------- Comment #1 from hubicka at gcc dot gnu dot org 2010-06-28 07:46 ------- It really seems that both patches (prefetching and partial inlining) are responsible for some of the regressions. I've now comited the patch to reduce autoinlining from 50 to 40. It solves size issues at least with -O3 non-FDO/LTO SPECint2000 build. Unfortunately the tester also shows some regressions that might turn out to be offnoise. Lets see how the other benchmarks go.
I've also switched FDO tester to native testing and plan to enable prefetching with FDO and -O2 once this change is tested. Honza -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44688
