[Bug target/84482] New: 437.leslie3d regresses on Haswell and SandyBridge at -O3 and -Ofast with generic march and tuning

[jamborm at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84482

            Bug ID: 84482
           Summary: 437.leslie3d regresses on Haswell and SandyBridge at
                    -O3 and -Ofast with generic march and tuning
           Product: gcc
           Version: 8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jamborm at gcc dot gnu.org
  Target Milestone: ---

It has been brought to my attention that the run-time of benchmark
437.leslie3d from SPEC 2006 regresses on Intel Haswell by about 8%
when compiled with -O3 and generic march and tuning using trunk,
compared to gcc 7.2.

I tried on my desktop Sandy Bridge system and actually observed a 13%
regression at -Ofast (and generic march and tuning).

For the record, this does not seem to be happening on Zen CPUs.