https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119044
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Resolution|--- |WONTFIX
Status|ASSIGNED |RESOLVED
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
I'll note there are larger changes in the past, I'll also note 436.cactusADM is
known bad with respect to spilling and prone to changes with minor changes to
optimization. It also tends to be noisy. I get, on a Zen4 machine, with
-O2 -march=znver3
436.cactusADM 11950 68.4 175 * 11950 65.6 182 S
436.cactusADM 11950 73.1 163 S 11950 58.4 205 S
436.cactusADM 11950 64.5 185 S 11950 63.4 188 *
comparing r15-7726 with the offending rev reverted to r15-7726 which shows
the change improves there.
So I'm not sure if there's anything to fix for the revision in question.
For -O2 generic I see
436.cactusADM 11950 94.8 126 S 11950 98.6 121 S
436.cactusADM 11950 100 119 S 11950 102 117 S
436.cactusADM 11950 99.9 120 * 11950 99.9 120 *
so maybe a small regression.
I agree to WONTFIX.
