Update 2026-02-07

CMakeLists.txt

@@ -1,13 +1,13 @@
 cmake_minimum_required(VERSION 3.20.0)  # RHEL 8
 
 project(benchmarks
-    VERSION 0.0.0.1 # PROJECT_VERSION_MAJOR, PROJECT_VERSION_MINOR, PROJECT_VERSION_PATCH, and PROJECT_VERSION_TWEAK.
+    VERSION 0.0.1.0 # PROJECT_VERSION_MAJOR, PROJECT_VERSION_MINOR, PROJECT_VERSION_PATCH, and PROJECT_VERSION_TWEAK.
     DESCRIPTION "Various benchmarks"
     HOMEPAGE_URL "https://github.com/gunchev/bench/"
     LANGUAGES C CXX)
 set(CMAKE_C_STANDARD 11)
 set(CMAKE_C_STANDARD_REQUIRED TRUE)
-set(CMAKE_CXX_STANDARD 20)
+set(CMAKE_CXX_STANDARD 23)
 set(CMAKE_CXX_STANDARD_REQUIRED TRUE)
 
 # set(CMAKE_VERBOSE_MAKEFILE TRUE) # or "cd build; make VERBOSE=1"
@@ -18,6 +18,9 @@ if(NOT CMAKE_BUILD_TYPE)
         "Choose the type of build, options are: Debug DebugCov RelWithDebInfo Release MinSizeRel." FORCE)
 endif()
 
+include(FetchContent)
+
+
 # SOVERSION is ignored if NO_SONAME property is set.
 # "${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR}.${PROJECT_VERSION_PATCH}.${PROJECT_VERSION_TWEAK}"
 set(${PROJECT_NAME}_SOVERSION "${PROJECT_VERSION_MAJOR}.${PROJECT_VERSION_MINOR}")
@@ -47,19 +50,14 @@ set(CMAKE_POSITION_INDEPENDENT_CODE ON)
 
 
 # Compiler/linker flags
-if(EXISTS "/etc/fedora-release")
-set(CMAKE_C_FLAGS   "-Wall -Wextra")
-set(CMAKE_CXX_FLAGS "-Wall -Wextra")
-else()
 set(CMAKE_C_FLAGS   "-Wall -Wextra -Werror")
 set(CMAKE_CXX_FLAGS "-Wall -Wextra -Werror")
-endif()
+
+# https://gcc.gnu.org/onlinedocs/gcc/Template-Instantiation.html#Template-Instantiation
 
 include(CheckCCompilerFlag)
 include(CheckCXXCompilerFlag)
 
-# https://gcc.gnu.org/onlinedocs/gcc/Template-Instantiation.html#Template-Instantiation
-
 set(CMAKE_REQUIRED_QUIET_OLD ${CMAKE_REQUIRED_QUIET})
 set(CMAKE_REQUIRED_QUIET ON)
 macro(CheckAndAddFlag flag)
@@ -120,20 +118,24 @@ set(CMAKE_C_FLAGS_MINSIZEREL                 "-Os -ggdb3 -D_FORTIFY_SOURCE=2 -DN
 set(CMAKE_CXX_FLAGS_MINSIZEREL               "-Os -ggdb3 -D_FORTIFY_SOURCE=2 -DNDEBUG")
 
 
+# Google benchmark
+find_package(benchmark 1.9.4)
+if (NOT benchmark_FOUND)
+    message(STATUS "Google Benchmark not found, fetching it...")
 
