Improvements to protocol splitting
Instead of generating new OT extensions, all yao protocols share a single extension box, and use GCC atomics to distribute nonces among themselves without collisions. This saves quite a bit of time (400ms on my machines). Updated the protocol splitting test to allow for testing and benchmarking of this change.
I'm realizing only now that I never notified you: I believe your concerns have been taken care of. I will handle all of the other issues we talked about in and around this PR (gate counter improvements, correlated OT) in a separate PR once this one is completed.
Btw, I still have two other largish PRs to submit after this one. Any chance we can get it merged before too long?
Ah, thanks for the bump. Setting a reminder on my calendar to do it by next weekend.