public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [PATCH] Ensure matching types in debug stmts created by tree-sra.c (PR tree-optimization/55920)
@ 2013-01-15 15:39 Jakub Jelinek
  2013-01-15 16:06 ` Richard Biener
  0 siblings, 1 reply; 2+ messages in thread
From: Jakub Jelinek @ 2013-01-15 15:39 UTC (permalink / raw)
  To: Richard Biener, Martin Jambor; +Cc: gcc-patches

Hi!

This patch fixes ICE on the attached testcase on ARM, where we ended up
creating invalid debug stmt, which had SImode var on the lhs, but BLKmode
MEM_REF on the rhs.  Fixed by matching (almost, with the build_ref_for_model
-> build_debug_ref_for_model change) what we do for normal stmt
modifications earlier.

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

2013-01-15  Jakub Jelinek  <jakub@redhat.com>

	PR tree-optimization/55920
	* tree-sra.c (sra_modify_assign): If for lacc->grp_to_be_debug_replaced
	there is non-useless type conversion needed from debug rhs to lhs,
	use build_debug_ref_for_model and/or VIEW_CONVERT_EXPR.

	* gcc.c-torture/compile/pr55920.c: New test.

--- gcc/tree-sra.c.jj	2013-01-11 09:02:55.000000000 +0100
+++ gcc/tree-sra.c	2013-01-15 12:52:01.161324202 +0100
@@ -3108,8 +3108,20 @@ sra_modify_assign (gimple *stmt, gimple_
 
   if (lacc && lacc->grp_to_be_debug_replaced)
     {
-      gimple ds = gimple_build_debug_bind (get_access_replacement (lacc),
-					   unshare_expr (rhs), *stmt);
+      tree debug_lhs = get_access_replacement (lacc);
+      tree debug_rhs = unshare_expr (rhs);
+      if (!useless_type_conversion_p (TREE_TYPE (debug_lhs),
+				      TREE_TYPE (debug_rhs)))
+	{
+	  if (AGGREGATE_TYPE_P (TREE_TYPE (debug_rhs))
+	      && !contains_vce_or_bfcref_p (debug_rhs))
+	    debug_rhs = build_debug_ref_for_model (loc, debug_rhs, 0, lacc);
+	  if (!useless_type_conversion_p (TREE_TYPE (debug_lhs),
+					  TREE_TYPE (debug_rhs)))
+	    debug_rhs = fold_build1_loc (loc, VIEW_CONVERT_EXPR,
+					 TREE_TYPE (debug_lhs), debug_rhs);
+	}
+      gimple ds = gimple_build_debug_bind (debug_lhs, debug_rhs, *stmt);
       gsi_insert_before (gsi, ds, GSI_SAME_STMT);
     }
 
--- gcc/testsuite/gcc.c-torture/compile/pr55920.c.jj	2013-01-15 12:58:01.782125163 +0100
+++ gcc/testsuite/gcc.c-torture/compile/pr55920.c	2013-01-15 12:57:23.000000000 +0100
@@ -0,0 +1,16 @@
+/* PR tree-optimization/55920 */
+
+struct A { unsigned a; } __attribute__((packed));
+struct B { int b; unsigned char c[16]; };
+void bar (struct A);
+
+void
+foo (struct B *x)
+{
+  struct A a;
+  if (x->b)
+    __builtin_memcpy (&a, x->c, sizeof a);
+  else
+    a.a = 0;
+  bar (a);
+}

	Jakub

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

* Re: [PATCH] Ensure matching types in debug stmts created by tree-sra.c (PR tree-optimization/55920)
  2013-01-15 15:39 [PATCH] Ensure matching types in debug stmts created by tree-sra.c (PR tree-optimization/55920) Jakub Jelinek
@ 2013-01-15 16:06 ` Richard Biener
  0 siblings, 0 replies; 2+ messages in thread
From: Richard Biener @ 2013-01-15 16:06 UTC (permalink / raw)
  To: Jakub Jelinek; +Cc: Martin Jambor, gcc-patches

On Tue, 15 Jan 2013, Jakub Jelinek wrote:

> Hi!
> 
> This patch fixes ICE on the attached testcase on ARM, where we ended up
> creating invalid debug stmt, which had SImode var on the lhs, but BLKmode
> MEM_REF on the rhs.  Fixed by matching (almost, with the build_ref_for_model
> -> build_debug_ref_for_model change) what we do for normal stmt
> modifications earlier.
> 
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

Works for me.

