetUp() override
-  {
-    params = ::testing::TestWithParam<HistInputs>::GetParam();
-    raft::random::RngState r(params.seed);
-    auto stream = resource::get_cuda_stream(handle);
-    int len     = params.nrows * params.ncols;
-    in.resize(len, stream);
-    if (params.isNormal) {
-      normalInt(handle, r, in.data(), len, params.start, params.end);
-    } else {
-      uniformInt(handle, r, in.data(), len, params.start, params.end);
-    }
-    bins.resize(params.nbins * params.ncols, stream);
-    ref_bins.resize(params.nbins * params.ncols, stream);
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(ref_bins.data(), 0, sizeof(int) * params.nbins * params.ncols, stream));
-    naiveHist(ref_bins.data(), params.nbins, in.data(), params.nrows, params.ncols, stream);
-    histogram(handle,
-              params.type,
-              raft::make_device_matrix_view<const int, int, raft::col_major>(
-                in.data(), params.nrows, params.ncols),
-              raft::make_device_matrix_view<int, int, raft::col_major>(
-                bins.data(), params.nbins, params.ncols));
-    resource::sync_stream(handle);
-  }
-
- protected:
-  raft::resources handle;
-  HistInputs params;
-  rmm::device_uvector<int> in, bins, ref_bins;
-};
-
-class HistMdspanTest : public ::testing::TestWithParam<HistInputs> {
- protected:
-  HistMdspanTest()
-    : in(0, resource::get_cuda_stream(handle)),
-      bins(0, resource::get_cuda_stream(handle)),
-      ref_bins(0, resource::get_cuda_stream(handle))
-  {
-  }
-
-  void SetUp() override
-  {
-    params = ::testing::TestWithParam<HistInputs>::GetParam();
-    raft::random::RngState r(params.seed);
-    auto stream = resource::get_cuda_stream(handle);
-    int len     = params.nrows * params.ncols;
-    in.resize(len, stream);
-
-    raft::device_vector_view<int, int> in_view(in.data(), in.size());
-    if (params.isNormal) {
-      normalInt(handle, r, in_view, params.start, params.end);
-    } else {
-      uniformInt(handle, r, in_view, params.start, params.end);
-    }
-    bins.resize(params.nbins * params.ncols, stream);
-    ref_bins.resize(params.nbins * params.ncols, stream);
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(ref_bins.data(), 0, sizeof(int) * params.nbins * params.ncols, stream));
-    naiveHist(ref_bins.data(), params.nbins, in.data(), params.nrows, params.ncols, stream);
-    histogram<int>(
-      params.type, bins.data(), params.nbins, in.data(), params.nrows, params.ncols, stream);
-    resource::sync_stream(handle);
-  }
-
- protected:
-  raft::resources handle;
-  HistInputs params;
-  rmm::device_uvector<int> in, bins, ref_bins;
-};
-
-static const int oneK                = 1024;
-static const int oneM                = oneK * oneK;
-const std::vector<HistInputs> inputs = {
-  {oneM, 1, 2 * oneM, false, HistTypeGmem, 0, 2 * oneM, 1234ULL},
-  {oneM, 1, 2 * oneM, true, HistTypeGmem, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneM, false, HistTypeGmem, 0, 2 * oneM, 1234ULL},
-  {oneM + 1, 1, 2 * oneM, true, HistTypeGmem, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneM, false, HistTypeGmem, 0, 2 * oneM, 1234ULL},
-  {oneM + 2, 1, 2 * oneM, true, HistTypeGmem, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneM, false, HistTypeGmem, 0, 2 * oneM, 1234ULL},
-  {oneM, 21, 2 * oneM, true, HistTypeGmem, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneM, false, HistTypeGmem, 0, 2 * oneM, 1234ULL},
-  {oneM + 1, 21, 2 * oneM, true, HistTypeGmem, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneM, false, HistTypeGmem, 0, 2 * oneM, 1234ULL},
-  {oneM + 2, 21, 2 * oneM, true, HistTypeGmem, 1000, 50, 1234ULL},
-
-  {oneM, 1, 2 * oneK, false, HistTypeSmem, 0, 2 * oneK, 1234ULL},
-  {oneM, 1, 2 * oneK, true, HistTypeSmem, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, false, HistTypeSmem, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, true, HistTypeSmem, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, false, HistTypeSmem, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, true, HistTypeSmem, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneK, false, HistTypeSmem, 0, 2 * oneK, 1234ULL},
-  {oneM, 21, 2 * oneK, true, HistTypeSmem, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, false, HistTypeSmem, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, true, HistTypeSmem, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, false, HistTypeSmem, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, true, HistTypeSmem, 1000, 50, 1234ULL},
-
-  {oneM, 1, 2 * oneK, false, HistTypeSmemMatchAny, 0, 2 * oneK, 1234ULL},
-  {oneM, 1, 2 * oneK, true, HistTypeSmemMatchAny, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, false, HistTypeSmemMatchAny, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, true, HistTypeSmemMatchAny, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, false, HistTypeSmemMatchAny, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, true, HistTypeSmemMatchAny, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneK, false, HistTypeSmemMatchAny, 0, 2 * oneK, 1234ULL},
-  {oneM, 21, 2 * oneK, true, HistTypeSmemMatchAny, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, false, HistTypeSmemMatchAny, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, true, HistTypeSmemMatchAny, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, false, HistTypeSmemMatchAny, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, true, HistTypeSmemMatchAny, 1000, 50, 1234ULL},
-
-  {oneM, 1, 2 * oneK, false, HistTypeSmemBits16, 0, 2 * oneK, 1234ULL},
-  {oneM, 1, 2 * oneK, true, HistTypeSmemBits16, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, false, HistTypeSmemBits16, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, true, HistTypeSmemBits16, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, false, HistTypeSmemBits16, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, true, HistTypeSmemBits16, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneK, false, HistTypeSmemBits16, 0, 2 * oneK, 1234ULL},
-  {oneM, 21, 2 * oneK, true, HistTypeSmemBits16, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, false, HistTypeSmemBits16, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, true, HistTypeSmemBits16, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, false, HistTypeSmemBits16, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, true, HistTypeSmemBits16, 1000, 50, 1234ULL},
-
-  {oneM, 1, 2 * oneK, false, HistTypeSmemBits8, 0, 2 * oneK, 1234ULL},
-  {oneM, 1, 2 * oneK, true, HistTypeSmemBits8, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, false, HistTypeSmemBits8, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, true, HistTypeSmemBits8, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, false, HistTypeSmemBits8, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, true, HistTypeSmemBits8, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneK, false, HistTypeSmemBits8, 0, 2 * oneK, 1234ULL},
-  {oneM, 21, 2 * oneK, true, HistTypeSmemBits8, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, false, HistTypeSmemBits8, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, true, HistTypeSmemBits8, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, false, HistTypeSmemBits8, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, true, HistTypeSmemBits8, 1000, 50, 1234ULL},
-
-  {oneM, 1, 2 * oneK, false, HistTypeSmemBits4, 0, 2 * oneK, 1234ULL},
-  {oneM, 1, 2 * oneK, true, HistTypeSmemBits4, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, false, HistTypeSmemBits4, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, true, HistTypeSmemBits4, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, false, HistTypeSmemBits4, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, true, HistTypeSmemBits4, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneK, false, HistTypeSmemBits4, 0, 2 * oneK, 1234ULL},
-  {oneM, 21, 2 * oneK, true, HistTypeSmemBits4, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, false, HistTypeSmemBits4, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, true, HistTypeSmemBits4, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, false, HistTypeSmemBits4, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, true, HistTypeSmemBits4, 1000, 50, 1234ULL},
-
-  {oneM, 1, 2 * oneK, false, HistTypeSmemBits2, 0, 2 * oneK, 1234ULL},
-  {oneM, 1, 2 * oneK, true, HistTypeSmemBits2, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, false, HistTypeSmemBits2, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, true, HistTypeSmemBits2, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, false, HistTypeSmemBits2, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, true, HistTypeSmemBits2, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneK, false, HistTypeSmemBits2, 0, 2 * oneK, 1234ULL},
-  {oneM, 21, 2 * oneK, true, HistTypeSmemBits2, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, false, HistTypeSmemBits2, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, true, HistTypeSmemBits2, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, false, HistTypeSmemBits2, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, true, HistTypeSmemBits2, 1000, 50, 1234ULL},
-
-  {oneM, 1, 2 * oneK, false, HistTypeSmemBits1, 0, 2 * oneK, 1234ULL},
-  {oneM, 1, 2 * oneK, true, HistTypeSmemBits1, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, false, HistTypeSmemBits1, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, true, HistTypeSmemBits1, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, false, HistTypeSmemBits1, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, true, HistTypeSmemBits1, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneK, false, HistTypeSmemBits1, 0, 2 * oneK, 1234ULL},
-  {oneM, 21, 2 * oneK, true, HistTypeSmemBits1, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, false, HistTypeSmemBits1, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, true, HistTypeSmemBits1, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, false, HistTypeSmemBits1, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, true, HistTypeSmemBits1, 1000, 50, 1234ULL},
-
-  {oneM, 1, 2 * oneM, false, HistTypeSmemHash, 0, 2 * oneM, 1234ULL},
-  {oneM, 1, 2 * oneM, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneM, false, HistTypeSmemHash, 0, 2 * oneM, 1234ULL},
-  {oneM + 1, 1, 2 * oneM, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneM, false, HistTypeSmemHash, 0, 2 * oneM, 1234ULL},
-  {oneM + 2, 1, 2 * oneM, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-  {oneM, 1, 2 * oneK, false, HistTypeSmemHash, 0, 2 * oneK, 1234ULL},
-  {oneM, 1, 2 * oneK, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, false, HistTypeSmemHash, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, false, HistTypeSmemHash, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneM, false, HistTypeSmemHash, 0, 2 * oneM, 1234ULL},
-  {oneM, 21, 2 * oneM, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneM, false, HistTypeSmemHash, 0, 2 * oneM, 1234ULL},
-  {oneM + 1, 21, 2 * oneM, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneM, false, HistTypeSmemHash, 0, 2 * oneM, 1234ULL},
-  {oneM + 2, 21, 2 * oneM, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneK, false, HistTypeSmemHash, 0, 2 * oneK, 1234ULL},
-  {oneM, 21, 2 * oneK, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, false, HistTypeSmemHash, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, false, HistTypeSmemHash, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, true, HistTypeSmemHash, 1000, 50, 1234ULL},
-
-  {oneM, 1, 2 * oneM, false, HistTypeAuto, 0, 2 * oneM, 1234ULL},
-  {oneM, 1, 2 * oneM, true, HistTypeAuto, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneM, false, HistTypeAuto, 0, 2 * oneM, 1234ULL},
-  {oneM + 1, 1, 2 * oneM, true, HistTypeAuto, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneM, false, HistTypeAuto, 0, 2 * oneM, 1234ULL},
-  {oneM + 2, 1, 2 * oneM, true, HistTypeAuto, 1000, 50, 1234ULL},
-  {oneM, 1, 2 * oneK, false, HistTypeAuto, 0, 2 * oneK, 1234ULL},
-  {oneM, 1, 2 * oneK, true, HistTypeAuto, 1000, 50, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, false, HistTypeAuto, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 1, 2 * oneK, true, HistTypeAuto, 1000, 50, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, false, HistTypeAuto, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 1, 2 * oneK, true, HistTypeAuto, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneM, false, HistTypeAuto, 0, 2 * oneM, 1234ULL},
-  {oneM, 21, 2 * oneM, true, HistTypeAuto, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneM, false, HistTypeAuto, 0, 2 * oneM, 1234ULL},
-  {oneM + 1, 21, 2 * oneM, true, HistTypeAuto, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneM, false, HistTypeAuto, 0, 2 * oneM, 1234ULL},
-  {oneM + 2, 21, 2 * oneM, true, HistTypeAuto, 1000, 50, 1234ULL},
-  {oneM, 21, 2 * oneK, false, HistTypeAuto, 0, 2 * oneK, 1234ULL},
-  {oneM, 21, 2 * oneK, true, HistTypeAuto, 1000, 50, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, false, HistTypeAuto, 0, 2 * oneK, 1234ULL},
-  {oneM + 1, 21, 2 * oneK, true, HistTypeAuto, 1000, 50, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, false, HistTypeAuto, 0, 2 * oneK, 1234ULL},
-  {oneM + 2, 21, 2 * oneK, true, HistTypeAuto, 1000, 50, 1234ULL},
-};
-
-TEST_P(HistTest, Result)
-{
-  ASSERT_TRUE(raft::devArrMatch(
-    ref_bins.data(), bins.data(), params.nbins * params.ncols, raft::Compare<int>()));
-}
-INSTANTIATE_TEST_CASE_P(HistTests, HistTest, ::testing::ValuesIn(inputs));
-
-TEST_P(HistMdspanTest, Result)
-{
-  ASSERT_TRUE(raft::devArrMatch(
-    ref_bins.data(), bins.data(), params.nbins * params.ncols, raft::Compare<int>()));
-}
-INSTANTIATE_TEST_CASE_P(HistMdspanTests, HistMdspanTest, ::testing::ValuesIn(inputs));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/homogeneity_score.cu b/cpp/test/stats/homogeneity_score.cu
deleted file mode 100644
index 88247f5b5..000000000
--- a/cpp/test/stats/homogeneity_score.cu
+++ /dev/null
@@ -1,134 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "../test_utils.cuh"
-#include <algorithm>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/stats/homogeneity_score.cuh>
-#include <raft/stats/mutual_info_score.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <random>
-
-namespace raft {
-namespace stats {
-
-// parameter structure definition
-struct homogeneityParam {
-  int nElements;
-  int lowerLabelRange;
-  int upperLabelRange;
-  bool sameArrays;
-  double tolerance;
-};
-
-// test fixture class
-template <typename T>
-class homogeneityTest : public ::testing::TestWithParam<homogeneityParam> {
- protected:
-  // the constructor
-  void SetUp() override
-  {
-    // getting the parameters
-    params = ::testing::TestWithParam<homogeneityParam>::GetParam();
-
-    nElements       = params.nElements;
-    lowerLabelRange = params.lowerLabelRange;
-    upperLabelRange = params.upperLabelRange;
-    stream          = resource::get_cuda_stream(handle);
-
-    // generating random value test input
-    std::vector<int> arr1(nElements, 0);
-    std::vector<int> arr2(nElements, 0);
-    std::random_device rd;
-    std::default_random_engine dre(rd());
-    std::uniform_int_distribution<int> intGenerator(lowerLabelRange, upperLabelRange);
-
-    std::generate(arr1.begin(), arr1.end(), [&]() { return intGenerator(dre); });
-    if (params.sameArrays) {
-      arr2 = arr1;
-    } else {
-      std::generate(arr2.begin(), arr2.end(), [&]() { return intGenerator(dre); });
-    }
-
-    // allocating and initializing memory to the GPU
-    rmm::device_uvector<T> truthClusterArray(nElements, stream);
-    rmm::device_uvector<T> predClusterArray(nElements, stream);
-    raft::update_device(truthClusterArray.data(), &arr1[0], (int)nElements, stream);
-    raft::update_device(predClusterArray.data(), &arr2[0], (int)nElements, stream);
-
-    // calculating the golden output
-    double truthMI, truthEntropy;
-
-    truthMI      = raft::stats::mutual_info_score(truthClusterArray.data(),
-                                             predClusterArray.data(),
-                                             nElements,
-                                             lowerLabelRange,
-                                             upperLabelRange,
-                                             stream);
-    truthEntropy = raft::stats::entropy(
-      truthClusterArray.data(), nElements, lowerLabelRange, upperLabelRange, stream);
-
-    if (truthEntropy) {
-      truthHomogeneity = truthMI / truthEntropy;
-    } else
-      truthHomogeneity = 1.0;
-
-    if (nElements == 0) truthHomogeneity = 1.0;
-
-    // calling the homogeneity CUDA implementation
