https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105930
--- Comment #15 from Richard Biener <rguenth at gcc dot gnu.org> --- So we feed DImode rotates into RA which constrains register allocation enough to require spills (all 4 DImode vals are live across the kernel, not even -fschedule-insn can do anything here). I wonder if it ever makes sense to not split wide ops before reload.