From: Jerin Jacob <jerin.jacob@caviumnetworks.com>
To: David Hunt <david.hunt@intel.com>
Cc: dev@dpdk.org, harry.van.haaren@intel.com,
Gage Eads <gage.eads@intel.com>,
Bruce Richardson <bruce.richardson@intel.com>
Subject: Re: [dpdk-dev] [PATCH v5 1/3] examples/eventdev_pipeline: added sample app
Date: Mon, 3 Jul 2017 09:27:57 +0530 [thread overview]
Message-ID: <20170703035755.GA6275@jerin> (raw)
In-Reply-To: <1498830673-56759-2-git-send-email-david.hunt@intel.com>
-----Original Message-----
>
> From: Harry van Haaren <harry.van.haaren@intel.com>
>
> This commit adds a sample app for the eventdev library.
> The app has been tested with DPDK 17.05-rc2, hence this
> release (or later) is recommended.
>
> The sample app showcases a pipeline processing use-case,
> with event scheduling and processing defined per stage.
> The application receives traffic as normal, with each
> packet traversing the pipeline. Once the packet has
> been processed by each of the pipeline stages, it is
> transmitted again.
>
> The app provides a framework to utilize cores for a single
> role or multiple roles. Examples of roles are the RX core,
> TX core, Scheduling core (in the case of the event/sw PMD),
> and worker cores.
>
> Various flags are available to configure numbers of stages,
> cycles of work at each stage, type of scheduling, number of
> worker cores, queue depths etc. For a full explaination,
> please refer to the documentation.
A few comments on bugs and "to avoid the future rework on base code when
HW PMD is introduced". As we agreed, We will keep the functionality intact to
provide an application to test ethdev + eventdev with _SW PMD_ for 17.08
> ---
> examples/Makefile | 2 +
> examples/eventdev_pipeline/Makefile | 49 ++
> examples/eventdev_pipeline/main.c | 999 ++++++++++++++++++++++++++++++++++++
> 3 files changed, 1050 insertions(+)
> create mode 100644 examples/eventdev_pipeline/Makefile
> create mode 100644 examples/eventdev_pipeline/main.c
Do we need to update the MAINTAINERS file?
>
> diff --git a/examples/Makefile b/examples/Makefile
> index 6298626..a6dcc2b 100644
> --- a/examples/Makefile
> +++ b/examples/Makefile
> @@ -100,4 +100,6 @@ $(info vm_power_manager requires libvirt >= 0.9.3)
> endif
> endif
>
> +DIRS-y += eventdev_pipeline
Can you change to eventdev_pipeline_sw_pmd to emphasis on the scope.
We will rename to eventdev_pipeline once it working effectively on both SW and HW
PMD with ethdev.
> +
> include $(RTE_SDK)/mk/rte.extsubdir.mk
> diff --git a/examples/eventdev_pipeline/Makefile b/examples/eventdev_pipeline/Makefile
> new file mode 100644
> index 0000000..4c26e15
> --- /dev/null
> +++ b/examples/eventdev_pipeline/Makefile
> @@ -0,0 +1,49 @@
> +# BSD LICENSE
> +#
> +# Copyright(c) 2016 Intel Corporation. All rights reserved.
2016-2017
> +#
> +# Redistribution and use in source and binary forms, with or without
> +# modification, are permitted provided that the following conditions
> +# are met:
> +#
> +
> +static unsigned int active_cores;
> +static unsigned int num_workers;
> +static long num_packets = (1L << 25); /* do ~32M packets */
> +static unsigned int num_fids = 512;
> +static unsigned int num_stages = 1;
> +static unsigned int worker_cq_depth = 16;
> +static int queue_type = RTE_EVENT_QUEUE_CFG_ATOMIC_ONLY;
> +static int16_t next_qid[MAX_NUM_STAGES+1] = {-1};
> +static int16_t qid[MAX_NUM_STAGES] = {-1};
> +static int worker_cycles;
> +static int enable_queue_priorities;
> +
> +struct prod_data {
> + uint8_t dev_id;
> + uint8_t port_id;
> + int32_t qid;
> + unsigned int num_nic_ports;
> +} __rte_cache_aligned;
> +
> +struct cons_data {
> + uint8_t dev_id;
> + uint8_t port_id;
> +} __rte_cache_aligned;
> +
> +static struct prod_data prod_data;
> +static struct cons_data cons_data;
> +
> +struct worker_data {
> + uint8_t dev_id;
> + uint8_t port_id;
> +} __rte_cache_aligned;
> +
> +static unsigned int *enqueue_cnt;
> +static unsigned int *dequeue_cnt;
Not been consumed. Remove it.
