diff options
author | Radu Nicolau <radu.nicolau@intel.com> | 2021-01-14 10:25:02 +0000 |
---|---|---|
committer | Damjan Marion <dmarion@me.com> | 2021-01-21 13:39:12 +0000 |
commit | e1480a2c12ff764622dd2ae1bc9bce6cd25bcbdd (patch) | |
tree | b3b35b537b2428ce32fcdefcb040e2f3878b6f9f /src/plugins/avf/input.c | |
parent | f35fef244682645762cdff14cf313a42dd2d750c (diff) |
avf: use write combining store for queues tail update
Performance improvement: on supported platforms, currently only Intel Tremont,
use a write combining store to update the tail pointers.
Also, Tremont node variant is added for all.
Type: improvement
Signed-off-by: Radu Nicolau <radu.nicolau@intel.com>
Change-Id: Ie9606e403b7d9655184f778e3ffee3027c8c9edd
Diffstat (limited to 'src/plugins/avf/input.c')
-rw-r--r-- | src/plugins/avf/input.c | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/src/plugins/avf/input.c b/src/plugins/avf/input.c index 5041f6ef4e7..221b54b19a9 100644 --- a/src/plugins/avf/input.c +++ b/src/plugins/avf/input.c @@ -125,7 +125,7 @@ avf_rxq_refill (vlib_main_t * vm, vlib_node_runtime_t * node, avf_rxq_t * rxq, n_alloc -= 8; } - clib_atomic_store_rel_n (rxq->qrx_tail, slot); + avf_tail_write (rxq->qrx_tail, slot); } |