http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60381
--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> --- I've measured (everything --enable-checking=release bootstrapped) vanilla trunk, that + my http://gcc.gnu.org/ml/gcc-patches/2014-03/msg00142.html patch, trunk with 208220 reverted and the same plus your WIP patch on insn-recog.ii (on a SNB box): vanilla real 2m11.855s user 2m9.917s sys 0m1.479s my real 2m11.760s user 2m9.823s sys 0m1.449s 208220revert real 2m11.632s user 2m9.847s sys 0m1.310s lxo real 2m12.491s user 2m10.699s sys 0m1.304s so I guess all the changes are pretty much in the noise when the r208221 change is in, so we might as well just revert r208220 and do nothing else.