> +
> +static volatile int done;
> +static int quiet;
> +static int dump_dev;
> +static int dump_dev_signal;
> +
> +static uint32_t rx_lock;
> +static uint32_t tx_lock;
> +static uint32_t sched_lock;
> +static bool rx_single;
> +static bool tx_single;
> +static bool sched_single;
> +
> +static unsigned int rx_core[MAX_NUM_CORE];
> +static unsigned int tx_core[MAX_NUM_CORE];
> +static unsigned int sched_core[MAX_NUM_CORE];
> +static unsigned int worker_core[MAX_NUM_CORE];
> +
> +static struct rte_eth_dev_tx_buffer *tx_buf[RTE_MAX_ETHPORTS];
Could you please remove this global variable and group under a structure
for "command line parsing specific" and "fast path specific"(anything comes
in producer(), worker() and consumer()). And please
allocate "fast path specific" structure variable from huge page area.
So that we can easily add new parsing and fastpath variable in future.
> +
> +static int
> +consumer(void)
> +{
> + const uint64_t freq_khz = rte_get_timer_hz() / 1000;
> + struct rte_event packets[BATCH_SIZE];
> +
> + static uint64_t received;
> + static uint64_t last_pkts;
> + static uint64_t last_time;
> + static uint64_t start_time;
> + unsigned int i, j;
> + uint8_t dev_id = cons_data.dev_id;
> + uint8_t port_id = cons_data.port_id;
> +
> + uint16_t n = rte_event_dequeue_burst(dev_id, port_id,
> + packets, RTE_DIM(packets), 0);
> +
> + if (n == 0) {
> + for (j = 0; j < rte_eth_dev_count(); j++)
> + rte_eth_tx_buffer_flush(j, 0, tx_buf[j]);
> + return 0;
> + }
> + if (start_time == 0)
> + last_time = start_time = rte_get_timer_cycles();
> +
> + received += n;
> + for (i = 0; i < n; i++) {
> + uint8_t outport = packets[i].mbuf->port;
> + rte_eth_tx_buffer(outport, 0, tx_buf[outport],
> + packets[i].mbuf);
> + }
> +
> + /* Print out mpps every 1<22 packets */
> + if (!quiet && received >= last_pkts + (1<<22)) {
> + const uint64_t now = rte_get_timer_cycles();
> + const uint64_t total_ms = (now - start_time) / freq_khz;
> + const uint64_t delta_ms = (now - last_time) / freq_khz;
> + uint64_t delta_pkts = received - last_pkts;
> +
> + printf("# consumer RX=%"PRIu64", time %"PRIu64 "ms, "
> + "avg %.3f mpps [current %.3f mpps]\n",
> + received,
> + total_ms,
> + received / (total_ms * 1000.0),
> + delta_pkts / (delta_ms * 1000.0));
> + last_pkts = received;
> + last_time = now;
> + }
> +
> + dequeue_cnt[0] += n;
Not really used.
> +
> + num_packets -= n;
> + if (num_packets <= 0)
> + done = 1;
> +
> + return 0;
> +}
> +
> +static int
> +producer(void)
> +{
> + static uint8_t eth_port;
> + struct rte_mbuf *mbufs[BATCH_SIZE+2];
> + struct rte_event ev[BATCH_SIZE+2];
> + uint32_t i, num_ports = prod_data.num_nic_ports;
> + int32_t qid = prod_data.qid;
> + uint8_t dev_id = prod_data.dev_id;
> + uint8_t port_id = prod_data.port_id;
> + uint32_t prio_idx = 0;
> +
> + const uint16_t nb_rx = rte_eth_rx_burst(eth_port, 0, mbufs, BATCH_SIZE);
> + if (++eth_port == num_ports)
> + eth_port = 0;
> + if (nb_rx == 0) {
> + rte_pause();
> + return 0;
> + }
> +
> + for (i = 0; i < nb_rx; i++) {
> + ev[i].flow_id = mbufs[i]->hash.rss;
> + ev[i].op = RTE_EVENT_OP_NEW;
> + ev[i].sched_type = queue_type;
> + ev[i].queue_id = qid;
> + ev[i].event_type = RTE_EVENT_TYPE_ETHDEV;
> + ev[i].sub_event_type = 0;
> + ev[i].priority = RTE_EVENT_DEV_PRIORITY_NORMAL;
> + ev[i].mbuf = mbufs[i];
> + RTE_SET_USED(prio_idx);
> + }
> +
> + const int nb_tx = rte_event_enqueue_burst(dev_id, port_id, ev, nb_rx);
> + if (nb_tx != nb_rx) {
> + for (i = nb_tx; i < nb_rx; i++)
> + rte_pktmbuf_free(mbufs[i]);
> + }
> + enqueue_cnt[0] += nb_tx;
Not really used.
