gperftools VS mimalloc

For profiling you are right clang has no -pg that works. But there are options, since clang supports PGO the fprofile flags could be what you need. they will generated a profraw file for you. There is also gperf tools which work for more than just linux. https://github.com/gperftools/gperftools

Because we couldn't identify the issue using the results we got from Callgrind, we reached for another profiler, gperftools. It's a sampling profiler and therefor it has a smaller impact on the application's performance in exchange for less accurate call statistics. After filtering out the unimportant parts and visualizing the rest with pprof, it was evident that something strange was happening with the send function. It took only 71 milliseconds with the previous implementation and more than 900 milliseconds with the new implementation of our Bolt server. It was very suspicious, but based on Callgrind, its cost was almost the same as before. We were confused as the two results seemed to conflict with each other.

gprftools (https://github.com/gperftools/gperftools) can be easily plugged in using LD_PRELOAD and signal, and has nice go implemented visualization tool https://github.com/google/pprof.

Google's tcmalloc

Switching to jemalloc or mimalloc doesn't improve the performance either.

mimalloc: mimalloc is an open source implementation of malloc, currently the best performing allocator.

Some compiler writers thought that was the case, if [0] is related to OP. I don't have a "modern" (after 6th gen) Intel CPU to test it on, but note that most programs are compiled for a relatively generic CPU.

[0]: https://github.com/microsoft/mimalloc/issues/807

Yes, it's very efficient and that's not where the main problem lies. However, small allocations with modern C heap allocators like mimalloc or snmalloc has gotten extremely efficient as well. Would be interesting to see a benchmark comparison with Java's G1 and ZGC.

Memory management for C is not itself a solved problem, not only is there a lot of performance to squeeze out of malloc itself (the benchmarks on https://github.com/microsoft/mimalloc exemplifies the variance between the implementations), but it's up to the programmer to implement memory management in the large in an efficient way, which is not an easy task. One sure mark of a slow C program is one with a ton of mallocs and frees strewn all over.

It's a very terse howto for replacing Factorio's memory allocator with Microsoft's mimalloc, and configuring mimalloc so that memory is always allocated on huge pages by using madvise().

I was wondering if Koka's perceus referencing counting style is any different from the reference counting that Lean4 implements? I understand that both rely upon the mimalloc (https://github.com/microsoft/mimalloc) library in the backend.

If you are fine with heap allocations then there are only few dozens operator new/delete to override to regain control over normal C++ code memory use. Allocators and STL all need to call those. At least that's what gaming does on all platforms. If you need examples you can check Mimalloc on github ( https://github.com/microsoft/mimalloc/blob/master/include/mimalloc-new-delete.h ).

What's the difference between this and Microsoft's? https://github.com/microsoft/mimalloc