https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128206
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Last reconfirmed| |2026-10-27
Blocks| |53947
Status|UNCONFIRMED |NEW
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
The issue is that the minimum vectorization factor we think we can use is 16:
t.c:8:23: note: Loop contains only SLP stmts
t.c:8:23: note: vectorization factor = 16
t.c:8:23: missed: bad data dependence.
t.c:8:23: note: ***** Analysis failed with vector mode V16QI
and with V8QI we decide to not to vectorize because we'd end up with V1DImode
(aka scalar) (but V8QI).
This is somewhat caused by our idea of forcing the same vector size rather
than forcing a VF of 8 and adjusting vector types.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53947
[Bug 53947] [meta-bug] vectorizer missed-optimizations
