On 24/09/2011, at 2:19 AM, Martin Jambor wrote: > However, both of these are really 4.8 material and since the patches > probably need only minor updates, it might be worthwhile to do that so > that gcc can handle the "embarrassing" simple cases. So I will do > that (though it might need to wait for about a week), re-try them on > Firefox and probably propose them for submission.
Great! Thank you. > > By the way, do you evaluate (your) devirtualization by compiling any > real software (other than Firefox)? Do the patches make any > difference there? The devirtualization and inlining improvements with your patches in the mix provided very good improvements on certain proprietary source base. Code size shrank by 8-10% and performance improved by 3-10% (depending on hardware). There was some amount of parameter tuning involved, so the improvements with default settings will not be as big, but they still should be noticeable. [The testing was done with a 4.6-based toolchain.] The optimizations that we implemented targeted coding patterns that are distilled in the 9 inline-devirt-*.C testcases that I posted. The actual code base was not micro-optimized, and this makes me confident that the improvements are general enough to be valuable to broad bodies of code. Regards, -- Maxim Kuvyrkov CodeSourcery / Mentor Graphics
