https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105493
--- Comment #5 from Martin Jambor <jamborm at gcc dot gnu.org> --- I can confirm that an Intel Cascade Lake Xeon, using -Ofast -flto -march=native, 538.imagick_r produced by GCC 12.1 is almost 20% slower than the benchmark built with GCC 11.2 and the same options (on Zen2 or 3 I can see only smaller regressios at -O2 and with PGO) I do not see the 525.x264_r regression using the same compilers and options on this machine.
