* [PATCH] net/null: Support bulk alloc and free. @ 2018-02-03 3:11 Mallesh Koujalagi 2018-03-05 15:24 ` Ferruh Yigit 2018-03-08 23:40 ` [PATCH v2] net/null: support bulk allocation Mallesh Koujalagi 0 siblings, 2 replies; 8+ messages in thread From: Mallesh Koujalagi @ 2018-02-03 3:11 UTC (permalink / raw) To: dev; +Cc: mtetsuyah, ferruh.yigit, malleshx.koujalagi After bulk allocation and freeing of multiple mbufs increase more than ~2% throughput on single core. Signed-off-by: Mallesh Koujalagi <malleshx.koujalagi@intel.com> --- drivers/net/null/rte_eth_null.c | 16 +++++++--------- 1 file changed, 7 insertions(+), 9 deletions(-) diff --git a/drivers/net/null/rte_eth_null.c b/drivers/net/null/rte_eth_null.c index 9385ffd..247ede0 100644 --- a/drivers/net/null/rte_eth_null.c +++ b/drivers/net/null/rte_eth_null.c @@ -130,10 +130,11 @@ eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) return 0; packet_size = h->internals->packet_size; + + if (rte_pktmbuf_alloc_bulk(h->mb_pool, bufs, nb_bufs) != 0) + return 0; + for (i = 0; i < nb_bufs; i++) { - bufs[i] = rte_pktmbuf_alloc(h->mb_pool); - if (!bufs[i]) - break; rte_memcpy(rte_pktmbuf_mtod(bufs[i], void *), h->dummy_packet, packet_size); bufs[i]->data_len = (uint16_t)packet_size; @@ -149,18 +150,15 @@ eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) static uint16_t eth_null_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) { - int i; struct null_queue *h = q; if ((q == NULL) || (bufs == NULL)) return 0; - for (i = 0; i < nb_bufs; i++) - rte_pktmbuf_free(bufs[i]); + rte_mempool_put_bulk(bufs[0]->pool, (void **)bufs, nb_bufs); + rte_atomic64_add(&h->tx_pkts, nb_bufs); - rte_atomic64_add(&(h->tx_pkts), i); - - return i; + return nb_bufs; } static uint16_t -- 2.7.4 ^ permalink raw reply related [flat|nested] 8+ messages in thread
* Re: [PATCH] net/null: Support bulk alloc and free. 2018-02-03 3:11 [PATCH] net/null: Support bulk alloc and free Mallesh Koujalagi @ 2018-03-05 15:24 ` Ferruh Yigit 2018-03-05 15:36 ` Ananyev, Konstantin 2018-03-08 23:40 ` [PATCH v2] net/null: support bulk allocation Mallesh Koujalagi 1 sibling, 1 reply; 8+ messages in thread From: Ferruh Yigit @ 2018-03-05 15:24 UTC (permalink / raw) To: Mallesh Koujalagi, dev; +Cc: mtetsuyah On 2/3/2018 3:11 AM, Mallesh Koujalagi wrote: > After bulk allocation and freeing of multiple mbufs increase more than ~2% > throughput on single core. > > Signed-off-by: Mallesh Koujalagi <malleshx.koujalagi@intel.com> > --- > drivers/net/null/rte_eth_null.c | 16 +++++++--------- > 1 file changed, 7 insertions(+), 9 deletions(-) > > diff --git a/drivers/net/null/rte_eth_null.c b/drivers/net/null/rte_eth_null.c > index 9385ffd..247ede0 100644 > --- a/drivers/net/null/rte_eth_null.c > +++ b/drivers/net/null/rte_eth_null.c > @@ -130,10 +130,11 @@ eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) > return 0; > > packet_size = h->internals->packet_size; > + > + if (rte_pktmbuf_alloc_bulk(h->mb_pool, bufs, nb_bufs) != 0) > + return 0; > + > for (i = 0; i < nb_bufs; i++) { > - bufs[i] = rte_pktmbuf_alloc(h->mb_pool); > - if (!bufs[i]) > - break; > rte_memcpy(rte_pktmbuf_mtod(bufs[i], void *), h->dummy_packet, > packet_size); > bufs[i]->data_len = (uint16_t)packet_size; > @@ -149,18 +150,15 @@ eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) > static uint16_t > eth_null_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) > { > - int i; > struct null_queue *h = q; > > if ((q == NULL) || (bufs == NULL)) > return 0; > > - for (i = 0; i < nb_bufs; i++) > - rte_pktmbuf_free(bufs[i]); > + rte_mempool_put_bulk(bufs[0]->pool, (void **)bufs, nb_bufs); Is it guarantied that all mbufs will be from same mempool? > + rte_atomic64_add(&h->tx_pkts, nb_bufs); > > - rte_atomic64_add(&(h->tx_pkts), i); > - > - return i; > + return nb_bufs; > } > > static uint16_t > ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] net/null: Support bulk alloc and free. 2018-03-05 15:24 ` Ferruh Yigit @ 2018-03-05 15:36 ` Ananyev, Konstantin 2018-03-07 10:57 ` Ferruh Yigit 0 siblings, 1 reply; 8+ messages in thread From: Ananyev, Konstantin @ 2018-03-05 15:36 UTC (permalink / raw) To: Yigit, Ferruh, Koujalagi, MalleshX, dev; +Cc: mtetsuyah > -----Original Message----- > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Ferruh Yigit > Sent: Monday, March 5, 2018 3:25 PM > To: Koujalagi, MalleshX <malleshx.koujalagi@intel.com>; dev@dpdk.org > Cc: mtetsuyah@gmail.com > Subject: Re: [dpdk-dev] [PATCH] net/null: Support bulk alloc and free. > > On 2/3/2018 3:11 AM, Mallesh Koujalagi wrote: > > After bulk allocation and freeing of multiple mbufs increase more than ~2% > > throughput on single core. > > > > Signed-off-by: Mallesh Koujalagi <malleshx.koujalagi@intel.com> > > --- > > drivers/net/null/rte_eth_null.c | 16 +++++++--------- > > 1 file changed, 7 insertions(+), 9 deletions(-) > > > > diff --git a/drivers/net/null/rte_eth_null.c b/drivers/net/null/rte_eth_null.c > > index 9385ffd..247ede0 100644 > > --- a/drivers/net/null/rte_eth_null.c > > +++ b/drivers/net/null/rte_eth_null.c > > @@ -130,10 +130,11 @@ eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) > > return 0; > > > > packet_size = h->internals->packet_size; > > + > > + if (rte_pktmbuf_alloc_bulk(h->mb_pool, bufs, nb_bufs) != 0) > > + return 0; > > + > > for (i = 0; i < nb_bufs; i++) { > > - bufs[i] = rte_pktmbuf_alloc(h->mb_pool); > > - if (!bufs[i]) > > - break; > > rte_memcpy(rte_pktmbuf_mtod(bufs[i], void *), h->dummy_packet, > > packet_size); > > bufs[i]->data_len = (uint16_t)packet_size; > > @@ -149,18 +150,15 @@ eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) > > static uint16_t > > eth_null_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) > > { > > - int i; > > struct null_queue *h = q; > > > > if ((q == NULL) || (bufs == NULL)) > > return 0; > > > > - for (i = 0; i < nb_bufs; i++) > > - rte_pktmbuf_free(bufs[i]); > > + rte_mempool_put_bulk(bufs[0]->pool, (void **)bufs, nb_bufs); > > Is it guarantied that all mbufs will be from same mempool? I don't think it does, plus rte_pktmbuf_free(mb) != rte_mempool_put_bulk(mb->pool, &mb, 1); Konstantin > > > + rte_atomic64_add(&h->tx_pkts, nb_bufs); > > > > - rte_atomic64_add(&(h->tx_pkts), i); > > - > > - return i; > > + return nb_bufs; > > } > > > > static uint16_t > > ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] net/null: Support bulk alloc and free. 2018-03-05 15:36 ` Ananyev, Konstantin @ 2018-03-07 10:57 ` Ferruh Yigit 2018-03-08 21:29 ` Koujalagi, MalleshX 0 siblings, 1 reply; 8+ messages in thread From: Ferruh Yigit @ 2018-03-07 10:57 UTC (permalink / raw) To: Ananyev, Konstantin, Koujalagi, MalleshX, dev; +Cc: mtetsuyah On 3/5/2018 3:36 PM, Ananyev, Konstantin wrote: > > >> -----Original Message----- >> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Ferruh Yigit >> Sent: Monday, March 5, 2018 3:25 PM >> To: Koujalagi, MalleshX <malleshx.koujalagi@intel.com>; dev@dpdk.org >> Cc: mtetsuyah@gmail.com >> Subject: Re: [dpdk-dev] [PATCH] net/null: Support bulk alloc and free. >> >> On 2/3/2018 3:11 AM, Mallesh Koujalagi wrote: >>> After bulk allocation and freeing of multiple mbufs increase more than ~2% >>> throughput on single core. >>> >>> Signed-off-by: Mallesh Koujalagi <malleshx.koujalagi@intel.com> >>> --- >>> drivers/net/null/rte_eth_null.c | 16 +++++++--------- >>> 1 file changed, 7 insertions(+), 9 deletions(-) >>> >>> diff --git a/drivers/net/null/rte_eth_null.c b/drivers/net/null/rte_eth_null.c >>> index 9385ffd..247ede0 100644 >>> --- a/drivers/net/null/rte_eth_null.c >>> +++ b/drivers/net/null/rte_eth_null.c >>> @@ -130,10 +130,11 @@ eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) >>> return 0; >>> >>> packet_size = h->internals->packet_size; >>> + >>> + if (rte_pktmbuf_alloc_bulk(h->mb_pool, bufs, nb_bufs) != 0) >>> + return 0; >>> + >>> for (i = 0; i < nb_bufs; i++) { >>> - bufs[i] = rte_pktmbuf_alloc(h->mb_pool); >>> - if (!bufs[i]) >>> - break; >>> rte_memcpy(rte_pktmbuf_mtod(bufs[i], void *), h->dummy_packet, >>> packet_size); >>> bufs[i]->data_len = (uint16_t)packet_size; >>> @@ -149,18 +150,15 @@ eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) >>> static uint16_t >>> eth_null_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) >>> { >>> - int i; >>> struct null_queue *h = q; >>> >>> if ((q == NULL) || (bufs == NULL)) >>> return 0; >>> >>> - for (i = 0; i < nb_bufs; i++) >>> - rte_pktmbuf_free(bufs[i]); >>> + rte_mempool_put_bulk(bufs[0]->pool, (void **)bufs, nb_bufs); >> >> Is it guarantied that all mbufs will be from same mempool? > > I don't think it does, plus > rte_pktmbuf_free(mb) != rte_mempool_put_bulk(mb->pool, &mb, 1); Perhaps we can just benefit from bulk alloc. Hi Mallesh, Does it give any performance improvement if we switch "rte_pktmbuf_alloc()" to "rte_pktmbuf_alloc_bulk()" but keep free functions untouched? Thanks, ferruh > Konstantin > >> >>> + rte_atomic64_add(&h->tx_pkts, nb_bufs); >>> >>> - rte_atomic64_add(&(h->tx_pkts), i); >>> - >>> - return i; >>> + return nb_bufs; >>> } >>> >>> static uint16_t >>> > ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] net/null: Support bulk alloc and free. 2018-03-07 10:57 ` Ferruh Yigit @ 2018-03-08 21:29 ` Koujalagi, MalleshX 0 siblings, 0 replies; 8+ messages in thread From: Koujalagi, MalleshX @ 2018-03-08 21:29 UTC (permalink / raw) To: Yigit, Ferruh, Ananyev, Konstantin, dev; +Cc: mtetsuyah Hi Ferruh, Bulk allocation gives benefit but how much, will check and provide patch. Best regards -/Mallesh -----Original Message----- From: Yigit, Ferruh Sent: Wednesday, March 7, 2018 2:57 AM To: Ananyev, Konstantin <konstantin.ananyev@intel.com>; Koujalagi, MalleshX <malleshx.koujalagi@intel.com>; dev@dpdk.org Cc: mtetsuyah@gmail.com Subject: Re: [dpdk-dev] [PATCH] net/null: Support bulk alloc and free. On 3/5/2018 3:36 PM, Ananyev, Konstantin wrote: > > >> -----Original Message----- >> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Ferruh Yigit >> Sent: Monday, March 5, 2018 3:25 PM >> To: Koujalagi, MalleshX <malleshx.koujalagi@intel.com>; dev@dpdk.org >> Cc: mtetsuyah@gmail.com >> Subject: Re: [dpdk-dev] [PATCH] net/null: Support bulk alloc and free. >> >> On 2/3/2018 3:11 AM, Mallesh Koujalagi wrote: >>> After bulk allocation and freeing of multiple mbufs increase more >>> than ~2% throughput on single core. >>> >>> Signed-off-by: Mallesh Koujalagi <malleshx.koujalagi@intel.com> >>> --- >>> drivers/net/null/rte_eth_null.c | 16 +++++++--------- >>> 1 file changed, 7 insertions(+), 9 deletions(-) >>> >>> diff --git a/drivers/net/null/rte_eth_null.c >>> b/drivers/net/null/rte_eth_null.c index 9385ffd..247ede0 