Thanks,
Richard.

> 2013-01-15  Jakub Jelinek  <jakub@redhat.com>
> 
> 	PR tree-optimization/55920
> 	* tree-sra.c (sra_modify_assign): If for lacc->grp_to_be_debug_replaced
> 	there is non-useless type conversion needed from debug rhs to lhs,
> 	use build_debug_ref_for_model and/or VIEW_CONVERT_EXPR.
> 
> 	* gcc.c-torture/compile/pr55920.c: New test.
> 
> --- gcc/tree-sra.c.jj	2013-01-11 09:02:55.000000000 +0100
> +++ gcc/tree-sra.c	2013-01-15 12:52:01.161324202 +0100
> @@ -3108,8 +3108,20 @@ sra_modify_assign (gimple *stmt, gimple_
>  
>    if (lacc && lacc->grp_to_be_debug_replaced)
>      {
> -      gimple ds = gimple_build_debug_bind (get_access_replacement (lacc),
> -					   unshare_expr (rhs), *stmt);
> +      tree debug_lhs = get_access_replacement (lacc);
> +      tree debug_rhs = unshare_expr (rhs);
> +      if (!useless_type_conversion_p (TREE_TYPE (debug_lhs),
> +				      TREE_TYPE (debug_rhs)))
> +	{
> +	  if (AGGREGATE_TYPE_P (TREE_TYPE (debug_rhs))
> +	      && !contains_vce_or_bfcref_p (debug_rhs))
> +	    debug_rhs = build_debug_ref_for_model (loc, debug_rhs, 0, lacc);
> +	  if (!useless_type_conversion_p (TREE_TYPE (debug_lhs),
> +					  TREE_TYPE (debug_rhs)))
> +	    debug_rhs = fold_build1_loc (loc, VIEW_CONVERT_EXPR,
> +					 TREE_TYPE (debug_lhs), debug_rhs);
> +	}
> +      gimple ds = gimple_build_debug_bind (debug_lhs, debug_rhs, *stmt);
>        gsi_insert_before (gsi, ds, GSI_SAME_STMT);
>      }
>  
> --- gcc/testsuite/gcc.c-torture/compile/pr55920.c.jj	2013-01-15 12:58:01.782125163 +0100
> +++ gcc/testsuite/gcc.c-torture/compile/pr55920.c	2013-01-15 12:57:23.000000000 +0100
> @@ -0,0 +1,16 @@
> +/* PR tree-optimization/55920 */
> +
> +struct A { unsigned a; } __attribute__((packed));
> +struct B { int b; unsigned char c[16]; };
> +void bar (struct A);
> +
> +void
> +foo (struct B *x)
> +{
> +  struct A a;
> +  if (x->b)
> +    __builtin_memcpy (&a, x->c, sizeof a);
> +  else
> +    a.a = 0;
> +  bar (a);
> +}
> 
> 	Jakub
> 
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE / SUSE Labs
SUSE LINUX Products GmbH - Nuernberg - AG Nuernberg - HRB 16746
GF: Jeff Hawn, Jennifer Guild, Felix Imend

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

end of thread, other threads:[~2013-01-15 16:06 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-01-15 15:39 [PATCH] Ensure matching types in debug stmts created by tree-sra.c (PR tree-optimization/55920) Jakub Jelinek
2013-01-15 16:06 ` 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).