https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70421
--- Comment #6 from Zdenek Sojka <zsojka at seznam dot cz> --- (In reply to Jakub Jelinek from comment #4) > (In reply to Zdenek Sojka from comment #3) > > Is the fact that the useless cast in: > > x ^= ((v16si)v)[u[0]]; > > changes the generated code (makes it significantly worse at all opt levels) > > worth creating a PR? > > I think so. But it is likely not a regression, so probably has to wait till > GCC 7. I have created PR70434 for that. It is a regression from 4.9, where (for all tested targets) the code is the same (faster) no matter whether the cast is used.