DPDK patches and discussions
 help / color / mirror / Atom feed
From: Anoob Joseph <anoobj@marvell.com>
To: Gagandeep Singh <G.Singh@nxp.com>
Cc: Akhil Goyal <gakhil@marvell.com>, "dev@dpdk.org" <dev@dpdk.org>,
	Hemant Agrawal <hemant.agrawal@nxp.com>
Subject: RE: [EXT] [PATCH 1/8] app/test-crypto-perf: improve dequeue logic
Date: Mon, 16 May 2022 07:26:46 +0000
Message-ID: <PH0PR18MB4672DAE9F6C112A21645D538DFCF9@PH0PR18MB4672.namprd18.prod.outlook.com> (raw)
In-Reply-To: <AS8PR04MB8198D0D70A95C1E4EA16DBD0E1CF9@AS8PR04MB8198.eurprd04.prod.outlook.com>

Hi Gagandeep,

Please see inline.

Thanks,
Anoob

> -----Original Message-----
> From: Gagandeep Singh <G.Singh@nxp.com>
> Sent: Monday, May 16, 2022 12:44 PM
> To: Akhil Goyal <gakhil@marvell.com>; dev@dpdk.org; Hemant Agrawal
> <hemant.agrawal@nxp.com>
> Cc: Anoob Joseph <anoobj@marvell.com>
> Subject: RE: [EXT] [PATCH 1/8] app/test-crypto-perf: improve dequeue logic
> 
> Hi
> 
> > -----Original Message-----
> > From: Akhil Goyal <gakhil@marvell.com>
> > Sent: Friday, May 13, 2022 3:17 PM
> > To: Gagandeep Singh <G.Singh@nxp.com>; dev@dpdk.org; Hemant Agrawal
> > <hemant.agrawal@nxp.com>
> > Cc: Anoob Joseph <anoobj@marvell.com>
> > Subject: RE: [EXT] [PATCH 1/8] app/test-crypto-perf: improve dequeue
> > logic
> >
> > Hi Gagan,
> > > Issue more dequeue commands if the gap between enqueued and dequeued
> > > packets is more than burst size *8
> > >
> > > Signed-off-by: Gagandeep Singh <g.singh@nxp.com>
> > > ---
> > Why is this change required? What gain are we getting?
> > I see a performance drop due to this patch.
> 
> Issue is, in case if security engine/driver is slow in processing the Jobs especially
> for larger packet sizes then in that case application will keep enqueuing packets
> with higher rate than dequeue which may results in buffer pool exhaustion.
> Application has option to increase pool size but that may not be Helpful for the
> platforms those with memory constraints.

[Anoob] Can you elaborate the issue that you are hitting? 

> 
> We can work on limiting the enqueue side instead of keeping the dequeue to
> avoid any performance drop due to any empty dequeue.

[Anoob] Shouldn't PMD take care of limiting the enqueue side? Application can specify the desired queue depth and when application enqueues beyond that, enqueue API can return 0. Wouldn't that good enough?
 
