https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114855
--- Comment #41 from Aldy Hernandez <aldyh at gcc dot gnu.org> --- Created attachment 59181 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=59181&action=edit unrtested patch sorting threadable paths The performance improvement with this patch is: ** mainline Time variable usr sys wall GGC dominator optimization : 161.07 ( 9%) 0.65 ( 8%) 162.00 ( 9%) 236M ( 9%) backwards jump threading :1386.84 ( 78%) 1.76 ( 22%)1393.97 ( 78%) 519M ( 20%) TOTAL :1775.41 7.88 1789.76 2636M ** with patchset torsion:~/bld/benchmark/tainted/gcc []$ Time variable usr sys wall GGC dominator optimization : 284.32 ( 46%) 1.39 ( 16%) 286.54 ( 46%) 447M ( 13%) backwards jump threading : 15.51 ( 3%) 0.87 ( 10%) 16.40 ( 3%) 510M ( 15%) TOTAL : 616.15 8.81 626.82 3334M As discussed in comment 39, it causes a regression in gcc.dg/tree-ssa/ssa-sink-13.c (and possibly elsewhere) that must be addressed.
