I use C++11 threading and need to set the affinity of a thread to a given processor core (windows' scheduler is not delegating to cores intelligently and I want control over what thread goes to what core).
For windows, there is
SetThreadAffinityMask(GetCurrentThread(), affinity);
What would a single cross-platform function look like that would work on windows, linux and OSX, with all the permutations of underlying threading libraries (windows threads, pthreads, etc.)?
If part of C++, I suspect it would look something like the pair of member function:
std::error_code std::thread::hardware_affinity(const std::vector<bool>&);
and free function:
std::error_code std::this_thread::hardware_affinity(const std::vector<bool>&);
on the assumption that we're talking about roughly the same sort of resources that std::thread::hardware_concurrency() counts, and that std::thread implementations must already take care of "all the permutations of underlying threading libraries".
Building one yourself, or using a utility library (I can't name an existing one right now) it would probably look like:
thread_utils::hardware_affinity(std::thread::native_handle_type, const std::vector<bool>&)
but it would have to support all the permutations internally. That's not hugely hard (many applications have done it out of necessity anyway), but it's more work than a Stack Overflow answer can hold. ^_^
One annoying limitation with this approach is that although there is the member function std::thread::native_handle(), there's no std::this_thread::native_handle() method. It's basically pending a write-up and a good use case to be standardised.
