Merge remote-tracking branch 'origin/dev' into insert_or_apply_identity

benchmarks/CMakeLists.txt

@@ -1,5 +1,5 @@
 #=============================================================================
-# Copyright (c) 2018-2023, NVIDIA CORPORATION.
+# Copyright (c) 2018-2024, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -49,52 +49,52 @@ endfunction(ConfigureBench)
 ###################################################################################################
 # - static_set benchmarks -------------------------------------------------------------------------
 ConfigureBench(STATIC_SET_BENCH
-  hash_table/static_set/contains_bench.cu
-  hash_table/static_set/find_bench.cu
-  hash_table/static_set/insert_bench.cu
-  hash_table/static_set/retrieve_all_bench.cu
-  hash_table/static_set/size_bench.cu
-  hash_table/static_set/rehash_bench.cu)
+  static_set/contains_bench.cu
+  static_set/find_bench.cu
+  static_set/insert_bench.cu
+  static_set/retrieve_all_bench.cu
+  static_set/size_bench.cu
+  static_set/rehash_bench.cu)
 
 ###################################################################################################
 # - static_map benchmarks -------------------------------------------------------------------------
 ConfigureBench(STATIC_MAP_BENCH
-  hash_table/static_map/insert_bench.cu
-  hash_table/static_map/find_bench.cu
-  hash_table/static_map/contains_bench.cu
-  hash_table/static_map/erase_bench.cu
-  hash_table/static_map/insert_or_apply_bench.cu)
+  static_map/insert_bench.cu
+  static_map/find_bench.cu
+  static_map/contains_bench.cu
+  static_map/erase_bench.cu
+  static_map/insert_or_apply_bench.cu)
 
 ###################################################################################################
 # - static_multiset benchmarks --------------------------------------------------------------------
 ConfigureBench(STATIC_MULTISET_BENCH
-  hash_table/static_multiset/contains_bench.cu
-  hash_table/static_multiset/count_bench.cu
-  hash_table/static_multiset/find_bench.cu
-  hash_table/static_multiset/insert_bench.cu)
+  static_multiset/contains_bench.cu
+  static_multiset/count_bench.cu
+  static_multiset/find_bench.cu
+  static_multiset/insert_bench.cu)
 
 ###################################################################################################
 # - static_multimap benchmarks --------------------------------------------------------------------
 ConfigureBench(STATIC_MULTIMAP_BENCH
-  hash_table/static_multimap/insert_bench.cu
-  hash_table/static_multimap/retrieve_bench.cu
-  hash_table/static_multimap/query_bench.cu
-  hash_table/static_multimap/count_bench.cu)
+  static_multimap/insert_bench.cu
+  static_multimap/retrieve_bench.cu
+  static_multimap/query_bench.cu
+  static_multimap/count_bench.cu)
 
 ###################################################################################################
 # - dynamic_map benchmarks ------------------------------------------------------------------------
 ConfigureBench(DYNAMIC_MAP_BENCH
-  hash_table/dynamic_map/insert_bench.cu
-  hash_table/dynamic_map/find_bench.cu
-  hash_table/dynamic_map/contains_bench.cu
-  hash_table/dynamic_map/erase_bench.cu)
+  dynamic_map/insert_bench.cu
+  dynamic_map/find_bench.cu
+  dynamic_map/contains_bench.cu
+  dynamic_map/erase_bench.cu)
 
 ###################################################################################################
 # - hash function benchmarks ----------------------------------------------------------------------
-ConfigureBench(HASH_BENCH
-  hash_bench.cu)
+ConfigureBench(HASH_FUNCTION_BENCH
+  hash_function/hash_function_bench.cu)
 
 ###################################################################################################
 # - distinct_count_estimator benchmarks -----------------------------------------------------------
 ConfigureBench(DISTINCT_COUNT_ESTIMATOR_BENCH
-  distinct_count_estimator_bench.cu)
+  distinct_count_estimator/distinct_count_estimator_bench.cu)

benchmarks/benchmark_defaults.hpp

@@ -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.
@@ -28,8 +28,8 @@ using VALUE_TYPE_RANGE = nvbench::type_list<nvbench::int32_t, nvbench::int64_t>;
 
 auto constexpr N             = 100'000'000;
 auto constexpr OCCUPANCY     = 0.5;
-auto constexpr MULTIPLICITY  = 8;
-auto constexpr MATCHING_RATE = 0.5;
+auto constexpr MULTIPLICITY  = 1;
+auto constexpr MATCHING_RATE = 1.0;
 auto constexpr MAX_NOISE     = 3;
 auto constexpr SKEW          = 0.5;
 auto constexpr BATCH_SIZE    = 1'000'000;

benchmarks/distinct_count_estimator_bench.cu → benchmarks/distinct_count_estimator/distinct_count_estimator_bench.cu

