High contention multithreaded performance using spin's Mutex vs galloc

[SFBdragon]

I don't know why, but Talc gets hit harder by extreme multithreaded contention than Galloc does using spin::Mutex. While this is not so relevant to most use cases, I'd like to resolve this issue if possible. If anyone has some tips for alleviating this, please open an issue!

I've tried aligning the bucket array to be on seperate cache lines, but no change in result. Other than that, I don't know what to try.