-    computedHomogeneity = raft::stats::homogeneity_score(
-      handle,
-      raft::make_device_vector_view<const T>(truthClusterArray.data(), nElements),
-      raft::make_device_vector_view<const T>(predClusterArray.data(), nElements),
-      lowerLabelRange,
-      upperLabelRange);
-  }
-
-  // declaring the data values
-  raft::resources handle;
-  homogeneityParam params;
-  T lowerLabelRange, upperLabelRange;
-  int nElements              = 0;
-  double truthHomogeneity    = 0;
-  double computedHomogeneity = 0;
-  cudaStream_t stream        = 0;
-};
-
-// setting test parameter values
-const std::vector<homogeneityParam> inputs = {{199, 1, 10, false, 0.000001},
-                                              {200, 15, 100, false, 0.000001},
-                                              {100, 1, 20, false, 0.000001},
-                                              {10, 1, 10, false, 0.000001},
-                                              {198, 1, 100, false, 0.000001},
-                                              {300, 3, 99, false, 0.000001},
-                                              {199, 1, 10, true, 0.000001},
-                                              {200, 15, 100, true, 0.000001},
-                                              {100, 1, 20, true, 0.000001},
-                                              {10, 1, 10, true, 0.000001},
-                                              {198, 1, 100, true, 0.000001},
-                                              {300, 3, 99, true, 0.000001}};
-
-// writing the test suite
-typedef homogeneityTest<int> homogeneityTestClass;
-TEST_P(homogeneityTestClass, Result)
-{
-  ASSERT_NEAR(computedHomogeneity, truthHomogeneity, params.tolerance);
-}
-INSTANTIATE_TEST_CASE_P(homogeneity, homogeneityTestClass, ::testing::ValuesIn(inputs));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/information_criterion.cu b/cpp/test/stats/information_criterion.cu
deleted file mode 100644
index 9e57f2c84..000000000
--- a/cpp/test/stats/information_criterion.cu
+++ /dev/null
@@ -1,151 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <raft/stats/information_criterion.cuh>
-
-#include <raft/core/resources.hpp>
-#include <raft/util/cudart_utils.hpp>
-#include <rmm/device_uvector.hpp>
-
-#include <gtest/gtest.h>
-
-#include <cmath>
-#include <random>
-#include <vector>
-
-namespace raft {
-namespace stats {
-
-template <typename T>
-void naive_ic(
-  T* h_ic, const T* h_loglike, IC_Type ic_type, int n_params, int batch_size, int n_samples)
-{
-  T ic_base{};
-  T N = static_cast<T>(n_params);
-  T M = static_cast<T>(n_samples);
-  switch (ic_type) {
-    case AIC: ic_base = (T)2 * N; break;
-    case AICc: ic_base = (T)2 * (N + (N * (N + (T)1)) / (M - N - (T)1)); break;
-    case BIC: ic_base = std::log(M) * N; break;
-  }
-#pragma omp parallel for
-  for (int bid = 0; bid < batch_size; bid++) {
-    h_ic[bid] = ic_base - (T)2.0 * h_loglike[bid];
-  }
-}
-
-template <typename T>
-struct BatchedICInputs {
-  int batch_size;
-  int n_params;
-  int n_samples;
-  IC_Type ic_type;
-  T tolerance;
-};
-
-template <typename T>
-class BatchedICTest : public ::testing::TestWithParam<BatchedICInputs<T>> {
- public:
-  BatchedICTest()
-    : params(::testing::TestWithParam<BatchedICInputs<T>>::GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      res_d(sizeof(T) * params.batch_size, stream)
-  {
-  }
-
- protected:
-  void SetUp() override
-  {
-    using std::vector;
-
-    // Create arrays
-    std::vector<T> loglike_h = std::vector<T>(params.batch_size);
-    res_h.resize(params.batch_size);
-    rmm::device_uvector<T> loglike_d(sizeof(T) * params.batch_size, stream);
-
-    // Generate random data
-    std::random_device rd;
-    std::mt19937 gen(rd());
-    std::uniform_real_distribution<T> udis(0.001, 1.0);  // 0 has no log
-    for (int i = 0; i < params.batch_size; i++)
-      loglike_h[i] = std::log(udis(gen));
-
-    // Copy the data to the device
-    raft::update_device(loglike_d.data(), loglike_h.data(), params.batch_size, stream);
-
-    // Compute the tested results
-    information_criterion_batched(
-      handle,
-      raft::make_device_vector_view<const T>(loglike_d.data(), params.batch_size),
-      raft::make_device_vector_view(res_d.data(), params.batch_size),
-      params.ic_type,
-      params.n_params,
-      params.n_samples);
-
-    // Compute the expected results
-    naive_ic(res_h.data(),
-             loglike_h.data(),
-             params.ic_type,
-             params.n_params,
-             params.batch_size,
-             params.n_samples);
-
-    RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
-  }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream = 0;
-  BatchedICInputs<T> params;
-  rmm::device_uvector<T> res_d;
-  std::vector<T> res_h;
-};
-
-// Test parameters (op, n_batches, m, n, p, q, tolerance)
-const std::vector<BatchedICInputs<double>> inputsd = {
-  {1, 5, 52, AIC, 1e-3}, {10, 7, 100, AICc, 1e-3}, {67, 2, 350, BIC, 1e-3}};
-
-// Test parameters (op, n_batches, m, n, p, q, tolerance)
-const std::vector<BatchedICInputs<float>> inputsf = {
-  {1, 5, 52, AIC, 1e-3}, {10, 7, 100, AICc, 1e-3}, {67, 2, 350, BIC, 1e-3}};
-
-using BatchedICTestD = BatchedICTest<double>;
-using BatchedICTestF = BatchedICTest<float>;
-TEST_P(BatchedICTestD, Result)
-{
-  ASSERT_TRUE(devArrMatchHost(res_h.data(),
-                              res_d.data(),
-                              params.batch_size,
-                              raft::CompareApprox<double>(params.tolerance),
-                              stream));
-}
-TEST_P(BatchedICTestF, Result)
-{
-  ASSERT_TRUE(devArrMatchHost(res_h.data(),
-                              res_d.data(),
-                              params.batch_size,
-                              raft::CompareApprox<float>(params.tolerance),
-                              stream));
-}
-
-INSTANTIATE_TEST_CASE_P(BatchedICTests, BatchedICTestD, ::testing::ValuesIn(inputsd));
-INSTANTIATE_TEST_CASE_P(BatchedICTests, BatchedICTestF, ::testing::ValuesIn(inputsf));
-
-}  // namespace stats
-}  // namespace raft
diff --git a/cpp/test/stats/kl_divergence.cu b/cpp/test/stats/kl_divergence.cu
deleted file mode 100644
index 571458367..000000000
--- a/cpp/test/stats/kl_divergence.cu
+++ /dev/null
@@ -1,107 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "../test_utils.cuh"
-#include <algorithm>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/stats/kl_divergence.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <random>
-
-namespace raft {
-namespace stats {
-
-// parameter structure definition
-struct klDivergenceParam {
-  int nElements;
-  double tolerance;
-};
-
-// test fixture class
-template <typename DataT>
-class klDivergenceTest : public ::testing::TestWithParam<klDivergenceParam> {
- protected:
-  // the constructor
-  void SetUp() override
-  {
-    // getting the parameters
-    params = ::testing::TestWithParam<klDivergenceParam>::GetParam();
-    stream = resource::get_cuda_stream(handle);
-
-    nElements = params.nElements;
-
-    // generating random value test input
-    std::vector<DataT> h_modelPDF(nElements, 0);
-    std::vector<DataT> h_candidatePDF(nElements, 0);
-    std::random_device rd;
-    std::default_random_engine dre(rd());
-    std::uniform_real_distribution<DataT> realGenerator(0.0, 1.0);
-
-    std::generate(h_modelPDF.begin(), h_modelPDF.end(), [&]() { return realGenerator(dre); });
-    std::generate(
-      h_candidatePDF.begin(), h_candidatePDF.end(), [&]() { return realGenerator(dre); });
-
-    // allocating and initializing memory to the GPU
-    rmm::device_uvector<DataT> d_modelPDF(nElements, stream);
-    rmm::device_uvector<DataT> d_candidatePDF(nElements, stream);
-    RAFT_CUDA_TRY(cudaMemset(d_modelPDF.data(), 0, d_modelPDF.size() * sizeof(DataT)));
-    RAFT_CUDA_TRY(cudaMemset(d_candidatePDF.data(), 0, d_candidatePDF.size() * sizeof(DataT)));
-
-    raft::update_device(d_modelPDF.data(), &h_modelPDF[0], (int)nElements, stream);
-    raft::update_device(d_candidatePDF.data(), &h_candidatePDF[0], (int)nElements, stream);
-
-    // generating the golden output
-    for (int i = 0; i < nElements; ++i) {
-      if (h_modelPDF[i] == 0.0)
-        truthklDivergence += 0;
-
-      else
-        truthklDivergence += h_modelPDF[i] * log(h_modelPDF[i] / h_candidatePDF[i]);
-    }
-
-    // calling the kl_divergence CUDA implementation
-    computedklDivergence = raft::stats::kl_divergence(
-      handle,
-      raft::make_device_vector_view<const DataT>(d_modelPDF.data(), nElements),
-      raft::make_device_vector_view<const DataT>(d_candidatePDF.data(), nElements));
-  }
-
-  // declaring the data values
-  raft::resources handle;
-  klDivergenceParam params;
-  int nElements              = 0;
-  DataT truthklDivergence    = 0;
-  DataT computedklDivergence = 0;
-  cudaStream_t stream        = 0;
-};
-
-// setting test parameter values
-const std::vector<klDivergenceParam> inputs = {
-  {500, 0.000001}, {200, 0.001}, {5000, 0.000001}, {500000, 0.000001}
-
-};
-
-// writing the test suite
-typedef klDivergenceTest<double> klDivergenceTestClass;
-TEST_P(klDivergenceTestClass, Result)
-{
-  ASSERT_NEAR(computedklDivergence, truthklDivergence, params.tolerance);
-}
-INSTANTIATE_TEST_CASE_P(klDivergence, klDivergenceTestClass, ::testing::ValuesIn(inputs));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/mean.cu b/cpp/test/stats/mean.cu
deleted file mode 100644
index 0cb90b6d4..000000000
--- a/cpp/test/stats/mean.cu
+++ /dev/null
@@ -1,149 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/random/rng.cuh>
-#include <raft/stats/mean.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <stdio.h>
-#include <stdlib.h>
-
-namespace raft {
-namespace stats {
-
-template <typename T>
-struct MeanInputs {
-  T tolerance, mean;
-  int rows, cols;
-  bool sample, rowMajor;
-  unsigned long long int seed;
-};
-
-template <typename T>
-::std::ostream& operator<<(::std::ostream& os, const MeanInputs<T>& dims)
-{
-  return os;
-}
-
-template <typename T>
-class MeanTest : public ::testing::TestWithParam<MeanInputs<T>> {
- public:
-  MeanTest()
-    : params(::testing::TestWithParam<MeanInputs<T>>::GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      rows(params.rows),
-      cols(params.cols),
-      data(rows * cols, stream),
-      mean_act(cols, stream)
-  {
-  }
-
- protected:
-  void SetUp() override
-  {
-    raft::random::RngState r(params.seed);
-    int len = rows * cols;
-    normal(handle, r, data.data(), len, params.mean, (T)1.0);
-    meanSGtest(data.data(), stream);
-  }
-
-  void meanSGtest(T* data, cudaStream_t stream)
-  {
-    int rows = params.rows, cols = params.cols;
-    if (params.rowMajor) {
-      using layout = raft::row_major;
-      mean(handle,
-           raft::make_device_matrix_view<const T, int, layout>(data, rows, cols),
-           raft::make_device_vector_view<T, int>(mean_act.data(), cols),
-           params.sample);
-    } else {
-      using layout = raft::col_major;
-      mean(handle,
-           raft::make_device_matrix_view<const T, int, layout>(data, rows, cols),
-           raft::make_device_vector_view<T, int>(mean_act.data(), cols),
-           params.sample);
-    }
-  }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream;
-
-  MeanInputs<T> params;
-  int rows, cols;
-  rmm::device_uvector<T> data, mean_act;
-};
-
-// Note: For 1024 samples, 256 experiments, a mean of 1.0 with stddev=1.0, the
-// measured mean (of a normal distribution) will fall outside of an epsilon of
-// 0.15 only 4/10000 times. (epsilon of 0.1 will fail 30/100 times)
-const std::vector<MeanInputs<float>> inputsf = {{0.15f, 1.f, 1024, 32, true, false, 1234ULL},
-                                                {0.15f, 1.f, 1024, 64, true, false, 1234ULL},
-                                                {0.15f, 1.f, 1024, 128, true, false, 1234ULL},
-                                                {0.15f, 1.f, 1024, 256, true, false, 1234ULL},
-                                                {0.15f, -1.f, 1024, 32, false, false, 1234ULL},
-                                                {0.15f, -1.f, 1024, 64, false, false, 1234ULL},
-                                                {0.15f, -1.f, 1024, 128, false, false, 1234ULL},
-                                                {0.15f, -1.f, 1024, 256, false, false, 1234ULL},
-                                                {0.15f, 1.f, 1024, 32, true, true, 1234ULL},
-                                                {0.15f, 1.f, 1024, 64, true, true, 1234ULL},
-                                                {0.15f, 1.f, 1024, 128, true, true, 1234ULL},
-                                                {0.15f, 1.f, 1024, 256, true, true, 1234ULL},
-                                                {0.15f, -1.f, 1024, 32, false, true, 1234ULL},
-                                                {0.15f, -1.f, 1024, 64, false, true, 1234ULL},
-                                                {0.15f, -1.f, 1024, 128, false, true, 1234ULL},
-                                                {0.15f, -1.f, 1024, 256, false, true, 1234ULL}};
-
-const std::vector<MeanInputs<double>> inputsd = {{0.15, 1.0, 1024, 32, true, false, 1234ULL},
-                                                 {0.15, 1.0, 1024, 64, true, false, 1234ULL},
-                                                 {0.15, 1.0, 1024, 128, true, false, 1234ULL},
-                                                 {0.15, 1.0, 1024, 256, true, false, 1234ULL},
-                                                 {0.15, -1.0, 1024, 32, false, false, 1234ULL},
-                                                 {0.15, -1.0, 1024, 64, false, false, 1234ULL},
-                                                 {0.15, -1.0, 1024, 128, false, false, 1234ULL},
-                                                 {0.15, -1.0, 1024, 256, false, false, 1234ULL},
-                                                 {0.15, 1.0, 1024, 32, true, true, 1234ULL},
-                                                 {0.15, 1.0, 1024, 64, true, true, 1234ULL},
-                                                 {0.15, 1.0, 1024, 128, true, true, 1234ULL},
-                                                 {0.15, 1.0, 1024, 256, true, true, 1234ULL},
-                                                 {0.15, -1.0, 1024, 32, false, true, 1234ULL},
-                                                 {0.15, -1.0, 1024, 64, false, true, 1234ULL},
-                                                 {0.15, -1.0, 1024, 128, false, true, 1234ULL},
-                                                 {0.15, -1.0, 1024, 256, false, true, 1234ULL}};
-
-typedef MeanTest<float> MeanTestF;
-TEST_P(MeanTestF, Result)
-{
-  ASSERT_TRUE(
-    devArrMatch(params.mean, mean_act.data(), params.cols, CompareApprox<float>(params.tolerance)));
-}
-
-typedef MeanTest<double> MeanTestD;
-TEST_P(MeanTestD, Result)
-{
-  ASSERT_TRUE(devArrMatch(
-    params.mean, mean_act.data(), params.cols, CompareApprox<double>(params.tolerance)));
-}
-
-INSTANTIATE_TEST_SUITE_P(MeanTests, MeanTestF, ::testing::ValuesIn(inputsf));
-
-INSTANTIATE_TEST_SUITE_P(MeanTests, MeanTestD, ::testing::ValuesIn(inputsd));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/meanvar.cu b/cpp/test/stats/meanvar.cu
deleted file mode 100644
index df3d9d9c0..000000000
--- a/cpp/test/stats/meanvar.cu
+++ /dev/null
@@ -1,161 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/matrix/math.cuh>
-#include <raft/random/rng.cuh>
-#include <raft/stats/meanvar.cuh>
-#include <raft/util/cudart_utils.hpp>
-
-#include <algorithm>
-
-namespace raft {
-namespace stats {
-
-template <typename T>
-struct MeanVarInputs {
-  T mean, stddev;
-  int rows, cols;
-  bool sample, rowMajor;
-  unsigned long long int seed;
-  static const int N_SIGMAS = 6;
-
-  T mean_tol() const { return T(N_SIGMAS) * stddev / sqrt(T(rows)); }
-
-  T var_tol() const
-  {
-    return T(N_SIGMAS) * stddev * stddev * sqrt(T(2.0) / T(std::max(1, rows - 1)));
-  }
-};
-
-template <typename T>
-::std::ostream& operator<<(::std::ostream& os, const MeanVarInputs<T>& ps)
-{
-  return os << "rows: " << ps.rows << "; cols: " << ps.cols << "; "
-            << (ps.rowMajor ? "row-major" : "col-major") << " (tolerance: mean = " << ps.mean_tol()
-            << ", var = " << ps.var_tol() << ")";
-}
-
-template <typename T>
-class MeanVarTest : public ::testing::TestWithParam<MeanVarInputs<T>> {
- public:
-  MeanVarTest()
-    : params(::testing::TestWithParam<MeanVarInputs<T>>::GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      data(params.rows * params.cols, stream),
-      mean_act(params.cols, stream),
-      vars_act(params.cols, stream)
-  {
-  }
-
- protected:
-  void SetUp() override