@@ -30,8 +30,8 @@
 #include <cmath>
 #include <cstddef>
 
-using namespace cuco::benchmark;
-using namespace cuco::utility;
+using namespace cuco::benchmark;  // defaults, dist_from_state
+using namespace cuco::utility;    // key_generator, distribution
 
 template <typename InputIt>
 [[nodiscard]] std::size_t exact_distinct_count(InputIt first, std::size_t n)
@@ -114,7 +114,7 @@ void distinct_count_estimator_e2e(nvbench::state& state, nvbench::type_list<T, D
 }
 
 /**
- * @brief A benchmark evaluating `cuco::distinct_count_estimator::add` performance
+ * @brief A benchmark evaluating `cuco::distinct_count_estimator::add_async` performance
  */
 template <typename T, typename Dist>
 void distinct_count_estimator_add(nvbench::state& state, nvbench::type_list<T, Dist>)
@@ -146,7 +146,7 @@ using TYPE_RANGE = nvbench::type_list<nvbench::int32_t, nvbench::int64_t, __int1
 
 NVBENCH_BENCH_TYPES(distinct_count_estimator_e2e,
                     NVBENCH_TYPE_AXES(TYPE_RANGE, nvbench::type_list<distribution::uniform>))
-  .set_name("distinct_count_estimator_e2e")
+  .set_name("distinct_count_estimator_e2e_uniform")
   .set_type_axes_names({"T", "Distribution"})
   .add_int64_power_of_two_axis("NumInputs", {28, 29, 30})
   .add_int64_axis("SketchSizeKB", {8, 16, 32, 64, 128, 256})  // 256KB uses gmem fallback kernel
@@ -155,7 +155,7 @@ NVBENCH_BENCH_TYPES(distinct_count_estimator_e2e,
 
 NVBENCH_BENCH_TYPES(distinct_count_estimator_add,
                     NVBENCH_TYPE_AXES(TYPE_RANGE, nvbench::type_list<distribution::uniform>))
-  .set_name("distinct_count_estimator::add_async")
+  .set_name("distinct_count_estimator_add_uniform")
   .set_type_axes_names({"T", "Distribution"})
   .add_int64_power_of_two_axis("NumInputs", {28, 29, 30})
   .add_int64_axis("SketchSizeKB", {8, 16, 32, 64, 128, 256})

benchmarks/hash_table/dynamic_map/contains_bench.cu → benchmarks/dynamic_map/contains_bench.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.
@@ -25,8 +25,8 @@
 #include <thrust/device_vector.h>
 #include <thrust/transform.h>
 
-using namespace cuco::benchmark;
-using namespace cuco::utility;
+using namespace cuco::benchmark;  // defaults, dist_from_state
+using namespace cuco::utility;    // key_generator, distribution
 
 /**
  * @brief A benchmark evaluating `cuco::dynamic_map::contains` performance
@@ -37,9 +37,9 @@ std::enable_if_t<(sizeof(Key) == sizeof(Value)), void> dynamic_map_contains(
 {
   using pair_type = cuco::pair<Key, Value>;
 
-  auto const num_keys      = state.get_int64_or_default("NumInputs", defaults::N);
-  auto const initial_size  = state.get_int64_or_default("InitSize", defaults::INITIAL_SIZE);
-  auto const matching_rate = state.get_float64_or_default("MatchingRate", defaults::MATCHING_RATE);
+  auto const num_keys      = state.get_int64("NumInputs");
+  auto const initial_size  = state.get_int64("InitSize");
+  auto const matching_rate = state.get_float64("MatchingRate");
 
   thrust::device_vector<Key> keys(num_keys);
 
