diff options
| author | Jonathan Peyton <jonathan.l.peyton@intel.com> | 2017-10-24 16:10:09 +0000 |
|---|---|---|
| committer | Jonathan Peyton <jonathan.l.peyton@intel.com> | 2017-10-24 16:10:09 +0000 |
| commit | dff0ee2f4ecfa953bb575ea6bd5df0280ca84d4b (patch) | |
| tree | d4b4102f5963c7a4d79c756456453295ba304674 /openmp/runtime/src | |
| parent | e3f6eb1a74cb23dd7872a613cb79fd7e40b2270b (diff) | |
| download | bcm5719-llvm-dff0ee2f4ecfa953bb575ea6bd5df0280ca84d4b.tar.gz bcm5719-llvm-dff0ee2f4ecfa953bb575ea6bd5df0280ca84d4b.zip | |
Disable threadprivate data cleanup if runtime is terminating
The problem is due to the runtime's threadprivate cleanup code which tries to
access data that was already destroyed by one of the root threads.
__kmp_init_gtid is used as a checker here since it is set to false before actual
resource cleanup is done in __kmp_cleanup().
Patch by Hansang Bae
llvm-svn: 316452
Diffstat (limited to 'openmp/runtime/src')
| -rw-r--r-- | openmp/runtime/src/kmp_threadprivate.cpp | 7 |
1 files changed, 7 insertions, 0 deletions
diff --git a/openmp/runtime/src/kmp_threadprivate.cpp b/openmp/runtime/src/kmp_threadprivate.cpp index bbcfadd7961..362c0757f98 100644 --- a/openmp/runtime/src/kmp_threadprivate.cpp +++ b/openmp/runtime/src/kmp_threadprivate.cpp @@ -226,6 +226,13 @@ void __kmp_common_destroy_gtid(int gtid) { struct private_common *tn; struct shared_common *d_tn; + if (!TCR_4(__kmp_init_gtid)) { + // This is possible when one of multiple roots initiates early library + // termination in a sequential region while other teams are active, and its + // child threads are about to end. + return; + } + KC_TRACE(10, ("__kmp_common_destroy_gtid: T#%d called\n", gtid)); if ((__kmp_foreign_tp) ? (!KMP_INITIAL_GTID(gtid)) : (!KMP_UBER_GTID(gtid))) { |
