Merge pull request #1042 from lplewa/benchmark_apply

reduce code duplication in Benchmark by using Apply() to set arguments

Author: lukaszstolarczuk

benchmark/benchmark.cpp

@@ -1,11 +1,13 @@
 /*
- * Copyright (C) 2024 Intel Corporation
+ * Copyright (C) 2024-2025 Intel Corporation
  *
  * Under the Apache License v2.0 with LLVM Exceptions. See LICENSE.TXT.
  * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
  *
  */
 
+#include <benchmark/benchmark.h>
+
 #include "benchmark.hpp"
 
 #define UMF_BENCHMARK_TEMPLATE_DEFINE(BaseClass, Method, ...)                  \
@@ -18,69 +20,65 @@
 
 #define UMF_BENCHMARK_REGISTER_F(BaseClass, Method)                            \
     BENCHMARK_REGISTER_F(BaseClass, Method)                                    \
-        ->ArgNames(                                                            \
-            BENCHMARK_PRIVATE_CONCAT_NAME(BaseClass, Method)::argsName())      \
-        ->Name(BENCHMARK_PRIVATE_CONCAT_NAME(BaseClass, Method)::name())       \
-        ->Iterations(                                                          \
-            BENCHMARK_PRIVATE_CONCAT_NAME(BaseClass, Method)::iterations())
-
-UMF_BENCHMARK_TEMPLATE_DEFINE(alloc_benchmark, glibc_fix, fixed_alloc_size,
-                              glibc_malloc);
+        ->Apply(                                                               \
+            &BENCHMARK_PRIVATE_CONCAT_NAME(BaseClass, Method)::defaultArgs)
 
 // Benchmarks scenarios:
 
 // The benchmark arguments specified in Args() are, in order:
 // benchmark arguments, allocator arguments, size generator arguments.
 // The exact meaning of each argument depends on the benchmark, allocator, and size components used.
 // Refer to the 'argsName()' function in each component to find detailed descriptions of these arguments.
+
+static void default_alloc_fix_size(benchmark::internal::Benchmark *benchmark) {
+    benchmark->Args({10000, 0, 4096});
+    benchmark->Args({10000, 100000, 4096});
+    benchmark->Threads(4);
+    benchmark->Threads(1);
+}
+
+static void
+default_alloc_uniform_size(benchmark::internal::Benchmark *benchmark) {
+    benchmark->Args({10000, 0, 8, 64 * 1024, 8});
+    benchmark->Threads(4);
+    benchmark->Threads(1);
+}
+
+UMF_BENCHMARK_TEMPLATE_DEFINE(alloc_benchmark, glibc_fix, fixed_alloc_size,
+                              glibc_malloc);
+
 UMF_BENCHMARK_REGISTER_F(alloc_benchmark, glibc_fix)
-    ->Args({10000, 0, 4096})
-    ->Args({10000, 100000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_alloc_fix_size);
 
 UMF_BENCHMARK_TEMPLATE_DEFINE(alloc_benchmark, glibc_uniform,
                               uniform_alloc_size, glibc_malloc);
 UMF_BENCHMARK_REGISTER_F(alloc_benchmark, glibc_uniform)
-    ->Args({10000, 0, 8, 64 * 1024, 8})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_alloc_uniform_size);
 
 UMF_BENCHMARK_TEMPLATE_DEFINE(alloc_benchmark, os_provider, fixed_alloc_size,
                               provider_allocator<os_provider>);
 UMF_BENCHMARK_REGISTER_F(alloc_benchmark, os_provider)
-    ->Args({10000, 0, 4096})
-    ->Args({10000, 100000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_alloc_fix_size);
 
 UMF_BENCHMARK_TEMPLATE_DEFINE(alloc_benchmark, proxy_pool, fixed_alloc_size,
                               pool_allocator<proxy_pool<os_provider>>);
 
 UMF_BENCHMARK_REGISTER_F(alloc_benchmark, proxy_pool)
-    ->Args({1000, 0, 4096})
-    ->Args({1000, 100000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_alloc_fix_size);
 
 #ifdef UMF_POOL_DISJOINT_ENABLED
 UMF_BENCHMARK_TEMPLATE_DEFINE(alloc_benchmark, disjoint_pool_fix,
                               fixed_alloc_size,
                               pool_allocator<disjoint_pool<os_provider>>);
 UMF_BENCHMARK_REGISTER_F(alloc_benchmark, disjoint_pool_fix)
-    ->Args({10000, 0, 4096})
-    ->Args({10000, 100000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_alloc_fix_size);
 
