I just isolated a truly bizarre performance bug - 64-bit size_t is 1.5x as fast as uint32_t in my simulation code. It's either a compiler codegen problem, CPU OoO/pipeline problem, or an integer overflow somewhere. :woozy_baa:
Follow
@niconiconi any divisions present? If not it's really bizarre, because casting down to u32 could be pushed off to the very end.