@@ -77,10 +77,12 @@ NVBENCH_BENCH_TYPES(dynamic_map_contains,
                     NVBENCH_TYPE_AXES(defaults::KEY_TYPE_RANGE,
                                       defaults::VALUE_TYPE_RANGE,
                                       nvbench::type_list<distribution::unique>))
-  .set_name("dynamic_map_contains_unique_num_inputs")
+  .set_name("dynamic_map_contains_unique_capacity")
   .set_type_axes_names({"Key", "Value", "Distribution"})
   .set_max_noise(defaults::MAX_NOISE)
-  .add_int64_axis("NumInputs", defaults::N_RANGE);
+  .add_int64_axis("NumInputs", defaults::N_RANGE)
+  .add_int64_axis("InitSize", {defaults::INITIAL_SIZE})
+  .add_float64_axis("MatchingRate", {defaults::MATCHING_RATE});
 
 NVBENCH_BENCH_TYPES(dynamic_map_contains,
                     NVBENCH_TYPE_AXES(defaults::KEY_TYPE_RANGE,
@@ -89,4 +91,6 @@ NVBENCH_BENCH_TYPES(dynamic_map_contains,
   .set_name("dynamic_map_contains_unique_matching_rate")
   .set_type_axes_names({"Key", "Value", "Distribution"})
   .set_max_noise(defaults::MAX_NOISE)
+  .add_int64_axis("NumInputs", {defaults::N})
+  .add_int64_axis("InitSize", {defaults::INITIAL_SIZE})
   .add_float64_axis("MatchingRate", defaults::MATCHING_RATE_RANGE);

benchmarks/hash_table/dynamic_map/erase_bench.cu → benchmarks/dynamic_map/erase_bench.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.
@@ -25,8 +25,8 @@
 #include <thrust/device_vector.h>
 #include <thrust/transform.h>
 
-using namespace cuco::benchmark;
-using namespace cuco::utility;
+using namespace cuco::benchmark;  // defaults, dist_from_state
+using namespace cuco::utility;    // key_generator, distribution
 
 /**
  * @brief A benchmark evaluating `cuco::dynamic_map::erase` performance
@@ -37,9 +37,9 @@ std::enable_if_t<(sizeof(Key) == sizeof(Value)), void> dynamic_map_erase(
 {
   using pair_type = cuco::pair<Key, Value>;
 
-  auto const num_keys      = state.get_int64_or_default("NumInputs", defaults::N);
-  auto const initial_size  = state.get_int64_or_default("InitSize", defaults::INITIAL_SIZE);
-  auto const matching_rate = state.get_float64_or_default("MatchingRate", defaults::MATCHING_RATE);
+  auto const num_keys      = state.get_int64("NumInputs");
+  auto const initial_size  = state.get_int64("InitSize");
+  auto const matching_rate = state.get_float64("MatchingRate");
 
   thrust::device_vector<Key> keys(num_keys);
 
@@ -80,10 +80,12 @@ NVBENCH_BENCH_TYPES(dynamic_map_erase,
                     NVBENCH_TYPE_AXES(defaults::KEY_TYPE_RANGE,
                                       defaults::VALUE_TYPE_RANGE,
                                       nvbench::type_list<distribution::unique>))
-  .set_name("dynamic_map_erase_unique_num_inputs")
+  .set_name("dynamic_map_erase_unique_capacity")
   .set_type_axes_names({"Key", "Value", "Distribution"})
   .set_max_noise(defaults::MAX_NOISE)
-  .add_int64_axis("NumInputs", defaults::N_RANGE);
+  .add_int64_axis("NumInputs", defaults::N_RANGE)
+  .add_int64_axis("InitSize", {defaults::INITIAL_SIZE})
+  .add_float64_axis("MatchingRate", {defaults::MATCHING_RATE});
 
 NVBENCH_BENCH_TYPES(dynamic_map_erase,
                     NVBENCH_TYPE_AXES(defaults::KEY_TYPE_RANGE,
@@ -92,4 +94,6 @@ NVBENCH_BENCH_TYPES(dynamic_map_erase,
   .set_name("dynamic_map_erase_unique_matching_rate")
   .set_type_axes_names({"Key", "Value", "Distribution"})
   .set_max_noise(defaults::MAX_NOISE)
+  .add_int64_axis("NumInputs", {defaults::N})
+  .add_int64_axis("InitSize", {defaults::INITIAL_SIZE})
   .add_float64_axis("MatchingRate", defaults::MATCHING_RATE_RANGE);