[PATCH] Fix up foperator_abs::op1_range [PR107569]

[PATCH] Fix up foperator_abs::op1_range [PR107569]

[Jakub Jelinek]

Hi!

foperator_abs::op1_range works except for the NaN handling,
from:
[frange] double [-Inf, 1.79769313486231570814527423731704356798070567525844996599e+308 (0x0.fffffffffffff8p+1024)]
lhs it computes r
[frange] double [-1.79769313486231570814527423731704356798070567525844996599e+308 (-0x0.fffffffffffff8p+1024), 1.79769313486231570814527423731704356798070567525844996599e+308 (0x0.fffffffffffff8p+1024)] +-NAN
which is correct except for the +-NAN part.
For r before the final step it makes sure to add -NAN if there is +NAN
in the lhs range, but the final r.union_ makes it unconditional +-NAN,
because the frange ctor sets +-NAN.
So, I think we need to clear it (or have some set variant which
says not to set NAN).

This patch fixes that, but isn't enough to fix the PR, something in
the assumptions handling is still broken (and the PR has other parts).

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

2022-11-09  Jakub Jelinek  <jakub@redhat.com>

	PR tree-optimization/107569
	* range-op-float.cc (foperator_abs::op1_range): Clear NaNs
	from the negatives frange before unioning it into r.

--- gcc/range-op-float.cc.jj	2022-11-06 11:56:27.138137781 +0100
+++ gcc/range-op-float.cc	2022-11-08 18:13:18.026974667 +0100
@@ -1280,9 +1280,10 @@ foperator_abs::op1_range (frange &r, tre
     return true;
   // Then add the negative of each pair:
   // ABS(op1) = [5,20] would yield op1 => [-20,-5][5,20].
-  r.union_ (frange (type,
-		    real_value_negate (&positives.upper_bound ()),
-		    real_value_negate (&positives.lower_bound ())));
+  frange negatives (type, real_value_negate (&positives.upper_bound ()),
+		    real_value_negate (&positives.lower_bound ()));
+  negatives.clear_nan ();
+  r.union_ (negatives);
   return true;
 }
 

	Jakub

Re: [PATCH] Fix up foperator_abs::op1_range [PR107569]

[Aldy Hernandez]

On Wed, Nov 9, 2022 at 11:55 AM Jakub Jelinek <jakub@redhat.com> wrote:
>
> Hi!
>
> foperator_abs::op1_range works except for the NaN handling,
> from:
> [frange] double [-Inf, 1.79769313486231570814527423731704356798070567525844996599e+308 (0x0.fffffffffffff8p+1024)]
> lhs it computes r
> [frange] double [-1.79769313486231570814527423731704356798070567525844996599e+308 (-0x0.fffffffffffff8p+1024), 1.79769313486231570814527423731704356798070567525844996599e+308 (0x0.fffffffffffff8p+1024)] +-NAN
> which is correct except for the +-NAN part.
> For r before the final step it makes sure to add -NAN if there is +NAN
> in the lhs range, but the final r.union_ makes it unconditional +-NAN,
> because the frange ctor sets +-NAN.
> So, I think we need to clear it (or have some set variant which
> says not to set NAN).
>
> This patch fixes that, but isn't enough to fix the PR, something in
> the assumptions handling is still broken (and the PR has other parts).

LGTM.

Sorry I haven't gotten to the PR yet.  I'll poke at it later today.

Thanks.
Aldy

>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
>
> 2022-11-09  Jakub Jelinek  <jakub@redhat.com>
>
>         PR tree-optimization/107569
>         * range-op-float.cc (foperator_abs::op1_range): Clear NaNs
>         from the negatives frange before unioning it into r.
>
> --- gcc/range-op-float.cc.jj    2022-11-06 11:56:27.138137781 +0100
> +++ gcc/range-op-float.cc       2022-11-08 18:13:18.026974667 +0100
> @@ -1280,9 +1280,10 @@ foperator_abs::op1_range (frange &r, tre
>      return true;
>    // Then add the negative of each pair:
>    // ABS(op1) = [5,20] would yield op1 => [-20,-5][5,20].
> -  r.union_ (frange (type,
> -                   real_value_negate (&positives.upper_bound ()),
> -                   real_value_negate (&positives.lower_bound ())));
> +  frange negatives (type, real_value_negate (&positives.upper_bound ()),
> +                   real_value_negate (&positives.lower_bound ()));
> +  negatives.clear_nan ();
> +  r.union_ (negatives);
>    return true;
>  }
>
>
>         Jakub
>