https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113359
--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> --- Created attachment 57061 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57061&action=edit SloppyCRCMap.ii.xz When using current GCC trunk instead of current GCC 13 branch (CrtStuff.i.xz is the same between both), the problematic optimization doesn't happen, both before SRA and after it the code has MEM[(struct pair *)_5 + 32B] = MEM[(const struct pair &)__args#0_4(D)];