https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81611
--- Comment #12 from Jeffrey A. Law <law at redhat dot com> --- You know, I wonder if we're missing something bigger here. ISTM we're potentially missing CSEs in memory addresses as well as forward propagation opportunities in MEM_REF expressions. I strongly suspect DOM doesn't look inside a MEM_REF to see if the computed address is lying around in hash table. Similarly I wouldn't be surprised if forward propagation misses the opportunity to eliminate instructions that compute addresses by propagating the constant offset part into a MEM_REF.