-  {
-    random::RngState r(params.seed);
-    normal(handle, r, data.data(), params.cols * params.rows, params.mean, params.stddev);
-
-    if (params.rowMajor) {
-      using layout = raft::row_major;
-      meanvar(
-        handle,
-        raft::make_device_matrix_view<const T, int, layout>(data.data(), params.rows, params.cols),
-        raft::make_device_vector_view<T, int>(mean_act.data(), params.cols),
-        raft::make_device_vector_view<T, int>(vars_act.data(), params.cols),
-        params.sample);
-    } else {
-      using layout = raft::col_major;
-      meanvar(
-        handle,
-        raft::make_device_matrix_view<const T, int, layout>(data.data(), params.rows, params.cols),
-        raft::make_device_vector_view<T, int>(mean_act.data(), params.cols),
-        raft::make_device_vector_view<T, int>(vars_act.data(), params.cols),
-        params.sample);
-    }
-    RAFT_CUDA_TRY(cudaStreamSynchronize(stream));
-  }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream;
-
-  MeanVarInputs<T> params;
-  rmm::device_uvector<T> data, mean_act, vars_act;
-};
-
-const std::vector<MeanVarInputs<float>> inputsf = {
-  {1.f, 2.f, 1024, 32, true, false, 1234ULL},    {1.f, 2.f, 1024, 64, true, false, 1234ULL},
-  {1.f, 2.f, 1024, 128, true, false, 1234ULL},   {1.f, 2.f, 1024, 256, true, false, 1234ULL},
-  {-1.f, 2.f, 1024, 32, false, false, 1234ULL},  {-1.f, 2.f, 1024, 64, false, false, 1234ULL},
-  {-1.f, 2.f, 1024, 128, false, false, 1234ULL}, {-1.f, 2.f, 1024, 256, false, false, 1234ULL},
-  {-1.f, 2.f, 1024, 256, false, false, 1234ULL}, {-1.f, 2.f, 1024, 257, false, false, 1234ULL},
-  {1.f, 2.f, 1024, 32, true, true, 1234ULL},     {1.f, 2.f, 1024, 64, true, true, 1234ULL},
-  {1.f, 2.f, 1024, 128, true, true, 1234ULL},    {1.f, 2.f, 1024, 256, true, true, 1234ULL},
-  {-1.f, 2.f, 1024, 32, false, true, 1234ULL},   {-1.f, 2.f, 1024, 64, false, true, 1234ULL},
-  {-1.f, 2.f, 1024, 128, false, true, 1234ULL},  {-1.f, 2.f, 1024, 256, false, true, 1234ULL},
-  {-1.f, 2.f, 1024, 257, false, true, 1234ULL},  {-1.f, 2.f, 700, 13, false, true, 1234ULL},
-  {10.f, 2.f, 500000, 811, false, true, 1234ULL}};
-
-const std::vector<MeanVarInputs<double>> inputsd = {{1.0, 2.0, 1024, 32, true, false, 1234ULL},
-                                                    {1.0, 2.0, 1024, 64, true, false, 1234ULL},
-                                                    {1.0, 2.0, 1024, 128, true, false, 1234ULL},
-                                                    {1.0, 2.0, 1024, 256, true, false, 1234ULL},
-                                                    {-1.0, 2.0, 1024, 32, false, false, 1234ULL},
-                                                    {-1.0, 2.0, 1024, 64, false, false, 1234ULL},
-                                                    {-1.0, 2.0, 1024, 128, false, false, 1234ULL},
-                                                    {-1.0, 2.0, 1024, 256, false, false, 1234ULL},
-                                                    {1.0, 2.0, 1024, 32, true, true, 1234ULL},
-                                                    {1.0, 2.0, 1024, 64, true, true, 1234ULL},
-                                                    {1.0, 2.0, 1024, 128, true, true, 1234ULL},
-                                                    {1.0, 2.0, 1024, 256, true, true, 1234ULL},
-                                                    {-1.0, 2.0, 1024, 32, false, true, 1234ULL},
-                                                    {-1.0, 2.0, 1024, 64, false, true, 1234ULL},
-                                                    {-1.0, 2.0, 1024, 128, false, true, 1234ULL},
-                                                    {-1.0, 2.0, 1024, 256, false, true, 1234ULL}};
-
-typedef MeanVarTest<float> MeanVarTestF;
-TEST_P(MeanVarTestF, Result)
-{
-  ASSERT_TRUE(devArrMatch(
-    params.mean, mean_act.data(), params.cols, CompareApprox<float>(params.mean_tol()), stream));
-
-  ASSERT_TRUE(devArrMatch(params.stddev * params.stddev,
-                          vars_act.data(),
-                          params.cols,
-                          CompareApproxNoScaling<float>(params.var_tol()),
-                          stream));
-}
-
-typedef MeanVarTest<double> MeanVarTestD;
-TEST_P(MeanVarTestD, Result)
-{
-  ASSERT_TRUE(devArrMatch(
-    params.mean, mean_act.data(), params.cols, CompareApprox<double>(params.mean_tol()), stream));
-
-  ASSERT_TRUE(devArrMatch(params.stddev * params.stddev,
-                          vars_act.data(),
-                          params.cols,
-                          CompareApproxNoScaling<double>(params.var_tol()),
-                          stream));
-}
-
-INSTANTIATE_TEST_SUITE_P(MeanVarTests, MeanVarTestF, ::testing::ValuesIn(inputsf));
-
-INSTANTIATE_TEST_SUITE_P(MeanVarTests, MeanVarTestD, ::testing::ValuesIn(inputsd));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/minmax.cu b/cpp/test/stats/minmax.cu
deleted file mode 100644
index 3715bc5bd..000000000
--- a/cpp/test/stats/minmax.cu
+++ /dev/null
@@ -1,208 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <limits>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/random/rng.cuh>
-#include <raft/stats/minmax.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <stdio.h>
-#include <stdlib.h>
-
-namespace raft {
-namespace stats {
-
-///@todo: need to add tests for verifying the column subsampling feature
-
-template <typename T>
-struct MinMaxInputs {
-  T tolerance;
-  int rows, cols;
-  unsigned long long int seed;
-};
-
-template <typename T>
-::std::ostream& operator<<(::std::ostream& os, const MinMaxInputs<T>& dims)
-{
-  return os;
-}
-
-template <typename T>
-RAFT_KERNEL naiveMinMaxInitKernel(int ncols, T* globalmin, T* globalmax, T init_val)
-{
-  int tid = threadIdx.x + blockIdx.x * blockDim.x;
-  if (tid >= ncols) return;
-  globalmin[tid] = init_val;
-  globalmax[tid] = -init_val;
-}
-
-template <typename T>
-RAFT_KERNEL naiveMinMaxKernel(const T* data, int nrows, int ncols, T* globalmin, T* globalmax)
-{
-  int tid = threadIdx.x + blockIdx.x * blockDim.x;
-  int col = tid / nrows;
-  if (col < ncols) {
-    T val = data[tid];
-    if (!isnan(val)) {
-      raft::myAtomicMin(&globalmin[col], val);
-      raft::myAtomicMax(&globalmax[col], val);
-    }
-  }
-}
-
-template <typename T>
-void naiveMinMax(
-  const T* data, int nrows, int ncols, T* globalmin, T* globalmax, cudaStream_t stream)
-{
-  const int TPB = 128;
-  int nblks     = raft::ceildiv(ncols, TPB);
-  T init_val    = std::numeric_limits<T>::max();
-  naiveMinMaxInitKernel<<<nblks, TPB, 0, stream>>>(ncols, globalmin, globalmax, init_val);
-  RAFT_CUDA_TRY(cudaGetLastError());
-  nblks = raft::ceildiv(nrows * ncols, TPB);
-  naiveMinMaxKernel<<<nblks, TPB, 0, stream>>>(data, nrows, ncols, globalmin, globalmax);
-  RAFT_CUDA_TRY(cudaGetLastError());
-}
-
-template <typename T>
-RAFT_KERNEL nanKernel(T* data, const bool* mask, int len, T nan)
-{
-  int tid = threadIdx.x + blockIdx.x * blockDim.x;
-  if (tid >= len) return;
-  if (!mask[tid]) data[tid] = nan;
-}
-
-template <typename T>
-class MinMaxTest : public ::testing::TestWithParam<MinMaxInputs<T>> {
- protected:
-  MinMaxTest()
-    : minmax_act(0, resource::get_cuda_stream(handle)),
-      minmax_ref(0, resource::get_cuda_stream(handle))
-  {
-  }
-
-  void SetUp() override
-  {
-    auto stream = resource::get_cuda_stream(handle);
-    params      = ::testing::TestWithParam<MinMaxInputs<T>>::GetParam();
-    raft::random::RngState r(params.seed);
-    int len = params.rows * params.cols;
-
-    rmm::device_uvector<T> data(len, stream);
-    rmm::device_uvector<bool> mask(len, stream);
-    minmax_act.resize(2 * params.cols, stream);
-    minmax_ref.resize(2 * params.cols, stream);
-
-    normal(handle, r, data.data(), len, (T)0.0, (T)1.0);
-    T nan_prob = 0.01;
-    bernoulli(handle, r, mask.data(), len, nan_prob);
-    const int TPB = 256;
-    nanKernel<<<raft::ceildiv(len, TPB), TPB, 0, stream>>>(
-      data.data(), mask.data(), len, std::numeric_limits<T>::quiet_NaN());
-    RAFT_CUDA_TRY(cudaPeekAtLastError());
-    naiveMinMax(data.data(),
-                params.rows,
-                params.cols,
-                minmax_ref.data(),
-                minmax_ref.data() + params.cols,
-                stream);
-    raft::stats::minmax<T, int>(
-      handle,
-      raft::make_device_matrix_view<const T, int, raft::layout_f_contiguous>(
-        data.data(), params.rows, params.cols),
-      std::nullopt,
-      std::nullopt,
-      raft::make_device_vector_view<T, int>(minmax_act.data(), params.cols),
-      raft::make_device_vector_view<T, int>(minmax_act.data() + params.cols, params.cols),
-      std::nullopt);
-  }
-
- protected:
-  raft::resources handle;
-  MinMaxInputs<T> params;
-  rmm::device_uvector<T> minmax_act;
-  rmm::device_uvector<T> minmax_ref;
-};
-
-const std::vector<MinMaxInputs<float>> inputsf = {{0.00001f, 1024, 32, 1234ULL},
-                                                  {0.00001f, 1024, 64, 1234ULL},
-                                                  {0.00001f, 1024, 128, 1234ULL},
-                                                  {0.00001f, 1024, 256, 1234ULL},
-                                                  {0.00001f, 1024, 512, 1234ULL},
-                                                  {0.00001f, 1024, 1024, 1234ULL},
-                                                  {0.00001f, 4096, 32, 1234ULL},
-                                                  {0.00001f, 4096, 64, 1234ULL},
-                                                  {0.00001f, 4096, 128, 1234ULL},
-                                                  {0.00001f, 4096, 256, 1234ULL},
-                                                  {0.00001f, 4096, 512, 1234ULL},
-                                                  {0.00001f, 4096, 1024, 1234ULL},
-                                                  {0.00001f, 8192, 32, 1234ULL},
-                                                  {0.00001f, 8192, 64, 1234ULL},
-                                                  {0.00001f, 8192, 128, 1234ULL},
-                                                  {0.00001f, 8192, 256, 1234ULL},
-                                                  {0.00001f, 8192, 512, 1234ULL},
-                                                  {0.00001f, 8192, 1024, 1234ULL},
-                                                  {0.00001f, 1024, 8192, 1234ULL}};
-
-const std::vector<MinMaxInputs<double>> inputsd = {{0.0000001, 1024, 32, 1234ULL},
-                                                   {0.0000001, 1024, 64, 1234ULL},
-                                                   {0.0000001, 1024, 128, 1234ULL},
-                                                   {0.0000001, 1024, 256, 1234ULL},
-                                                   {0.0000001, 1024, 512, 1234ULL},
-                                                   {0.0000001, 1024, 1024, 1234ULL},
-                                                   {0.0000001, 4096, 32, 1234ULL},
-                                                   {0.0000001, 4096, 64, 1234ULL},
-                                                   {0.0000001, 4096, 128, 1234ULL},
-                                                   {0.0000001, 4096, 256, 1234ULL},
-                                                   {0.0000001, 4096, 512, 1234ULL},
-                                                   {0.0000001, 4096, 1024, 1234ULL},
-                                                   {0.0000001, 8192, 32, 1234ULL},
-                                                   {0.0000001, 8192, 64, 1234ULL},
-                                                   {0.0000001, 8192, 128, 1234ULL},
-                                                   {0.0000001, 8192, 256, 1234ULL},
-                                                   {0.0000001, 8192, 512, 1234ULL},
-                                                   {0.0000001, 8192, 1024, 1234ULL},
-                                                   {0.0000001, 1024, 8192, 1234ULL}};
-
-typedef MinMaxTest<float> MinMaxTestF;
-TEST_P(MinMaxTestF, Result)
-{
-  ASSERT_TRUE(raft::devArrMatch(minmax_ref.data(),
-                                minmax_act.data(),
-                                2 * params.cols,
-                                raft::CompareApprox<float>(params.tolerance)));
-}
-
-typedef MinMaxTest<double> MinMaxTestD;
-TEST_P(MinMaxTestD, Result)
-{
-  ASSERT_TRUE(raft::devArrMatch(minmax_ref.data(),
-                                minmax_act.data(),
-                                2 * params.cols,
-                                raft::CompareApprox<double>(params.tolerance)));
-}
-
-INSTANTIATE_TEST_CASE_P(MinMaxTests, MinMaxTestF, ::testing::ValuesIn(inputsf));
-
-INSTANTIATE_TEST_CASE_P(MinMaxTests, MinMaxTestD, ::testing::ValuesIn(inputsd));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/mutual_info_score.cu b/cpp/test/stats/mutual_info_score.cu
deleted file mode 100644
index 9f3135084..000000000
--- a/cpp/test/stats/mutual_info_score.cu
+++ /dev/null
@@ -1,162 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "../test_utils.cuh"
-#include <algorithm>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resources.hpp>
-#include <raft/stats/mutual_info_score.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <random>
-
-namespace raft {
-namespace stats {
-
-// parameter structure definition
-struct mutualInfoParam {
-  int nElements;
-  int lowerLabelRange;
-  int upperLabelRange;
-  bool sameArrays;
-  double tolerance;
-};
-
-// test fixture class
-template <typename T>
-class mutualInfoTest : public ::testing::TestWithParam<mutualInfoParam> {
- protected:
-  // the constructor
-  void SetUp() override
-  {
-    // getting the parameters
-    params = ::testing::TestWithParam<mutualInfoParam>::GetParam();
-
-    nElements       = params.nElements;
-    lowerLabelRange = params.lowerLabelRange;
-    upperLabelRange = params.upperLabelRange;
-
-    // generating random value test input
-    std::vector<int> arr1(nElements, 0);
-    std::vector<int> arr2(nElements, 0);
-    std::random_device rd;
-    std::default_random_engine dre(rd());
-    std::uniform_int_distribution<int> intGenerator(lowerLabelRange, upperLabelRange);
-
-    std::generate(arr1.begin(), arr1.end(), [&]() { return intGenerator(dre); });
-    if (params.sameArrays) {
-      arr2 = arr1;
-    } else {
-      std::generate(arr2.begin(), arr2.end(), [&]() { return intGenerator(dre); });
-    }
-
-    // generating the golden output
-    // calculating the contingency matrix
-    int numUniqueClasses = upperLabelRange - lowerLabelRange + 1;
-    size_t sizeOfMat     = numUniqueClasses * numUniqueClasses * sizeof(int);
-    int* hGoldenOutput   = (int*)malloc(sizeOfMat);
-    memset(hGoldenOutput, 0, sizeOfMat);
-    int i, j;
-    for (i = 0; i < nElements; i++) {
-      int row    = arr1[i] - lowerLabelRange;
-      int column = arr2[i] - lowerLabelRange;
-
-      hGoldenOutput[row * numUniqueClasses + column] += 1;
-    }
-
-    int* a = (int*)malloc(numUniqueClasses * sizeof(int));
-    int* b = (int*)malloc(numUniqueClasses * sizeof(int));
-    memset(a, 0, numUniqueClasses * sizeof(int));
-    memset(b, 0, numUniqueClasses * sizeof(int));
-
-    // and also the reducing contingency matrix along row and column
-    for (i = 0; i < numUniqueClasses; ++i) {
-      for (j = 0; j < numUniqueClasses; ++j) {
-        a[i] += hGoldenOutput[i * numUniqueClasses + j];
-        b[i] += hGoldenOutput[j * numUniqueClasses + i];
-      }
-    }
-
-    // calculating the truth mutual information
-    for (int i = 0; i < numUniqueClasses; ++i) {
-      for (int j = 0; j < numUniqueClasses; ++j) {
-        if (a[i] * b[j] != 0 && hGoldenOutput[i * numUniqueClasses + j] != 0) {
-          truthmutualInfo +=
-            (double)(hGoldenOutput[i * numUniqueClasses + j]) *
-            (log((double)(double(nElements) * hGoldenOutput[i * numUniqueClasses + j])) -
-             log((double)(a[i] * b[j])));
-        }
-      }
-    }
-
-    truthmutualInfo /= nElements;
-
-    // allocating and initializing memory to the GPU
-    stream = resource::get_cuda_stream(handle);
-
-    rmm::device_uvector<T> firstClusterArray(nElements, stream);
-    rmm::device_uvector<T> secondClusterArray(nElements, stream);
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(firstClusterArray.data(), 0, firstClusterArray.size() * sizeof(T), stream));
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(secondClusterArray.data(), 0, secondClusterArray.size() * sizeof(T), stream));
-
-    raft::update_device(firstClusterArray.data(), &arr1[0], (int)nElements, stream);
-    raft::update_device(secondClusterArray.data(), &arr2[0], (int)nElements, stream);
-
-    // calling the mutualInfo CUDA implementation
