Dan Sugalski <[EMAIL PROTECTED]> wrote: > What're we looking at for performance impact on mops.pasm and the > other simple benchmarks?
I didn't look yet. Compiling optimized takes ages here :) Anyway: Estimate performance impact: JIT mops.pasm 0 (i386, PPC) CGP mops.pasm 0 - 1% plain mops.pasm 0.5% (PPC) - 3% (i386) The numbers for the plain function core are from docs/nanoparrot.c with and w/o -DINDIRECT. > ... (We've apparently been slowly slowing down > over the past months, which I'm not happy about) Pure function call speed decreased by ~300%. Taking now max. 3% and catching up with function call speed is worth the effort. If anyone wants to try: all run cores except JIT should be able to run the mops benchmark with INDIRECT_REGS enabled. leo
