public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [PATCH] Fix PR 110554: vec lowering introduces scalar signed-boolean:32 comparisons
@ 2023-07-05 17:01 Andrew Pinski
  2023-07-06  6:08 ` Richard Biener
  0 siblings, 1 reply; 2+ messages in thread
From: Andrew Pinski @ 2023-07-05 17:01 UTC (permalink / raw)
  To: gcc-patches; +Cc: Andrew Pinski

So the problem is vector generic decided to do comparisons in signed-boolean:32
types but the rest of the middle-end was not ready for that. Since we are building
the comparison which will feed into a cond_expr here, using boolean_type_node is
better and also correct. The rest of the compiler thinks the ranges for
comparison is always [0,1] too.

Note this code does not currently lowers bigger vector sizes into smaller
vector sizes so using boolean_type_node here is better.

OK? bootstrapped and tested on x86_64-linux-gnu with no regressions.

gcc/ChangeLog:

	PR middle-end/110554
	* tree-vect-generic.cc (expand_vector_condition): For comparisons,
	just build using boolean_type_node instead of the cond_type.
	For non-comparisons/non-scalar-bitmask, build a ` != 0` gimple
	that will feed into the COND_EXPR.
---
 gcc/tree-vect-generic.cc | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/gcc/tree-vect-generic.cc b/gcc/tree-vect-generic.cc
index df04a0db68d..a7e6cb87a5e 100644
--- a/gcc/tree-vect-generic.cc
+++ b/gcc/tree-vect-generic.cc
@@ -1121,7 +1121,7 @@ expand_vector_condition (gimple_stmt_iterator *gsi, bitmap dce_ssa_names)
 				       comp_width, comp_index);
 	  tree aa2 = tree_vec_extract (gsi, comp_inner_type, a2,
 				       comp_width, comp_index);
-	  aa = gimplify_build2 (gsi, code, cond_type, aa1, aa2);
+	  aa = gimplify_build2 (gsi, code, boolean_type_node, aa1, aa2);
 	}
       else if (a_is_scalar_bitmask)
 	{
@@ -1132,7 +1132,11 @@ expand_vector_condition (gimple_stmt_iterator *gsi, bitmap dce_ssa_names)
 				build_zero_cst (TREE_TYPE (a)));
 	}
       else
-	aa = tree_vec_extract (gsi, cond_type, a, comp_width, comp_index);
+	{
+	  result = tree_vec_extract (gsi, cond_type, a, comp_width, comp_index);
+	  aa = gimplify_build2 (gsi, NE_EXPR, boolean_type_node, result,
+				build_zero_cst (cond_type));
+	}
       result = gimplify_build3 (gsi, COND_EXPR, inner_type, aa, bb, cc);
       if (!CONSTANT_CLASS_P (result))
 	constant_p = false;
-- 
2.31.1


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH] Fix PR 110554: vec lowering introduces scalar signed-boolean:32 comparisons
  2023-07-05 17:01 [PATCH] Fix PR 110554: vec lowering introduces scalar signed-boolean:32 comparisons Andrew Pinski
@ 2023-07-06  6:08 ` Richard Biener
  0 siblings, 0 replies; 2+ messages in thread
From: Richard Biener @ 2023-07-06  6:08 UTC (permalink / raw)
  To: Andrew Pinski; +Cc: gcc-patches

On Wed, Jul 5, 2023 at 7:02 PM Andrew Pinski via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> So the problem is vector generic decided to do comparisons in signed-boolean:32
> types but the rest of the middle-end was not ready for that. Since we are building
> the comparison which will feed into a cond_expr here, using boolean_type_node is
> better and also correct. The rest of the compiler thinks the ranges for
> comparison is always [0,1] too.
>
> Note this code does not currently lowers bigger vector sizes into smaller
> vector sizes so using boolean_type_node here is better.
>
> OK? bootstrapped and tested on x86_64-linux-gnu with no regressions.

OK.

> gcc/ChangeLog:
>
>         PR middle-end/110554
>         * tree-vect-generic.cc (expand_vector_condition): For comparisons,
>         just build using boolean_type_node instead of the cond_type.
>         For non-comparisons/non-scalar-bitmask, build a ` != 0` gimple
>         that will feed into the COND_EXPR.
> ---
>  gcc/tree-vect-generic.cc | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/gcc/tree-vect-generic.cc b/gcc/tree-vect-generic.cc
> index df04a0db68d..a7e6cb87a5e 100644
> --- a/gcc/tree-vect-generic.cc
> +++ b/gcc/tree-vect-generic.cc
> @@ -1121,7 +1121,7 @@ expand_vector_condition (gimple_stmt_iterator *gsi, bitmap dce_ssa_names)
>                                        comp_width, comp_index);
>           tree aa2 = tree_vec_extract (gsi, comp_inner_type, a2,
>                                        comp_width, comp_index);
> -         aa = gimplify_build2 (gsi, code, cond_type, aa1, aa2);
> +         aa = gimplify_build2 (gsi, code, boolean_type_node, aa1, aa2);
>         }
>        else if (a_is_scalar_bitmask)
>         {
> @@ -1132,7 +1132,11 @@ expand_vector_condition (gimple_stmt_iterator *gsi, bitmap dce_ssa_names)
>                                 build_zero_cst (TREE_TYPE (a)));
>         }
>        else
> -       aa = tree_vec_extract (gsi, cond_type, a, comp_width, comp_index);
> +       {
> +         result = tree_vec_extract (gsi, cond_type, a, comp_width, comp_index);
> +         aa = gimplify_build2 (gsi, NE_EXPR, boolean_type_node, result,
> +                               build_zero_cst (cond_type));
> +       }
>        result = gimplify_build3 (gsi, COND_EXPR, inner_type, aa, bb, cc);
>        if (!CONSTANT_CLASS_P (result))
>         constant_p = false;
> --
> 2.31.1
>

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2023-07-06  6:08 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-07-05 17:01 [PATCH] Fix PR 110554: vec lowering introduces scalar signed-boolean:32 comparisons Andrew Pinski
2023-07-06  6:08 ` Richard Biener

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).