-    computedmutualInfo = raft::stats::mutual_info_score(
-      handle,
-      raft::make_device_vector_view<const T>(firstClusterArray.data(), nElements),
-      raft::make_device_vector_view<const T>(secondClusterArray.data(), nElements),
-      lowerLabelRange,
-      upperLabelRange);
-  }
-
-  // declaring the data values
-  raft::resources handle;
-  mutualInfoParam params;
-  T lowerLabelRange, upperLabelRange;
-  int nElements             = 0;
-  double truthmutualInfo    = 0;
-  double computedmutualInfo = 0;
-  cudaStream_t stream       = 0;
-};
-
-// setting test parameter values
-const std::vector<mutualInfoParam> inputs = {{199, 1, 10, false, 0.000001},
-                                             {200, 15, 100, false, 0.000001},
-                                             {100, 1, 20, false, 0.000001},
-                                             {10, 1, 10, false, 0.000001},
-                                             {198, 1, 100, false, 0.000001},
-                                             {300, 3, 99, false, 0.000001},
-                                             {199, 1, 10, true, 0.000001},
-                                             {200, 15, 100, true, 0.000001},
-                                             {100, 1, 20, true, 0.000001},
-                                             {10, 1, 10, true, 0.000001},
-                                             {198, 1, 100, true, 0.000001},
-                                             {300, 3, 99, true, 0.000001}};
-
-// writing the test suite
-typedef mutualInfoTest<int> mutualInfoTestClass;
-TEST_P(mutualInfoTestClass, Result)
-{
-  ASSERT_NEAR(computedmutualInfo, truthmutualInfo, params.tolerance);
-}
-INSTANTIATE_TEST_CASE_P(mutualInfo, mutualInfoTestClass, ::testing::ValuesIn(inputs));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/neighborhood_recall.cu b/cpp/test/stats/neighborhood_recall.cu
deleted file mode 100644
index a911e2f33..000000000
--- a/cpp/test/stats/neighborhood_recall.cu
+++ /dev/null
@@ -1,178 +0,0 @@
-/*
- * Copyright (c) 2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../neighbors/ann_utils.cuh"
-#include "../test_utils.h"
-
-#include <raft/core/host_mdarray.hpp>
-#include <raft/core/mdspan.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/core/resources.hpp>
-
-#include <cuvs_internal/neighbors/naive_knn.cuh>
-
-#include <raft/stats/neighborhood_recall.cuh>
-#include <raft/util/itertools.hpp>
-
-#include <gtest/gtest.h>
-
-namespace raft::stats {
-
-struct NeighborhoodRecallInputs {
-  int n_rows;
-  int n_cols;
-  int k;
-};
-
-template <typename DistanceT, typename IdxT>
-class NeighborhoodRecallTest : public ::testing::TestWithParam<NeighborhoodRecallInputs> {
- public:
-  NeighborhoodRecallTest()
-    : ps{::testing::TestWithParam<NeighborhoodRecallInputs>::GetParam()},
-      data_1{raft::make_device_matrix<DistanceT, IdxT>(res, ps.n_rows, ps.n_cols)},
-      data_2{raft::make_device_matrix<DistanceT, IdxT>(res, ps.n_rows, ps.n_cols)}
-  {
-  }
-
- protected:
-  void test_recall()
-  {
-    size_t queries_size = ps.n_rows * ps.k;
-
-    // calculate nn for dataset 1
-    auto distances_1 = raft::make_device_matrix<DistanceT, IdxT>(res, ps.n_rows, ps.k);
-    auto indices_1   = raft::make_device_matrix<IdxT, IdxT>(res, ps.n_rows, ps.k);
-    cuvs::neighbors::naive_knn<DistanceT, DistanceT, IdxT>(
-      res,
-      distances_1.data_handle(),
-      indices_1.data_handle(),
-      data_1.data_handle(),
-      data_1.data_handle(),
-      ps.n_rows,
-      ps.n_rows,
-      ps.n_cols,
-      ps.k,
-      cuvs::distance::DistanceType::L2Expanded);
-    std::vector<DistanceT> distances_1_h(queries_size);
-    std::vector<IdxT> indices_1_h(queries_size);
-    raft::copy(distances_1_h.data(),
-               distances_1.data_handle(),
-               ps.n_rows * ps.k,
-               raft::resource::get_cuda_stream(res));
-    raft::copy(indices_1_h.data(),
-               indices_1.data_handle(),
-               ps.n_rows * ps.k,
-               raft::resource::get_cuda_stream(res));
-
-    // calculate nn for dataset 2
-    auto distances_2 = raft::make_device_matrix<DistanceT, IdxT>(res, ps.n_rows, ps.k);
-    auto indices_2   = raft::make_device_matrix<IdxT, IdxT>(res, ps.n_rows, ps.k);
-    cuvs::neighbors::naive_knn<DistanceT, DistanceT, IdxT>(
-      res,
-      distances_2.data_handle(),
-      indices_2.data_handle(),
-      data_2.data_handle(),
-      data_2.data_handle(),
-      ps.n_rows,
-      ps.n_rows,
-      ps.n_cols,
-      ps.k,
-      cuvs::distance::DistanceType::L2Expanded);
-    std::vector<DistanceT> distances_2_h(queries_size);
-    std::vector<IdxT> indices_2_h(queries_size);
-    raft::copy(distances_2_h.data(),
-               distances_2.data_handle(),
-               ps.n_rows * ps.k,
-               raft::resource::get_cuda_stream(res));
-    raft::copy(indices_2_h.data(),
-               indices_2.data_handle(),
-               ps.n_rows * ps.k,
-               raft::resource::get_cuda_stream(res));
-
-    raft::resource::sync_stream(res);
-
-    // find CPU recall scores
-    [[maybe_unused]] auto [indices_only_recall_h, mc1, tc1] =
-      cuvs::neighbors::calc_recall(indices_1_h, indices_2_h, ps.n_rows, ps.k);
-    [[maybe_unused]] auto [recall_h, mc2, tc2] = cuvs::neighbors::calc_recall(
-      indices_1_h, indices_2_h, distances_1_h, distances_2_h, ps.n_rows, ps.k, 0.001);
-
-    // find GPU recall scores
-    auto s1                         = 0;
-    auto indices_only_recall_scalar = raft::make_host_scalar<double>(s1);
-    neighborhood_recall(res,
-                        raft::make_const_mdspan(indices_1.view()),
-                        raft::make_const_mdspan(indices_2.view()),
-                        indices_only_recall_scalar.view());
-
-    auto s2            = 0;
-    auto recall_scalar = raft::make_host_scalar<double>(s2);
-    DistanceT s3       = 0.001;
-    auto eps_mda       = raft::make_host_scalar<DistanceT>(s3);
-
-    neighborhood_recall<IdxT, IdxT, double, DistanceT>(res,
-                                                       raft::make_const_mdspan(indices_1.view()),
-                                                       raft::make_const_mdspan(indices_2.view()),
-                                                       recall_scalar.view(),
-                                                       raft::make_const_mdspan(distances_1.view()),
-                                                       raft::make_const_mdspan(distances_2.view()));
-
-    // assert correctness
-    ASSERT_TRUE(raft::match(indices_only_recall_h,
-                            *indices_only_recall_scalar.data_handle(),
-                            raft::CompareApprox<double>(0.01)));
-    ASSERT_TRUE(
-      raft::match(recall_h, *recall_scalar.data_handle(), raft::CompareApprox<double>(0.01)));
-  }
-
-  void SetUp() override
-  {
-    // form two random datasets
-    raft::random::Rng r1(1234ULL);
-    r1.normal(data_1.data_handle(),
-              ps.n_rows * ps.n_cols,
-              DistanceT(0.1),
-              DistanceT(2.0),
-              raft::resource::get_cuda_stream(res));
-    raft::random::Rng r2(21111ULL);
-    r2.normal(data_2.data_handle(),
-              ps.n_rows * ps.n_cols,
-              DistanceT(0.1),
-              DistanceT(2.0),
-              raft::resource::get_cuda_stream(res));
-    resource::sync_stream(res);
-  }
-
- private:
-  raft::resources res;
-  NeighborhoodRecallInputs ps;
-  raft::device_matrix<DistanceT, IdxT> data_1;
-  raft::device_matrix<DistanceT, IdxT> data_2;
-};
-
-const std::vector<NeighborhoodRecallInputs> inputs =
-  raft::util::itertools::product<NeighborhoodRecallInputs>({10, 50, 100},  // n_rows
-                                                           {80, 100},      // n_cols
-                                                           {32, 64});      // k
-
-using NeighborhoodRecallTestF_U32 = NeighborhoodRecallTest<float, std::uint32_t>;
-TEST_P(NeighborhoodRecallTestF_U32, AnnCagra) { this->test_recall(); }
-
-INSTANTIATE_TEST_CASE_P(NeighborhoodRecallTest,
-                        NeighborhoodRecallTestF_U32,
-                        ::testing::ValuesIn(inputs));
-
-}  // end namespace raft::stats
diff --git a/cpp/test/stats/r2_score.cu b/cpp/test/stats/r2_score.cu
deleted file mode 100644
index aa4f069f0..000000000
--- a/cpp/test/stats/r2_score.cu
+++ /dev/null
@@ -1,114 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <optional>
-#include <raft/core/interruptible.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/random/rng.cuh>
-#include <raft/stats/r2_score.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <rmm/device_uvector.hpp>
-#include <stdio.h>
-#include <stdlib.h>
-#include <vector>
-
-namespace raft {
-namespace stats {
-
-template <typename T>
-struct R2_scoreInputs {
-  T tolerance;
-  int nrows;
-  unsigned long long int seed;
-};
-
-template <typename T>
-::std::ostream& operator<<(::std::ostream& os, const R2_scoreInputs<T>& dims)
-{
-  return os;
-}
-
-template <typename T>
-class R2_scoreTest : public ::testing::TestWithParam<R2_scoreInputs<T>> {
- protected:
-  R2_scoreTest() : stream(resource::get_cuda_stream(handle)) {}
-
-  void SetUp() override
-  {
-    params = ::testing::TestWithParam<R2_scoreInputs<T>>::GetParam();
-    raft::random::RngState r(params.seed);
-    rmm::device_uvector<T> y(params.nrows, stream);
-    rmm::device_uvector<T> y_hat(params.nrows, stream);
-    uniform(handle, r, y.data(), params.nrows, (T)-1.0, (T)1.0);
-    uniform(handle, r, y_hat.data(), params.nrows, (T)-1.0, (T)1.0);
-
-    actualVal   = r2_score(handle,
-                         raft::make_device_vector_view<const T>(y.data(), params.nrows),
-                         raft::make_device_vector_view<const T>(y_hat.data(), params.nrows));
-    expectedVal = T(0);
-    std::vector<T> h_y(params.nrows, 0);
-    std::vector<T> h_y_hat(params.nrows, 0);
-    raft::update_host(h_y.data(), y.data(), params.nrows, stream);
-    raft::update_host(h_y_hat.data(), y_hat.data(), params.nrows, stream);
-    T mean = T(0);
-    for (int i = 0; i < params.nrows; ++i) {
-      mean += h_y[i];
-    }
-    mean /= params.nrows;
-
-    std::vector<T> sse_arr(params.nrows, 0);
-    std::vector<T> ssto_arr(params.nrows, 0);
-    T sse  = T(0);
-    T ssto = T(0);
-    for (int i = 0; i < params.nrows; ++i) {
-      sse += (h_y[i] - h_y_hat[i]) * (h_y[i] - h_y_hat[i]);
-      ssto += (h_y[i] - mean) * (h_y[i] - mean);
-    }
-    expectedVal = 1.0 - sse / ssto;
-    raft::interruptible::synchronize(stream);
-  }
-
- protected:
-  R2_scoreInputs<T> params;
-  raft::resources handle;
-  cudaStream_t stream = 0;
-  T expectedVal, actualVal;
-};
-
-const std::vector<R2_scoreInputs<float>> inputsf = {
-  {0.001f, 30, 1234ULL}, {0.001f, 100, 1234ULL}, {0.001f, 1000, 1234ULL}};
-typedef R2_scoreTest<float> R2_scoreTestF;
-TEST_P(R2_scoreTestF, Result)
-{
-  auto eq = raft::CompareApprox<float>(params.tolerance);
-  ASSERT_TRUE(match(expectedVal, actualVal, eq));
-}
-INSTANTIATE_TEST_CASE_P(R2_scoreTests, R2_scoreTestF, ::testing::ValuesIn(inputsf));
-
-const std::vector<R2_scoreInputs<double>> inputsd = {
-  {0.001, 30, 1234ULL}, {0.001, 100, 1234ULL}, {0.001, 1000, 1234ULL}};
-typedef R2_scoreTest<double> R2_scoreTestD;
-TEST_P(R2_scoreTestD, Result)
-{
-  auto eq = raft::CompareApprox<double>(params.tolerance);
-  ASSERT_TRUE(match(expectedVal, actualVal, eq));
-}
-INSTANTIATE_TEST_CASE_P(R2_scoreTests, R2_scoreTestD, ::testing::ValuesIn(inputsd));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/rand_index.cu b/cpp/test/stats/rand_index.cu
deleted file mode 100644
index 41b0823e7..000000000
--- a/cpp/test/stats/rand_index.cu
+++ /dev/null
@@ -1,129 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <raft/util/cudart_utils.hpp>
-
-#include <gtest/gtest.h>
-
-#include <algorithm>
-#include <iostream>
-#include <raft/core/resources.hpp>
-#include <raft/stats/rand_index.cuh>
-#include <random>
-
-namespace raft {
-namespace stats {
-
-// parameter structure definition
-struct randIndexParam {
-  uint64_t nElements;
-  int lowerLabelRange;
-  int upperLabelRange;
-  double tolerance;
-};
-
-// test fixture class
-template <typename T>
-class randIndexTest : public ::testing::TestWithParam<randIndexParam> {
- protected:
-  // the constructor
-  void SetUp() override
-  {
-    // getting the parameters
-    params = ::testing::TestWithParam<randIndexParam>::GetParam();
-
-    size            = params.nElements;
-    lowerLabelRange = params.lowerLabelRange;
-    upperLabelRange = params.upperLabelRange;
-
-    // generating random value test input
-    std::vector<int> arr1(size, 0);
-    std::vector<int> arr2(size, 0);
-    std::random_device rd;
-    std::default_random_engine dre(rd());
-    std::uniform_int_distribution<int> intGenerator(lowerLabelRange, upperLabelRange);
-
-    std::generate(arr1.begin(), arr1.end(), [&]() { return intGenerator(dre); });
-    std::generate(arr2.begin(), arr2.end(), [&]() { return intGenerator(dre); });
-
-    // generating the golden output
-    int64_t a_truth = 0;
-    int64_t b_truth = 0;
-
-    for (uint64_t iter = 0; iter < size; ++iter) {
-      for (uint64_t jiter = 0; jiter < iter; ++jiter) {
-        if (arr1[iter] == arr1[jiter] && arr2[iter] == arr2[jiter]) {
-          ++a_truth;
-        } else if (arr1[iter] != arr1[jiter] && arr2[iter] != arr2[jiter]) {
-          ++b_truth;
-        }
-      }
-    }
-    uint64_t nChooseTwo = (size * (size - 1)) / 2;
-    truthRandIndex      = (double)(((double)(a_truth + b_truth)) / (double)nChooseTwo);
-
-    // allocating and initializing memory to the GPU
-    stream = resource::get_cuda_stream(handle);
-
-    rmm::device_uvector<T> firstClusterArray(size, stream);
-    rmm::device_uvector<T> secondClusterArray(size, stream);
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(firstClusterArray.data(), 0, firstClusterArray.size() * sizeof(T), stream));
-    RAFT_CUDA_TRY(
-      cudaMemsetAsync(secondClusterArray.data(), 0, secondClusterArray.size() * sizeof(T), stream));
-
-    raft::update_device(firstClusterArray.data(), &arr1[0], (int)size, stream);
-    raft::update_device(secondClusterArray.data(), &arr2[0], (int)size, stream);
-
-    // calling the rand_index CUDA implementation
-    computedRandIndex = raft::stats::rand_index(
-      handle,
-      raft::make_device_vector_view<const T>(firstClusterArray.data(), size),
-      raft::make_device_vector_view<const T>(secondClusterArray.data(), size));
-  }
-
-  // declaring the data values
-  raft::resources handle;
-  randIndexParam params;
-  int lowerLabelRange = 0, upperLabelRange = 2;
-  uint64_t size            = 0;
-  double truthRandIndex    = 0;
-  double computedRandIndex = 0;
-  cudaStream_t stream      = 0;
-};
-
-// setting test parameter values
-const std::vector<randIndexParam> inputs = {{199, 1, 10, 0.000001},
-                                            {200, 1, 100, 0.000001},
-                                            {10, 1, 1200, 0.000001},
-                                            {100, 1, 10000, 0.000001},
-                                            {198, 1, 100, 0.000001},
-                                            {300, 3, 99, 0.000001},
-                                            {2, 0, 0, 0.00001}};
-
-// writing the test suite
-typedef randIndexTest<int> randIndexTestClass;
-TEST_P(randIndexTestClass, Result)
-{
-  ASSERT_NEAR(computedRandIndex, truthRandIndex, params.tolerance);
-}
-INSTANTIATE_TEST_CASE_P(randIndex, randIndexTestClass, ::testing::ValuesIn(inputs));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/regression_metrics.cu b/cpp/test/stats/regression_metrics.cu
deleted file mode 100644
index b0c4cca53..000000000
--- a/cpp/test/stats/regression_metrics.cu
+++ /dev/null
@@ -1,146 +0,0 @@
-/*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <algorithm>
-#include <gtest/gtest.h>
-#include <optional>
-#include <raft/core/interruptible.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/random/rng.cuh>
-#include <raft/stats/regression_metrics.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <rmm/device_uvector.hpp>
-#include <stdio.h>
-#include <stdlib.h>
-#include <vector>
-
-namespace raft {
-namespace stats {
-
-template <typename T>
-struct RegressionInputs {
-  T tolerance;
-  int len;
-  unsigned long long int seed;
-};
-
-template <typename T>
-::std::ostream& operator<<(::std::ostream& os, const RegressionInputs<T>& dims)
-{
-  return os;
-}
-
-template <typename T>
-void naive_reg_metrics(std::vector<T>& predictions,
-                       std::vector<T>& ref_predictions,
-                       double& mean_abs_error,
-                       double& mean_squared_error,
