[IDEA] Dispatch to emplace{,_exists}

Would need a dedicated config element later

include/hibf/build/insert_into_ibf.hpp

@@ -21,7 +21,8 @@ namespace seqan::hibf::build
  * \details
  * Automatically does naive splitting if number_of_bins > 1.
  */
-void insert_into_ibf(robin_hood::unordered_flat_set<uint64_t> const & kmers,
+void insert_into_ibf(build_data const & data,
+                     robin_hood::unordered_flat_set<uint64_t> const & kmers,
                      size_t const number_of_bins,
                      size_t const bin_index,
                      seqan::hibf::interleaved_bloom_filter & ibf,

src/build/construct_ibf.cpp

@@ -55,7 +55,7 @@ seqan::hibf::interleaved_bloom_filter construct_ibf(robin_hood::unordered_flat_s
     local_index_allocation_timer.stop();
     data.index_allocation_timer += local_index_allocation_timer;
 
-    insert_into_ibf(kmers, number_of_bins, ibf_node.max_bin_index, ibf, data.fill_ibf_timer);
+    insert_into_ibf(data, kmers, number_of_bins, ibf_node.max_bin_index, ibf, data.fill_ibf_timer);
     if (!is_root)
         update_parent_kmers(parent_kmers, kmers, data.merge_kmers_timer);
 

src/build/insert_into_ibf.cpp

@@ -23,8 +23,25 @@
 namespace seqan::hibf::build
 {
 
+template <bool use_exists>
+inline void
+dispatch_emplace(seqan::hibf::interleaved_bloom_filter & ibf, auto && values, seqan::hibf::bin_index const bin_index)
+{
+    if constexpr (use_exists)
+    {
+        for (auto && value : values)
+            ibf.emplace_exists(value, bin_index);
+    }
+    else
+    {
+        for (auto && value : values)
+            ibf.emplace(value, bin_index);
+    }
+}
+
 // automatically does naive splitting if number_of_bins > 1
-void insert_into_ibf(robin_hood::unordered_flat_set<uint64_t> const & kmers,
+void insert_into_ibf(build_data const & data,
+                     robin_hood::unordered_flat_set<uint64_t> const & kmers,
                      size_t const number_of_bins,
                      size_t const bin_index,
                      seqan::hibf::interleaved_bloom_filter & ibf,
@@ -40,8 +57,10 @@ void insert_into_ibf(robin_hood::unordered_flat_set<uint64_t> const & kmers,
         assert(chunk_number < number_of_bins);
         seqan::hibf::bin_index const bin_idx{bin_index + chunk_number};
         ++chunk_number;
-        for (size_t const value : chunk)
-            ibf.emplace_exists(value, bin_idx);
+        if (data.config.empty_bin_fraction > 0.0)
+            dispatch_emplace<true>(ibf, std::move(chunk), bin_idx);
+        else
+            dispatch_emplace<false>(ibf, std::move(chunk), bin_idx);
     }
     local_fill_ibf_timer.stop();
     fill_ibf_timer += local_fill_ibf_timer;
@@ -62,8 +81,10 @@ void insert_into_ibf(build_data const & data,
 
     serial_timer local_fill_ibf_timer{};
     local_fill_ibf_timer.start();
-    for (auto && value : values)
-        ibf.emplace_exists(value, bin_index);
+    if (data.config.empty_bin_fraction > 0.0)
+        dispatch_emplace<true>(ibf, std::move(values), bin_index);
+    else
+        dispatch_emplace<false>(ibf, std::move(values), bin_index);
     local_fill_ibf_timer.stop();
     data.fill_ibf_timer += local_fill_ibf_timer;
 }

src/hierarchical_interleaved_bloom_filter.cpp

@@ -133,7 +133,7 @@ size_t hierarchical_build(hierarchical_interleaved_bloom_filter & hibf,
                 size_t const mutex_id{parent_bin_index / 64};
                 std::lock_guard<std::mutex> guard{local_ibf_mutex[mutex_id]};
                 technical_bin_to_ibf_id[parent_bin_index] = new_ibf_pos;
-                build::insert_into_ibf(kmers, 1, parent_bin_index, ibf, data.fill_ibf_timer);
+                build::insert_into_ibf(data, kmers, 1, parent_bin_index, ibf, data.fill_ibf_timer);
                 if (!is_root)
                     build::update_parent_kmers(parent_kmers, kmers, data.merge_kmers_timer);
             }
@@ -155,7 +155,8 @@ size_t hierarchical_build(hierarchical_interleaved_bloom_filter & hibf,
         else
         {
             compute_kmers(kmers, data, record);
-            build::insert_into_ibf(kmers,
+            build::insert_into_ibf(data,
+                                   kmers,
                                    record.number_of_technical_bins,
                                    record.storage_TB_id,
                                    ibf,