> Ran a bench run on this morning's MAME v0.202 (mame0202-95-ge4d039bc07) > > Have a couple of big double digit performance regressions, wanted to check if > expected w/ code changes. > > Pong dropped ~24% > SFIII dropped ~20% >
yes, CPS3 is now drawing per-line instead of per-tile for one of the layers which makes it a bit slower as the per-line code isn't as tightly optimized (might remove the optimizations entirely as it's well above 100% on all systems anyway)
|