From what I see, the PR contains two changes: One, changing int out for a long, on the basis that itâs more consistent with the functions being used, which is true enough, but itâs simultaneously less consistent with CDDAâs codebase, which uses ints all over the place. I suppose this sates Muffinâs personal sense of order, but doesnât fix anything and as jbytheway points out, may cause problems on some platforms. I donât see a pressing need for that change, but itâs also a minor detail.
The big change is to swap out the RNG to a mersenne twister, on what I can only see as the blind ass-pull assumption that the RNG is causing slowness. I gotta say, Kevinâs absolutely right on this, the right way to fix slowness is to first profile your code and figure out where the problem actually is. (In my experience, itâs often in a place you never would have guessed.) Then having identified it, you can work on the actual source of the problem without wasting time and potentially introducing bugs by working on things that wonât help.
Just throwing popular algorithms into your codebase willy-nilly because somebody told you they were fast, and hoping for the best, is absolutely the wrong way to go about optimization.
Asking for benchmarks to show that this would help isnât exactly a high bar to hurdle, either. Any n00b coder with an afternoon free can run a benchmark. Itâs not rocket science or even a big ask. Maybe the mersenne twister would help, maybe not, but we donât know without some numbers to show what it will do.
Besides all of that, if you think that post of yours wasnât insulting or likely to result in censure, youâve got way bigger problems than your buddyâs PR getting denied.