-                       double& median_abs_error)
-{
-  auto len        = predictions.size();
-  double abs_diff = 0;
-  double sq_diff  = 0;
-  std::vector<double> abs_errors(len);
-  for (std::size_t i = 0; i < len; ++i) {
-    auto diff = predictions[i] - ref_predictions[i];
-    abs_diff += abs(diff);
-    sq_diff += diff * diff;
-    abs_errors[i] = abs(diff);
-  }
-  mean_abs_error     = abs_diff / len;
-  mean_squared_error = sq_diff / len;
-
-  std::sort(abs_errors.begin(), abs_errors.end());
-  auto middle = len / 2;
-  if (len % 2 == 1) {
-    median_abs_error = abs_errors[middle];
-  } else {
-    median_abs_error = (abs_errors[middle] + abs_errors[middle - 1]) / 2;
-  }
-}
-
-template <typename T>
-class RegressionTest : public ::testing::TestWithParam<RegressionInputs<T>> {
- protected:
-  RegressionTest() : stream(resource::get_cuda_stream(handle)) {}
-
-  void SetUp() override
-  {
-    params = ::testing::TestWithParam<RegressionInputs<T>>::GetParam();
-    raft::random::RngState r(params.seed);
-    rmm::device_uvector<T> predictions(params.len, stream);
-    rmm::device_uvector<T> ref_predictions(params.len, stream);
-    uniform(handle, r, predictions.data(), params.len, T(-10.0), T(10.0));
-    uniform(handle, r, ref_predictions.data(), params.len, T(-10.0), T(10.0));
-
-    regression_metrics(handle,
-                       raft::make_device_vector_view<const T>(predictions.data(), params.len),
-                       raft::make_device_vector_view<const T>(ref_predictions.data(), params.len),
-                       raft::make_host_scalar_view(&mean_abs_error),
-                       raft::make_host_scalar_view(&mean_squared_error),
-                       raft::make_host_scalar_view(&median_abs_error));
-    std::vector<T> h_predictions(params.len, 0);
-    std::vector<T> h_ref_predictions(params.len, 0);
-    raft::update_host(h_predictions.data(), predictions.data(), params.len, stream);
-    raft::update_host(h_ref_predictions.data(), ref_predictions.data(), params.len, stream);
-
-    naive_reg_metrics(h_predictions,
-                      h_ref_predictions,
-                      ref_mean_abs_error,
-                      ref_mean_squared_error,
-                      ref_median_abs_error);
-    raft::interruptible::synchronize(stream);
-  }
-
- protected:
-  raft::resources handle;
-  RegressionInputs<T> params;
-  cudaStream_t stream           = 0;
-  double mean_abs_error         = 0;
-  double mean_squared_error     = 0;
-  double median_abs_error       = 0;
-  double ref_mean_abs_error     = 0;
-  double ref_mean_squared_error = 0;
-  double ref_median_abs_error   = 0;
-};
-
-const std::vector<RegressionInputs<float>> inputsf = {
-  {0.001f, 30, 1234ULL}, {0.001f, 100, 1234ULL}, {0.001f, 4000, 1234ULL}};
-typedef RegressionTest<float> RegressionTestF;
-TEST_P(RegressionTestF, Result)
-{
-  auto eq = raft::CompareApprox<float>(params.tolerance);
-  ASSERT_TRUE(match(ref_mean_abs_error, mean_abs_error, eq));
-  ASSERT_TRUE(match(ref_mean_squared_error, mean_squared_error, eq));
-  ASSERT_TRUE(match(ref_median_abs_error, median_abs_error, eq));
-}
-INSTANTIATE_TEST_CASE_P(RegressionTests, RegressionTestF, ::testing::ValuesIn(inputsf));
-
-const std::vector<RegressionInputs<double>> inputsd = {
-  {0.001, 30, 1234ULL}, {0.001, 100, 1234ULL}, {0.001, 4000, 1234ULL}};
-typedef RegressionTest<double> RegressionTestD;
-TEST_P(RegressionTestD, Result)
-{
-  auto eq = raft::CompareApprox<double>(params.tolerance);
-  ASSERT_TRUE(match(ref_mean_abs_error, mean_abs_error, eq));
-  ASSERT_TRUE(match(ref_mean_squared_error, mean_squared_error, eq));
-  ASSERT_TRUE(match(ref_median_abs_error, median_abs_error, eq));
-}
-INSTANTIATE_TEST_CASE_P(RegressionTests, RegressionTestD, ::testing::ValuesIn(inputsd));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/silhouette_score.cu b/cpp/test/stats/silhouette_score.cu
deleted file mode 100644
index 7c99dbf4b..000000000
--- a/cpp/test/stats/silhouette_score.cu
+++ /dev/null
@@ -1,227 +0,0 @@
-/*
- * Copyright (c) 2021-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "../test_utils.cuh"
-#include <algorithm>
-#include <cuvs/distance/distance_types.hpp>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/util/cudart_utils.hpp>
-
-#include <raft/stats/silhouette_score.cuh>
-#include <random>
-#include <rmm/device_uvector.hpp>
-
-namespace raft {
-namespace stats {
-
-// parameter structure definition
-struct silhouetteScoreParam {
-  int nRows;
-  int nCols;
-  int nLabels;
-  cuvs::distance::DistanceType metric;
-  int chunk;
-  double tolerance;
-};
-
-// test fixture class
-template <typename LabelT, typename DataT>
-class silhouetteScoreTest : public ::testing::TestWithParam<silhouetteScoreParam> {
- protected:
-  silhouetteScoreTest()
-    : d_X(0, resource::get_cuda_stream(handle)),
-      sampleSilScore(0, resource::get_cuda_stream(handle)),
-      d_labels(0, resource::get_cuda_stream(handle))
-  {
-  }
-
-  void host_silhouette_score()
-  {
-    // generating random value test input
-    std::vector<double> h_X(nElements, 0.0);
-    std::vector<int> h_labels(nRows, 0);
-    std::random_device rd;
-    std::default_random_engine dre(nElements * nLabels);
-    std::uniform_int_distribution<int> intGenerator(0, nLabels - 1);
-    std::uniform_real_distribution<double> realGenerator(0, 100);
-
-    std::generate(h_X.begin(), h_X.end(), [&]() { return realGenerator(dre); });
-    std::generate(h_labels.begin(), h_labels.end(), [&]() { return intGenerator(dre); });
-
-    // allocating and initializing memory to the GPU
-    auto stream = resource::get_cuda_stream(handle);
-    d_X.resize(nElements, stream);
-    d_labels.resize(nElements, stream);
-    RAFT_CUDA_TRY(cudaMemsetAsync(d_X.data(), 0, d_X.size() * sizeof(DataT), stream));
-    RAFT_CUDA_TRY(cudaMemsetAsync(d_labels.data(), 0, d_labels.size() * sizeof(LabelT), stream));
-    sampleSilScore.resize(nElements, stream);
-
-    raft::update_device(d_X.data(), &h_X[0], (int)nElements, stream);
-    raft::update_device(d_labels.data(), &h_labels[0], (int)nElements, stream);
-
-    // finding the distance matrix
-
-    rmm::device_uvector<double> d_distanceMatrix(nRows * nRows, stream);
-    double* h_distanceMatrix = (double*)malloc(nRows * nRows * sizeof(double*));
-
-    cuvs::distance::pairwise_distance(
-      handle, d_X.data(), d_X.data(), d_distanceMatrix.data(), nRows, nRows, nCols, params.metric);
-
-    resource::sync_stream(handle, stream);
-
-    raft::update_host(h_distanceMatrix, d_distanceMatrix.data(), nRows * nRows, stream);
-
-    // finding the bincount array
-
-    double* binCountArray = (double*)malloc(nLabels * sizeof(double*));
-    memset(binCountArray, 0, nLabels * sizeof(double));
-
-    for (int i = 0; i < nRows; ++i) {
-      binCountArray[h_labels[i]] += 1;
-    }
-
-    // finding the average intra cluster distance for every element
-
-    double* a = (double*)malloc(nRows * sizeof(double*));
-
-    for (int i = 0; i < nRows; ++i) {
-      int myLabel               = h_labels[i];
-      double sumOfIntraClusterD = 0;
-
-      for (int j = 0; j < nRows; ++j) {
-        if (h_labels[j] == myLabel) { sumOfIntraClusterD += h_distanceMatrix[i * nRows + j]; }
-      }
-
-      if (binCountArray[myLabel] <= 1)
-        a[i] = -1;
-      else
-        a[i] = sumOfIntraClusterD / (binCountArray[myLabel] - 1);
-    }
-
-    // finding the average inter cluster distance for every element
-
-    double* b = (double*)malloc(nRows * sizeof(double*));
-
-    for (int i = 0; i < nRows; ++i) {
-      int myLabel          = h_labels[i];
-      double minAvgInterCD = ULLONG_MAX;
-
-      for (int j = 0; j < nLabels; ++j) {
-        int curClLabel = j;
-        if (curClLabel == myLabel) continue;
-        double avgInterCD = 0;
-
-        for (int k = 0; k < nRows; ++k) {
-          if (h_labels[k] == curClLabel) { avgInterCD += h_distanceMatrix[i * nRows + k]; }
-        }
-
-        if (binCountArray[curClLabel])
-          avgInterCD /= binCountArray[curClLabel];
-        else
-          avgInterCD = ULLONG_MAX;
-        minAvgInterCD = min(minAvgInterCD, avgInterCD);
-      }
-
-      b[i] = minAvgInterCD;
-    }
-
-    // finding the silhouette score for every element
-
-    double* truthSampleSilScore = (double*)malloc(nRows * sizeof(double*));
-    for (int i = 0; i < nRows; ++i) {
-      if (a[i] == -1)
-        truthSampleSilScore[i] = 0;
-      else if (a[i] == 0 && b[i] == 0)
-        truthSampleSilScore[i] = 0;
-      else
-        truthSampleSilScore[i] = (b[i] - a[i]) / max(a[i], b[i]);
-      truthSilhouetteScore += truthSampleSilScore[i];
-    }
-
-    truthSilhouetteScore /= nRows;
-  }
-
-  // the constructor
-  void SetUp() override
-  {
-    // getting the parameters
-    params = ::testing::TestWithParam<silhouetteScoreParam>::GetParam();
-
-    nRows     = params.nRows;
-    nCols     = params.nCols;
-    nLabels   = params.nLabels;
-    chunk     = params.chunk;
-    nElements = nRows * nCols;
-
-    host_silhouette_score();
-
-    // calling the silhouette_score CUDA implementation
-    computedSilhouetteScore = raft::stats::silhouette_score(
-      handle,
-      raft::make_device_matrix_view<const DataT>(d_X.data(), nRows, nCols),
-      raft::make_device_vector_view<const LabelT>(d_labels.data(), nRows),
-      std::make_optional(raft::make_device_vector_view(sampleSilScore.data(), nRows)),
-      nLabels,
-      params.metric);
-
-    batchedSilhouetteScore = raft::stats::silhouette_score_batched(
-      handle,
-      raft::make_device_matrix_view<const DataT>(d_X.data(), nRows, nCols),
-      raft::make_device_vector_view<const LabelT>(d_labels.data(), nRows),
-      std::make_optional(raft::make_device_vector_view(sampleSilScore.data(), nRows)),
-      nLabels,
-      chunk,
-      params.metric);
-  }
-
-  // declaring the data values
-  raft::resources handle;
-  silhouetteScoreParam params;
-  int nLabels;
-  rmm::device_uvector<DataT> d_X;
-  rmm::device_uvector<DataT> sampleSilScore;
-  rmm::device_uvector<LabelT> d_labels;
-  int nRows;
-  int nCols;
-  int nElements;
-  double truthSilhouetteScore    = 0;
-  double computedSilhouetteScore = 0;
-  double batchedSilhouetteScore  = 0;
-  int chunk;
-};
-
-// setting test parameter values
-const std::vector<silhouetteScoreParam> inputs = {
-  {4, 2, 3, cuvs::distance::DistanceType::L2Expanded, 4, 0.00001},
-  {4, 2, 2, cuvs::distance::DistanceType::L2SqrtUnexpanded, 2, 0.00001},
-  {8, 8, 3, cuvs::distance::DistanceType::L2Unexpanded, 4, 0.00001},
-  {11, 2, 5, cuvs::distance::DistanceType::L2Expanded, 3, 0.00001},
-  {40, 2, 8, cuvs::distance::DistanceType::L2Expanded, 10, 0.00001},
-  {12, 7, 3, cuvs::distance::DistanceType::CosineExpanded, 8, 0.00001},
-  {7, 5, 5, cuvs::distance::DistanceType::L1, 2, 0.00001}};
-
-// writing the test suite
-typedef silhouetteScoreTest<int, double> silhouetteScoreTestClass;
-TEST_P(silhouetteScoreTestClass, Result)
-{
-  ASSERT_NEAR(computedSilhouetteScore, truthSilhouetteScore, params.tolerance);
-  ASSERT_NEAR(batchedSilhouetteScore, truthSilhouetteScore, params.tolerance);
-}
-INSTANTIATE_TEST_CASE_P(silhouetteScore, silhouetteScoreTestClass, ::testing::ValuesIn(inputs));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/stddev.cu b/cpp/test/stats/stddev.cu
deleted file mode 100644
index 998f7a88e..000000000
--- a/cpp/test/stats/stddev.cu
+++ /dev/null
@@ -1,196 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <gtest/gtest.h>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/matrix/math.cuh>
-#include <raft/random/rng.cuh>
-#include <raft/stats/mean.cuh>
-#include <raft/stats/stddev.cuh>
-#include <raft/util/cudart_utils.hpp>
-
-namespace raft {
-namespace stats {
-
-template <typename T>
-struct StdDevInputs {
-  T tolerance, mean, stddev;
-  int rows, cols;
-  bool sample, rowMajor;
-  unsigned long long int seed;
-};
-
-template <typename T>
-::std::ostream& operator<<(::std::ostream& os, const StdDevInputs<T>& dims)
-{
-  return os;
-}
-
-template <typename T>
-class StdDevTest : public ::testing::TestWithParam<StdDevInputs<T>> {
- public:
-  StdDevTest()
-    : params(::testing::TestWithParam<StdDevInputs<T>>::GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      rows(params.rows),
-      cols(params.cols),
-      data(rows * cols, stream),
-      mean_act(cols, stream),
-      stddev_act(cols, stream),
-      vars_act(cols, stream)
-  {
-  }
-
- protected:
-  void SetUp() override
-  {
-    random::RngState r(params.seed);
-    int len = rows * cols;
-
-    data.resize(len, stream);
-    mean_act.resize(cols, stream);
-    stddev_act.resize(cols, stream);
-    vars_act.resize(cols, stream);
-    normal(handle, r, data.data(), len, params.mean, params.stddev);
-    stdVarSGtest(data.data(), stream);
-    resource::sync_stream(handle, stream);
-  }
-
-  void stdVarSGtest(T* data, cudaStream_t stream)
-  {
-    int rows = params.rows, cols = params.cols;
-
-    if (params.rowMajor) {
-      using layout_t = raft::row_major;
-      mean(handle,
-           raft::make_device_matrix_view<const T, int, layout_t>(data, rows, cols),
-           raft::make_device_vector_view<T, int>(mean_act.data(), cols),
-           params.sample);
-
-      stddev(handle,
-             raft::make_device_matrix_view<const T, int, layout_t>(data, rows, cols),
-             raft::make_device_vector_view<const T, int>(mean_act.data(), cols),
-             raft::make_device_vector_view<T, int>(stddev_act.data(), cols),
-             params.sample);
-
-      vars(handle,
-           raft::make_device_matrix_view<const T, int, layout_t>(data, rows, cols),
-           raft::make_device_vector_view<const T, int>(mean_act.data(), cols),
-           raft::make_device_vector_view<T, int>(vars_act.data(), cols),
-           params.sample);
-    } else {
-      using layout_t = raft::col_major;
-      mean(handle,
-           raft::make_device_matrix_view<const T, int, layout_t>(data, rows, cols),
-           raft::make_device_vector_view<T>(mean_act.data(), cols),
-           params.sample);
-
-      stddev(handle,
-             raft::make_device_matrix_view<const T, int, layout_t>(data, rows, cols),
-             raft::make_device_vector_view<const T, int>(mean_act.data(), cols),
-             raft::make_device_vector_view<T, int>(stddev_act.data(), cols),
-             params.sample);
-
-      vars(handle,
-           raft::make_device_matrix_view<const T, int, layout_t>(data, rows, cols),
-           raft::make_device_vector_view<const T, int>(mean_act.data(), cols),
-           raft::make_device_vector_view<T, int>(vars_act.data(), cols),
-           params.sample);
-    }
-    raft::matrix::seqRoot(vars_act.data(), T(1), cols, stream);
-  }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream;
-
-  StdDevInputs<T> params;
-  int rows, cols;
-  rmm::device_uvector<T> data, mean_act, stddev_act, vars_act;
-};
-
-const std::vector<StdDevInputs<float>> inputsf = {
-  {0.1f, 1.f, 2.f, 1024, 32, true, false, 1234ULL},
-  {0.1f, 1.f, 2.f, 1024, 64, true, false, 1234ULL},
-  {0.1f, 1.f, 2.f, 1024, 128, true, false, 1234ULL},
-  {0.1f, 1.f, 2.f, 1024, 256, true, false, 1234ULL},
-  {0.1f, -1.f, 2.f, 1024, 32, false, false, 1234ULL},
-  {0.1f, -1.f, 2.f, 1024, 64, false, false, 1234ULL},
-  {0.1f, -1.f, 2.f, 1024, 128, false, false, 1234ULL},
-  {0.1f, -1.f, 2.f, 1024, 256, false, false, 1234ULL},
-  {0.1f, 1.f, 2.f, 1024, 32, true, true, 1234ULL},
-  {0.1f, 1.f, 2.f, 1024, 64, true, true, 1234ULL},
-  {0.1f, 1.f, 2.f, 1024, 128, true, true, 1234ULL},
-  {0.1f, 1.f, 2.f, 1024, 256, true, true, 1234ULL},
-  {0.1f, -1.f, 2.f, 1024, 32, false, true, 1234ULL},
-  {0.1f, -1.f, 2.f, 1024, 64, false, true, 1234ULL},
-  {0.1f, -1.f, 2.f, 1024, 128, false, true, 1234ULL},
-  {0.1f, -1.f, 2.f, 1024, 256, false, true, 1234ULL}};
-
-const std::vector<StdDevInputs<double>> inputsd = {
-  {0.1, 1.0, 2.0, 1024, 32, true, false, 1234ULL},
-  {0.1, 1.0, 2.0, 1024, 64, true, false, 1234ULL},
-  {0.1, 1.0, 2.0, 1024, 128, true, false, 1234ULL},
-  {0.1, 1.0, 2.0, 1024, 256, true, false, 1234ULL},
-  {0.1, -1.0, 2.0, 1024, 32, false, false, 1234ULL},
-  {0.1, -1.0, 2.0, 1024, 64, false, false, 1234ULL},
-  {0.1, -1.0, 2.0, 1024, 128, false, false, 1234ULL},
-  {0.1, -1.0, 2.0, 1024, 256, false, false, 1234ULL},
-  {0.1, 1.0, 2.0, 1024, 32, true, true, 1234ULL},
-  {0.1, 1.0, 2.0, 1024, 64, true, true, 1234ULL},
-  {0.1, 1.0, 2.0, 1024, 128, true, true, 1234ULL},
-  {0.1, 1.0, 2.0, 1024, 256, true, true, 1234ULL},
-  {0.1, -1.0, 2.0, 1024, 32, false, true, 1234ULL},
-  {0.1, -1.0, 2.0, 1024, 64, false, true, 1234ULL},
-  {0.1, -1.0, 2.0, 1024, 128, false, true, 1234ULL},
-  {0.1, -1.0, 2.0, 1024, 256, false, true, 1234ULL}};
-
-typedef StdDevTest<float> StdDevTestF;
-TEST_P(StdDevTestF, Result)
