https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100045
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Target| |x86_64-*-* i?86-*-*
Component|target |middle-end
Status|UNCONFIRMED |NEW
Keywords| |missed-optimization
Last reconfirmed| |2021-04-12
Ever confirmed|0 |1
Version|unknown |11.0
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed. This would be RTL expansion-time at the moment but we can of course
do such optimizations elsewhere (we have the recip pass which could be seen
doing similar things). Rather than relying on invariant motion I'd say we
want to do LCM placement of the invariant parts though.
And indeed costing is target dependent.
