From: <pbhagavatula@marvell.com>
To: <jerinj@marvell.com>, Nithin Dabilpuram <ndabilpuram@marvell.com>,
"Kiran Kumar K" <kirankumark@marvell.com>
Cc: <dev@dpdk.org>, Pavan Nikhilesh <pbhagavatula@marvell.com>
Subject: [dpdk-dev] [PATCH] net/octeontx2: perf improvement to tx vector func
Date: Sat, 11 Jan 2020 19:24:40 +0530 [thread overview]
Message-ID: <20200111135440.6720-1-pbhagavatula@marvell.com> (raw)
From: Pavan Nikhilesh <pbhagavatula@marvell.com>
Use scalar loads instead of vector loads for fields
that don't need any vector operations.
Change-Id: I846731cd731d347a7f7e6a2e01d89dfa189ac244
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Reviewed-on: https://sj1git1.cavium.com/21243
Tested-by: sa_ip-toolkits-Jenkins
Tested-by: sa_ip-sw-jenkins <sa_ip-sw-jenkins@marvell.com>
Reviewed-by: Nithin Kumar Dabilpuram <ndabilpuram@marvell.com>
Reviewed-by: Jerin Jacob Kollanukkaran <jerinj@marvell.com>
---
drivers/net/octeontx2/otx2_tx.c | 15 +++++----------
1 file changed, 5 insertions(+), 10 deletions(-)
diff --git a/drivers/net/octeontx2/otx2_tx.c b/drivers/net/octeontx2/otx2_tx.c
index fa533000e..9f5926560 100644
--- a/drivers/net/octeontx2/otx2_tx.c
+++ b/drivers/net/octeontx2/otx2_tx.c
@@ -112,7 +112,6 @@ nix_xmit_pkts_vector(void *tx_queue, struct rte_mbuf **tx_pkts,
uint64x2_t ltypes01, ltypes23;
uint64x2_t xtmp128, ytmp128;
uint64x2_t xmask01, xmask23;
- uint64x2_t mbuf01, mbuf23;
uint64x2_t cmd00, cmd01;
uint64x2_t cmd10, cmd11;
uint64x2_t cmd20, cmd21;
@@ -137,9 +136,6 @@ nix_xmit_pkts_vector(void *tx_queue, struct rte_mbuf **tx_pkts,
sgdesc23_w0 = sgdesc01_w0;
for (i = 0; i < pkts; i += NIX_DESCS_PER_LOOP) {
- mbuf01 = vld1q_u64((uint64_t *)tx_pkts);
- mbuf23 = vld1q_u64((uint64_t *)(tx_pkts + 2));
-
/* Clear lower 32bit of SEND_HDR_W0 and SEND_SG_W0 */
senddesc01_w0 = vbicq_u64(senddesc01_w0,
vdupq_n_u64(0xFFFFFFFF));
@@ -149,13 +145,11 @@ nix_xmit_pkts_vector(void *tx_queue, struct rte_mbuf **tx_pkts,
senddesc23_w0 = senddesc01_w0;
sgdesc23_w0 = sgdesc01_w0;
- tx_pkts = tx_pkts + NIX_DESCS_PER_LOOP;
-
/* Move mbufs to iova */
- mbuf0 = (uint64_t *)vgetq_lane_u64(mbuf01, 0);
- mbuf1 = (uint64_t *)vgetq_lane_u64(mbuf01, 1);
- mbuf2 = (uint64_t *)vgetq_lane_u64(mbuf23, 0);
- mbuf3 = (uint64_t *)vgetq_lane_u64(mbuf23, 1);
+ mbuf0 = (uint64_t *)tx_pkts[0];
+ mbuf1 = (uint64_t *)tx_pkts[1];
+ mbuf2 = (uint64_t *)tx_pkts[2];
+ mbuf3 = (uint64_t *)tx_pkts[3];
mbuf0 = (uint64_t *)((uintptr_t)mbuf0 +
offsetof(struct rte_mbuf, buf_iova));
@@ -927,6 +921,7 @@ nix_xmit_pkts_vector(void *tx_queue, struct rte_mbuf **tx_pkts,
lmt_status = otx2_lmt_submit(io_addr);
} while (lmt_status == 0);
+ tx_pkts = tx_pkts + NIX_DESCS_PER_LOOP;
}
return pkts;
--
2.17.1
next reply other threads:[~2020-01-11 13:54 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-01-11 13:54 pbhagavatula [this message]
2020-01-11 13:57 ` [dpdk-dev] [PATCH v2] net/octeontx2: perf improvement to Tx " pbhagavatula
2020-01-14 6:27 ` Jerin Jacob
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=20200111135440.6720-1-pbhagavatula@marvell.com \
--to=pbhagavatula@marvell.com \
--cc=dev@dpdk.org \
--cc=jerinj@marvell.com \
--cc=kirankumark@marvell.com \
--cc=ndabilpuram@marvell.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).