public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r13-2829] openmp: Fix ICE with taskgroup at -O0 -fexceptions [PR107001]
@ 2022-09-24 7:31 Jakub Jelinek
0 siblings, 0 replies; only message in thread
From: Jakub Jelinek @ 2022-09-24 7:31 UTC (permalink / raw)
To: gcc-cvs
https://gcc.gnu.org/g:ad2aab5c816a6fd56b46210c0a4a4c6243da1de9
commit r13-2829-gad2aab5c816a6fd56b46210c0a4a4c6243da1de9
Author: Jakub Jelinek <jakub@redhat.com>
Date: Sat Sep 24 09:24:26 2022 +0200
openmp: Fix ICE with taskgroup at -O0 -fexceptions [PR107001]
The following testcase ICEs because with -O0 -fexceptions GOMP_taskgroup_end
call isn't directly followed by GOMP_RETURN statement, but there are some
conditionals to handle exceptions and we fail to find the correct GOMP_RETURN.
The fix is to treat taskgroup similarly to target data, both of these constructs
emit a try { body } finally { end_call } around the construct's body during
gimplification and we need to see proper construct nesting during gimplification
and omp lowering (including nesting of regions checks), but during omp expansion
we don't really need their nesting anymore, all we need is emit something at
the start of the region and the end of the region is the end API call we've
already emitted during gimplification. For target data, we weren't adding
GOMP_RETURN statement during omp lowering, so after that pass it is treated
merely like stand-alone omp directives. This patch does the same for
taskgroup too.
2022-09-24 Jakub Jelinek <jakub@redhat.com>
PR c/107001
* omp-low.cc (lower_omp_taskgroup): Don't add GOMP_RETURN statement
at the end.
* omp-expand.cc (build_omp_regions_1): Clarify GF_OMP_TARGET_KIND_DATA
is not stand-alone directive. For GIMPLE_OMP_TASKGROUP, also don't
update parent.
(omp_make_gimple_edges) <case GIMPLE_OMP_TASKGROUP>: Reset
cur_region back after new_omp_region.
* c-c++-common/gomp/pr107001.c: New test.
Diff:
---
gcc/omp-expand.cc | 17 +++++++++++++++--
gcc/omp-low.cc | 1 -
gcc/testsuite/c-c++-common/gomp/pr107001.c | 14 ++++++++++++++
3 files changed, 29 insertions(+), 3 deletions(-)
diff --git a/gcc/omp-expand.cc b/gcc/omp-expand.cc
index 5cac8dfe634..5dc0bf16e17 100644
--- a/gcc/omp-expand.cc
+++ b/gcc/omp-expand.cc
@@ -10712,7 +10712,10 @@ build_omp_regions_1 (basic_block bb, struct omp_region *parent,
case GF_OMP_TARGET_KIND_OACC_ENTER_DATA:
case GF_OMP_TARGET_KIND_OACC_EXIT_DATA:
case GF_OMP_TARGET_KIND_OACC_DECLARE:
- /* ..., other than for those stand-alone directives... */
+ /* ..., other than for those stand-alone directives...
+ To be precise, target data isn't stand-alone, but
+ gimplifier put the end API call into try finally block
+ for it, so omp expansion can treat it as such. */
region = NULL;
break;
default:
@@ -10728,6 +10731,11 @@ build_omp_regions_1 (basic_block bb, struct omp_region *parent,
&& gimple_omp_task_taskwait_p (stmt))
/* #pragma omp taskwait depend(...) is a stand-alone directive. */
region = NULL;
+ else if (code == GIMPLE_OMP_TASKGROUP)
+ /* #pragma omp taskgroup isn't a stand-alone directive, but
+ gimplifier put the end API call into try finall block
+ for it, so omp expansion can treat it as such. */
+ region = NULL;
/* ..., this directive becomes the parent for a new region. */
if (region)
parent = region;
@@ -10927,13 +10935,18 @@ omp_make_gimple_edges (basic_block bb, struct omp_region **region,
case GIMPLE_OMP_MASTER:
case GIMPLE_OMP_MASKED:
case GIMPLE_OMP_SCOPE:
- case GIMPLE_OMP_TASKGROUP:
case GIMPLE_OMP_CRITICAL:
case GIMPLE_OMP_SECTION:
cur_region = new_omp_region (bb, code, cur_region);
fallthru = true;
break;
+ case GIMPLE_OMP_TASKGROUP:
+ cur_region = new_omp_region (bb, code, cur_region);
+ fallthru = true;
+ cur_region = cur_region->outer;
+ break;
+
case GIMPLE_OMP_TASK:
cur_region = new_omp_region (bb, code, cur_region);
fallthru = true;
diff --git a/gcc/omp-low.cc b/gcc/omp-low.cc
index f0469d20b3d..d9f9aaebc0b 100644
--- a/gcc/omp-low.cc
+++ b/gcc/omp-low.cc
@@ -9724,7 +9724,6 @@ lower_omp_taskgroup (gimple_stmt_iterator *gsi_p, omp_context *ctx)
gimple_bind_add_seq (bind, gimple_omp_body (stmt));
gimple_omp_set_body (stmt, NULL);
- gimple_bind_add_stmt (bind, gimple_build_omp_return (true));
gimple_bind_add_seq (bind, dseq);
pop_gimplify_context (bind);
diff --git a/gcc/testsuite/c-c++-common/gomp/pr107001.c b/gcc/testsuite/c-c++-common/gomp/pr107001.c
new file mode 100644
index 00000000000..9c19d9b8745
--- /dev/null
+++ b/gcc/testsuite/c-c++-common/gomp/pr107001.c
@@ -0,0 +1,14 @@
+/* PR c/107001 */
+/* { dg-do compile } */
+/* { dg-options "-O0 -fopenmp -fexceptions" } */
+/* { dg-require-effective-target exceptions } */
+
+void bar (void);
+void foo (void)
+{
+ #pragma omp taskgroup
+ {
+ #pragma omp taskgroup
+ bar ();
+ }
+}
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2022-09-24 7:31 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-09-24 7:31 [gcc r13-2829] openmp: Fix ICE with taskgroup at -O0 -fexceptions [PR107001] 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).