 // TODO: debug why this crashes
 /*UMF_BENCHMARK_TEMPLATE_DEFINE(alloc_benchmark, disjoint_pool_uniform,
                               uniform_alloc_size,
                               pool_allocator<disjoint_pool<os_provider>>);
 UMF_BENCHMARK_REGISTER_F(alloc_benchmark, disjoint_pool_uniform)
-    ->Args({10000, 0, 8, 64 * 1024, 8})
-    //    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_alloc_uniform_size);
 */
 #endif
 
@@ -89,18 +87,13 @@ UMF_BENCHMARK_TEMPLATE_DEFINE(alloc_benchmark, jemalloc_pool_fix,
                               fixed_alloc_size,
                               pool_allocator<jemalloc_pool<os_provider>>);
 UMF_BENCHMARK_REGISTER_F(alloc_benchmark, jemalloc_pool_fix)
-    ->Args({10000, 0, 4096})
-    ->Args({10000, 100000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_alloc_fix_size);
 
 UMF_BENCHMARK_TEMPLATE_DEFINE(alloc_benchmark, jemalloc_pool_uniform,
                               uniform_alloc_size,
                               pool_allocator<jemalloc_pool<os_provider>>);
 UMF_BENCHMARK_REGISTER_F(alloc_benchmark, jemalloc_pool_uniform)
-    ->Args({10000, 0, 8, 64 * 1024, 8})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_alloc_uniform_size);
 
 #endif
 #ifdef UMF_POOL_SCALABLE_ENABLED
@@ -109,71 +102,67 @@ UMF_BENCHMARK_TEMPLATE_DEFINE(alloc_benchmark, scalable_pool_fix,
                               pool_allocator<scalable_pool<os_provider>>);
 
 UMF_BENCHMARK_REGISTER_F(alloc_benchmark, scalable_pool_fix)
-    ->Args({10000, 0, 4096})
-    ->Args({10000, 100000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_alloc_fix_size);
 
 UMF_BENCHMARK_TEMPLATE_DEFINE(alloc_benchmark, scalable_pool_uniform,
                               uniform_alloc_size,
                               pool_allocator<scalable_pool<os_provider>>);
 
 UMF_BENCHMARK_REGISTER_F(alloc_benchmark, scalable_pool_uniform)
-    ->Args({10000, 0, 8, 64 * 1024, 8})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_alloc_uniform_size);
 #endif
 // Multiple allocs/free
+static void
+default_multiple_alloc_fix_size(benchmark::internal::Benchmark *benchmark) {
+    benchmark->Args({10000, 4096});
+    benchmark->Threads(4);
+    benchmark->Threads(1);
+}
+
+static void
+default_multiple_alloc_uniform_size(benchmark::internal::Benchmark *benchmark) {
+    benchmark->Args({10000, 8, 64 * 1024, 8});
+    benchmark->Threads(4);
+    benchmark->Threads(1);
+}
 
 UMF_BENCHMARK_TEMPLATE_DEFINE(multiple_malloc_free_benchmark, glibc_fix,
                               fixed_alloc_size, glibc_malloc);
 
 UMF_BENCHMARK_REGISTER_F(multiple_malloc_free_benchmark, glibc_fix)
-    ->Args({10000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_multiple_alloc_fix_size);
 
 UMF_BENCHMARK_TEMPLATE_DEFINE(multiple_malloc_free_benchmark, glibc_uniform,
                               uniform_alloc_size, glibc_malloc);
 UMF_BENCHMARK_REGISTER_F(multiple_malloc_free_benchmark, glibc_uniform)
-    ->Args({10000, 8, 64 * 1024, 8})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_multiple_alloc_uniform_size);
 
 UMF_BENCHMARK_TEMPLATE_DEFINE(multiple_malloc_free_benchmark, proxy_pool,
                               fixed_alloc_size,
                               pool_allocator<proxy_pool<os_provider>>);
 
 UMF_BENCHMARK_REGISTER_F(multiple_malloc_free_benchmark, proxy_pool)
-    ->Args({10000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_multiple_alloc_fix_size);
 
 UMF_BENCHMARK_TEMPLATE_DEFINE(multiple_malloc_free_benchmark, os_provider,
                               fixed_alloc_size,
                               provider_allocator<os_provider>);
 UMF_BENCHMARK_REGISTER_F(multiple_malloc_free_benchmark, os_provider)
-    ->Args({10000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_multiple_alloc_fix_size);
 
 #ifdef UMF_POOL_DISJOINT_ENABLED
 UMF_BENCHMARK_TEMPLATE_DEFINE(multiple_malloc_free_benchmark, disjoint_pool_fix,
                               fixed_alloc_size,
                               pool_allocator<disjoint_pool<os_provider>>);
 UMF_BENCHMARK_REGISTER_F(multiple_malloc_free_benchmark, disjoint_pool_fix)
-    ->Args({10000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_multiple_alloc_fix_size);
 
