This project shows the performance comparsion between system built-in malloc and tcmalloc.
Besides that, it is also shows a example to integrated tcmalloc via CMake build system.
Hardward environment:
Notebook: Dell G3 3590 CPU: Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz RAM: 32GB Software environment:
Windows: Windows 10.0.19045 Compiler: Visual Studio 2022 Linux: Linux xu-G3-3590 5.15.0-60-generic #66~20.04.1-Ubuntu SMP Wed Jan 25 09:41:30 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux Compiler: gcc version 9.4.0 * Note: Windows and Linux is due boot in the same SATA disk. Benchmark code:
#define COUNT 1000*1000 void bench_malloc_func() { for (size_t i = 0; i < COUNT; ++i) { int malloc_size = i * sizeof(int); int * pInt = (int*)malloc(malloc_size); free(pInt); } } void bench_malloc_and_access_func() { for (size_t i = 0; i < COUNT; ++i) { int malloc_size = i * sizeof(int); int * pInt = (int*)malloc(malloc_size); memset(pInt, 0, malloc_size); free(pInt); } }
Windows built-in malloc consumer a lot of time during this benchmark.
git clone --recursive https://github.com/xuhancn/bench_malloc cd bench_malloc # if you are updating an existing checkout git submodule sync git submodule update --init --recursivemkdir build cd build cmake ..On Linux
makeOn Windows
Need run command in Visual Studio Developer Command Prompt
msbuild bench_malloc.sln /property:Configuration=Release