The initial goal of zimalloc is to provide a good Zig-native general purpose allocator. For this, it would be best to have benchmarks written in Zig that use zimalloc as a normal Zig module, rather than using libzimalloc to override malloc in other pre-existing benchmarks. Benchmarking against other libc malloc replacements by using libzimalloc does not directly measure performance for the primary intended use case and should be negatively impacting performance due to the malloc API requiring the allocator to track additional allocation metadata.
The simplest solution would be to port existing allocator benchmarks to Zig. These can then be used to compare zimalloc to other allocators using c_allocator and LD_PRELOAD.
The initial goal of zimalloc is to provide a good Zig-native general purpose allocator. For this, it would be best to have benchmarks written in Zig that use zimalloc as a normal Zig module, rather than using
libzimallocto override malloc in other pre-existing benchmarks. Benchmarking against other libc malloc replacements by usinglibzimallocdoes not directly measure performance for the primary intended use case and should be negatively impacting performance due to the malloc API requiring the allocator to track additional allocation metadata.The simplest solution would be to port existing allocator benchmarks to Zig. These can then be used to compare zimalloc to other allocators using
c_allocatorandLD_PRELOAD.