From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 6E11341C4D; Thu, 9 Feb 2023 10:34:21 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 570F440EDC; Thu, 9 Feb 2023 10:34:21 +0100 (CET) Received: from smartserver.smartsharesystems.com (smartserver.smartsharesystems.com [77.243.40.215]) by mails.dpdk.org (Postfix) with ESMTP id 1BEE440DF8 for ; Thu, 9 Feb 2023 10:34:19 +0100 (CET) X-MimeOLE: Produced By Microsoft Exchange V6.5 Content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Subject: RE: [PATCH 1/2] net/i40e: replace put function Date: Thu, 9 Feb 2023 10:34:13 +0100 Message-ID: <98CBD80474FA8B44BF855DF32C47DC35D8771B@smartserver.smartshare.dk> In-Reply-To: <20230209062501.142828-1-kamalakshitha.aligeri@arm.com> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: [PATCH 1/2] net/i40e: replace put function Thread-Index: Adk8T0tP6Z/TGGoHR9asfClsgvDFXAAGJ/rA References: <20230209062501.142828-1-kamalakshitha.aligeri@arm.com> From: =?iso-8859-1?Q?Morten_Br=F8rup?= To: "Kamalakshitha Aligeri" , , , , , , , Cc: , , "Ruifeng Wang" , "Feifei Wang" X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org > From: Kamalakshitha Aligeri [mailto:kamalakshitha.aligeri@arm.com] > Sent: Thursday, 9 February 2023 07.25 >=20 > Integrated zero-copy put API in mempool cache in i40e PMD. > On Ampere Altra server, l3fwd single core's performance improves by 5% > with the new API >=20 > Signed-off-by: Kamalakshitha Aligeri > Reviewed-by: Ruifeng Wang > Reviewed-by: Feifei Wang > --- > Link: > https://patchwork.dpdk.org/project/dpdk/patch/20221227151700.80887-1- > mb@smartsharesystems.com/ >=20 > .mailmap | 1 + > drivers/net/i40e/i40e_rxtx_vec_common.h | 34 = ++++++++++++++++++++----- > 2 files changed, 28 insertions(+), 7 deletions(-) >=20 > diff --git a/.mailmap b/.mailmap > index 75884b6fe2..05a42edbcf 100644 > --- a/.mailmap > +++ b/.mailmap > @@ -670,6 +670,7 @@ Kai Ji > Kaiwen Deng > Kalesh AP > Kamalakannan R > +Kamalakshitha Aligeri > Kamil Bednarczyk > Kamil Chalupnik > Kamil Rytarowski > diff --git a/drivers/net/i40e/i40e_rxtx_vec_common.h > b/drivers/net/i40e/i40e_rxtx_vec_common.h > index fe1a6ec75e..80d4a159e6 100644 > --- a/drivers/net/i40e/i40e_rxtx_vec_common.h > +++ b/drivers/net/i40e/i40e_rxtx_vec_common.h > @@ -95,17 +95,37 @@ i40e_tx_free_bufs(struct i40e_tx_queue *txq) >=20 > n =3D txq->tx_rs_thresh; >=20 > - /* first buffer to free from S/W ring is at index > - * tx_next_dd - (tx_rs_thresh-1) > - */ > + /* first buffer to free from S/W ring is at index > + * tx_next_dd - (tx_rs_thresh-1) > + */ > txep =3D &txq->sw_ring[txq->tx_next_dd - (n - 1)]; >=20 > if (txq->offloads & RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE) { > - for (i =3D 0; i < n; i++) { > - free[i] =3D txep[i].mbuf; > - /* no need to reset txep[i].mbuf in vector path */ > + struct rte_mempool *mp =3D txep[0].mbuf->pool; > + struct rte_mempool_cache *cache =3D > rte_mempool_default_cache(mp, rte_lcore_id()); > + > + if (!cache || n > RTE_MEMPOOL_CACHE_MAX_SIZE) { If the mempool has a cache, do not compare n to = RTE_MEMPOOL_CACHE_MAX_SIZE. Instead, call = rte_mempool_cache_zc_put_bulk() to determine if n is acceptable for = zero-copy. It looks like this patch behaves incorrectly if the cache is configured = to be smaller than RTE_MEMPOOL_CACHE_MAX_SIZE. Let's say the cache size = is 8, which will make the flush threshold 12. If n is 32, your code will = not enter this branch, but proceed to call = rte_mempool_cache_zc_put_bulk(), which will return NULL, and then you = will goto done. Obviously, if there is no cache, fall back to the standard = rte_mempool_put_bulk(). > + for (i =3D 0; i < n ; i++) > + free[i] =3D txep[i].mbuf; > + if (!cache) { > + rte_mempool_generic_put(mp, (void **)free, n, > cache); > + goto done; > + } > + if (n > RTE_MEMPOOL_CACHE_MAX_SIZE) { > + rte_mempool_ops_enqueue_bulk(mp, (void **)free, > n); > + goto done; > + } > + } > + void **cache_objs; > + > + cache_objs =3D rte_mempool_cache_zc_put_bulk(cache, mp, n); > + if (cache_objs) { > + for (i =3D 0; i < n; i++) { > + cache_objs[i] =3D txep->mbuf; > + /* no need to reset txep[i].mbuf in vector path > */ > + txep++; > + } > } > - rte_mempool_put_bulk(free[0]->pool, (void **)free, n); > goto done; > } >=20 > -- > 2.25.1 >=20