https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88440
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |missed-optimization
Status|UNCONFIRMED |NEW
Last reconfirmed| |2018-12-11
CC| |mliska at suse dot cz,
| |rguenth at gcc dot gnu.org
Ever confirmed|0 |1
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
I think distributing patterns is reasonable for -O[2s]. I think we kept it at
-O3 because of compile-time concerns (dependence analysis is quadratic).
But then we still don't do basic vectorization at -O[2s] either... (same
compile-time for not too much gain issue).
So if somebody is willing to do some compile-time / effect numbers
(bootstrap / SPEC?) then we can consider enabling loop-distribute-patterns
for -O[2s] for GCC 10.