> +
> + return 0;
> +}
> +
> +
> +static inline void
> +work(struct rte_mbuf *m)
> +{
> + struct ether_hdr *eth;
> + struct ether_addr addr;
> +
> + /* change mac addresses on packet (to use mbuf data) */
> + eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
> + ether_addr_copy(ð->d_addr, &addr);
> + ether_addr_copy(ð->s_addr, ð->d_addr);
> + ether_addr_copy(&addr, ð->s_addr);
If it is even number of stages(say 2), Will mac swap be negated? as we are
swapping on each stage NOT in consumer?
> +
> + /* do a number of cycles of work per packet */
> + volatile uint64_t start_tsc = rte_rdtsc();
> + while (rte_rdtsc() < start_tsc + worker_cycles)
> + rte_pause();
> +}
> +
> +static int
> +worker(void *arg)
Looks good.
> +/*
> + * Initializes a given port using global settings and with the RX buffers
> + * coming from the mbuf_pool passed as a parameter.
> + */
> +static inline int
> +port_init(uint8_t port, struct rte_mempool *mbuf_pool)
Looks good.
> +static int
> +setup_eventdev(struct prod_data *prod_data,
> + struct cons_data *cons_data,
> + struct worker_data *worker_data)
> +{
> + /* final queue for sending to TX core */
> + if (rte_event_queue_setup(dev_id, i, &tx_q_conf) < 0) {
> + printf("%d: error creating qid %d\n", __LINE__, i);
> + return -1;
> + }
> + tx_queue.queue_id = i;
> + tx_queue.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
> +
> + if (wkr_p_conf.dequeue_depth > config.nb_event_port_dequeue_depth)
> + tx_p_conf.dequeue_depth = config.nb_event_port_dequeue_depth;
s/tx_p_conf.dequeue_depth/wkr_p_conf.dequeue_depth
> + if (wkr_p_conf.dequeue_depth > config.nb_event_port_enqueue_depth)
s/wkr_p_conf.dequeue_depth/wkr_p_conf.enqueue_depth
> + tx_p_conf.enqueue_depth = config.nb_event_port_enqueue_depth;
s/tx_p_conf.enqueue_depth/wkr_p_conf.enqueue_depth
> +
> + /* set up one port per worker, linking to all stage queues */
> + for (i = 0; i < num_workers; i++) {
> + struct worker_data *w = &worker_data[i];
> + w->dev_id = dev_id;
> + if (rte_event_port_setup(dev_id, i, &wkr_p_conf) < 0) {
> + printf("Error setting up port %d\n", i);
> + return -1;
> + }
> +
> + uint32_t s;
> + for (s = 0; s < num_stages; s++) {
> + if (rte_event_port_link(dev_id, i,
> + &worker_queues[s].queue_id,
> + &worker_queues[s].priority,
> + 1) != 1) {
> + printf("%d: error creating link for port %d\n",
> + __LINE__, i);
> + return -1;
> + }
> + }
> + w->port_id = i;
> + }
> +
> + if (tx_p_conf.dequeue_depth > config.nb_event_port_dequeue_depth)
> + tx_p_conf.dequeue_depth = config.nb_event_port_dequeue_depth;
> + if (tx_p_conf.dequeue_depth > config.nb_event_port_enqueue_depth)
s/dequeue_depth/enqueue_depth
> + tx_p_conf.enqueue_depth = config.nb_event_port_enqueue_depth;
> +
> + /* port for consumer, linked to TX queue */
> + if (rte_event_port_setup(dev_id, i, &tx_p_conf) < 0) {
> + printf("Error setting up port %d\n", i);
> + return -1;
> + }
> + if (rte_event_port_link(dev_id, i, &tx_queue.queue_id,
> + &tx_queue.priority, 1) != 1) {
> + printf("%d: error creating link for port %d\n",