-# Enable testing?
+    FetchContent_Declare(
-include(FindPkgConfig)
+        benchmark
-pkg_check_modules(CPPUNIT cppunit>=1.14.0)  # REQUIRED
+        GIT_REPOSITORY https://github.com/google/benchmark.git
-if(CPPUNIT_FOUND)
+        GIT_TAG v1.9.4
-    enable_testing()
+    )
-    include(CTest)
+
-    find_program(CTEST_COVERAGE_COMMAND NAMES gcov)
+    # Disable benchmark's own tests
-    find_program(CTEST_MEMORYCHECK_COMMAND NAMES valgrind)
+    set(BENCHMARK_ENABLE_TESTING OFF CACHE BOOL "" FORCE)
-    add_custom_target(check COMMAND ${CMAKE_CTEST_COMMAND})
+    set(BENCHMARK_ENABLE_GTEST_TESTS OFF CACHE BOOL "" FORCE)
+
+    FetchContent_MakeAvailable(benchmark)
 endif()
-
+message(STATUS "Found Google benchmark version: ${benchmark_VERSION}")
-
-find_package(benchmark REQUIRED)
 
 # Threads library
 # set(THREADS_PREFER_PTHREAD_FLAG TRUE)
@@ -145,8 +147,29 @@ find_package(benchmark REQUIRED)
 
 # pkg_check_modules(HIREDIS REQUIRED hiredis>=1.0.2)
 
+# Enable testing?
+#include(FindPkgConfig)
+#pkg_check_modules(CPPUNIT cppunit>=1.14.0)  # REQUIRED
+# set(GTEST_MIN_VERSION 1.13.0)
+# find_package(GTest ${GTEST_MIN_VERSION} QUIET)
+# if (NOT GTest_FOUND)
+#     message(STATUS "GoogleTest not found, fetching it...")
+#     FetchContent_Declare(
+#         googletest
+#         GIT_REPOSITORY https://github.com/google/googletest.git
+#         GIT_TAG v1.14.0
+#     )
+#     # Don't install or build gtest's own tests
+#     set(INSTALL_GTEST OFF CACHE BOOL "" FORCE)
+#     set(BUILD_GMOCK OFF CACHE BOOL "" FORCE)
+
+#     FetchContent_MakeAvailable(googletest)
+# endif()
+
+
 # include_directories(src)
 
 add_executable(transform src/transform.cpp)
 target_link_libraries(transform PRIVATE benchmark::benchmark) # fmt::fmt Threads::Threads)
+target_compile_features(transform PRIVATE cxx_std_17)
 install(TARGETS transform DESTINATION bin)

README.md

@@ -4,4 +4,59 @@
 Various C++ benchmarks:
 
 - transform.cpp - benchmark std::transform vs loop and custom lowercase and tolower
-![](bench_lowercase.png)
+
+---
+```shell
+    $ make releasec # clang++ 21.1.8
+    $ ./build/releasec/transform
+```
+
+            2026-02-07T15:43:00+02:00
+            Running ./build/releasec/transform
+            Run on (32 X 5187.5 MHz CPU s)
+            CPU Caches:
+              L1 Data 48 KiB (x16)
+              L1 Instruction 32 KiB (x16)
+              L2 Unified 1024 KiB (x16)
+              L3 Unified 32768 KiB (x2)
+            Load Average: 0.58, 0.53, 0.60
+            ***WARNING*** CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.
+            ***WARNING*** ASLR is enabled, the results may have unreproducible noise in them.
+            ----------------------------------------------------------------------
+            Benchmark                            Time             CPU   Iterations
+            ----------------------------------------------------------------------
+            algorithm_lowercase_my            17.8 ns         17.7 ns     39664595
+            loop_lowercase_my                 31.9 ns         31.9 ns     22028280
+            algorithm_lowercase_toupper       79.3 ns         79.2 ns      8838048
+            loop_lowercase_toupper            69.7 ns         69.5 ns      9977740
+---
+```shell
+    $ make release # g++ 15.2.1
+    $ ./build/release/transform 
+```
+
+            2026-02-07T15:43:03+02:00
+            Running ./build/release/transform
+            Run on (32 X 5187.5 MHz CPU s)
+            CPU Caches:
+              L1 Data 48 KiB (x16)
+              L1 Instruction 32 KiB (x16)
+              L2 Unified 1024 KiB (x16)
+              L3 Unified 32768 KiB (x2)
+            Load Average: 0.61, 0.54, 0.60
+            ***WARNING*** CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.
+            ***WARNING*** ASLR is enabled, the results may have unreproducible noise in them.
+            ----------------------------------------------------------------------
+            Benchmark                            Time             CPU   Iterations
+            ----------------------------------------------------------------------
+            algorithm_lowercase_my            17.7 ns         17.7 ns     33607332
+            loop_lowercase_my                 28.2 ns         28.2 ns     24781013
+            algorithm_lowercase_toupper       77.1 ns         77.0 ns      9116098
+            loop_lowercase_toupper            65.5 ns         65.4 ns     10687636
+---
+
+<img src="bench_lowercase_gcc13_cpp23.png" alt="gcc13" style="width:98%;"/>
+
+<img src="bench_lowercase_clang17_cpp23.png" alt="clang17" style="width:98%;"/>
+
+---

