RE: patch 'examples/l3fwd: add Tx burst size configuration option' has been queued to stable release 24.11.4

["Ande, Venkat Kumar" <VenkatKumar.Ande@amd.com>]

[AMD Official Use Only - AMD Internal Distribution Only]

Hi Kevin,


If the commit d5c4897ecfb2 ("examples/l3fwd: add option to set Rx burst > size"), is taken in stable branch, then this patch also should be taken. As the performance of few of our platforms drastically came down for l3fwd test scenario with that commit.


Regards,
Venkat

-----Original Message-----
From: Kevin Traynor <ktraynor@redhat.com>
Sent: Wednesday, November 26, 2025 4:39 PM
To: Tummala, Sivaprasad <Sivaprasad.Tummala@amd.com>
Cc: Ande, Venkat Kumar <VenkatKumar.Ande@amd.com>; Dengdui Huang <huangdengdui@huawei.com>; Pavan Nikhilesh <pbhagavatula@marvell.com>; Chengwen Feng <fengchengwen@huawei.com>; dpdk stable <stable@dpdk.org>; Shani Peretz <shperetz@nvidia.com>
Subject: Re: patch 'examples/l3fwd: add Tx burst size configuration option' has been queued to stable release 24.11.4

Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding.


On 25/11/2025 15:05, Kevin Traynor wrote:
> Hi,
>
> FYI, your patch has been queued to stable release 24.11.4
>
> Note it hasn't been pushed to http://dpdk.org/browse/dpdk-stable yet.
> It will be pushed if I get no objections before 11/28/25. So please
> shout if anyone has objections.
>

Re-reading this patch. It doesn't have stable tags and I don't think we should take it to LTS branches, as it is changing defaults and seems to be a small performance improvement.

