Make futex and condition variable waiting interfaces more similar.

[VoxSciurorum]

Also, move some code out of header files. The implementations already require function calls and system calls. Inlining them adds little value.

This adds support for the FreeBSD equivalent of futex.

Apple's operating system has a similar feature, not yet enabled because the interface is not meant to be used directly by applications. If we switch to using C++20 for the runtime we can use std::atomic::wait.

[kyle-singer]

Why not add a condvar.h?

[kyle-singer]

atomic-interop.tar.gz
My understanding is that C11 and C++ atomics are bit compatible for at least clang, gcc, and msvc (I think one of the standards recommends, but does not require it?). I tested with the attached code on x86_64 with clang and gcc, and it seems to suggest we could make just condvar.c -> condvar.cpp to use std::atomic::wait with this interface.

(This was meant to just be a quick and simple test, so doesn't directly test std::atomic::wait, just checks that storing in c and loading in c++ works).

[VoxSciurorum]

atomic-interop.tar.gz My understanding is that C11 and C++ atomics are bit compatible for at least clang, gcc, and msvc (I think one of the standards recommends, but does not require it?). I tested with the attached code on x86_64 with clang and gcc, and it seems to suggest we could make just condvar.c -> condvar.cpp to use std::atomic::wait with this interface.

(This was meant to just be a quick and simple test, so doesn't directly test std::atomic::wait, just checks that storing in c and loading in c++ works).

I would rather be type safe than rely on that behavior. TB and I have talked about changing the runtime from C to simple C++. If we do that we can switch everything to std::atomic, assuming the implementations are fast.