On Thu, Aug 11, 2011 at 08:23, Michael Matz <m...@suse.de> wrote: > The obvious question: have you measured performance (with an optimized > cc1) on a medium sized exectutable, say cc1 itself? (I'm worrying about > inlining that doesn't happen anymore).
I compared three things: profiled bootstrap time for C, LTO compilation of insn-attrtab.o and LTO link of cc1. I found no significant differences: Bootstrap: before = 3,177 secs. after = 3,133 secs (-1% difference) insn-attrtab.o: before = 83.85 secs. after = 83.69 secs (0% difference) cc1 link: before = 57.55 secs. after = 57.19 secs (-1% difference) I also tested on the stage1 compiler (to remove the effects of the LTO optimizations). Again, little to no differences: insn-attrtab.o: before = 345.65 secs. after = 347.29 secs (0% difference) cc1 link: before = 99.30 secs. after = 99.39 secs (0% difference) The effect of the callbacks is negligible. Diego.