> Also note that after the patch there's a diff of the upstream commit
> vs the patch applied to the branch. This will indicate if there was
> any rebasing needed to apply to the stable branch. If there were code
> changes for rebasing
> (ie: not only metadata diffs), please double check that the rebase was
> correctly done.
>
> Queued patches are on a temporary branch at:
> https://github.com/kevintraynor/dpdk-stable
>
> This queued commit can be viewed at:
> https://github.com/kevintraynor/dpdk-stable/commit/d150d3359e9f7cbd22b
> 2d34c0fdec7b101400e98
>
> Thanks.
>
> Kevin
>
> ---
>>From d150d3359e9f7cbd22b2d34c0fdec7b101400e98 Mon Sep 17 00:00:00 2001
> From: Sivaprasad Tummala <sivaprasad.tummala@amd.com>
> Date: Thu, 6 Nov 2025 14:16:31 +0000
> Subject: [PATCH] examples/l3fwd: add Tx burst size configuration
> option
>
> [ upstream commit 79375d1015b308234e8b6955671a296394249f9b ]
>
> Previously, the Tx burst size in l3fwd was fixed at 256, which could
> lead to suboptimal performance in certain scenarios.
>
> This patch introduces separate --rx-burst and --tx-burst options to
> explicitly configure Rx and Tx burst sizes. By default, the Tx burst
> size now matches the Rx burst size for better efficiency and pipeline
> balance.
>
> Fixes: d5c4897ecfb2 ("examples/l3fwd: add option to set Rx burst
> size")
>
> Signed-off-by: Sivaprasad Tummala <sivaprasad.tummala@amd.com>
> Tested-by: Venkat Kumar Ande <venkatkumar.ande@amd.com>
> Tested-by: Dengdui Huang <huangdengdui@huawei.com>
> Tested-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
> Acked-by: Chengwen Feng <fengchengwen@huawei.com>
> ---
>  doc/guides/sample_app_ug/l3_forward.rst |  6 ++
>  examples/l3fwd/l3fwd.h                  | 10 +---
>  examples/l3fwd/l3fwd_acl.c              |  2 +-
>  examples/l3fwd/l3fwd_common.h           |  5 +-
>  examples/l3fwd/l3fwd_em.c               |  2 +-
>  examples/l3fwd/l3fwd_fib.c              |  2 +-
>  examples/l3fwd/l3fwd_lpm.c              |  2 +-
>  examples/l3fwd/main.c                   | 80 +++++++++++++++----------
>  8 files changed, 67 insertions(+), 42 deletions(-)
>
> diff --git a/doc/guides/sample_app_ug/l3_forward.rst
> b/doc/guides/sample_app_ug/l3_forward.rst
> index de347eeeb1..aba671d2b7 100644
> --- a/doc/guides/sample_app_ug/l3_forward.rst
> +++ b/doc/guides/sample_app_ug/l3_forward.rst
> @@ -78,4 +78,6 @@ The application has a number of command line options::
>                               [--lookup LOOKUP_METHOD]
>
> --config(port,queue,lcore)[,(port,queue,lcore)]
> +                             [--rx-burst NPKTS]
> +                             [--tx-burst NPKTS]
>                               [--eth-dest=X,MM:MM:MM:MM:MM:MM]
>                               [--max-pkt-len PKTLEN] @@ -114,4 +116,8
> @@ Where,
>  * ``--config (port,queue,lcore)[,(port,queue,lcore)]:`` Determines which queues from which ports are mapped to which cores.
>
> +* ``--rx-burst NPKTS:`` Optional, Rx burst size in decimal (default 32).
> +
> +* ``--tx-burst NPKTS:`` Optional, Tx burst size in decimal (default 32).
> +
>  * ``--eth-dest=X,MM:MM:MM:MM:MM:MM:`` Optional, ethernet destination for port X.
>
> diff --git a/examples/l3fwd/l3fwd.h b/examples/l3fwd/l3fwd.h index
> 0cce3406ee..471e3b488f 100644
> --- a/examples/l3fwd/l3fwd.h
> +++ b/examples/l3fwd/l3fwd.h
> @@ -33,8 +33,4 @@
>  #define VECTOR_SIZE_DEFAULT   MAX_PKT_BURST
>  #define VECTOR_TMO_NS_DEFAULT 1E6 /* 1ms */
> -/*
> - * Try to avoid TX buffering if we have at least MAX_TX_BURST packets to send.
> - */
