There is a slight negative performance impact, by zeroing out the flow
batch pointers in dp_netdev_input__ ahead of packet_batch_execute(). The
issue has been observed with multiple batches test scenario.
This patch fixes the problem by removing the extra for loop and clear
the flow batches inside the packet_batch_per_flow_execute(). Also the
vtune analysis showed that the overall no. of instructions retired for
dp_netdev_input__ reduced by 1,273,800,000 with this patch.
Fixes: 603f2ce04d00 ("dpif-netdev: Clear flow batches before execute.")
Signed-off-by: Bhanuprakash Bodireddy <bhanuprakash.bodire...@intel.com>
Signed-off-by: Antonio Fischetti <antonio.fische...@intel.com>
---
lib/dpif-netdev.c | 5 +----
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index c002dd3..d0bb191 100644
--- a/lib/dpif-netdev.c
+++ b/lib/dpif-netdev.c
@@ -3878,6 +3878,7 @@ packet_batch_per_flow_execute(struct
packet_batch_per_flow *batch,
{
struct dp_netdev_actions *actions;
struct dp_netdev_flow *flow = batch->flow;
+ flow->batch = NULL;
dp_netdev_flow_used(flow, batch->array.count, batch->byte_count,
batch->tcp_flags, now);
@@ -4173,10 +4174,6 @@ dp_netdev_input__(struct dp_netdev_pmd_thread *pmd,
}
for (i = 0; i < n_batches; i++) {
- batches[i].flow->batch = NULL;
- }
-
- for (i = 0; i < n_batches; i++) {
packet_batch_per_flow_execute(&batches[i], pmd, now);
}
}
--
2.4.11
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