100644 >>> --- a/drivers/net/null/rte_eth_null.c >>> +++ b/drivers/net/null/rte_eth_null.c >>> @@ -130,10 +130,11 @@ eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) >>> return 0; >>> >>> packet_size = h->internals->packet_size; >>> + >>> + if (rte_pktmbuf_alloc_bulk(h->mb_pool, bufs, nb_bufs) != 0) >>> + return 0; >>> + >>> for (i = 0; i < nb_bufs; i++) { >>> - bufs[i] = rte_pktmbuf_alloc(h->mb_pool); >>> - if (!bufs[i]) >>> - break; >>> rte_memcpy(rte_pktmbuf_mtod(bufs[i], void *), h->dummy_packet, >>> packet_size); >>> bufs[i]->data_len = (uint16_t)packet_size; @@ -149,18 +150,15 @@ >>> eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) >>> static uint16_t eth_null_tx(void *q, struct rte_mbuf **bufs, >>> uint16_t nb_bufs) { >>> - int i; >>> struct null_queue *h = q; >>> >>> if ((q == NULL) || (bufs == NULL)) >>> return 0; >>> >>> - for (i = 0; i < nb_bufs; i++) >>> - rte_pktmbuf_free(bufs[i]); >>> + rte_mempool_put_bulk(bufs[0]->pool, (void **)bufs, nb_bufs); >> >> Is it guarantied that all mbufs will be from same mempool? > > I don't think it does, plus > rte_pktmbuf_free(mb) != rte_mempool_put_bulk(mb->pool, &mb, 1); Perhaps we can just benefit from bulk alloc. Hi Mallesh, Does it give any performance improvement if we switch "rte_pktmbuf_alloc()" to "rte_pktmbuf_alloc_bulk()" but keep free functions untouched? Thanks, ferruh > Konstantin > >> >>> + rte_atomic64_add(&h->tx_pkts, nb_bufs); >>> >>> - rte_atomic64_add(&(h->tx_pkts), i); >>> - >>> - return i; >>> + return nb_bufs; >>> } >>> >>> static uint16_t >>> > ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH v2] net/null: support bulk allocation 2018-02-03 3:11 [PATCH] net/null: Support bulk alloc and free Mallesh Koujalagi 2018-03-05 15:24 ` Ferruh Yigit @ 2018-03-08 23:40 ` Mallesh Koujalagi 2018-03-09 11:09 ` Ferruh Yigit 1 sibling, 1 reply; 8+ messages in thread From: Mallesh Koujalagi @ 2018-03-08 23:40 UTC (permalink / raw) To: dev, ferruh.yigit, konstantin.ananyev; +Cc: mtetsuyah, Mallesh Koujalagi Bulk allocation of multiple mbufs increased more than ~2% and less than 8% throughput on single core (1.8 GHz), based on usage for example 1: Testpmd case: Two null devices with copy 8% improvement. testpmd -c 0x3 -n 4 --socket-mem 1024,1024 --vdev 'eth_null0,size=64,copy=1' --vdev 'eth_null1,size=64,copy=1' -- -i -a --coremask=0x2 --txrst=64 --txfreet=64 --txd=256 --rxd=256 --rxfreet=64 --burst=64 --txpt=64 --txq=1 --rxq=1 --numa 2. Ovs switch case: 2% improvement. $VSCTL add-port ovs-br dpdk1 -- set Interface dpdk1 type=dpdk \ options:dpdk-devargs=eth_null0,size=64,copy=1 $VSCTL add-port ovs-br dpdk2 -- set Interface dpdk2 type=dpdk \ options:dpdk-devargs=eth_null1,size=64,copy=1 Signed-off-by: Mallesh Koujalagi <malleshx.koujalagi@intel.com> --- drivers/net/null/rte_eth_null.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/drivers/net/null/rte_eth_null.c b/drivers/net/null/rte_eth_null.c index 9385ffd..c019d2d 100644 --- a/drivers/net/null/rte_eth_null.c +++ b/drivers/net/null/rte_eth_null.c @@ -105,10 +105,10 @@ eth_null_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) return 0; packet_size = h->internals->packet_size; + if (rte_pktmbuf_alloc_bulk(h->mb_pool, bufs, nb_bufs) != 0) + return 0; + for (i = 0; i < nb_bufs; i++) { - bufs[i] = rte_pktmbuf_alloc(h->mb_pool); - if (!bufs[i]) - break; bufs[i]->data_len = (uint16_t)packet_size; bufs[i]->pkt_len = packet_size; bufs[i]->port = h->internals->port_id; @@ -130,10 +130,10 @@ eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) return 0; packet_size = h->internals->packet_size; + if (rte_pktmbuf_alloc_bulk(h->mb_pool, bufs, nb_bufs) != 0) + return 0; + for (i = 0; i < nb_bufs; i++) { - bufs[i] = rte_pktmbuf_alloc(h->mb_pool); - if (!bufs[i]) - break; rte_memcpy(rte_pktmbuf_mtod(bufs[i], void *), h->dummy_packet, packet_size); bufs[i]->data_len = (uint16_t)packet_size; -- 2.7.4 ^ permalink raw reply related [flat|nested] 8+ messages in thread
