https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126914
--- Comment #7 from Hongtao Liu <liuhongt at gcc dot gnu.org> --- (In reply to Jeffrey A. Law from comment #6) > Unfortunately, it appears that the x86 cost model is broken and if we go Possibly, one thing I noticed is that nice_parallel_costed_insn_p relies on set_rtx_cost(*set, speed_p) == COSTS_N_INSNS(1), which doesn't work well for x86. Many x86 patterns have costs greater than COSTS_N_INSNS(1), causing them to be incorrectly excluded.
