From: Henry Kao <henry.kao@huawei.com>
Date: Fri, 27 Mar 2020 20:29:05 +0000 (-0400)
Subject: [OpenMP] Add memory barrier to solve data race
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=236ac68fa5b173e54b181e0a03216dfd8fd4f95e;p=platform%2Fupstream%2Fllvm.git

[OpenMP] Add memory barrier to solve data race

Data race occurs when acquiring lock for critical section
triggering assertion failure. Added barrier to ensure
all memory is commited before checking assertion.

Reviewed By: Hahnfeld

Differential Revision: https://reviews.llvm.org/D76780
---

diff --git a/openmp/runtime/src/kmp_lock.cpp b/openmp/runtime/src/kmp_lock.cpp
index 2cc9e08..8e20c6c 100644
--- a/openmp/runtime/src/kmp_lock.cpp
+++ b/openmp/runtime/src/kmp_lock.cpp
@@ -1239,6 +1239,9 @@ __kmp_acquire_queuing_lock_timed_template(kmp_queuing_lock_t *lck,
       KMP_MB();
       // ToDo: Use __kmp_wait_sleep or similar when blocktime != inf
       KMP_WAIT(spin_here_p, FALSE, KMP_EQ, lck);
+      // Synchronize writes to both runtime thread structures
+      // and writes in user code.
+      KMP_MB();
 
 #ifdef DEBUG_QUEUING_LOCKS
       TRACE_LOCK(gtid + 1, "acq spin");