Re: [PATCH] Fold (a > 0 ? 1.0 : -1.0) into copysign (1.0, a) and a * copysign (1.0, a) into abs(a)

[Uros Bizjak <ubizjak at gmail dot com>]

On Wed, Jun 28, 2017 at 9:37 AM, Jakub Jelinek <jakub@redhat.com> wrote:
> On Tue, Jun 27, 2017 at 10:52:47AM -0700, Andrew Pinski wrote:
>> On Tue, Jun 27, 2017 at 7:56 AM, Richard Biener
>> <richard.guenther@gmail.com> wrote:
>> > On June 27, 2017 4:52:28 PM GMT+02:00, Tamar Christina <Tamar.Christina@arm.com> wrote:
>> >>> >> +(for cmp (gt ge lt le)
>> >>> >> +     outp (convert convert negate negate)
>> >>> >> +     outn (negate negate convert convert)
>> >>> >> + /* Transform (X > 0.0 ? 1.0 : -1.0) into copysign(1, X). */
>> >>> >> + /* Transform (X >= 0.0 ? 1.0 : -1.0) into copysign(1, X). */
>> >>> >> + /* Transform (X < 0.0 ? 1.0 : -1.0) into copysign(1,-X). */
>> >>> >> + /* Transform (X <= 0.0 ? 1.0 : -1.0) into copysign(1,-X). */
>> >>> >> +(simplify
>> >>> >> +  (cond (cmp @0 real_zerop) real_onep real_minus_onep)
>> >>> >> +  (if (!HONOR_NANS (type) && !HONOR_SIGNED_ZEROS (type)
>> >>> >> +       && types_match (type, TREE_TYPE (@0)))
>> >>> >> +   (switch
>> >>> >> +    (if (types_match (type, float_type_node))
>> >>> >> +     (BUILT_IN_COPYSIGNF { build_one_cst (type); } (outp @0)))
>> >>> >> +    (if (types_match (type, double_type_node))
>> >>> >> +     (BUILT_IN_COPYSIGN { build_one_cst (type); } (outp @0)))
>> >>> >> +    (if (types_match (type, long_double_type_node))
>> >>> >> +     (BUILT_IN_COPYSIGNL { build_one_cst (type); } (outp @0))))))
>
> The patch regressed the gcc.target/i386/cmov7.c testcase.
> From the description in the testcase it was testing the combiner, so I've
> transformed it into something that also tests the combiner the same way,
> ok for trunk?  That said, for copysign if we match it we don't emit a fcmov
> while it might be a good idea.
>
> 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
>
>         * gcc.target/i386/cmov7.c (sgn): Renamed to ...
>         (foo): ... this.  Change constants such that it isn't matched
>         as __builtin_copysign, yet tests the combiner the same.

OK.

Thanks,
Uros.

> --- gcc/testsuite/gcc.target/i386/cmov7.c.jj    2016-05-22 12:20:23.000000000 +0200
> +++ gcc/testsuite/gcc.target/i386/cmov7.c       2017-06-28 09:20:24.000000000 +0200
> @@ -10,7 +10,7 @@
>     (set (reg:DF) (float_extend:DF (mem:SF (symbol_ref...)))).  */
>
>  double
> -sgn (double __x)
> +foo (double __x)
>  {
> -  return __x >= 0.0 ? 1.0 : -1.0;
> +  return __x >= 1.0 ? 0.0 : -1.0;
>  }
>
>
>         Jakub