-{
-  ASSERT_TRUE(devArrMatch(
-    params.stddev, stddev_act.data(), params.cols, CompareApprox<float>(params.tolerance), stream));
-
-  ASSERT_TRUE(devArrMatch(stddev_act.data(),
-                          vars_act.data(),
-                          params.cols,
-                          CompareApprox<float>(params.tolerance),
-                          stream));
-}
-
-typedef StdDevTest<double> StdDevTestD;
-TEST_P(StdDevTestD, Result)
-{
-  ASSERT_TRUE(devArrMatch(params.stddev,
-                          stddev_act.data(),
-                          params.cols,
-                          CompareApprox<double>(params.tolerance),
-                          stream));
-
-  ASSERT_TRUE(devArrMatch(stddev_act.data(),
-                          vars_act.data(),
-                          params.cols,
-                          CompareApprox<double>(params.tolerance),
-                          stream));
-}
-
-INSTANTIATE_TEST_SUITE_P(StdDevTests, StdDevTestF, ::testing::ValuesIn(inputsf));
-
-INSTANTIATE_TEST_SUITE_P(StdDevTests, StdDevTestD, ::testing::ValuesIn(inputsd));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/sum.cu b/cpp/test/stats/sum.cu
deleted file mode 100644
index 040b662c4..000000000
--- a/cpp/test/stats/sum.cu
+++ /dev/null
@@ -1,111 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <raft/core/resource/cuda_stream.hpp>
-
-#include <raft/core/resources.hpp>
-#include <raft/linalg/eltwise.cuh>
-#include <raft/stats/sum.cuh>
-#include <raft/util/cudart_utils.hpp>
-
-#include <rmm/device_uvector.hpp>
-
-#include <gtest/gtest.h>
-
-namespace raft {
-namespace stats {
-
-template <typename T>
-struct SumInputs {
-  T tolerance;
-  int rows, cols;
-  unsigned long long int seed;
-};
-
-template <typename T>
-::std::ostream& operator<<(::std::ostream& os, const SumInputs<T>& dims)
-{
-  return os;
-}
-
-template <typename T>
-class SumTest : public ::testing::TestWithParam<SumInputs<T>> {
- public:
-  SumTest()
-    : params(::testing::TestWithParam<SumInputs<T>>::GetParam()),
-      stream(resource::get_cuda_stream(handle)),
-      rows(params.rows),
-      cols(params.cols),
-      data(rows * cols, stream),
-      sum_act(cols, stream)
-  {
-  }
-
- protected:
-  void SetUp() override
-  {
-    int len = rows * cols;
-
-    T data_h[len];
-    for (int i = 0; i < len; i++) {
-      data_h[i] = T(1);
-    }
-
-    raft::update_device(data.data(), data_h, len, stream);
-    sum(handle,
-        raft::make_device_matrix_view<const T>(data.data(), rows, cols),
-        raft::make_device_vector_view(sum_act.data(), cols));
-    resource::sync_stream(handle, stream);
-  }
-
- protected:
-  raft::resources handle;
-  cudaStream_t stream;
-
-  SumInputs<T> params;
-  int rows, cols;
-  rmm::device_uvector<T> data, sum_act;
-};
-
-const std::vector<SumInputs<float>> inputsf = {{0.05f, 1024, 32, 1234ULL},
-                                               {0.05f, 1024, 256, 1234ULL}};
-
-const std::vector<SumInputs<double>> inputsd = {{0.05, 1024, 32, 1234ULL},
-                                                {0.05, 1024, 256, 1234ULL}};
-
-typedef SumTest<float> SumTestF;
-TEST_P(SumTestF, Result)
-{
-  ASSERT_TRUE(raft::devArrMatch(
-    float(params.rows), sum_act.data(), params.cols, raft::CompareApprox<float>(params.tolerance)));
-}
-
-typedef SumTest<double> SumTestD;
-TEST_P(SumTestD, Result)
-{
-  ASSERT_TRUE(raft::devArrMatch(double(params.rows),
-                                sum_act.data(),
-                                params.cols,
-                                raft::CompareApprox<double>(params.tolerance)));
-}
-
-INSTANTIATE_TEST_CASE_P(SumTests, SumTestF, ::testing::ValuesIn(inputsf));
-
-INSTANTIATE_TEST_CASE_P(SumTests, SumTestD, ::testing::ValuesIn(inputsd));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/trustworthiness.cu b/cpp/test/stats/trustworthiness.cu
deleted file mode 100644
index b0a21691f..000000000
--- a/cpp/test/stats/trustworthiness.cu
+++ /dev/null
@@ -1,352 +0,0 @@
-/*
- * Copyright (c) 2018-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <cuvs/distance/distance.cuh>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/util/cudart_utils.hpp>
-
-#include <raft/stats/trustworthiness_score.cuh>
-#include <vector>
-
-namespace raft {
-namespace stats {
-
-class TrustworthinessScoreTest : public ::testing::Test {
- public:
-  TrustworthinessScoreTest()
-    : d_X(0, resource::get_cuda_stream(handle)), d_X_embedded(0, resource::get_cuda_stream(handle))
-  {
-  }
-
- protected:
-  void basicTest()
-  {
-    std::vector<float> X = {
-      5.6142087,   8.59787,     -4.382763,   -3.6452143,  -5.8816037,  -0.6330313,  4.6920023,
-      -0.79210913, 0.6106314,   2.1210914,   5.919943,    -8.43784,    -6.4819884,  0.41001374,
-      -6.1052523,  -4.0825715,  -5.314755,   -2.834671,   5.751696,    -6.5012555,  -0.4719201,
-      -7.53353,    7.6789393,   -1.4959852,  -5.5977287,  -9.564147,   1.2902534,   3.559834,
-      -6.7659483,  8.265964,    4.595404,    9.133477,    -6.1553917,  -6.319754,   -2.9039452,
-      4.4150834,   -3.094395,   -4.426273,   9.584571,    -5.64133,    6.6209483,   7.4044604,
-      3.9620576,   5.639907,    10.33007,    -0.8792053,  5.143776,    -7.464049,   1.2448754,
-      -5.6300974,  5.4518576,   4.119535,    6.749645,    7.627064,    -7.2298336,  1.9681473,
-      -6.9083176,  6.404673,    0.07186685,  9.0994835,   8.51037,     -8.986389,   0.40534487,
-      2.115397,    4.086756,    1.2284287,   -2.6272132,  0.06527536,  -9.587425,   -7.206078,
-      7.864875,    7.4397306,   -6.9233336,  -2.6643622,  3.3466153,   7.0408177,   -3.6069896,
-      -9.971769,   4.4075623,   7.9063697,   2.559074,    4.323717,    1.6867131,   -1.1576937,
-      -9.893141,   -3.251416,   -7.4889135,  -4.0588717,  -2.73338,    -7.4852257,  3.4460473,
-      9.759119,    -5.4680476,  -4.722435,   -8.032619,   -1.4598992,  4.227361,    3.135568,
-      1.1950601,   1.1982028,   6.998856,    -6.131138,   -6.6921015,  0.5361224,   -7.1213965,
-      -5.6104236,  -7.2212887,  -2.2710054,  8.544764,    -6.0254574,  1.4582269,   -5.5587835,
-      8.031556,    -0.26328218, -5.2591386,  -9.262641,   2.8691363,   5.299787,    -9.209455,
-      8.523085,    5.180329,    10.655528,   -5.7171874,  -6.7739563,  -3.6306462,  4.067106,
-      -1.5912259,  -3.2345476,  8.042973,    -3.6364832,  4.1242137,   9.886953,    5.4743724,
-      6.3058076,   9.369645,    -0.5175337,  4.9859877,   -7.879498,   1.358422,    -4.147944,
-      3.8984218,   5.894656,    6.4903927,   8.702036,    -8.023722,   2.802145,    -7.748032,
-      5.8461113,   -0.34215945, 11.298865,   1.4107164,   -9.949621,   -1.6257563,  -10.655836,
-      2.4528909,   1.1570255,   5.170669,    2.8398793,   7.1838694,   9.088459,    2.631155,
-      3.964414,    2.8769252,   0.04198391,  -0.16993195, 3.6747139,   -2.8377378,  6.1782537,
-      10.759618,   -4.5642614,  -8.522967,   0.8614642,   6.623416,    -1.029324,   5.5488334,
-      -7.804511,   2.128833,    7.9042315,   7.789576,    -2.7944536,  0.72271067,  -10.511495,
-      -0.78634536, -10.661714,  2.9376361,   1.9148129,   6.22859,     0.26264945,  8.028384,
-      6.8743043,   0.9351067,   7.0690722,   4.2846055,   1.4134506,   -0.18144785, 5.2778087,
-      -1.7140163,  9.217541,    8.602799,    -2.6537218,  -7.8377395,  1.1244944,   5.4540544,
-      -0.38506773, 3.9885726,   -10.76455,   1.4440702,   9.136163,    6.664117,    -5.7046547,
-      8.038592,    -9.229767,   -0.2799413,  3.6064725,   4.187257,    1.0516582,   -2.0707326,
-      -0.7615968,  -8.561018,   -3.7831352,  10.300297,   5.332594,    -6.5880876,  -4.2508664,
-      1.7985519,   5.7226253,   -4.1223383,  -9.6697855,  1.4885283,   7.524974,    1.7206005,
-      4.890457,    3.7264557,   0.4428284,   -9.922455,   -4.250455,   -6.4410596,  -2.107994,
-      -1.4109765,  -6.1325397,  0.32883006,  6.0489736,   7.7257385,   -8.281174,   1.0129383,
-      -10.792166,  8.378851,    10.802716,   9.848448,    -9.188757,   1.3151443,   1.9971865,
-      -2.521849,   4.3268294,   -7.775683,   -2.2902298,  3.0824065,   -7.17559,    9.6100855,
-      7.3965735,   -10.476525,  5.895973,    -3.6974669,  -7.6688933,  1.7354839,   -7.4045196,
-      -1.7992063,  -4.0394845,  5.2471714,   -2.250571,   2.528036,    -8.343515,   -2.2374575,
-      -10.019771,  0.73371273,  3.1853926,   2.7994921,   2.6637669,   7.620401,    7.515571,
-      0.68636256,  5.834537,    4.650282,    -1.0362619,  0.4461701,   3.7870514,   -4.1340904,
-      7.202998,    9.736904,    -3.005512,   -8.920467,   1.1228397,   6.2598724,   1.2812365,
-      4.5442104,   -8.791537,   0.92113096,  8.464749,    8.359035,    -4.3923397,  1.2252625,
-      -10.1986475, -1.4409319,  -10.013967,  3.9071581,   1.683064,    4.877419,    1.6570637,
-      9.559105,    7.3546534,   0.36635467,  5.220211,    4.6303267,   0.6601065,   0.16149978,
-      3.8818731,   -3.4438233,  8.42085,     8.659159,    -3.0935583,  -8.039611,   2.3060374,
-      5.134666,    1.0458113,   6.0190983,   -9.143728,   0.99048865,  9.210842,    6.670241,
-      -5.9614363,  0.8747396,   7.078824,    8.067469,    -10.314754,  0.45977542,  -9.28306,
-      9.1838665,   9.318644,    7.189082,    -11.092555,  1.0320464,   3.882163,    0.10953151,
-      7.9029684,   -6.9068265,  -1.3526366,  5.3996363,   -8.430931,   11.452577,   6.39663,
-      -11.090514,  4.6662245,   -3.1268113,  -8.357452,   2.2276728,   -10.357126,  -0.9291848,
-      -3.4193344,  3.1289792,   -2.5030103,  6.772719,    11.457757,   -4.2125936,  -6.684548,
-      -4.7611327,  3.6960156,   -2.3030636,  -3.0591488,  10.452471,   -4.1267314,  5.66614,
-      7.501461,    5.072407,    6.636537,    8.990381,    -0.2559256,  4.737867,    -6.2149944,
-      2.535682,    -5.5484023,  5.7113924,   3.4742818,   7.9915137,   7.0052586,   -7.156467,
-      1.4354781,   -8.286235,   5.7523417,   -2.4175215,  9.678009,    0.05066403,  -9.645226,
-      -2.2658763,  -9.518178,   4.493372,    2.3232365,   2.1659086,   0.42507997,  8.360246,
-      8.23535,     2.6878164,   5.236947,    3.4924245,   -0.6089895,  0.8884741,   4.359464,
-      -4.6073823,  7.83441,     8.958755,    -3.4690795,  -9.182282,   1.2478025,   5.6311107,
-      -1.2408862,  3.6316886,   -8.684654,   2.1078515,   7.2813864,   7.9265943,   -3.6135032,
-      0.4571511,   8.493568,    10.496853,   -7.432897,   0.8625995,   -9.607528,   7.2899456,
-      8.83158,     8.908199,    -10.300263,  1.1451302,   3.7871468,   -0.97040755, 5.7664757,
-      -8.9688,     -2.146672,   5.9641485,   -6.2908535,  10.126465,   6.1553903,   -12.066902,
-      6.301596,    -5.0419583,  -8.228695,   2.4879954,   -8.918582,   -3.7434099,  -4.1593685,
-      3.7431836,   -1.1704745,  0.5524103,   9.109399,    9.571567,    -11.209955,  1.2462777,
-      -9.554555,   9.091726,    11.477966,   7.630937,    -10.450911,  1.9205878,   5.358983,
-      -0.44546837, 6.7611346,   -9.74753,    -0.5939732,  3.8892255,   -6.437991,   10.294727,
-      5.6723895,   -10.7883,    6.192348,    -5.293862,   -10.811491,  1.0194173,   -7.074576,
-      -3.192368,   -2.5231771,  4.2791643,   -0.53309685, 0.501366,    9.636625,    7.710316,
-      -6.4219728,  1.0975566,   -8.218886,   6.9011984,   9.873679,    8.903804,    -9.316832,
-      1.2404599,   4.9039655,   1.2272617,   4.541515,    -5.2753224,  -3.2196746,  3.1303136,
-      -7.285681,   9.041425,    5.6417427,   -9.93667,    5.7548947,   -5.113397,   -8.544622,
-      4.182665,    -7.7709813,  -3.2810235,  -3.312072,   3.8900535,   -2.0604856,  6.709082,
-      -8.461194,   1.2666026,   4.8770437,   2.6955879,   3.0340345,   -1.1614609,  -3.536341,
-      -7.090382,   -5.36146,    9.072544,    6.4554095,   -4.4728956,  -1.88395,    3.1095037,
-      8.782348,    -3.316743,   -8.65248,    1.6802986,   8.186188,    2.1783829,   4.931278,
-      4.158475,    1.4033595,   -11.320101,  -3.7084908,  -6.740436,   -2.5555193,  -1.0451177,
-      -6.5569925,  0.82810307,  8.505919,    8.332857,    -9.488569,   -0.21588463, -8.056692,
-      8.493993,    7.6401625,   8.812983,    -9.377281,   2.4369764,   3.1766508,   0.6300803,
-      5.6666765,   -7.913654,   -0.42301777, 4.506412,    -7.8954244,  10.904591,   5.042256,
-      -9.626183,   8.347351,    -3.605006,   -7.923387,   1.1024277,   -8.705793,   -2.5151258,
-      -2.5066147,  4.0515003,   -2.060757,   6.2635093,   8.286584,    -6.0509276,  -6.76452,
-      -3.1158175,  1.6578803,   -1.4608748,  -1.24211,    8.151246,    -4.2970877,  6.093071,
-      7.4911637,   4.51018,     4.8425875,   9.211085,    -2.4386222,  4.5830803,   -5.6079445,
-      2.3713675,   -4.0707507,  3.1787417,   5.462342,    6.915912,    6.3928423,   -7.2970796,
-      5.0112796,   -9.140893,   4.9990606,   0.38391754,  7.7088532,   1.9340848,   8.18833,
-      8.16617,     -9.42086,    -0.3388326,  -9.659727,   8.243045,    8.099073,    8.439428,
-      -7.038694,   2.1077902,   3.3866816,   -1.9975324,  7.4972878,   -7.2525196,  -1.553731,
-      4.08758,     -6.6922374,  9.50525,     4.026735,    -9.243538,   7.2740564,   -3.9319072,
-      -6.3228955,  1.6693478,   -7.923119,   -3.7423058,  -2.2813146,  5.3469067,   -1.8285407,
-      3.3118162,   8.826356,    -4.4641976,  -6.4751124,  -9.200089,   -2.519147,   4.225298,
-      2.4105988,   -0.4344186,  0.53441775,  5.2836394,   -8.2816105,  -4.996147,   -1.6870759,
-      -7.8543897,  -3.9788852,  -7.0346904,  -3.1289773,  7.4567637,   -5.6227813,  1.0709786,
-      -8.866012,   8.427324,    -1.1755563,  -5.789216,   -8.197835,   5.3342214,   6.0646234,
-      -6.8975716,  7.717031,    3.480355,    8.312151,    -3.6645212,  -3.0976524,  -8.090359,
-      -1.9176173,  2.4257212,   1.9700835,   0.4098958,   2.1341088,   7.652741,    -9.9595585,
-      -5.989757,   0.10119354,  -7.935407,   -5.792786,   -5.22783,    -4.318978,   5.414037,
-      -6.4621663,  1.670883,    -6.9224787,  8.696932,    -2.0214002,  -6.6681314,  -8.326418,
-      4.9049683,   5.4442496,   -6.403739,   7.5822453,   7.0972915,   -9.072851,   -0.23897195,
-      1.7662339,   5.3096304,   1.983179,    -2.222645,   -0.34700772, -9.094717,   -6.107907,
-      9.525174,    8.1550665,   -5.6940084,  -4.1636486,  1.7360662,   8.528821,    -3.7299833,
-      -9.341266,   2.608542,    9.108706,    0.7978509,   4.2488184,   2.454484,    0.9446999,
-      -10.106636,  -3.8973773,  -6.6566644,  -4.5647273,  -0.99837756, -6.568582,   9.324853,
-      -7.9020953,  2.0910501,   2.2896829,   1.6790711,   1.3159255,   -3.5258796,  1.8898442,
-      -8.105812,   -4.924962,   8.771129,    7.1202874,   -5.991957,   -3.4106019,  2.4450088,
-      7.796387,    -3.055946,   -7.8971434,  1.9856719,   9.001636,    1.8511922,   3.019749,
-      3.1227696,   0.4822102,   -10.021213,  -3.530504,   -6.225959,   -3.0029628,  -1.7881511,
-      -7.3879776,  1.3925704,   9.499782,    -3.7318087,  -3.7074296,  -7.7466836,  -1.5284524,
-      4.0535855,   3.112011,    0.10340207,  -0.5429599,  6.67026,     -9.155924,   -4.924038,
-      0.64248866,  -10.0103655, -3.2742946,  -4.850029,   -3.6707063,  8.586258,    -5.855605,
-      4.906918,    -6.7813993,  7.9938135,   -2.5473144,  -5.688948,   -7.822478,   2.1421318,
-      4.66659,     -9.701272,   9.549149,    0.8998125,   -8.651497,   -0.56899565, -8.639817,
-      2.3088377,   2.1264515,   3.2764478,   2.341989,    8.594338,    8.630639,    2.8440373,
-      6.2043204,   4.433932,    0.6320018,   -1.8179281,  5.09452,     -1.5741565,  8.153934,
-      8.744339,    -3.6945698,  -8.883078,   1.5329908,   5.2745943,   0.44716078,  4.8809066,
-      -7.9594903,  1.134374,    9.233994,    6.5528665,   -4.520542,   9.477355,    -8.622195,
-      -0.23191702, 2.0485356,   3.9379985,   1.5916302,   -1.4516805,  -0.0843819,  -7.8554378,
-      -5.88308,    7.999766,    6.2572145,   -5.585321,   -4.0097756,  0.42382592,  6.160884,
-      -3.631315,   -8.333449,   2.770595,    7.8495173,   3.3331623,   4.940415,    3.6207345,
-      -0.037517,   -11.034698,  -3.185103,   -6.614664,   -3.2177854,  -2.0792234,  -6.8879867,
-      7.821685,    -8.455084,   1.0784642,   4.0033927,   2.7343264,   2.6052725,   -4.1224284,
-      -0.89305353, -6.8267674,  -4.9715133,  8.880253,    5.6994023,   -5.9695024,  -4.9181266,
-      1.3017995,   7.972617,    -3.9452884,  -10.424556,  2.4504194,   6.21529,     0.93840516,
-      4.2070026,   6.159839,    0.91979957,  -8.706724,   -4.317946,   -6.6823545,  -3.0388,
-      -2.464262,   -7.3716645,  1.3926703,   6.544412,    -5.6251183,  -5.122411,   -8.622049,
-      -2.3905911,  3.9138813,   1.9779967,   -0.05011125, 0.13310997,  7.229751,    -9.742043,
-      -8.08724,    1.2426697,   -7.9230795,  -3.3162494,  -7.129571,   -3.5488048,  7.4701195,
-      -5.2357526,  0.5917681,   -6.272206,   6.342328,    -2.909731,   -4.991607,   -8.845513,
