https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113832
--- Comment #3 from Filip Kastl <pheeck at gcc dot gnu.org> --- It seems that the benchmark runtime returned to normal for a while but then went back to 6% slowdown. Somewhere between these two commits the benchmark sped up g:eb7a8f213d59e0cf g:15d1dae0d4d1be88 And between these two commits the benchmark regressed once again g:c7a9883663a88861 g:84fc8f4f3263d665