[Bug tree-optimization/21001] New: VRP is weak when the tested variable in a COND_EXPR is used in the COND_EXPR.

["kazu at cs dot umass dot edu" <gcc-bugzilla at gcc dot gnu dot org>]

Consider:

int
foo (int a)
{
  int b = a != 0;
  if (b)
    if (a != 0)
      return 1;
  return 0;
}

With "-O2 -ftree-no-dominator-opts", VRP generates:

foo (a)
{
  int b;
  int D.1153;

<bb 0>:
  b_3 = a_2 != 0;
  if (b_3 != 0) goto <L0>; else goto <L2>;

<L0>:;
  if (a_2 != 0) goto <L1>; else goto <L2>;

<L1>:;
  D.1153_6 = 1;
  goto <bb 4> (<L4>);

<L2>:;
  D.1153_5 = 0;

  # D.1153_1 = PHI <D.1153_6(2), D.1153_5(3)>;
<L4>:;
  return D.1153_1;

}

Note that the second "if" statement is always true,
but VRP does not notice that.

Forward-propagating "a_2 != 0" into the first "if" statement will do
the desired optimization.

-- 
           Summary: VRP is weak when the tested variable in a COND_EXPR is
                    used in the COND_EXPR.
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: enhancement
          Priority: P2
         Component: tree-optimization
        AssignedTo: kazu at cs dot umass dot edu
        ReportedBy: kazu at cs dot umass dot edu
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21001