[Bug tree-optimization/66739] [6 regression] FAIL: gcc.target/aarch64/subs.c scan-assembler subs\tw[0-9]

["rguenth at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66739

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2015-07-02
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
   Target Milestone|---                         |6.0
     Ever confirmed|0                           |1

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
This is

  x = a - b;
  if (x != 0)

vs.

  if (a != b)

which we now more aggressively produce (the choice is not obvious and
we are missing the reverse transform).

The usual kind of action is to stick a single_use guard on the minus for

/* Transform comparisons of the form X - Y CMP 0 to X CMP Y.
   ??? The transformation is valid for the other operators if overflow
   is undefined for the type, but performing it here badly interacts
   with the transformation in fold_cond_expr_with_comparison which
   attempts to synthetize ABS_EXPR.  */
(for cmp (eq ne)
 (simplify
  (cmp (minus @0 @1) integer_zerop)
  (cmp @0 @1)))

but I really don't like that solution (it will cause SCCVN regressions).

value-numbering can also perform the reverse transform (though late
forwprop will kill that again).  I suppose we should look into
finding a more general solution for the forwprop issues from inside
forwprop.