https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100866
--- Comment #12 from Segher Boessenkool <segher at gcc dot gnu.org> --- (In reply to Bill Schmidt from comment #11) > Segher, does this fit naturally in combine? This is just constant folding, combine won't have much to do with it. It is always better (namely, lower latency) to use one vector permute than to have multiple dependent permutation-class instructions. Combine will automatically pick this up when it gets the chance. Does it here though, or are there still some unspecs here that make all this non-clear?