-      3.3228495,   7.033246,    -7.8180246,  8.214469,    6.3910093,   9.185153,    -6.20472,
-      -7.713809,   -3.8481297,  3.5579286,   0.7078448,   -3.2893546,  7.384514,    -4.448121,
-      3.0104196,   9.492943,    8.024847,    4.9114385,   9.965594,    -3.014036,   5.182494,
-      -5.8806014,  2.5312455,   -5.9926524,  4.474469,    6.3717875,   6.993105,    6.493093,
-      -8.935534,   3.004074,    -8.055647,   8.315765,    -1.3026813,  8.250377,    0.02606229,
-      6.8508425,   9.655665,    -7.0116496,  -0.41060972, -10.049198,  7.897801,    6.7791023,
-      8.3362,      -9.821014,   2.491157,    3.5160472,   -1.6228812,  7.398063,    -8.769123,
-      -3.1743705,  3.2827861,   -6.497855,   10.831924,   5.2761307,   -9.704417,   4.3817043,
-      -3.9841619,  -8.111647,   1.1883026,   -8.115312,   -2.9240117,  -5.8879666,  4.20928,
-      -0.3587938,  6.935672,    -10.177582,  0.48819053,  3.1250648,   2.9306343,   3.082544,
-      -3.477687,   -1.3768549,  -7.4922366,  -3.756631,   10.039836,   3.6670392,   -5.9761434,
-      -4.4728765,  3.244255,    7.027899,    -2.3806512,  -10.4100685, 1.605716,    7.7953773,
-      0.5408159,   1.7156523,   3.824097,    -1.0604783,  -10.142124,  -5.246805,   -6.5283823,
-      -4.579547,   -2.42714,    -6.709197,   2.7782338,   7.33353,     -6.454507,   -2.9929368,
-      -7.8362985,  -2.695445,   2.4900775,   1.6682367,   0.4641757,   -1.0495365,  6.9631333,
-      -9.291356,   -8.23837,    -0.34263706, -8.275113,   -2.8454232,  -5.0864096,  -2.681942,
-      7.5450225,   -6.2517986,  0.06810654,  -6.470652,   4.9042645,   -1.8369255,  -6.6937943,
-      -7.9625087,  2.8510258,   6.180508,    -8.282598,   7.919079,    1.4897474,   6.7217417,
-      -4.2459426,  -4.114431,   -8.375707,   -2.143264,   5.6972933,   1.5574739,   0.39375135,
-      1.7930849,   5.1737595,   -7.826241,   -5.160268,   -0.80433255, -7.839536,   -5.2620406,
-      -5.4643164,  -3.185536,   6.620315,    -7.065227,   1.0524757,   -6.125088,   5.7126627,
-      -1.6161644,  -3.852159,   -9.164279,   2.7005782,   5.946544,    -8.468236,   8.2145405,
-      1.1035942,   6.590157,    -4.0461283,  -4.8090615,  -7.6702685,  -2.1121511,  5.1147075,
-      1.6128504,   2.0064135,   1.0544407,   6.0038295,   -7.8282537,  -4.801278,   0.32349443,
-      -8.0649805,  -4.372714,   -5.61336,    -5.21394,    8.176595,    -5.4753284,  1.7800134,
-      -8.267283,   7.2133374,   -0.16594432, -6.317046,   -9.490406,   4.1261597,   5.473317,
-      -7.7551675,  7.007468,    7.478628,    -8.801905,   0.10975724,  3.5478222,   4.797803,
-      1.3825226,   -3.357369,   0.99262005,  -6.94877,    -5.4781394,  9.632604,    5.7492557,
-      -5.9014316,  -3.1632116,  2.340859,    8.708098,    -3.1255999,  -8.848661,   4.5612836,
-      8.455157,    0.73460823,  4.112301,    4.392744,    -0.30759293, -6.8036823,  -3.0331545,
-      -8.269506,   -2.82415,    -0.9411246,  -5.993506,   2.1618164,   -8.716055,   -0.7432543,
-      -10.255819,  3.095418,    2.5131428,   4.752442,    0.9907621,   7.8279433,   7.85814,
-      0.50430876,  5.2840405,   4.457291,    0.03330028,  -0.40692952, 3.9244103,   -2.117118,
-      7.6977615,   8.759009,    -4.2157164,  -9.136053,   3.247858,    4.668686,    0.76162136,
-      5.3833632,   -9.231471,   0.44309422,  8.380872,    6.7211227,   -3.091507,   2.173508,
-      -9.038242,   -1.3666698,  -9.819077,   0.37825826,  2.3898845,   4.2440815,   1.9161536,
-      7.24787,     6.9124637,   1.6238527,   5.1140285,   3.1935842,   1.02845,     -1.1273454,
-      5.638998,    -2.497932,   8.342559,    8.586319,    -2.9069402,  -7.6387944,  3.5975037,
-      4.4115705,   0.41506064,  4.9078383,   -9.68327,    1.8159529,   9.744613,    8.40622,
-      -4.495336,   9.244892,    -8.789869,   1.3158468,   4.018167,    3.3922846,   2.652022,
-      -2.7495477,  0.2528986,   -8.268324,   -6.004913,   10.428784,   6.6580734,   -5.537176,
-      -1.7177434,  2.7504628,   6.7735,      -2.4454272,  -9.998361,   2.9483433,   6.8266654,
-      2.3787718,   4.472637,    2.5871701,   0.7355365,   -7.7027745,  -4.1879907,  -7.172832,
-      -4.1843605,  -0.03646783, -5.419406,   6.958486,    11.011111,   -7.1821184,  -7.956423,
-      -3.408451,   4.6850276,   -2.348787,   -4.398289,   6.9787564,   -3.8324208,  5.967827,
-      8.433518,    4.660108,    5.5657144,   9.964243,    -1.3515275,  6.404833,    -6.4805903,
-      2.4379845,   -6.0816774,  1.752272,    5.3771873,   6.9613523,   6.9788294,   -6.3894596,
-      3.7521114,   -6.8034263,  6.4458385,   -0.7233525,  10.512529,   4.362273,    9.231461,
-      -6.3382263,  -7.659,      -3.461823,   4.71463,     0.17817476,  -3.685746,   7.2962036,
-      -4.6489477,  5.218017,    11.546999,   4.7218375,   6.8498397,   9.281103,    -3.900459,
-      6.844054,    -7.0886965,  -0.05019227, -8.233724,   5.5808983,   6.374517,    8.321048,
-      7.969449,    -7.3478637,  1.4917561,   -8.003144,   4.780668,    -1.1981848,  7.753739,
-      2.0260844,   -8.880096,   -3.4258451,  -7.141975,   1.9637157,   1.814725,    5.311151,
-      1.4831505,   7.8483663,   7.257948,    1.395786,    6.417756,    5.376912,    0.59505713,
-      0.00062552,  3.6634305,   -4.159713,   7.3571978,   10.966816,   -2.5419605,  -8.466229,
-      1.904205,    5.6338267,   -0.52567476, 5.59736,     -8.361799,   0.5009981,   8.460681,
-      7.3891273,   -3.5272243,  5.0552278,   9.921456,    -7.69693,    -7.286378,   -1.9198836,
-      3.1666567,   -2.5832257,  -2.2445817,  9.888111,    -5.076563,   5.677401,    7.497946,
-      5.662994,    5.414262,    8.566503,    -2.5530663,  7.1032815,   -6.0612082,  1.3419591,
-      -4.9595256,  4.3377542,   4.3790717,   6.793512,    8.383502,    -7.1278043,  3.3240774,
-      -9.379446,   6.838661,    -0.81241214, 8.694813,    0.79141915,  7.632467,    8.575382,
-      -8.533798,   0.28954387,  -7.5675836,  5.8653326,   8.97235,     7.1649346,   -10.575289,
-      0.9359381,   5.02381,     -0.5609511,  5.543464,    -7.69131,    -2.1792977,  2.4729247,
-      -6.1917787,  10.373678,   7.6549597,   -8.809486,   5.5657206,   -3.3169382,  -8.042887,
-      2.0874746,   -7.079005,   -3.33398,    -3.6843317,  4.0172358,   -2.0754814,  1.1726758,
-      7.4618697,   6.9483604,   -8.469206,   0.7401797,   -10.318176,  8.384557,    10.5476265,
-      9.146971,    -9.250223,   0.6290606,   4.4941425,   -0.7514017,  7.2271705,   -8.309598,
-      -1.4761636,  4.0140634,   -6.021102,   9.132852,    5.6610966,   -11.249811,  8.359293,
-      -1.9445792,  -7.7393436,  -0.3931331,  -8.824441,   -2.5995944,  -2.5714035,  4.140213,
-      -3.6863053,  5.517265,    9.020411,    -4.9286127,  -7.871219,   -3.7446704,  2.5179656,
-      -1.4543481,  -2.2703636,  7.010597,    -3.6436229,  6.753862,    7.4129915,   7.1406755,
-      5.653706,    9.5445175,   0.15698843,  4.761813,    -7.698002,   1.6870106,   -4.5410123,
-      4.171763,    5.3747005,   6.341021,    7.456738,    -8.231657,   2.763487,    -9.208167,
-      6.676799,    -1.1957736,  10.062605,   4.0975976,   7.312957,    -2.4981596,  -2.9658387,
-      -8.150425,   -2.1075552,  2.64375,     1.6636052,   1.1483809,   0.09276015,  5.8556347,
-      -7.8481026,  -5.9913163,  -0.02840613, -9.937289,   -1.0486673,  -5.2340155,  -3.83912,
-      7.7165728,   -8.409944,   0.80863273,  -6.9119215,  7.5712357,   0.36031485,  -6.056131,
-      -8.470033,   1.8678337,   3.0121377,   -7.3096333,  8.205484,    5.262654,    8.774514,
-      -4.7603083,  -7.2096143,  -4.437014,   3.6080024,   -1.624254,   -4.2787876,  8.880863,
-      -4.8984556,  5.1782074,   9.944454,    3.911282,    3.5396595,   8.867042,    -1.2006199,
-      5.393288,    -5.6455317,  0.7829499,   -4.0338907,  2.479272,    6.5080743,   8.582535,
-      7.0097537,   -6.9823785,  3.984318,    -7.225381,   5.3135114,   -1.0391048,  8.951443,
-      -0.70119005, -8.510742,   -0.42949116, -10.9224825, 2.8176029,   1.6800792,   5.778404,
-      1.7269998,   7.1975236,   7.7258267,   2.7632928,   5.3399253,   3.4650044,   0.01971426,
-      -1.6468811,  4.114996,    -1.5110453,  6.8689218,   8.269899,    -3.1568048,  -7.0344677,
-      1.2911975,   5.950357,    0.19028673,  4.657226,    -8.199647,   2.246055,    8.989509,
-      5.3101015,   -4.2400866};
-
-    std::vector<float> X_embedded = {
-      -0.41849962, -0.53906363, 0.46958843,  -0.35832694, -0.23779503, -0.29751351, -0.01072748,
-      -0.21353109, -0.54769957, -0.55086273, 0.37093949,  -0.12714292, -0.06639574, -0.36098689,
-      -0.13060696, -0.07362658, -1.01205945, -0.39285606, 0.2864089,   -0.32031146, -0.19595343,
-      0.08900568,  -0.04813879, -0.06563424, -0.42655188, -0.69014251, 0.51459783,  -0.1942696,
-      -0.07767916, -0.6119386,  0.04813685,  -0.22557008, -0.56890118, -0.60293794, 0.43429622,
-      -0.09240723, -0.00624062, -0.25800395, -0.1886092,  0.01655941,  -0.01961523, -0.14147359,
-      0.41414487,  -0.8512944,  -0.61199242, -0.18586016, 0.14024924,  -0.41635606, -0.02890144,
-      0.1065347,   0.39700791,  -1.14060664, -0.95313865, 0.14416681,  0.17306046,  -0.53189689,
-      -0.98987544, -0.67918193, 0.41787854,  -0.20878236, -0.06612862, 0.03502904,  -0.03765266,
-      -0.0980606,  -0.00971657, 0.29432917,  0.36575687,  -1.1645509,  -0.89094597, 0.03718805,
-      0.2310573,   -0.38345811, -0.10401925, -0.10653082, 0.38469055,  -0.88302094, -0.80197543,
-      0.03548668,  0.02775662,  -0.54374295, 0.03379983,  0.00923623,  0.29320273,  -1.05263519,
-      -0.93360096, 0.03778313,  0.12360487,  -0.56437284, 0.0644429,   0.33432651,  0.36450726,
-      -1.22978747, -0.83822101, -0.18796451, 0.34888434,  -0.3801491,  -0.45327303, -0.59747899,
-      0.39697698,  -0.15616602, -0.06159166, -0.40301991, -0.11725303, -0.11913263, -0.12406619,
-      -0.11227967, 0.43083835,  -0.90535849, -0.81646025, 0.10012121,  -0.0141237,  -0.63747931,
-      0.04805023,  0.34190539,  0.50725192,  -1.17861414, -0.74641538, -0.09333111, 0.27992678,
-      -0.56214809, 0.04970971,  0.36249384,  0.57705611,  -1.16913795, -0.69849908, 0.10957897,
-      0.27983218,  -0.62088525, 0.0410459,   0.23973398,  0.40960434,  -1.14183664, -0.83321381,
-      0.02149482,  0.21720445,  -0.49869928, -0.95655465, -0.51680422, 0.45761383,  -0.08351214,
-      -0.12151554, 0.00819737,  -0.20813803, -0.01055793, 0.25319234,  0.36154974,  0.1822421,
-      -1.15837133, -0.92209691, -0.0501582,  0.08535917,  -0.54003763, -1.08675635, -1.04009593,
-      0.09408128,  0.07009826,  -0.01762833, -0.19180447, -0.18029785, -0.20342001, 0.04034991,
-      0.1814747,   0.36906669,  -1.13532007, -0.8852452,  0.0782818,   0.16825101,  -0.50301319,
-      -0.29128098, -0.65341312, 0.51484352,  -0.38758236, -0.22531103, -0.55021971, 0.10804344,
-      -0.3521522,  -0.38849035, -0.74110794, 0.53761131,  -0.25142813, -0.1118066,  -0.47453368,
-      0.06347904,  -0.23796193, -1.02682328, -0.47594091, 0.39515916,  -0.2782529,  -0.16566519,
-      0.08063579,  0.00810116,  -0.06213913, -1.059654,   -0.62496334, 0.53698546,  -0.11806234,
-      0.00356161,  0.11513405,  -0.14213292, 0.04102662,  -0.36622161, -0.73686272, 0.48323864,
-      -0.27338892, -0.14203401, -0.41736352, 0.03332564,  -0.21907479, -0.06396769, 0.01831361,
-      0.46263444,  -1.01878166, -0.86486858, 0.17622118,  -0.01249686, -0.74530888, -0.9354887,
-      -0.5027945,  0.38170099,  -0.15547098, 0.00677824,  -0.04677663, -0.13541745, 0.07253501,
-      -0.97933143, -0.58001202, 0.48235369,  -0.18836913, -0.02430783, 0.07572441,  -0.08101331,
-      0.00630076,  -0.16881248, -0.67989182, 0.46083611,  -0.43910736, -0.29321918, -0.38735861,
-      0.07669903,  -0.29749861, -0.40047669, -0.56722462, 0.33168188,  -0.13118173, -0.06672747,
-      -0.56856316, -0.26269144, -0.14236671, 0.10651901,  0.4962585,   0.38848072,  -1.06653547,
-      -0.64079332, -0.47378591, 0.43195483,  -0.04856951, -0.9840439,  -0.70610428, 0.34028092,
-      -0.2089237,  -0.05382041, 0.01625874,  -0.02080803, -0.12535211, -0.04146428, -1.24533033,
-      0.48944879,  0.0578458,   0.26708388,  -0.90321028, 0.35377088,  -0.36791429, -0.35382384,
-      -0.52748734, 0.42854419,  -0.31744713, -0.19174226, -0.39073724, -0.03258846, -0.19978228,
-      -0.36185205, -0.57412046, 0.43681973,  -0.25414538, -0.12904905, -0.46334973, -0.03123853,
-      -0.11303604, -0.87073672, -0.45441297, 0.41825858,  -0.25303507, -0.21845073, 0.10248682,
-      -0.11045569, -0.10002795, -0.00572806, 0.16519061,  0.42651513,  -1.11417019, -0.83789682,
-      0.02995787,  0.16843079,  -0.53874511, 0.03056994,  0.17877036,  0.49632853,  -1.03276777,
-      -0.74778616, -0.03971953, 0.10907949,  -0.67385727, -0.9523471,  -0.56550741, 0.40409449,
-      -0.2703723,  -0.10175014, 0.13605487,  -0.06306008, -0.01768126, -0.4749442,  -0.56964815,
-      0.39389887,  -0.19248079, -0.04161081, -0.38728487, -0.20341556, -0.12656988, -0.35949609,
-      -0.46137866, 0.28798422,  -0.06603147, -0.04363992, -0.60343552, -0.23565227, -0.10242701,
-      -0.06792886, 0.09689897,  0.33259571,  -0.98854214, -0.84444433, 0.00673901,  0.13457057,
-      -0.43145794, -0.51500046, -0.50821936, 0.38000089,  0.0132636,   0.0580942,   -0.40157595,
-      -0.11967677, 0.02549113,  -0.10350953, 0.22918226,  0.40411913,  -1.05619383, -0.71218503,
-      -0.02197581, 0.26422262,  -0.34765676, 0.06601537,  0.21712676,  0.34723559,  -1.20982027,
-      -0.95646334, 0.00793948,  0.27620381,  -0.43475035, -0.67326003, -0.6137197,  0.43724492,
-      -0.17666136, -0.06591748, -0.18937394, -0.07400128, -0.06881691, -0.5201112,  -0.61088628,
-      0.4225319,   -0.18969463, -0.06921366, -0.33993208, -0.06990873, -0.10288513, -0.70659858,
-      -0.56003648, 0.46628812,  -0.16090363, -0.0185108,  -0.1431348,  -0.1128775,  -0.0078648,
-      -0.02323332, 0.04292452,  0.39291084,  -0.94897962, -0.63863206, -0.16546988, 0.23698957,
-      -0.30633628};
-
-    auto stream = resource::get_cuda_stream(handle);
-
-    d_X.resize(X.size(), stream);
-    d_X_embedded.resize(X_embedded.size(), stream);
-    raft::update_device(d_X.data(), X.data(), X.size(), stream);
-    raft::update_device(d_X_embedded.data(), X_embedded.data(), X_embedded.size(), stream);
-    auto n_sample            = 50;
-    auto n_features_origin   = 30;
-    auto n_features_embedded = 8;
-
-    // euclidean test
-    score = trustworthiness_score<cuvs::distance::DistanceType::L2SqrtUnexpanded, float>(
-      handle,
-      raft::make_device_matrix_view<const float>(d_X.data(), n_sample, n_features_origin),
-      raft::make_device_matrix_view<const float>(
-        d_X_embedded.data(), n_sample, n_features_embedded),
-      5);
-  }
-
-  void SetUp() override { basicTest(); }
-
-  void TearDown() override {}
-
- protected:
-  raft::resources handle;
-
-  rmm::device_uvector<float> d_X;
-  rmm::device_uvector<float> d_X_embedded;
-
-  double score;
-};
-
-typedef TrustworthinessScoreTest TrustworthinessScoreTestF;
-TEST_F(TrustworthinessScoreTestF, Result) { ASSERT_TRUE(0.9375 < score && score < 0.9379); }
-};  // namespace stats
-};  // namespace raft
diff --git a/cpp/test/stats/v_measure.cu b/cpp/test/stats/v_measure.cu
deleted file mode 100644
index 0cc164f27..000000000
--- a/cpp/test/stats/v_measure.cu
+++ /dev/null
@@ -1,139 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-#include "../test_utils.cuh"
-#include <algorithm>
-#include <gtest/gtest.h>
-#include <iostream>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/stats/homogeneity_score.cuh>
-#include <raft/stats/v_measure.cuh>
-#include <raft/util/cudart_utils.hpp>
-#include <random>
-
-namespace raft {
-namespace stats {
-
-// parameter structure definition
-struct vMeasureParam {
-  int nElements;
-  int lowerLabelRange;
-  int upperLabelRange;
-  double beta;
-  bool sameArrays;
-  double tolerance;
-};
-
-// test fixture class
-template <typename T>
-class vMeasureTest : public ::testing::TestWithParam<vMeasureParam> {
- protected:
-  // the constructor
-  void SetUp() override
-  {
-    // getting the parameters
-    params = ::testing::TestWithParam<vMeasureParam>::GetParam();
-
-    nElements       = params.nElements;
-    lowerLabelRange = params.lowerLabelRange;
-    upperLabelRange = params.upperLabelRange;
-
-    // generating random value test input
-    std::vector<int> arr1(nElements, 0);
-    std::vector<int> arr2(nElements, 0);
-    std::random_device rd;
-    std::default_random_engine dre(rd());
-    std::uniform_int_distribution<int> intGenerator(lowerLabelRange, upperLabelRange);
-
-    std::generate(arr1.begin(), arr1.end(), [&]() { return intGenerator(dre); });
-    if (params.sameArrays) {
-      arr2 = arr1;
-    } else {
-      std::generate(arr2.begin(), arr2.end(), [&]() { return intGenerator(dre); });
-    }
