https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118959
--- Comment #10 from Filip Kastl <pheeck at gcc dot gnu.org> --- Ok, so these two benchmark configurations are back to their original speed: https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=465.10.0 https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=993.10.0 That's nice. 15% and 8% slowdown gone. These two didn't improve very much. The slowdowns here might not be connected to the same issue as the first two. They also aren't necessarily regressions against GCC 14: https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=992.10.0 https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=1119.10.0 So I propose leaving these two be. However, I found two configurations which were fine before and got worse with r15-7852-ge836d80374aa03 (I bisected the Zen4 slowdown to the commit): https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=693.10.0 https://lnt.opensuse.org/db_default/v4/SPEC/graph?plot.0=961.10.0 The slowdowns are around 5%. Richard, do you think anything can be done with that?