> -#define      MAX_TX_BURST      (MAX_PKT_BURST / 2)
>
>  #define NB_SOCKETS        8
> @@ -117,5 +113,5 @@ extern struct acl_algorithms acl_alg[];  extern
> uint32_t max_pkt_len;
>
> -extern uint32_t nb_pkt_per_burst;
> +extern uint32_t rx_burst_size;
>  extern uint32_t mb_mempool_cache_size;
>
> @@ -153,6 +149,6 @@ send_single_packet(struct lcore_conf *qconf,
>
>       /* enough pkts to be sent */
> -     if (unlikely(len == MAX_PKT_BURST)) {
> -             send_burst(qconf, MAX_PKT_BURST, port);
> +     if (unlikely(len == rx_burst_size)) {
> +             send_burst(qconf, rx_burst_size, port);
>               len = 0;
>       }
> diff --git a/examples/l3fwd/l3fwd_acl.c b/examples/l3fwd/l3fwd_acl.c
> index 4fc4b986cc..a1275e18bc 100644
> --- a/examples/l3fwd/l3fwd_acl.c
> +++ b/examples/l3fwd/l3fwd_acl.c
> @@ -1137,5 +1137,5 @@ acl_main_loop(__rte_unused void *dummy)
>                       queueid = qconf->rx_queue_list[i].queue_id;
>                       nb_rx = rte_eth_rx_burst(portid, queueid,
> -                             pkts_burst, nb_pkt_per_burst);
> +                             pkts_burst, rx_burst_size);
>
>                       if (nb_rx > 0) { diff --git
> a/examples/l3fwd/l3fwd_common.h b/examples/l3fwd/l3fwd_common.h index
> d94e5f1357..ccaf0cd402 100644
> --- a/examples/l3fwd/l3fwd_common.h
> +++ b/examples/l3fwd/l3fwd_common.h
> @@ -26,4 +26,7 @@
>  #define SENDM_PORT_OVERHEAD(x) (x)
>
> +extern uint32_t rx_burst_size;
> +extern uint32_t tx_burst_size;
> +
>  /*
>   * From http://www.rfc-editor.org/rfc/rfc1812.txt section 5.2.2:
> @@ -72,5 +75,5 @@ send_packetsx4(struct lcore_conf *qconf, uint16_t port, struct rte_mbuf *m[],
>        * then send them straightway.
>        */
> -     if (num >= MAX_TX_BURST && len == 0) {
> +     if (num >= tx_burst_size && len == 0) {
>               n = rte_eth_tx_burst(port, qconf->tx_queue_id[port], m, num);
>               if (unlikely(n < num)) { diff --git
> a/examples/l3fwd/l3fwd_em.c b/examples/l3fwd/l3fwd_em.c index
> da9c45e3a4..58c54ed77e 100644
> --- a/examples/l3fwd/l3fwd_em.c
> +++ b/examples/l3fwd/l3fwd_em.c
> @@ -645,5 +645,5 @@ em_main_loop(__rte_unused void *dummy)
>                       queueid = qconf->rx_queue_list[i].queue_id;
>                       nb_rx = rte_eth_rx_burst(portid, queueid, pkts_burst,
> -                             nb_pkt_per_burst);
> +                             rx_burst_size);
>                       if (nb_rx == 0)
>                               continue; diff --git
> a/examples/l3fwd/l3fwd_fib.c b/examples/l3fwd/l3fwd_fib.c index
> 82f1739df7..4fc6bf90d5 100644
> --- a/examples/l3fwd/l3fwd_fib.c
> +++ b/examples/l3fwd/l3fwd_fib.c
> @@ -240,5 +240,5 @@ fib_main_loop(__rte_unused void *dummy)
>                       queueid = qconf->rx_queue_list[i].queue_id;
>                       nb_rx = rte_eth_rx_burst(portid, queueid, pkts_burst,
> -                                     nb_pkt_per_burst);
> +                                     rx_burst_size);
>                       if (nb_rx == 0)
>                               continue; diff --git
> a/examples/l3fwd/l3fwd_lpm.c b/examples/l3fwd/l3fwd_lpm.c index
> fec0aeb79c..a71eee69ec 100644
> --- a/examples/l3fwd/l3fwd_lpm.c
> +++ b/examples/l3fwd/l3fwd_lpm.c
> @@ -206,5 +206,5 @@ lpm_main_loop(__rte_unused void *dummy)
>                       queueid = qconf->rx_queue_list[i].queue_id;
>                       nb_rx = rte_eth_rx_burst(portid, queueid, pkts_burst,
> -                             nb_pkt_per_burst);
