On Tue, Feb 20, 2007 at 03:53:55PM -0800, Mark Mitchell wrote: > Richard Guenther wrote: > > >> This is 4.7% drop of SPECfp_base2006 ratio (geomean of individual FP > >> ratios). > > Clearly, 4.7% is significant. Grigory, thanks for the measurements! > > >> Here is the full set of changes in cpu2k6/fp performance of GCC 4.2 > >> compiler between r116799 and r120817, measured on Intel Core2 Duo at -O2 > >> optimization level. > > > > Do you happen to have a 4.1.x baseline to compare those against? > > Excellent question; I should have asked for that as well. If 4.2 has > gained on 4.1 in other respects, the 4.7% drop might represent a smaller > regression relative to 4.1. >
Comparing gcc 4.1 performance on 64bit Intel against 4.2 isn't very meaningful since the -mtune=generic patch isn't in gcc 4.1. However, gcc 4.1 from many Linux vendors do include it. I suggest we do one or all of the followings: 1. Limit gcc 4.1 on 32bit. 2. Limit gcc 4.1 on non-Intel platforms. 3. Use of one of gcc 4.1 from Linux vendor as a reference. gcc 4.1 from Red Hat and Novell are OK. H.J.