[Bug tree-optimization/65207] New: [5 regression] performance: tree level loop optimization does not peel SpecCPU href loop

[krebbel at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65207

            Bug ID: 65207
           Summary: [5 regression] performance: tree level loop
                    optimization does not peel SpecCPU href loop
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: krebbel at gcc dot gnu.org

mv-search.c
SetupFastFullPelSearch

   for (blky = 0; blky < 4; blky++)
    {
      LineSadBlk0 = LineSadBlk1 = LineSadBlk2 = LineSadBlk3 = 0;
      for (y = 0; y < 4; y++)
      {
        refptr = PelYline_11 (ref_pic, abs_y++, abs_x, img_height, img_width);
...

After the loop peeling has been removed from RTL level this loop does not get
peeled completely as before on S/390 which results in a performance regression
(-5%).

https://gcc.gnu.org/ml/gcc-patches/2014-10/msg01211.html

The tree level loop peeling refuses to peel the loop because it contains a
function call.