> +                             rx_burst_size);
>                       if (nb_rx == 0)
>                               continue; diff --git
> a/examples/l3fwd/main.c b/examples/l3fwd/main.c index
> ae3b4f6439..11aed26f94 100644
> --- a/examples/l3fwd/main.c
> +++ b/examples/l3fwd/main.c
> @@ -58,6 +58,7 @@ static_assert(MEMPOOL_CACHE_SIZE >= MAX_PKT_BURST,
> "MAX_PKT_BURST should be at m  uint16_t nb_rxd = RX_DESC_DEFAULT;
> uint16_t nb_txd = TX_DESC_DEFAULT; -uint32_t nb_pkt_per_burst =
> DEFAULT_PKT_BURST;
> +uint32_t rx_burst_size = DEFAULT_PKT_BURST;
>  uint32_t mb_mempool_cache_size = MEMPOOL_CACHE_SIZE;
> +uint32_t tx_burst_size = DEFAULT_PKT_BURST;
>
>  /**< Ports set in promiscuous mode off by default. */ @@ -401,5
> +402,6 @@ print_usage(const char *prgname)
>               " [--rx-queue-size NPKTS]"
>               " [--tx-queue-size NPKTS]"
> -             " [--burst NPKTS]"
> +             " [--rx-burst NPKTS]"
> +             " [--tx-burst NPKTS]"
>               " [--mbcache CACHESZ]"
>               " [--eth-dest=X,MM:MM:MM:MM:MM:MM]"
> @@ -428,5 +430,7 @@ print_usage(const char *prgname)
>               "  --tx-queue-size NPKTS: Tx queue size in decimal\n"
>               "            Default: %d\n"
> -             "  --burst NPKTS: Burst size in decimal\n"
> +             "  --rx-burst NPKTS: RX Burst size in decimal\n"
> +             "            Default: %d\n"
> +             "  --tx-burst NPKTS: TX Burst size in decimal\n"
>               "            Default: %d\n"
>               "  --mbcache CACHESZ: Mbuf cache size in decimal\n"
> @@ -461,6 +465,6 @@ print_usage(const char *prgname)
>               "  --rule_ipv6=FILE: Specify the ipv6 rules entries file.\n"
>               "  --alg: ACL classify method to use, one of: %s.\n\n",
> -             prgname, RX_DESC_DEFAULT, TX_DESC_DEFAULT, DEFAULT_PKT_BURST, MEMPOOL_CACHE_SIZE,
> -             ACL_LEAD_CHAR, ROUTE_LEAD_CHAR, alg);
> +             prgname, RX_DESC_DEFAULT, TX_DESC_DEFAULT, DEFAULT_PKT_BURST, DEFAULT_PKT_BURST,
> +             MEMPOOL_CACHE_SIZE, ACL_LEAD_CHAR, ROUTE_LEAD_CHAR,
> + alg);
>  }
>
> @@ -696,5 +700,5 @@ parse_mbcache_size(const char *optarg)
>
>  static void
> -parse_pkt_burst(const char *optarg)
> +parse_pkt_burst(const char *optarg, bool is_rx_burst, uint32_t
> +*burst_sz)
>  {
>       struct rte_eth_dev_info dev_info; @@ -711,29 +715,36 @@
> parse_pkt_burst(const char *optarg)
>       if (pkt_burst > MAX_PKT_BURST) {
>               RTE_LOG(INFO, L3FWD, "User provided burst must be <= %d. Using default value %d\n",
> -                     MAX_PKT_BURST, nb_pkt_per_burst);
> +                     MAX_PKT_BURST, *burst_sz);
>               return;
>       } else if (pkt_burst > 0) {
> -             nb_pkt_per_burst = (uint32_t)pkt_burst;
> +             *burst_sz = (uint32_t)pkt_burst;
>               return;
>       }
>
> -     /* If user gives a value of zero, query the PMD for its recommended Rx burst size. */
> -     ret = rte_eth_dev_info_get(0, &dev_info);
> -     if (ret != 0)
> -             return;
> -     burst_size = dev_info.default_rxportconf.burst_size;
> -     if (burst_size == 0) {
> -             RTE_LOG(INFO, L3FWD, "PMD does not recommend a burst size. Using default value %d. "
> -                     "User provided value must be in [1, %d]\n",
> -                     nb_pkt_per_burst, MAX_PKT_BURST);
> -             return;
> -     } else if (burst_size > MAX_PKT_BURST) {
> -             RTE_LOG(INFO, L3FWD, "PMD recommended burst size %d exceeds maximum value %d. "
