[Bug tree-optimization/70046] New: [6 Regression] 410.bwaves regression on Haswell

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70046

            Bug ID: 70046
           Summary: [6 Regression] 410.bwaves regression on Haswell
           Product: gcc
           Version: 6.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: rguenth at gcc dot gnu.org
                CC: amker at gcc dot gnu.org
  Target Milestone: ---

http://gcc.opensuse.org/SPEC/CFP/sb-czerny-head-64-2006/index.html

shows a 10% performance regression for BASE flags, -Ofast -march=native (but
not PEAK which adds -flto)

Last known good rev. is r230634 first known bad is r230689.

Past bugreports about bwaves suggest IVOPTs sensitivity and thus

r230647 (fix for PR52272) is my primary suspect in that revision range
(which incidentially was supposed to fix a bwaves regression).  There is
also the old PR50480 also hinting at IVOPTs (bug has seen no updates).

No negative effect seen on AMD bdver2.