> 
> Dropping this patch from this series. I will update the logic and will try to send as
> separate patch.
> 
> >
> > >  app/test-crypto-perf/cperf_test_throughput.c | 42
> > > +++++++++++---------
> > >  1 file changed, 23 insertions(+), 19 deletions(-)
> > >
> > > diff --git a/app/test-crypto-perf/cperf_test_throughput.c
> > > b/app/test-crypto- perf/cperf_test_throughput.c index
> > > cecf30e470..5cd8919c91 100644
> > > --- a/app/test-crypto-perf/cperf_test_throughput.c
> > > +++ b/app/test-crypto-perf/cperf_test_throughput.c
> > > @@ -223,26 +223,30 @@ cperf_throughput_test_runner(void *test_ctx)
> > >  			ops_unused = burst_size - ops_enqd;
> > >  			ops_enqd_total += ops_enqd;
> > >
> > > -
> > >  			/* Dequeue processed burst of ops from crypto device
> */
> > > -			ops_deqd = rte_cryptodev_dequeue_burst(ctx->dev_id,
> > > ctx->qp_id,
> > > -					ops_processed, test_burst_size);
> > > -
> > > -			if (likely(ops_deqd))  {
> > > -				/* Free crypto ops so they can be reused. */
> > > -				rte_mempool_put_bulk(ctx->pool,
> > > -						(void **)ops_processed,
> > > ops_deqd);
> > > -
> > > -				ops_deqd_total += ops_deqd;
> > > -			} else {
> > > -				/**
> > > -				 * Count dequeue polls which didn't return any
> > > -				 * processed operations. This statistic is mainly
> > > -				 * relevant to hw accelerators.
> > > -				 */
> > > -				ops_deqd_failed++;
> > > -			}
> > > -
> > > +			do {
> > > +				ops_deqd = rte_cryptodev_dequeue_burst(
> > > +						ctx->dev_id, ctx->qp_id,
> > > +						ops_processed,
> > > test_burst_size);
> > > +
> > > +				if (likely(ops_deqd))  {
> > > +					/* Free crypto ops for reuse */
> > > +					rte_mempool_put_bulk(ctx->pool,
> > > +							(void
> > > **)ops_processed,
> > > +							ops_deqd);
> > > +
> > > +					ops_deqd_total += ops_deqd;
> > > +				} else {
> > > +					/**
> > > +					 * Count dequeue polls which didn't
> > > +					 * return any processed operations.
> > > +					 * This statistic is mainly relevant
> > > +					 * to hw accelerators.
> > > +					 */
> > > +					ops_deqd_failed++;
> > > +				}
> > > +			} while (ops_enqd_total - ops_deqd_total >
> > > +					test_burst_size * 8);
> > >  		}
> > >
> > >  		/* Dequeue any operations still in the crypto device */
> > > --
> > > 2.25.1


  reply	other threads:[~2022-05-16  7:26 UTC|newest]

