DPDK patches and discussions
 help / color / mirror / Atom feed
From: Ajit Khaparde <ajit.khaparde@broadcom.com>
To: dev@dpdk.org
Cc: ferruh.yigit@intel.com,
	Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>,
	Ajit Kumar Khaparde <ajit.khaparde@broadcom.com>
Subject: [dpdk-dev] [PATCH 5/6] net/bnxt: optimize Tx batching code
Date: Wed, 15 May 2019 11:08:16 -0700	[thread overview]
Message-ID: <20190515180817.71523-6-ajit.khaparde@broadcom.com> (raw)
In-Reply-To: <20190515180817.71523-1-ajit.khaparde@broadcom.com>

From: Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>

This patch adds a few optimizations to Tx batching. Instead of
specifying 'cmpl_next' to bnxt_start_xmit() as a hint to enable
completion for a given packet, request for completion on the last
successfully enqueued TxBD. This takes care of both success and
error cases. It eliminates 'cmpl_next' arg and related per-packet
condition checks around it.

Fixes: 5735eb241947 ("net/bnxt: support Tx batching")
Signed-off-by: Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>
Reviewed-by: Ajit Kumar Khaparde <ajit.khaparde@broadcom.com>
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
---
 drivers/net/bnxt/bnxt_txr.c | 32 +++++++++-----------------------
 1 file changed, 9 insertions(+), 23 deletions(-)

diff --git a/drivers/net/bnxt/bnxt_txr.c b/drivers/net/bnxt/bnxt_txr.c
index 186934136..b15778b39 100644
--- a/drivers/net/bnxt/bnxt_txr.c
+++ b/drivers/net/bnxt/bnxt_txr.c
@@ -121,7 +121,6 @@ static inline uint32_t bnxt_tx_avail(struct bnxt_tx_queue *txq)
 static uint16_t bnxt_start_xmit(struct rte_mbuf *tx_pkt,
 				struct bnxt_tx_queue *txq,
 				uint16_t *coal_pkts,
-				uint16_t *cmpl_next,
 				struct tx_bd_long **last_txbd)
 {
 	struct bnxt_tx_ring_info *txr = txq->tx_ring;
@@ -184,12 +183,7 @@ static uint16_t bnxt_start_xmit(struct rte_mbuf *tx_pkt,
 	txbd->opaque = *coal_pkts;
 	txbd->flags_type = nr_bds << TX_BD_LONG_FLAGS_BD_CNT_SFT;
 	txbd->flags_type |= TX_BD_SHORT_FLAGS_COAL_NOW;
-	if (!*cmpl_next) {
-		txbd->flags_type |= TX_BD_LONG_FLAGS_NO_CMPL;
-	} else {
-		*coal_pkts = 0;
-		*cmpl_next = false;
-	}
+	txbd->flags_type |= TX_BD_LONG_FLAGS_NO_CMPL;
 	txbd->len = tx_pkt->data_len;
 	if (tx_pkt->pkt_len >= 2014)
 		txbd->flags_type |= TX_BD_LONG_FLAGS_LHINT_GTE2K;
@@ -415,10 +409,10 @@ static int bnxt_handle_tx_cp(struct bnxt_tx_queue *txq)
 uint16_t bnxt_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
 			       uint16_t nb_pkts)
 {
-	struct bnxt_tx_queue *txq = tx_queue;
+	int rc;
 	uint16_t nb_tx_pkts = 0;
 	uint16_t coal_pkts = 0;
-	uint16_t cmpl_next = 0;
+	struct bnxt_tx_queue *txq = tx_queue;
 	struct tx_bd_long *last_txbd = NULL;
 
 	/* Handle TX completions */
@@ -432,27 +426,19 @@ uint16_t bnxt_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
 
 	/* Handle TX burst request */
 	for (nb_tx_pkts = 0; nb_tx_pkts < nb_pkts; nb_tx_pkts++) {
-		int rc;
-
-		/* Request a completion on the last packet */
-		cmpl_next |= (nb_pkts == nb_tx_pkts + 1);
 		coal_pkts++;
 		rc = bnxt_start_xmit(tx_pkts[nb_tx_pkts], txq,
-				     &coal_pkts, &cmpl_next, &last_txbd);
+				     &coal_pkts, &last_txbd);
 
-		if (unlikely(rc)) {
-			/* Request a completion on the last successfully
-			 * enqueued packet
-			 */
-			if (last_txbd)
-				last_txbd->flags_type &=
-					~TX_BD_LONG_FLAGS_NO_CMPL;
+		if (unlikely(rc))
 			break;
-		}
 	}
 
-	if (nb_tx_pkts)
+	if (likely(nb_tx_pkts)) {
+		/* Request a completion on the last packet */
+		last_txbd->flags_type &= ~TX_BD_LONG_FLAGS_NO_CMPL;
 		B_TX_DB(txq->tx_ring->tx_doorbell, txq->tx_ring->tx_prod);
+	}
 
 	return nb_tx_pkts;
 }
-- 
2.20.1 (Apple Git-117)

  parent reply	other threads:[~2019-05-15 18:08 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-05-15 18:08 [dpdk-dev] [PATCH 0/6] bnxt patchset for Tx performance optimization Ajit Khaparde
2019-05-15 18:08 ` Ajit Khaparde
2019-05-15 18:08 ` [dpdk-dev] [PATCH 1/6] net/bnxt: fix an issue seen with TSO Ajit Khaparde
2019-05-15 18:08   ` Ajit Khaparde
2019-05-15 18:08 ` [dpdk-dev] [PATCH 2/6] net/bnxt: add support to update ipid Ajit Khaparde
2019-05-15 18:08   ` Ajit Khaparde
2019-05-15 18:08 ` [dpdk-dev] [PATCH 3/6] net/bnxt: check for some error conditions in Tx path Ajit Khaparde
2019-05-15 18:08   ` Ajit Khaparde
2019-05-15 18:08 ` [dpdk-dev] [PATCH 4/6] net/bnxt: fix a couple of issues with Tx batching Ajit Khaparde
2019-05-15 18:08   ` Ajit Khaparde
2019-05-15 18:08 ` Ajit Khaparde [this message]
2019-05-15 18:08   ` [dpdk-dev] [PATCH 5/6] net/bnxt: optimize Tx batching code Ajit Khaparde
2019-05-15 18:08 ` [dpdk-dev] [PATCH 6/6] net/bnxt: support bulk free of Tx mbufs Ajit Khaparde
2019-05-15 18:08   ` Ajit Khaparde
2019-05-27 16:16 ` [dpdk-dev] [PATCH 0/6] bnxt patchset for Tx performance optimization Ferruh Yigit

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190515180817.71523-6-ajit.khaparde@broadcom.com \
    --to=ajit.khaparde@broadcom.com \
    --cc=dev@dpdk.org \
    --cc=ferruh.yigit@intel.com \
    --cc=sriharsha.basavapatna@broadcom.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).