[Bug tree-optimization/125750] New: Excessive vector unrolling leads to excessive spilling

[ktkachov at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125750

            Bug ID: 125750
           Summary: Excessive vector unrolling leads to excessive spilling
           Product: gcc
           Version: 16.0
            Status: UNCONFIRMED
          Keywords: aarch64-sve
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ktkachov at gcc dot gnu.org
  Target Milestone: ---
            Target: aarch64

Created attachment 64712
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=64712&action=edit
Reproducer

This may be aarch64-specific but not sure. Once PR125672 is addressed, the rest
of the astcenc (SPEC2026) gap with LLVM on aarch64 seems to be to excessive
unrolling in one kernel when vectorising. I'm attaching what I hope is a
reasonable reproducer where GCC generates much more code at -O3
-march=armv8.2-a+sve -ffast-math .