[Bug regression/68695] New: Performance regression related to ssa patch / ifcvt

["vogt at linux dot vnet.ibm.com" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68695

            Bug ID: 68695
           Summary: Performance regression related to ssa patch / ifcvt
           Product: gcc
           Version: 6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: regression
          Assignee: unassigned at gcc dot gnu.org
          Reporter: vogt at linux dot vnet.ibm.com
                CC: aoliva at gcc dot gnu.org, krebbel at gcc dot gnu.org
  Target Milestone: ---
              Host: s390x
            Target: s390x
             Build: s390x

This commit introduced a performance regression on s390/s390x:
https://gcc.gnu.org/viewcvs/gcc?view=revision&revision=224262
See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64164#c35

With the change committed the ifcvt-4.c test program (current HEAD) results in
much less efficient assembly language code.

-- snip --
int
foo (int x, int y, int a)
{
  int i = x;
  int j = y;
  /* Try to make taking the branch likely.  */
  __builtin_expect (x > y, 1);
  if (x > y)
    {
      i = a;
      j = i;
    }
  return i * j;
}
-- snip --

Before change:
$ gcc -O2 -S -m64 ifcvt-4.c
-- snip --
foo:
.LFB0:
        .cfi_startproc
        cr      %r2,%r3
        jle     .L2
        lgr     %r3,%r4
        lgr     %r2,%r4
.L2:
        msr     %r2,%r3
        lgfr    %r2,%r2
        br      %r14
-- snip --

After change:
$ gcc -O2 -S -m64 ifcvt-4.c
-- snip --
foo:
.LFB0:
        cr      %r2,%r3
        jle     .L3
        lr      %r3,%r4
        lr      %r2,%r4
        msr     %r2,%r3
        lgfr    %r2,%r2
        br      %r14
.L3:
        lr      %r4,%r2
        lr      %r2,%r4
        msr     %r2,%r3
        lgfr    %r2,%r2
        br      %r14
-- snip --

I can provide specific rtl dumps or other information if you tell what you
need.