Home

what's been tried, what works, what doesn't:

building with modern gcc, optimizations, lto

tried.

result: works great. up to 8% speedup on hardware, 7% in emulators.

letting gcc optimize the c versions of everything instead of using the asm

tried, just to satisfy the people that would surely ask.

result: horrible failure. if you really want to see how bad it is you can go build yourself with OASM=no

rewrite asm snippets to avoid the fxch renaming optimization

in progress.

expected result of up to 5% speedup on 486 hardware, up to 12% slowdown on pentiums (best/worst case scenario) if all cases are rewritten well