https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82438
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jakub at gcc dot gnu.org
--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
That is called vectorization and gcc performs it only with -O3 or if
-ftree-vectorize is used, so in this case you need to use -O2 -ftree-vectorize.
Then gcc handles 16 original iterations in every loop iterations (for SSE2, or
32 for AVX2, or 64 for AVX512BW).