-
-    // allocating and initializing memory to the GPU
-
-    stream = resource::get_cuda_stream(handle);
-    rmm::device_uvector<T> truthClusterArray(nElements, stream);
-    rmm::device_uvector<T> predClusterArray(nElements, stream);
-    raft::update_device(truthClusterArray.data(), &arr1[0], (int)nElements, stream);
-    raft::update_device(predClusterArray.data(), &arr2[0], (int)nElements, stream);
-
-    // calculating the golden output
-    double truthHomogeity, truthCompleteness;
-
-    truthHomogeity    = raft::stats::homogeneity_score(truthClusterArray.data(),
-                                                    predClusterArray.data(),
-                                                    nElements,
-                                                    lowerLabelRange,
-                                                    upperLabelRange,
-                                                    stream);
-    truthCompleteness = raft::stats::homogeneity_score(predClusterArray.data(),
-                                                       truthClusterArray.data(),
-                                                       nElements,
-                                                       lowerLabelRange,
-                                                       upperLabelRange,
-                                                       stream);
-
-    if (truthCompleteness + truthHomogeity == 0.0)
-      truthVMeasure = 0.0;
-    else
-      truthVMeasure = ((1 + params.beta) * truthHomogeity * truthCompleteness /
-                       (params.beta * truthHomogeity + truthCompleteness));
-    // calling the v_measure CUDA implementation
-    computedVMeasure = raft::stats::v_measure(
-      handle,
-      raft::make_device_vector_view<const T>(truthClusterArray.data(), nElements),
-      raft::make_device_vector_view<const T>(predClusterArray.data(), nElements),
-      lowerLabelRange,
-      upperLabelRange,
-      params.beta);
-  }
-
-  // declaring the data values
-  raft::resources handle;
-  vMeasureParam params;
-  T lowerLabelRange, upperLabelRange;
-  int nElements           = 0;
-  double truthVMeasure    = 0;
-  double computedVMeasure = 0;
-  cudaStream_t stream     = 0;
-};
-
-// setting test parameter values
-const std::vector<vMeasureParam> inputs = {{199, 1, 10, 1.0, false, 0.000001},
-                                           {200, 15, 100, 1.0, false, 0.000001},
-                                           {100, 1, 20, 1.0, false, 0.000001},
-                                           {10, 1, 10, 1.0, false, 0.000001},
-                                           {198, 1, 100, 1.0, false, 0.000001},
-                                           {300, 3, 99, 1.0, false, 0.000001},
-                                           {199, 1, 10, 1.0, true, 0.000001},
-                                           {200, 15, 100, 1.0, true, 0.000001},
-                                           {100, 1, 20, 1.0, true, 0.000001},
-                                           {10, 1, 10, 1.0, true, 0.000001},
-                                           {198, 1, 100, 1.0, true, 0.000001},
-                                           {300, 3, 99, 1.0, true, 0.000001}};
-
-// writing the test suite
-typedef vMeasureTest<int> vMeasureTestClass;
-TEST_P(vMeasureTestClass, Result)
-{
-  ASSERT_NEAR(computedVMeasure, truthVMeasure, params.tolerance);
-}
-INSTANTIATE_TEST_CASE_P(vMeasure, vMeasureTestClass, ::testing::ValuesIn(inputs));
-
-}  // end namespace stats
-}  // end namespace raft
diff --git a/cpp/test/stats/weighted_mean.cu b/cpp/test/stats/weighted_mean.cu
deleted file mode 100644
index da1a825da..000000000
--- a/cpp/test/stats/weighted_mean.cu
+++ /dev/null
@@ -1,339 +0,0 @@
-/*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "../test_utils.cuh"
-#include <cstdint>
-#include <gtest/gtest.h>
-#include <raft/core/device_mdspan.hpp>
-#include <raft/core/resource/cuda_stream.hpp>
-#include <raft/random/rng.cuh>
-#include <raft/stats/weighted_mean.cuh>
-#include <raft/util/cuda_utils.cuh>
-#include <thrust/device_vector.h>
-#include <thrust/host_vector.h>
-
-namespace raft {
-namespace stats {
-
-template <typename T>
-struct WeightedMeanInputs {
-  T tolerance;
-  int M, N;
-  unsigned long long int seed;
-  bool along_rows;  // Used only for the weightedMean test function
-  bool row_major;
-};
-
-template <typename T>
-::std::ostream& operator<<(::std::ostream& os, const WeightedMeanInputs<T>& I)
-{
-  return os << "{ " << I.tolerance << ", " << I.M << ", " << I.N << ", " << I.seed << ", "
-            << I.along_rows << "}" << std::endl;
-}
-
-///// weighted row-wise mean test and support functions
-template <typename T>
-void naiveRowWeightedMean(T* R, T* D, T* W, int M, int N, bool rowMajor)
-{
-  int istr = rowMajor ? 1 : M;
-  int jstr = rowMajor ? N : 1;
-
-  // sum the weights
-  T WS = 0;
-  for (int i = 0; i < N; i++)
-    WS += W[i];
-
-  for (int j = 0; j < M; j++) {
-    R[j] = (T)0;
-    for (int i = 0; i < N; i++) {
-      // R[j] += (W[i]*D[i*istr + j*jstr] - R[j])/(T)(i+1);
-      R[j] += (W[i] * D[i * istr + j * jstr]) / WS;
-    }
-  }
-}
-
-template <typename T>
-class RowWeightedMeanTest : public ::testing::TestWithParam<WeightedMeanInputs<T>> {
- protected:
-  void SetUp() override
-  {
-    params = ::testing::TestWithParam<WeightedMeanInputs<T>>::GetParam();
-    raft::random::RngState r(params.seed);
-    int rows = params.M, cols = params.N, len = rows * cols;
-    auto stream = resource::get_cuda_stream(handle);
-    // device-side data
-    din.resize(len);
-    dweights.resize(cols);
-    dexp.resize(rows);
-    dact.resize(rows);
-
-    // create random matrix and weights
-    uniform(handle, r, din.data().get(), len, T(-1.0), T(1.0));
-    uniform(handle, r, dweights.data().get(), cols, T(-1.0), T(1.0));
-
-    // host-side data
-    thrust::host_vector<T> hin      = din;
-    thrust::host_vector<T> hweights = dweights;
-    thrust::host_vector<T> hexp(rows);
-
-    // compute naive result & copy to GPU
-    naiveRowWeightedMean(hexp.data(), hin.data(), hweights.data(), rows, cols, params.row_major);
-    dexp        = hexp;
-    auto output = raft::make_device_vector_view<T, std::uint32_t>(dact.data().get(), rows);
-    auto weights =
-      raft::make_device_vector_view<const T, std::uint32_t>(dweights.data().get(), cols);
-
-    if (params.row_major) {
-      auto input = raft::make_device_matrix_view<const T, std::uint32_t, raft::row_major>(
-        din.data().get(), rows, cols);
-      // compute result
-      row_weighted_mean(handle, input, weights, output);
-    } else {
-      auto input = raft::make_device_matrix_view<const T, std::uint32_t, raft::col_major>(
-        din.data().get(), rows, cols);
-      // compute result
-      row_weighted_mean(handle, input, weights, output);
-    }
-
-    // adjust tolerance to account for round-off accumulation
-    params.tolerance *= params.N;
-  }
-
- protected:
-  raft::resources handle;
-  WeightedMeanInputs<T> params;
-  thrust::host_vector<T> hin, hweights;
-  thrust::device_vector<T> din, dweights, dexp, dact;
-};
-
-///// weighted column-wise mean test and support functions
-template <typename T>
-void naiveColWeightedMean(T* R, T* D, T* W, int M, int N, bool rowMajor)
-{
-  int istr = rowMajor ? 1 : M;
-  int jstr = rowMajor ? N : 1;
-
-  // sum the weights
-  T WS = 0;
-  for (int j = 0; j < M; j++)
-    WS += W[j];
-
-  for (int i = 0; i < N; i++) {
-    R[i] = (T)0;
-    for (int j = 0; j < M; j++) {
-      // R[i] += (W[j]*D[i*istr + j*jstr] - R[i])/(T)(j+1);
-      R[i] += (W[j] * D[i * istr + j * jstr]) / WS;
-    }
-  }
-}
-
-template <typename T>
-class ColWeightedMeanTest : public ::testing::TestWithParam<WeightedMeanInputs<T>> {
-  void SetUp() override
-  {
-    params = ::testing::TestWithParam<WeightedMeanInputs<T>>::GetParam();
-    raft::random::RngState r(params.seed);
-    int rows = params.M, cols = params.N, len = rows * cols;
-
-    auto stream = resource::get_cuda_stream(handle);
-    // device-side data
-    din.resize(len);
-    dweights.resize(rows);
-    dexp.resize(cols);
-    dact.resize(cols);
-
-    // create random matrix and weights
-    uniform(handle, r, din.data().get(), len, T(-1.0), T(1.0));
-    uniform(handle, r, dweights.data().get(), rows, T(-1.0), T(1.0));
-
-    // host-side data
-    thrust::host_vector<T> hin      = din;
-    thrust::host_vector<T> hweights = dweights;
-    thrust::host_vector<T> hexp(cols);
-
-    // compute naive result & copy to GPU
-    naiveColWeightedMean(hexp.data(), hin.data(), hweights.data(), rows, cols, params.row_major);
-    dexp = hexp;
-
-    auto output = raft::make_device_vector_view<T, std::uint32_t>(dact.data().get(), cols);
-    auto weights =
-      raft::make_device_vector_view<const T, std::uint32_t>(dweights.data().get(), rows);
-    if (params.row_major) {
-      auto input = raft::make_device_matrix_view<const T, std::uint32_t, raft::row_major>(
-        din.data().get(), rows, cols);
-      // compute result
-      col_weighted_mean(handle, input, weights, output);
-    } else {
-      auto input = raft::make_device_matrix_view<const T, std::uint32_t, raft::col_major>(
-        din.data().get(), rows, cols);
-      // compute result
-      col_weighted_mean(handle, input, weights, output);
-    }
-    // adjust tolerance to account for round-off accumulation
-    params.tolerance *= params.M;
-  }
-
- protected:
-  raft::resources handle;
-  WeightedMeanInputs<T> params;
-  thrust::host_vector<T> hin, hweights;
-  thrust::device_vector<T> din, dweights, dexp, dact;
-};
-
-template <typename T>
-class WeightedMeanTest : public ::testing::TestWithParam<WeightedMeanInputs<T>> {
- protected:
-  void SetUp() override
-  {
-    params = ::testing::TestWithParam<WeightedMeanInputs<T>>::GetParam();
-    raft::random::RngState r(params.seed);
-    auto stream = resource::get_cuda_stream(handle);
-    int rows = params.M, cols = params.N, len = rows * cols;
-    auto weight_size = params.along_rows ? cols : rows;
-    auto mean_size   = params.along_rows ? rows : cols;
-    // device-side data
-    din.resize(len);
-    dweights.resize(weight_size);
-    dexp.resize(mean_size);
-    dact.resize(mean_size);
-
-    // create random matrix and weights
-    uniform(handle, r, din.data().get(), len, T(-1.0), T(1.0));
-    uniform(handle, r, dweights.data().get(), weight_size, T(-1.0), T(1.0));
-
-    // host-side data
-    thrust::host_vector<T> hin      = din;
-    thrust::host_vector<T> hweights = dweights;
-    thrust::host_vector<T> hexp(mean_size);
-
-    // compute naive result & copy to GPU
-    if (params.along_rows)
-      naiveRowWeightedMean(hexp.data(), hin.data(), hweights.data(), rows, cols, params.row_major);
-    else
-      naiveColWeightedMean(hexp.data(), hin.data(), hweights.data(), rows, cols, params.row_major);
-    dexp = hexp;
-
-    auto output = raft::make_device_vector_view<T, std::uint32_t>(dact.data().get(), mean_size);
-    auto weights =
-      raft::make_device_vector_view<const T, std::uint32_t>(dweights.data().get(), weight_size);
-    if (params.row_major) {
-      auto input = raft::make_device_matrix_view<const T, std::uint32_t, raft::row_major>(
-        din.data().get(), rows, cols);
-      // compute result
-      weighted_mean(handle, input, weights, output, params.along_rows);
-    } else {
-      auto input = raft::make_device_matrix_view<const T, std::uint32_t, raft::col_major>(
-        din.data().get(), rows, cols);
-      // compute result
-      weighted_mean(handle, input, weights, output, params.along_rows);
-    }
-    // adjust tolerance to account for round-off accumulation
-    params.tolerance *= params.N;
-  }
-
- protected:
-  raft::resources handle;
-  WeightedMeanInputs<T> params;
-  thrust::host_vector<T> hin, hweights;
-  thrust::device_vector<T> din, dweights, dexp, dact;
-};
-
-////// Parameter sets and test instantiation
-static const float tolF  = 128 * std::numeric_limits<float>::epsilon();
-static const double tolD = 256 * std::numeric_limits<double>::epsilon();
-
-const std::vector<WeightedMeanInputs<float>> inputsf = {{tolF, 4, 4, 1234, true, true},
-                                                        {tolF, 32, 32, 1234, true, false},
-                                                        {tolF, 32, 64, 1234, false, false},
-                                                        {tolF, 32, 256, 1234, true, true},
-                                                        {tolF, 32, 256, 1234, false, false},
-                                                        {tolF, 1024, 32, 1234, true, false},
-                                                        {tolF, 1024, 64, 1234, true, true},
-                                                        {tolF, 1024, 128, 1234, true, false},
-                                                        {tolF, 1024, 256, 1234, true, true},
-                                                        {tolF, 1024, 32, 1234, false, false},
-                                                        {tolF, 1024, 64, 1234, false, true},
-                                                        {tolF, 1024, 128, 1234, false, false},
-                                                        {tolF, 1024, 256, 1234, false, true}};
-
-const std::vector<WeightedMeanInputs<double>> inputsd = {{tolD, 4, 4, 1234, true, true},
-                                                         {tolD, 32, 32, 1234, true, false},
-                                                         {tolD, 32, 64, 1234, false, false},
-                                                         {tolD, 32, 256, 1234, true, true},
-                                                         {tolD, 32, 256, 1234, false, false},
-                                                         {tolD, 1024, 32, 1234, true, false},
-                                                         {tolD, 1024, 64, 1234, true, true},
-                                                         {tolD, 1024, 128, 1234, true, false},
-                                                         {tolD, 1024, 256, 1234, true, true},
-                                                         {tolD, 1024, 32, 1234, false, false},
-                                                         {tolD, 1024, 64, 1234, false, true},
-                                                         {tolD, 1024, 128, 1234, false, false},
-                                                         {tolD, 1024, 256, 1234, false, true}};
-
-using RowWeightedMeanTestF = RowWeightedMeanTest<float>;
-TEST_P(RowWeightedMeanTestF, Result)
-{
-  ASSERT_TRUE(devArrMatch(
-    dexp.data().get(), dact.data().get(), params.M, raft::CompareApprox<float>(params.tolerance)));
-}
-INSTANTIATE_TEST_CASE_P(RowWeightedMeanTest, RowWeightedMeanTestF, ::testing::ValuesIn(inputsf));
-
-using RowWeightedMeanTestD = RowWeightedMeanTest<double>;
-TEST_P(RowWeightedMeanTestD, Result)
-{
-  ASSERT_TRUE(devArrMatch(
-    dexp.data().get(), dact.data().get(), params.M, raft::CompareApprox<double>(params.tolerance)));
-}
-INSTANTIATE_TEST_CASE_P(RowWeightedMeanTest, RowWeightedMeanTestD, ::testing::ValuesIn(inputsd));
-
-using ColWeightedMeanTestF = ColWeightedMeanTest<float>;
-TEST_P(ColWeightedMeanTestF, Result)
-{
-  ASSERT_TRUE(devArrMatch(
-    dexp.data().get(), dact.data().get(), params.N, raft::CompareApprox<float>(params.tolerance)));
-}
-INSTANTIATE_TEST_CASE_P(ColWeightedMeanTest, ColWeightedMeanTestF, ::testing::ValuesIn(inputsf));
-
-using ColWeightedMeanTestD = ColWeightedMeanTest<double>;
-TEST_P(ColWeightedMeanTestD, Result)
-{
-  ASSERT_TRUE(devArrMatch(
-    dexp.data().get(), dact.data().get(), params.N, raft::CompareApprox<double>(params.tolerance)));
-}
-INSTANTIATE_TEST_CASE_P(ColWeightedMeanTest, ColWeightedMeanTestD, ::testing::ValuesIn(inputsd));
-
-using WeightedMeanTestF = WeightedMeanTest<float>;
-TEST_P(WeightedMeanTestF, Result)
-{
-  auto mean_size = params.along_rows ? params.M : params.N;
-  ASSERT_TRUE(devArrMatch(
-    dexp.data().get(), dact.data().get(), mean_size, raft::CompareApprox<float>(params.tolerance)));
-}
-INSTANTIATE_TEST_CASE_P(WeightedMeanTest, WeightedMeanTestF, ::testing::ValuesIn(inputsf));
-
-using WeightedMeanTestD = WeightedMeanTest<double>;
-TEST_P(WeightedMeanTestD, Result)
-{
-  auto mean_size = params.along_rows ? params.M : params.N;
-  ASSERT_TRUE(devArrMatch(dexp.data().get(),
-                          dact.data().get(),
-                          mean_size,
-                          raft::CompareApprox<double>(params.tolerance)));
-}
-INSTANTIATE_TEST_CASE_P(WeightedMeanTest, WeightedMeanTestD, ::testing::ValuesIn(inputsd));
-
-};  // end namespace stats
-};  // end namespace raft
\ No newline at end of file