src/transform.cpp

@@ -3,13 +3,16 @@
 #include <benchmark/benchmark.h>
 #include <string>
 #include <sysexits.h>
+#include <cstring>
 
 
 const char* data = "Hello world, to be or not to be a long string, that is the question!";
 
 static void algorithm_lowercase_my(benchmark::State& state) {
+    std::string name;
+    name.reserve(strlen(data));
     for (auto _ : state) {  // Code inside this loop is measured repeatedly
-        std::string name = data;
+        name = data;
         std::transform(name.cbegin(), name.cend(), name.begin(), [](char c) noexcept {
             return (c >= 'a' && c <= 'z') ? c = c - 'a' + 'A' : c;
         });
@@ -19,8 +22,10 @@ static void algorithm_lowercase_my(benchmark::State& state) {
 BENCHMARK(algorithm_lowercase_my);  // Register the function as a benchmark
 
 static void loop_lowercase_my(benchmark::State& state) {
+    std::string name;
+    name.reserve(strlen(data));
     for (auto _ : state) {  // Code before the loop is not measured
-        std::string name = data;
+        name = data;
         for (char& c : name) {
             if (c >= 'a' && c <= 'z') {
                 c = c - 'a' + 'A';
@@ -33,8 +38,10 @@ BENCHMARK(loop_lowercase_my);  // Register the function as a benchmark
 
 
 static void algorithm_lowercase_toupper(benchmark::State& state) {
+    std::string name;
+    name.reserve(strlen(data));
     for (auto _ : state) {  // Code inside this loop is measured repeatedly
-        std::string name = data;
+        name = data;
         std::transform(name.cbegin(), name.cend(), name.begin(), [](char c) noexcept {
             return std::toupper(static_cast<unsigned char>(c));
         });
@@ -44,6 +51,21 @@ static void algorithm_lowercase_toupper(benchmark::State& state) {
 BENCHMARK(algorithm_lowercase_toupper);  // Register the function as a benchmark
 
 
+static void loop_lowercase_toupper(benchmark::State& state) {
+    std::string name;
+    name.reserve(strlen(data));
+    for (auto _ : state) {  // Code inside this loop is measured repeatedly
+        name = data;
+        for (char& c : name) {
+            if (c >= 'a' && c <= 'z') {
+                c = static_cast<char>(std::toupper(c));
+            }
+        }
+        benchmark::DoNotOptimize(name);  // Make sure the variable is not optimized away by compiler
+    }
+}
+BENCHMARK(loop_lowercase_toupper);  // Register the function as a benchmark
+
 int main(int argc, char** argv) {
     benchmark::Initialize(&argc, argv);
     if (benchmark::ReportUnrecognizedArguments(argc, argv)) {
@@ -53,16 +75,3 @@ int main(int argc, char** argv) {
     benchmark::RunSpecifiedBenchmarks();
     return EX_OK;
 }
-
-static void loop_lowercase_toupper(benchmark::State& state) {
-    for (auto _ : state) {  // Code inside this loop is measured repeatedly
-        std::string name = data;
-        for (char& c : name) {
-            if (c >= 'a' && c <= 'z') {
-                c = static_cast<char>(std::toupper(c));
-            }
-        }
-        benchmark::DoNotOptimize(name);  // Make sure the variable is not optimized away by compiler
-    }
-}
-BENCHMARK(loop_lowercase_toupper);  // Register the function as a benchmark