https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121662
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |rguenth at gcc dot gnu.org
Keywords| |missed-optimization
Component|target |tree-optimization
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
This is done to avoid all-masked stores which can repeatedly trigger fault
suppression, causing 1000x slowdown. For this case consider *a mapped
read-only
and all zero and time the loop with and without the branch (or make the
testcase to check a different object for the flag to get the unmapped
write-object case).
So, it works as designed, see tree-vect-loop.cc:optimize_mask_stores
