From: Bruce Richardson <bruce.richardson@intel.com>
To: dev@dpdk.org
Cc: Bruce Richardson <bruce.richardson@intel.com>
Subject: [RFC PATCH 19/27] net/intel: avoid writing the final pkt descriptor twice
Date: Fri, 19 Dec 2025 17:25:36 +0000 [thread overview]
Message-ID: <20251219172548.2660777-20-bruce.richardson@intel.com> (raw)
In-Reply-To: <20251219172548.2660777-1-bruce.richardson@intel.com>
In the scalar datapath, there is a loop to handle multi-segment, and
multi-descriptor packets on Tx. After that loop, the end-of-packet bit
was written to the descriptor separately, meaning that for each
single-descriptor packet there were two writes to the second quad-word -
basically 3 x 64-bit writes rather than just 2. Adjusting the code to
compute the EOP bit inside the loop saves that extra write per packet
and so improves performance.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
drivers/net/intel/common/tx_scalar_fns.h | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/drivers/net/intel/common/tx_scalar_fns.h b/drivers/net/intel/common/tx_scalar_fns.h
index 6079a558e4..7b643fcf44 100644
--- a/drivers/net/intel/common/tx_scalar_fns.h
+++ b/drivers/net/intel/common/tx_scalar_fns.h
@@ -378,6 +378,10 @@ ci_xmit_pkts(struct ci_tx_queue *txq,
txn = &sw_ring[txe->next_id];
}
+ /* fill the last descriptor with End of Packet (EOP) bit */
+ if (m_seg->next == NULL)
+ td_cmd |= CI_TX_DESC_CMD_EOP;
+
txd->buffer_addr = rte_cpu_to_le_64(buf_dma_addr);
txd->cmd_type_offset_bsz = rte_cpu_to_le_64(CI_TX_DESC_DTYPE_DATA |
((uint64_t)td_cmd << CI_TXD_QW1_CMD_S) |
@@ -390,21 +394,17 @@ ci_xmit_pkts(struct ci_tx_queue *txq,
txe = txn;
m_seg = m_seg->next;
} while (m_seg);
-
- /* fill the last descriptor with End of Packet (EOP) bit */
- td_cmd |= CI_TX_DESC_CMD_EOP;
txq->nb_tx_used = (uint16_t)(txq->nb_tx_used + nb_used);
txq->nb_tx_free = (uint16_t)(txq->nb_tx_free - nb_used);
/* set RS bit on the last descriptor of one packet */
if (txq->nb_tx_used >= txq->tx_rs_thresh) {
- td_cmd |= CI_TX_DESC_CMD_RS;
+ txd->cmd_type_offset_bsz |=
+ rte_cpu_to_le_64(CI_TX_DESC_CMD_RS << CI_TXD_QW1_CMD_S);
/* Update txq RS bit counters */
txq->nb_tx_used = 0;
}
- txd->cmd_type_offset_bsz |=
- rte_cpu_to_le_64(((uint64_t)td_cmd) << CI_TXD_QW1_CMD_S);
if (ts_fns != NULL)
ts_id = ts_fns->write_ts_desc(txq, tx_pkt, tx_id, ts_id);
--
2.51.0
next prev parent reply other threads:[~2025-12-19 17:28 UTC|newest]
Thread overview: 30+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-12-19 17:25 [RFC PATCH 00/27] combine multiple Intel scalar Tx paths Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 01/27] net/intel: create common Tx descriptor structure Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 02/27] net/intel: use common tx ring structure Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 03/27] net/intel: create common post-Tx cleanup function Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 04/27] net/intel: consolidate definitions for Tx desc fields Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 05/27] net/intel: create separate header for Tx scalar fns Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 06/27] net/intel: add common fn to calculate needed descriptors Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 07/27] net/ice: refactor context descriptor handling Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 08/27] net/i40e: " Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 09/27] net/idpf: " Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 10/27] net/intel: consolidate checksum mask definition Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 11/27] net/intel: create common checksum Tx offload function Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 12/27] net/intel: create a common scalar Tx function Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 13/27] net/i40e: use " Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 14/27] net/intel: add IPSec hooks to common " Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 15/27] net/intel: support configurable VLAN tag insertion on Tx Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 16/27] net/iavf: use common scalar Tx function Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 17/27] net/i40e: document requirement for QinQ support Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 18/27] net/idpf: use common scalar Tx function Bruce Richardson
2025-12-19 17:25 ` Bruce Richardson [this message]
2025-12-19 17:25 ` [RFC PATCH 20/27] net/intel: write descriptors using non-volatile pointers Bruce Richardson
2025-12-20 8:43 ` Morten Brørup
2025-12-19 17:25 ` [RFC PATCH 21/27] net/intel: remove unnecessary flag clearing Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 22/27] net/intel: mark mid-burst ring cleanup as unlikely Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 23/27] net/intel: add special handling for single desc packets Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 24/27] net/intel: use separate array for desc status tracking Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 25/27] net/ixgbe: " Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 26/27] net/intel: drop unused Tx queue used count Bruce Richardson
2025-12-19 17:25 ` [RFC PATCH 27/27] net/intel: remove index for tracking end of packet Bruce Richardson
2025-12-20 9:05 ` Morten Brørup
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=20251219172548.2660777-20-bruce.richardson@intel.com \
--to=bruce.richardson@intel.com \
--cc=dev@dpdk.org \
/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).