Merge remote-tracking branch 'upstream/branch-25.02' into cccl-2.7.0-rc2

.github/copy-pr-bot.yaml

@@ -2,3 +2,4 @@
 # https://docs.gha-runners.nvidia.com/apps/copy-pr-bot/
 
 enabled: true
+auto_sync_draft: false

.github/workflows/trigger-breaking-change-alert.yaml

@@ -0,0 +1,26 @@
+name: Trigger Breaking Change Notifications
+
+on:
+  pull_request_target:
+    types:
+      - closed
+      - reopened
+      - labeled
+      - unlabeled
+
+jobs:
+  trigger-notifier:
+    if: contains(github.event.pull_request.labels.*.name, 'breaking')
+    secrets: inherit
+    uses: rapidsai/shared-workflows/.github/workflows/[email protected]
+    with:
+      sender_login: ${{ github.event.sender.login }}
+      sender_avatar: ${{ github.event.sender.avatar_url }}
+      repo: ${{ github.repository }}
+      pr_number: ${{ github.event.pull_request.number }}
+      pr_title: "${{ github.event.pull_request.title }}"
+      pr_body: "${{ github.event.pull_request.body || '_Empty PR description_' }}"
+      pr_base_ref: ${{ github.event.pull_request.base.ref }}
+      pr_author: ${{ github.event.pull_request.user.login }}
+      event_action: ${{ github.event.action }}
+      pr_merged: ${{ github.event.pull_request.merged }}

ci/build_wheel_pylibraft.sh

@@ -18,4 +18,4 @@ esac
 export SKBUILD_CMAKE_ARGS="-DDETECT_CONDA_ENV=OFF;-DFIND_RAFT_CPP=OFF${EXTRA_CMAKE_ARGS}"
 
 ci/build_wheel.sh pylibraft ${package_dir}
-ci/validate_wheel.sh ${package_dir} final_dist
+ci/validate_wheel.sh ${package_dir} final_dist pylibraft

ci/build_wheel_raft_dask.sh

@@ -9,4 +9,4 @@ package_dir="python/raft-dask"
 export SKBUILD_CMAKE_ARGS="-DDETECT_CONDA_ENV=OFF;-DFIND_RAFT_CPP=OFF"
 
 ci/build_wheel.sh raft-dask ${package_dir}
-ci/validate_wheel.sh ${package_dir} final_dist
+ci/validate_wheel.sh ${package_dir} final_dist raft-dask

ci/validate_wheel.sh

@@ -5,13 +5,37 @@ set -euo pipefail
 
 package_dir=$1
 wheel_dir_relative_path=$2
+package_name=$3
+
+RAPIDS_CUDA_MAJOR="${RAPIDS_CUDA_VERSION%%.*}"
+
+# some packages are much larger on CUDA 11 than on CUDA 12
+if [[ "${package_name}" == "raft-dask" ]]; then
+    PYDISTCHECK_ARGS=(
+        --max-allowed-size-compressed '200M'
+    )
+elif [[ "${package_name}" == "pylibraft" ]]; then
+    if [[ "${RAPIDS_CUDA_MAJOR}" == "11" ]]; then
+        PYDISTCHECK_ARGS=(
+            --max-allowed-size-compressed '600M'
+        )
+    else
+        PYDISTCHECK_ARGS=(
+            --max-allowed-size-compressed '100M'
+        )
+    fi
+else
+    echo "Unsupported package name: ${package_name}"
+    exit 1
+fi
 
 cd "${package_dir}"
 
 rapids-logger "validate packages with 'pydistcheck'"
 
 pydistcheck \
     --inspect \
+    "${PYDISTCHECK_ARGS[@]}" \
     "$(echo ${wheel_dir_relative_path}/*.whl)"
 
 rapids-logger "validate packages with 'twine'"

cpp/CMakeLists.txt

@@ -180,7 +180,10 @@ target_include_directories(
 )
 
 # Keep RAFT as lightweight as possible. Only CUDA libs and rmm should be used in global target.
-target_link_libraries(raft INTERFACE rmm::rmm cuco::cuco nvidia::cutlass::cutlass CCCL::CCCL)
+target_link_libraries(
+  raft INTERFACE rmm::rmm rmm::rmm_logger spdlog::spdlog_header_only cuco::cuco
+                 nvidia::cutlass::cutlass CCCL::CCCL
+)
 
 target_compile_features(raft INTERFACE cxx_std_17 $<BUILD_INTERFACE:cuda_std_17>)
 target_compile_options(
@@ -288,8 +291,10 @@ if(RAFT_COMPILE_LIBRARY)
                       "$<$<COMPILE_LANGUAGE:CUDA>:${RAFT_CUDA_FLAGS}>"
   )
 