 // TODO: debug why this crashes
 /*UMF_BENCHMARK_TEMPLATE_DEFINE(multiple_malloc_free_benchmark,
                               disjoint_pool_uniform, uniform_alloc_size,
                               pool_allocator<disjoint_pool<os_provider>>);
 UMF_BENCHMARK_REGISTER_F(multiple_malloc_free_benchmark, disjoint_pool_uniform)
-    ->Args({10000, 0, 8, 64 * 1024, 8})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_multiple_alloc_uniform_size);
 */
 #endif
 
@@ -182,17 +171,13 @@ UMF_BENCHMARK_TEMPLATE_DEFINE(multiple_malloc_free_benchmark, jemalloc_pool_fix,
                               fixed_alloc_size,
                               pool_allocator<jemalloc_pool<os_provider>>);
 UMF_BENCHMARK_REGISTER_F(multiple_malloc_free_benchmark, jemalloc_pool_fix)
-    ->Args({10000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_multiple_alloc_fix_size);
 
 UMF_BENCHMARK_TEMPLATE_DEFINE(multiple_malloc_free_benchmark,
                               jemalloc_pool_uniform, uniform_alloc_size,
                               pool_allocator<jemalloc_pool<os_provider>>);
 UMF_BENCHMARK_REGISTER_F(multiple_malloc_free_benchmark, jemalloc_pool_uniform)
-    ->Args({1000, 8, 64 * 1024, 8})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_multiple_alloc_uniform_size);
 
 #endif
 
@@ -202,18 +187,14 @@ UMF_BENCHMARK_TEMPLATE_DEFINE(multiple_malloc_free_benchmark, scalable_pool_fix,
                               pool_allocator<scalable_pool<os_provider>>);
 
 UMF_BENCHMARK_REGISTER_F(multiple_malloc_free_benchmark, scalable_pool_fix)
-    ->Args({10000, 4096})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_multiple_alloc_fix_size);
 
 UMF_BENCHMARK_TEMPLATE_DEFINE(multiple_malloc_free_benchmark,
                               scalable_pool_uniform, uniform_alloc_size,
                               pool_allocator<scalable_pool<os_provider>>);
 
 UMF_BENCHMARK_REGISTER_F(multiple_malloc_free_benchmark, scalable_pool_uniform)
-    ->Args({10000, 8, 64 * 1024, 8})
-    ->Threads(4)
-    ->Threads(1);
+    ->Apply(&default_multiple_alloc_uniform_size);
 
 #endif
 BENCHMARK_MAIN();

benchmark/benchmark.hpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2024 Intel Corporation
+ * Copyright (C) 2024-2025 Intel Corporation
  *
  * Under the Apache License v2.0 with LLVM Exceptions. See LICENSE.TXT.
  * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
@@ -162,8 +162,15 @@ struct benchmark_interface : public benchmark::Fixture {
         return res;
     }
 
-    static std::string name() { return Allocator::name(); }
-    static int64_t iterations() { return 10000; }
+    virtual std::string name() { return Allocator::name(); }
+    virtual int64_t iterations() { return 10000; }
+    static void defaultArgs(Benchmark *benchmark) {
+        auto *bench =
+            static_cast<benchmark_interface<Size, Allocator> *>(benchmark);
+        benchmark->ArgNames(bench->argsName())
+            ->Name(bench->name())
+            ->Iterations(bench->iterations());
+    }
     Size alloc_size;
     Allocator allocator;
 };
@@ -260,15 +267,15 @@ class alloc_benchmark : public benchmark_interface<Size, Alloc> {
         }
     }
 
-    static std::vector<std::string> argsName() {
+    virtual std::vector<std::string> argsName() {
         auto n = benchmark_interface<Size, Alloc>::argsName();
         std::vector<std::string> res = {"max_allocs", "pre_allocs"};
         res.insert(res.end(), n.begin(), n.end());
         return res;
     }
 
-    static std::string name() { return base::name() + "/alloc"; }
-    static int64_t iterations() { return 200000; }
+    virtual std::string name() { return base::name() + "/alloc"; }
+    virtual int64_t iterations() { return 200000; }
 
   protected:
     using base = benchmark_interface<Size, Alloc>;
@@ -346,18 +353,18 @@ class multiple_malloc_free_benchmark : public alloc_benchmark<Size, Alloc> {
         }
     }
 
-    static std::string name() {
+    virtual std::string name() {
         return base::base::name() + "/multiple_malloc_free";
     }
 
-    static std::vector<std::string> argsName() {
+    virtual std::vector<std::string> argsName() {
         auto n = benchmark_interface<Size, Alloc>::argsName();
         std::vector<std::string> res = {"max_allocs"};
         res.insert(res.end(), n.begin(), n.end());
         return res;
     }
 
-    static int64_t iterations() { return 2000; }
+    virtual int64_t iterations() { return 2000; }
 
     std::default_random_engine generator;
     distribution dist;