Closed GabrielRavier closed 4 years ago
I just made a simple program that compared the speed of the routines ; my versions were ~40% quicker (mostly due to you pushing almost all registers at each call). Instructions like lodsb (who nobody uses anymore, thus they aren't made better by CPU manufacturers, thus nobody uses them, etc.) didn't help too
Ah, I guess push/pop stack was a bad idea. I've been a bit busy lately, but I'll get around to this eventually.
Ah, I guess push/pop stack was a bad idea. I've been a bit busy lately, but I'll get around to this eventually.
I don't think it was a bad idea, in fact it just makes it easier to manage. When defining individual push/pop we need to be careful to preserve the stack.
Hello!
Curious, how did you determine if the patches improve performance?