https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77468
James Greenhalgh <jgreenhalgh at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|REOPENED |RESOLVED Resolution|--- |FIXED --- Comment #23 from James Greenhalgh <jgreenhalgh at gcc dot gnu.org> --- I see a 13% performance improvement between trunk with ktkachov's patch applied and yesterday's trunk when compiling with -mcpu=cortex-a53 . In my tests -mcpu=cortex-a53 is equivalent to -mcpu=thunderx (-mcpu=cortex-a53 comes out slightly ahead, but not in a significant way, and not in a way I'd expect to reproduce across platforms). My only guess (from the magnitude of the difference you still see, and the improvement expected from ktkachov's patch) is that you didn't pick up the changes (perhaps they were not in the 20170214 snapshot you used?). We're going to be at an impasse here, as I can't validate your toolchain or environment, but everything I can see, and our analysis of the bug to this point, suggests that this can be considered fixed on trunk.