It appears that miniflow_extract() in emc_processing() spends a lot of
cycles waiting for the packet's data to be read.
Prefetching the next packet's data while parsing removes this delay.
For a single flow pipeline the throughput improves by ~10%. With a
more realistic pipeline the change has a much smaller effect (~0.5%
improvement)
Signed-off-by: Daniele Di Proietto <diproiet...@vmware.com>
---
lib/dpif-netdev.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index 5b82c8b..f13169c 100644
--- a/lib/dpif-netdev.c
+++ b/lib/dpif-netdev.c
@@ -3150,6 +3150,11 @@ emc_processing(struct dp_netdev_pmd_thread *pmd, struct
dp_packet **packets,
continue;
}
+ if (i != cnt - 1) {
+ /* Prefetch next packet data */
+ OVS_PREFETCH(dp_packet_data(packets[i+1]));
+ }
+
miniflow_extract(packets[i], &key.mf);
key.len = 0; /* Not computed yet. */
key.hash = dpif_netdev_packet_get_rss_hash(packets[i], &key.mf);
--
2.1.4
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