> -                     "Using default value %d\n",
> -                     burst_size, MAX_PKT_BURST, nb_pkt_per_burst);
> -             return;
> +     if (is_rx_burst) {
> +             /* If user gives a value of zero, query the PMD for its recommended
> +              * Rx burst size.
> +              */
> +             ret = rte_eth_dev_info_get(0, &dev_info);
> +             if (ret != 0)
> +                     return;
> +             burst_size = dev_info.default_rxportconf.burst_size;
> +             if (burst_size == 0) {
> +                     RTE_LOG(INFO, L3FWD, "PMD does not recommend a burst size. Using default value %d. "
> +                             "User provided value must be in [1, %d]\n",
> +                             rx_burst_size, MAX_PKT_BURST);
> +                     return;
> +             } else if (burst_size > MAX_PKT_BURST) {
> +                     RTE_LOG(INFO, L3FWD, "PMD recommended burst size %d exceeds maximum value %d. "
> +                             "Using default value %d\n",
> +                             burst_size, MAX_PKT_BURST, rx_burst_size);
> +                     return;
> +             }
> +             *burst_sz = burst_size;
> +             RTE_LOG(INFO, L3FWD, "Using PMD-provided RX burst value %d\n", burst_size);
> +     } else {
> +             RTE_LOG(INFO, L3FWD, "User provided TX burst is 0. Using default value %d\n",
> +                     *burst_sz);
>       }
> -     nb_pkt_per_burst = burst_size;
> -     RTE_LOG(INFO, L3FWD, "Using PMD-provided burst value %d\n", burst_size);
>  }
>
> @@ -769,5 +780,6 @@ static const char short_options[] =  #define
> CMD_LINE_OPT_RULE_IPV6 "rule_ipv6"
>  #define CMD_LINE_OPT_ALG "alg"
> -#define CMD_LINE_OPT_PKT_BURST "burst"
> +#define CMD_LINE_OPT_PKT_RX_BURST "rx-burst"
> +#define CMD_LINE_OPT_PKT_TX_BURST "tx-burst"
>  #define CMD_LINE_OPT_MB_CACHE_SIZE "mbcache"
>
> @@ -800,5 +812,6 @@ enum {
>       CMD_LINE_OPT_VECTOR_SIZE_NUM,
>       CMD_LINE_OPT_VECTOR_TMO_NS_NUM,
> -     CMD_LINE_OPT_PKT_BURST_NUM,
> +     CMD_LINE_OPT_PKT_RX_BURST_NUM,
> +     CMD_LINE_OPT_PKT_TX_BURST_NUM,
>       CMD_LINE_OPT_MB_CACHE_SIZE_NUM,
>  };
> @@ -828,5 +841,6 @@ static const struct option lgopts[] = {
>       {CMD_LINE_OPT_RULE_IPV6,   1, 0, CMD_LINE_OPT_RULE_IPV6_NUM},
>       {CMD_LINE_OPT_ALG,   1, 0, CMD_LINE_OPT_ALG_NUM},
> -     {CMD_LINE_OPT_PKT_BURST,   1, 0, CMD_LINE_OPT_PKT_BURST_NUM},
> +     {CMD_LINE_OPT_PKT_RX_BURST,   1, 0, CMD_LINE_OPT_PKT_RX_BURST_NUM},
> +     {CMD_LINE_OPT_PKT_TX_BURST,   1, 0, CMD_LINE_OPT_PKT_TX_BURST_NUM},
>       {CMD_LINE_OPT_MB_CACHE_SIZE,   1, 0, CMD_LINE_OPT_MB_CACHE_SIZE_NUM},
>       {NULL, 0, 0, 0}
> @@ -918,6 +932,10 @@ parse_args(int argc, char **argv)
>                       break;
>
> -             case CMD_LINE_OPT_PKT_BURST_NUM:
> -                     parse_pkt_burst(optarg);
> +             case CMD_LINE_OPT_PKT_RX_BURST_NUM:
> +                     parse_pkt_burst(optarg, true, &rx_burst_size);
> +                     break;
> +
> +             case CMD_LINE_OPT_PKT_TX_BURST_NUM:
> +                     parse_pkt_burst(optarg, false, &tx_burst_size);
>                       break;
>
> @@ -1655,4 +1673,6 @@ main(int argc, char **argv)
>               rte_exit(EXIT_FAILURE, "Invalid L3FWD parameters\n");
>
> +     RTE_LOG(INFO, L3FWD, "Using Rx burst %u Tx burst %u\n",
> + rx_burst_size, tx_burst_size);
> +
>       /* Setup function pointers for lookup method. */
>       setup_l3fwd_lookup_tables();