Thread overview: 49+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-04-25  4:14 Gagandeep Singh
2022-04-25  4:14 ` [PATCH 2/8] app/test-crypto-perf: support SDAP for PDCP operations Gagandeep Singh
2022-05-13  9:49   ` [EXT] " Akhil Goyal
2022-04-25  4:14 ` [PATCH 3/8] examples/ipsec-secgw: support XCBC-MAC/DES-CBC Gagandeep Singh
2022-05-13  9:54   ` [EXT] " Akhil Goyal
2022-05-16  9:45     ` Gagandeep Singh
2022-04-25  4:14 ` [PATCH 4/8] examples/l2fwd-crypto: add signal handler for exit Gagandeep Singh
2022-04-25  4:14 ` [PATCH 5/8] test/crypto: fix pmd name for dpaa raw buffer crypto driver Gagandeep Singh
2022-04-25  4:14 ` [PATCH 6/8] test/crypto: support raw buffer APIs for PDCP Gagandeep Singh
2022-04-25  4:14 ` [PATCH 7/8] test/crypto: add external buffer test case Gagandeep Singh
2022-05-13 10:26   ` [EXT] " Akhil Goyal
2022-05-16  7:25     ` Gagandeep Singh
2022-04-25  4:14 ` [PATCH 8/8] test/crypto: add short MAC-I test vector for zuc Gagandeep Singh
2022-05-13 10:46   ` [EXT] " Akhil Goyal
2022-05-16  7:27     ` Gagandeep Singh
2022-05-13  9:46 ` [EXT] [PATCH 1/8] app/test-crypto-perf: improve dequeue logic Akhil Goyal
2022-05-16  7:14   ` Gagandeep Singh
2022-05-16  7:26     ` Anoob Joseph [this message]
2022-05-16  7:54       ` Gagandeep Singh
2022-05-17  3:38 ` [PATCH v2 0/7] Crypto related changes in sample/test apps Gagandeep Singh
2022-05-17  3:38   ` [PATCH v2 1/7] app/test-crypto-perf: support SDAP for PDCP operations Gagandeep Singh
2022-05-20  4:20     ` [PATCH v3 0/7] Crypto related changes in sample/test apps Gagandeep Singh
2022-05-20  4:20       ` [PATCH v3 1/7] app/test-crypto-perf: support SDAP for PDCP operations Gagandeep Singh
2022-05-26 13:52         ` [EXT] " Akhil Goyal
2022-05-30  4:31           ` Gagandeep Singh
2022-05-30  6:04             ` Akhil Goyal
2022-05-31 18:14               ` Akhil Goyal
2022-06-06  4:00         ` [PATCH v4] " Gagandeep Singh
2022-06-15 16:06           ` [EXT] " Akhil Goyal
2022-05-20  4:20       ` [PATCH v3 2/7] examples/ipsec-secgw: support XCBC-MAC/DES-CBC Gagandeep Singh
2022-05-26 13:54         ` [EXT] " Akhil Goyal
2022-05-20  4:21       ` [PATCH v3 3/7] examples/l2fwd-crypto: add signal handler for exit Gagandeep Singh
2022-05-26 13:55         ` [EXT] " Akhil Goyal
2022-05-20  4:21       ` [PATCH v3 4/7] test/crypto: fix PMD name for dpaa raw buffer crypto driver Gagandeep Singh
2022-05-26 13:56         ` [EXT] " Akhil Goyal
2022-05-20  4:21       ` [PATCH v3 5/7] test/crypto: support raw buffer APIs for PDCP Gagandeep Singh
2022-05-26 13:58         ` [EXT] " Akhil Goyal
2022-05-20  4:21       ` [PATCH v3 6/7] test/crypto: add short MAC-I test vector for zuc Gagandeep Singh
2022-05-26 13:59         ` [EXT] " Akhil Goyal
2022-05-20  4:21       ` [PATCH v3 7/7] doc: add missing authentication algorithm Gagandeep Singh
2022-05-26 14:00         ` [EXT] " Akhil Goyal
2022-05-17  3:38   ` [PATCH v2 2/7] examples/ipsec-secgw: support XCBC-MAC/DES-CBC Gagandeep Singh
2022-05-17  3:38   ` [PATCH v2 3/7] examples/l2fwd-crypto: add signal handler for exit Gagandeep Singh
2022-05-17 16:32     ` Stephen Hemminger
2022-05-18  4:23       ` Gagandeep Singh
2022-05-17  3:38   ` [PATCH v2 4/7] test/crypto: fix PMD name for dpaa raw buffer crypto driver Gagandeep Singh
2022-05-17  3:38   ` [PATCH v2 5/7] test/crypto: support raw buffer APIs for PDCP Gagandeep Singh
2022-05-17  3:38   ` [PATCH v2 6/7] test/crypto: add short MAC-I test vector for zuc Gagandeep Singh
2022-05-17  3:38   ` [PATCH v2 7/7] doc: add missing authentication algorithm Gagandeep Singh

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=PH0PR18MB4672DAE9F6C112A21645D538DFCF9@PH0PR18MB4672.namprd18.prod.outlook.com \
    --to=anoobj@marvell.com \
    --cc=G.Singh@nxp.com \
    --cc=dev@dpdk.org \
    --cc=gakhil@marvell.com \
    --cc=hemant.agrawal@nxp.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

DPDK patches and discussions

This inbox may be cloned and mirrored by anyone:

	git clone --mirror http://inbox.dpdk.org/dev/0 dev/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 dev dev/ http://inbox.dpdk.org/dev \
		dev@dpdk.org
	public-inbox-index dev

Example config snippet for mirrors.
Newsgroup available over NNTP:
	nntp://inbox.dpdk.org/inbox.dpdk.dev


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git