https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118959
--- Comment #5 from Filip Kastl <pheeck at gcc dot gnu.org> --- I measured (-O2 -march=native -flto on an AMD Zen3 machine) r15-7400-gd3ff498c478ace and r15-7852-ge836d80374aa03 and there is an 11% speedup which means we're back to the execution time before r15-7400-gd3ff498c478ace. Seems like you fixed it :). I'll wait for our automated benchmarks to confirm that this slowdown disappeared on all configurations and then I'll close this PR.