* Re: [PATCH v2] net/null: support bulk allocation 2018-03-08 23:40 ` [PATCH v2] net/null: support bulk allocation Mallesh Koujalagi @ 2018-03-09 11:09 ` Ferruh Yigit 2018-03-16 14:08 ` Ferruh Yigit 0 siblings, 1 reply; 8+ messages in thread From: Ferruh Yigit @ 2018-03-09 11:09 UTC (permalink / raw) To: Mallesh Koujalagi, dev, konstantin.ananyev; +Cc: mtetsuyah On 3/8/2018 11:40 PM, Mallesh Koujalagi wrote: > Bulk allocation of multiple mbufs increased more than ~2% and less > than 8% throughput on single core (1.8 GHz), based on usage for example > 1: Testpmd case: Two null devices with copy 8% improvement. > testpmd -c 0x3 -n 4 --socket-mem 1024,1024 > --vdev 'eth_null0,size=64,copy=1' --vdev 'eth_null1,size=64,copy=1' > -- -i -a --coremask=0x2 --txrst=64 --txfreet=64 --txd=256 > --rxd=256 --rxfreet=64 --burst=64 --txpt=64 --txq=1 --rxq=1 --numa > 2. Ovs switch case: 2% improvement. > $VSCTL add-port ovs-br dpdk1 -- set Interface dpdk1 type=dpdk \ > options:dpdk-devargs=eth_null0,size=64,copy=1 > $VSCTL add-port ovs-br dpdk2 -- set Interface dpdk2 type=dpdk \ > options:dpdk-devargs=eth_null1,size=64,copy=1 > > Signed-off-by: Mallesh Koujalagi <malleshx.koujalagi@intel.com> Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com> ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH v2] net/null: support bulk allocation 2018-03-09 11:09 ` Ferruh Yigit @ 2018-03-16 14:08 ` Ferruh Yigit 0 siblings, 0 replies; 8+ messages in thread From: Ferruh Yigit @ 2018-03-16 14:08 UTC (permalink / raw) To: Mallesh Koujalagi, dev, konstantin.ananyev; +Cc: mtetsuyah On 3/9/2018 11:09 AM, Ferruh Yigit wrote: > On 3/8/2018 11:40 PM, Mallesh Koujalagi wrote: >> Bulk allocation of multiple mbufs increased more than ~2% and less >> than 8% throughput on single core (1.8 GHz), based on usage for example >> 1: Testpmd case: Two null devices with copy 8% improvement. >> testpmd -c 0x3 -n 4 --socket-mem 1024,1024 >> --vdev 'eth_null0,size=64,copy=1' --vdev 'eth_null1,size=64,copy=1' >> -- -i -a --coremask=0x2 --txrst=64 --txfreet=64 --txd=256 >> --rxd=256 --rxfreet=64 --burst=64 --txpt=64 --txq=1 --rxq=1 --numa >> 2. Ovs switch case: 2% improvement. >> $VSCTL add-port ovs-br dpdk1 -- set Interface dpdk1 type=dpdk \ >> options:dpdk-devargs=eth_null0,size=64,copy=1 >> $VSCTL add-port ovs-br dpdk2 -- set Interface dpdk2 type=dpdk \ >> options:dpdk-devargs=eth_null1,size=64,copy=1 >> >> Signed-off-by: Mallesh Koujalagi <malleshx.koujalagi@intel.com> > > Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com> Applied to dpdk-next-net/master, thanks. ^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2018-03-16 14:08 UTC | newest] Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2018-02-03 3:11 [PATCH] net/null: Support bulk alloc and free Mallesh Koujalagi 2018-03-05 15:24 ` Ferruh Yigit 2018-03-05 15:36 ` Ananyev, Konstantin 2018-03-07 10:57 ` Ferruh Yigit 2018-03-08 21:29 ` Koujalagi, MalleshX 2018-03-08 23:40 ` [PATCH v2] net/null: support bulk allocation Mallesh Koujalagi 2018-03-09 11:09 ` Ferruh Yigit 2018-03-16 14:08 ` Ferruh Yigit
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.