https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91320
Vladimir Makarov <vmakarov at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |vmakarov at gcc dot gnu.org
--- Comment #5 from Vladimir Makarov <vmakarov at gcc dot gnu.org> ---
(In reply to Jeffrey A. Law from comment #3)
>
> One might consider raising the priority of allocnos which are more sensitive
> to what register they're allocated to. But that's a risky proposition and
> would have to be extensively benchmarked to determine its impact, assuming
> we could even come up proposed heuristics in this space.
>
>
> Anyway, I'm cc-ing Vlad as our register allocation expert so that he can
> take a looksie. I'll also attach a usable testcase.
Jeff, thank you for the test and the PR analysis.
I've been working on analogous PRs (e.g. PR91333).
I've tried many new heuristics, mostly on sorting colorable allocno stack
(taking allocno live ranges, allocno (or thread) hard preferences, conflicting
(or thread) allocnos hard reg preferences for any allocnos, allocnos living in
BB, etc). The problem is that the heuristics worsen SPEC benchmarks in many
cases.
But I think I found simple heuristics which are improving the PRs code and do
not change SPEC. I hope the patch will be ready on this week or the next week.
