public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [committed] nested, openmp: Wrap OMP_CLAUSE_*_GIMPLE_SEQ into GIMPLE_BIND for declare_vars [PR108435]
@ 2023-02-02  9:00 Jakub Jelinek
  0 siblings, 0 replies; only message in thread
From: Jakub Jelinek @ 2023-02-02  9:00 UTC (permalink / raw)
  To: gcc-patches; +Cc: Tobias Burnus

Hi!

When gimplifying OMP_CLAUSE_{LASTPRIVATE,LINEAR}_STMT, we wrap it always
into a GIMPLE_BIND, but when putting statements directly into
OMP_CLAUSE_{LASTPRIVATE,LINEAR}_GIMPLE_SEQ, we do it only if needed (there
are any temporaries that need to be declared in the sequence).
convert_nonlocal_omp_clauses was relying on the GIMPLE_BIND to be there always
because it called declare_vars on it.

The following patch wraps it into GIMPLE_BIND in tree-nested if we need to
declare_vars on it on demand.

Bootstrapped/regtested on x86_64-linux and i686-linux, committed to trunk.

2023-02-02  Jakub Jelinek  <jakub@redhat.com>

	PR middle-end/108435
	* tree-nested.cc (convert_nonlocal_omp_clauses)
	<case OMP_CLAUSE_LASTPRIVATE>: If info->new_local_var_chain and
	*seq is not a GIMPLE_BIND, wrap the sequence into a new GIMPLE_BIND
	before calling declare_vars.
	(convert_nonlocal_omp_clauses) <case OMP_CLAUSE_LINEAR>: Merge
	with the OMP_CLAUSE_LASTPRIVATE handling except for whether
	seq is initialized to &OMP_CLAUSE_LASTPRIVATE_GIMPLE_SEQ (clause)
	or &OMP_CLAUSE_LINEAR_GIMPLE_SEQ (clause).

	* gcc.dg/gomp/pr108435.c: New test.

--- gcc/tree-nested.cc.jj	2023-01-02 09:32:40.110029184 +0100
+++ gcc/tree-nested.cc	2023-02-01 16:22:33.272248502 +0100
@@ -1566,29 +1566,29 @@ convert_nonlocal_omp_clauses (tree *pcla
 	  break;
 
 	case OMP_CLAUSE_LASTPRIVATE:
-	  {
-	    tree save_local_var_chain = info->new_local_var_chain;
-	    info->new_local_var_chain = NULL;
-	    gimple_seq *seq = &OMP_CLAUSE_LASTPRIVATE_GIMPLE_SEQ (clause);
-	    walk_body (convert_nonlocal_reference_stmt,
-		       convert_nonlocal_reference_op, info, seq);
-	    if (info->new_local_var_chain)
-	      declare_vars (info->new_local_var_chain,
-			    gimple_seq_first_stmt (*seq), false);
-	    info->new_local_var_chain = save_local_var_chain;
-	  }
-	  break;
-
 	case OMP_CLAUSE_LINEAR:
 	  {
 	    tree save_local_var_chain = info->new_local_var_chain;
 	    info->new_local_var_chain = NULL;
-	    gimple_seq *seq = &OMP_CLAUSE_LINEAR_GIMPLE_SEQ (clause);
+	    gimple_seq *seq;
+	    if (OMP_CLAUSE_CODE (clause) == OMP_CLAUSE_LASTPRIVATE)
+	      seq = &OMP_CLAUSE_LASTPRIVATE_GIMPLE_SEQ (clause);
+	    else
+	      seq = &OMP_CLAUSE_LINEAR_GIMPLE_SEQ (clause);
 	    walk_body (convert_nonlocal_reference_stmt,
 		       convert_nonlocal_reference_op, info, seq);
 	    if (info->new_local_var_chain)
-	      declare_vars (info->new_local_var_chain,
-			    gimple_seq_first_stmt (*seq), false);
+	      {
+		gimple *g = gimple_seq_first_stmt (*seq);
+		if (gimple_code (g) != GIMPLE_BIND)
+		  {
+		    g = gimple_build_bind (NULL_TREE, *seq, NULL_TREE);
+		    *seq = NULL;
+		    gimple_seq_add_stmt_without_update (seq, g);
+		  }
+		declare_vars (info->new_local_var_chain,
+			      gimple_seq_first_stmt (*seq), false);
+	      }
 	    info->new_local_var_chain = save_local_var_chain;
 	  }
 	  break;
--- gcc/testsuite/gcc.dg/gomp/pr108435.c.jj	2023-02-01 16:24:56.180152165 +0100
+++ gcc/testsuite/gcc.dg/gomp/pr108435.c	2023-02-01 16:24:36.150445983 +0100
@@ -0,0 +1,18 @@
+/* PR middle-end/108435 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fopenmp" } */
+
+int
+main ()
+{
+  int i, j;
+  void
+  bar (void)
+  {
+    #pragma omp for simd collapse(2)
+     for (i = 1; i <= 16; i++)
+       for (j = 1; j <= 16; j++)
+	 ;
+  }
+  bar ();
+}

	Jakub


^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2023-02-02  9:00 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-02-02  9:00 [committed] nested, openmp: Wrap OMP_CLAUSE_*_GIMPLE_SEQ into GIMPLE_BIND for declare_vars [PR108435] Jakub Jelinek

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).