-  add_library(raft_lib SHARED $<TARGET_OBJECTS:raft_objs>)
-  add_library(raft_lib_static STATIC $<TARGET_OBJECTS:raft_objs>)
+  # Make sure not to add the rmm logger twice since it will be brought in as an interface source by
+  # the rmm::rmm_logger_impl target.
+  add_library(raft_lib SHARED $<FILTER:$<TARGET_OBJECTS:raft_objs>,EXCLUDE,rmm.*logger>)
+  add_library(raft_lib_static STATIC $<FILTER:$<TARGET_OBJECTS:raft_objs>,EXCLUDE,rmm.*logger>)
 
   set_target_properties(
     raft_lib raft_lib_static
@@ -313,6 +318,8 @@ if(RAFT_COMPILE_LIBRARY)
     # ensure CUDA symbols aren't relocated to the middle of the debug build binaries
     target_link_options(${target} PRIVATE "${CMAKE_CURRENT_BINARY_DIR}/fatbin.ld")
   endforeach()
+  target_link_libraries(raft_lib PRIVATE rmm::rmm_logger_impl)
+  target_link_libraries(raft_lib_static PRIVATE rmm::rmm_logger_impl)
 endif()
 
 if(TARGET raft_lib AND (NOT TARGET raft::raft_lib))

cpp/bench/prims/CMakeLists.txt

@@ -32,6 +32,7 @@ function(ConfigureBench)
     PRIVATE raft::raft
             raft_internal
             $<$<BOOL:${ConfigureBench_LIB}>:raft::compiled>
+            $<$<NOT:$<BOOL:${ConfigureBench_LIB}>>:bench_rmm_logger>
             ${RAFT_CTK_MATH_DEPENDENCIES}
             benchmark::benchmark
             Threads::Threads
@@ -73,6 +74,9 @@ function(ConfigureBench)
 
 endfunction()
 
+add_library(bench_rmm_logger OBJECT)
+target_link_libraries(bench_rmm_logger PRIVATE rmm::rmm_logger_impl)
+
 if(BUILD_PRIMS_BENCH)
   ConfigureBench(NAME CORE_BENCH PATH core/bitset.cu core/copy.cu main.cpp)
 

cpp/cmake/thirdparty/get_spdlog.cmake

@@ -16,9 +16,9 @@
 function(find_and_configure_spdlog)
 
     include(${rapids-cmake-dir}/cpm/spdlog.cmake)
-    rapids_cpm_spdlog(FMT_OPTION "EXTERNAL_FMT_HO" INSTALL_EXPORT_SET rmm-exports)
-    rapids_export_package(BUILD spdlog rmm-exports)
+    rapids_cpm_spdlog(FMT_OPTION "EXTERNAL_FMT_HO" INSTALL_EXPORT_SET raft-exports)
+    rapids_export_package(BUILD spdlog raft-exports)
 
 endfunction()
 
-find_and_configure_spdlog()
+find_and_configure_spdlog()

cpp/include/raft/core/device_mdspan.hpp

@@ -210,7 +210,7 @@ auto constexpr make_device_strided_matrix_view(ElementType* ptr,
   constexpr auto is_row_major = std::is_same_v<LayoutPolicy, layout_c_contiguous>;
   constexpr auto is_col_major = std::is_same_v<LayoutPolicy, layout_f_contiguous>;
 
-  assert(is_row_major || is_col_major);
+  static_assert(is_row_major || is_col_major, "Unsupported layout policy for strided matrix view");
 
   IndexType stride0 = is_row_major ? (stride > 0 ? stride : n_cols) : 1;
   IndexType stride1 = is_row_major ? 1 : (stride > 0 ? stride : n_rows);

python/pylibraft/pyproject.toml

@@ -132,12 +132,10 @@ matrix-entry = "cuda_suffixed=true;use_cuda_wheels=true"
 
 [tool.pydistcheck]
 select = [
+    # NOTE: size threshold is managed via CLI args in CI scripts
     "distro-too-large-compressed",
 ]
 
-# detect when package size grows significantly
-max_allowed_size_compressed = '825M'
-
 [tool.pytest.ini_options]
 filterwarnings = [
     "error",