[Bug tree-optimization/46726] New: x*x has different cost than pow(x,2) with -ffast-math

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46726

           Summary: x*x has different cost than pow(x,2) with -ffast-math
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassig...@gcc.gnu.org
        ReportedBy: rgue...@gcc.gnu.org
                CC: hubi...@gcc.gnu.org


double 
foo(double x, double y)
{
  return x*x + y;
}

is inlined at -O2 but not at -O2 -ffast-math (which folds x*x to pow(x, 2)).
MUL_EXPR has const 1 but __builtin_pow isn't is_inexpensive_builtin.

We probably should special-case pow(, 2) and pow(, 0.5) and sqrt where
target support is available.