On Fri, May 31, 2019 at 11:43:29AM +0200, Richard Biener wrote:
> Hmm. Is it enough to set TREE_THIS_NOTRAP on the ARRAY_REF instead?
That works too.
> At least it's documented the ARRAY_REF then isn't the issue.
> As for conditional vs. unconditional I'm not so sure but automatic
> vars are excepmted from that anyways?
Only if they are guaranteed not to have out of bound accesses.
Anyway, I'll go with this, those ARRAY_REFs indeed should never trap.
2019-05-31 Jakub Jelinek <ja...@redhat.com>
* omp-low.c (lower_rec_simd_input_clauses): Set TREE_THIS_NOTRAP on
ivar and lvar.
* gcc.dg/vect/vect-cond-12.c: New test.
--- gcc/omp-low.c.jj 2019-05-30 23:19:14.469931759 +0200
+++ gcc/omp-low.c 2019-05-31 11:52:20.491195088 +0200
@@ -3728,6 +3728,8 @@ lower_rec_simd_input_clauses (tree new_v
NULL_TREE, NULL_TREE);
lvar = build4 (ARRAY_REF, TREE_TYPE (new_var), avar, sctx->lane,
NULL_TREE, NULL_TREE);
+ TREE_THIS_NOTRAP (ivar) = 1;
+ TREE_THIS_NOTRAP (lvar) = 1;
}
if (DECL_P (new_var))
{
--- gcc/testsuite/gcc.dg/vect/vect-cond-12.c.jj 2019-05-31 11:25:33.203577504
+0200
+++ gcc/testsuite/gcc.dg/vect/vect-cond-12.c 2019-05-31 11:26:58.616174115
+0200
@@ -0,0 +1,14 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-fopenmp-simd" } */
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { target
vect_condition } } } */
+
+int x;
+
+void
+foo (int *a)
+{
+ #pragma omp simd lastprivate (x)
+ for (int i = 0; i < 1024; ++i)
+ if (a[i])
+ x = i;
+}
Jakub
