[Bug optimization/14846] New: [tree-ssa] don't use a shift in A & CST_POWER_OF_2 == 0 until very late in tree-ssa optimizations

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

I wonder if the shift in (a >> 3) & 1 == 0 does any good for tree-ssa.

Consider:

void bar (void);

void
foo (unsigned int a)
{
  if ((a & 8) != 0)
    bar ();
}

I get:

  T.0_2 = a_1 >> 3;
  T.1_3 = (int)T.0_2;
  T.2_4 = T.1_3 & 1;
  T.3_5 = (_Bool)T.2_4;
  if (T.3_5) goto <L0>; else goto <L1>;

Even if Andrew Pinski removes the _Bool and int cast,

  T.0_2 = a_1 >> 3;
  T.2_4 = T.0_2 & 1;
  if (T.2_4 != 0) goto <L0>; else goto <L1>;

Without the shift,

  T.2_4 = a_1 & 1;
  if (T.2_4 != 0) goto <L0>; else goto <L1>;

Without the shift, it's a lot easier to combine two consecutive bit tests like

  if (a & 1) goto there;
  if (a & 4) goto there;

into

  if (a & 5) goto there;

Removing the shift should also make it easier to solve PR 14752
because we don't have to look as many statements to do the optimization.

If we need the shift for some machine-specific reasons,
we should introduce it very late in tree-ssa optimizations or at expand time.

-- 
           Summary: [tree-ssa] don't use a shift in A & CST_POWER_OF_2 == 0
                    until very late in tree-ssa optimizations
           Product: gcc
           Version: tree-ssa
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P2
         Component: optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        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=14846