https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128408
Josef Melcr <jmelcr at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Resolution|--- |INVALID
Status|UNCONFIRMED |RESOLVED
--- Comment #2 from Josef Melcr <jmelcr at gcc dot gnu.org> ---
(In reply to Hongtao Liu from comment #1)
> Could it be related to code alignment?
>
> ------------cut from [1] ---------------
> Note for lbm on AMD Zen 5 with GCC 16
> We discovered 782.lbm_r may have 2x performance difference on AMD Zen 5
> depending on PC alignment. When comparing the performance of AMD Zen 5 with
> other CPUs using GCC-16 with -O3 -march=x86-64-v4. It's recommand to use
> patch -p1 <patches/optional/782-lbm-zen5-prefetch-conflict.patch to avoid
> the prefetch conflict issue which can cause significant performance
> degradation. For other compilers or different flags on Zen 5, the align
> issue may be different and need further manual tuning.
> ---------------cut end--------------------------------
>
> [1]
> https://github.com/OpenXiangShan/CPU2026LiteWrapper#note-for-lbm-on-amd-zen-
> 5-with-gcc-16
The issue does get resolved with the linked patch. Closing this as invalid as
the commit is not at fault. Thanks :)
