aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorBrian Brooks <brian.brooks@arm.com>2017-08-26 00:40:08 -0500
committerMaxim Uvarov <maxim.uvarov@linaro.org>2017-09-21 23:20:02 +0300
commitc2ed3f71a8d61449c5e35ce201f020a74a8ea244 (patch)
tree67855c4ec799422ca2f815aa5f24d33e61525687
parentcbb7f52e28d2e1c20c8eac662aa6135242d072e8 (diff)
downloadodp-master.tar.gz
linux-gen: barrier: Use correct memory orderingHEADmaster
Memory accesses that happen-before, in program order, a call to odp_barrier_wait() cannot be reordered to after the call. Similarly, memory accesses that happen-after, in program order, a call to odp_barrier_wait() cannot be reordered to before the call. The current implementation of barriers uses sequentially consistent fences on either side of odp_barrier_wait(). The correct memory ordering for barriers is release upon entering odp_barrier_wait(), to prevent reordering to after the barrier, and acquire upon exiting odp_barrier_wait(), to prevent reordering to before the barrier. The measurable performance difference is negligible on weakly ordered architectures such as ARM, so the highlight of this change is correctness. Signed-off-by: Brian Brooks <brian.brooks@arm.com> Reviewed-by: Bill Fischofer <bill.fischofer@linaro.org> Signed-off-by: Maxim Uvarov <maxim.uvarov@linaro.org>
-rw-r--r--platform/linux-generic/odp_barrier.c4
1 files changed, 2 insertions, 2 deletions
diff --git a/platform/linux-generic/odp_barrier.c b/platform/linux-generic/odp_barrier.c
index 5eb354d..f70bdbf 100644
--- a/platform/linux-generic/odp_barrier.c
+++ b/platform/linux-generic/odp_barrier.c
@@ -34,7 +34,7 @@ void odp_barrier_wait(odp_barrier_t *barrier)
uint32_t count;
int wasless;
- odp_mb_full();
+ odp_mb_release();
count = odp_atomic_fetch_inc_u32(&barrier->bar);
wasless = count < barrier->count;
@@ -48,5 +48,5 @@ void odp_barrier_wait(odp_barrier_t *barrier)
odp_cpu_pause();
}
- odp_mb_full();
+ odp_mb_acquire();
}