[dpdk-dev] [PATCH 1/3] mlx4: improve RX performance with better prefetching

[Adrien Mazarguil <adrien.mazarguil@6wind.com>]

From: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>

Prefetching initial bytes of mbuf structures earlier and in two cache lines
instead of one improves performance of mlx4_rx_burst(), which accesses the
mbuf->next field not present in the first 128 bytes.

Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
---
 drivers/net/mlx4/mlx4.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/drivers/net/mlx4/mlx4.c b/drivers/net/mlx4/mlx4.c
index cb08ee8..4cf2f7d 100644
--- a/drivers/net/mlx4/mlx4.c
+++ b/drivers/net/mlx4/mlx4.c
@@ -2824,6 +2824,12 @@ mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n)
 		assert(wr->num_sge == 1);
 		assert(elts_head < rxq->elts_n);
 		assert(rxq->elts_head < rxq->elts_n);
+		/*
+		 * Fetch initial bytes of packet descriptor into a
+		 * cacheline while allocating rep.
+		 */
+		rte_prefetch0(seg);
+		rte_prefetch0(&seg->cacheline1);
 		ret = rxq->if_cq->poll_length_flags(rxq->cq, NULL, NULL,
 						    &flags);
 		if (unlikely(ret < 0)) {
@@ -2861,11 +2867,6 @@ mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n)
 		if (ret == 0)
 			break;
 		len = ret;
-		/*
-		 * Fetch initial bytes of packet descriptor into a
-		 * cacheline while allocating rep.
-		 */
-		rte_prefetch0(seg);
 		rep = __rte_mbuf_raw_alloc(rxq->mp);
 		if (unlikely(rep == NULL)) {
 			/*
-- 
2.1.0