> + __LINE__, i);
> + return -1;
> + }
> + /* port for producer, no links */
> + struct rte_event_port_conf rx_p_conf = {
> + .dequeue_depth = 8,
> + .enqueue_depth = 8,
> + .new_event_threshold = 1200,
> + };
> +
> + if (rx_p_conf.dequeue_depth > config.nb_event_port_dequeue_depth)
> + rx_p_conf.dequeue_depth = config.nb_event_port_dequeue_depth;
> + if (rx_p_conf.dequeue_depth > config.nb_event_port_enqueue_depth)
s/dequeue_depth/enqueue_depth
> + rx_p_conf.enqueue_depth = config.nb_event_port_enqueue_depth;
> +
> + if (rte_event_port_setup(dev_id, i + 1, &rx_p_conf) < 0) {
> + printf("Error setting up port %d\n", i);
> + return -1;
> + }
> +
> + *prod_data = (struct prod_data){.dev_id = dev_id,
> + .port_id = i + 1,
> + .qid = qid[0] };
> + *cons_data = (struct cons_data){.dev_id = dev_id,
> + .port_id = i };
> +
> + enqueue_cnt = rte_calloc(0,
> + RTE_CACHE_LINE_SIZE/(sizeof(enqueue_cnt[0])),
> + sizeof(enqueue_cnt[0]), 0);
> + dequeue_cnt = rte_calloc(0,
> + RTE_CACHE_LINE_SIZE/(sizeof(dequeue_cnt[0])),
> + sizeof(dequeue_cnt[0]), 0);
Debugging stuff. Remove this
> +
> + if (rte_event_dev_start(dev_id) < 0) {
> + printf("Error starting eventdev\n");
> + return -1;
> + }
> +
> + return dev_id;
> +}
> +
> +static void
> +signal_handler(int signum)
> +{
> + if (done)
> + rte_exit(1, "Exiting on signal %d\n", signum);
> + if (signum == SIGINT || signum == SIGTERM) {
> + printf("\n\nSignal %d received, preparing to exit...\n",
> + signum);
> + done = 1;
> + }
> + if (signum == SIGTSTP)
> + rte_event_dev_dump(0, stdout);
> +}
> +
> +int
> +main(int argc, char **argv)
> +{
> + struct worker_data *worker_data;
> + unsigned int num_ports;
> + int lcore_id;
> + int err;
> +
> + signal(SIGINT, signal_handler);
> + signal(SIGTERM, signal_handler);
> + signal(SIGTSTP, signal_handler);
> +
> + if (!quiet) {
> + printf("\nPort Workload distribution:\n");
> + uint32_t i;
> + uint64_t tot_pkts = 0;
> + uint64_t pkts_per_wkr[RTE_MAX_LCORE] = {0};
> + for (i = 0; i < num_workers; i++) {
> + char statname[64];
> + snprintf(statname, sizeof(statname), "port_%u_rx",
> + worker_data[i].port_id);
> + pkts_per_wkr[i] = rte_event_dev_xstats_by_name_get(
> + dev_id, statname, NULL);
As discussed, Check the the given xstat supported on the PMD first.
> + tot_pkts += pkts_per_wkr[i];
> + }
> + for (i = 0; i < num_workers; i++) {
> + float pc = pkts_per_wkr[i] * 100 /
> + ((float)tot_pkts);
> + printf("worker %i :\t%.1f %% (%"PRIu64" pkts)\n",
> + i, pc, pkts_per_wkr[i]);
> + }
> +
> + }
> +
> + return 0;
> +}
With above changes,
Jerin Jacob <jerin.jacob@caviumnetworks.com>
> --
> 2.7.4
>
next prev parent reply other threads:[~2017-07-03 3:58 UTC|newest]
Thread overview: 64+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-04-21 9:51 [dpdk-dev] [PATCH 0/3] next-eventdev: RFC evendev pipeline " Harry van Haaren
2017-04-21 9:51 ` [dpdk-dev] [PATCH 1/3] examples/eventdev_pipeline: added " Harry van Haaren
2017-05-10 14:12 ` Jerin Jacob
2017-05-10 16:40 ` Eads, Gage
2017-05-10 20:16 ` Eads, Gage
2017-06-26 14:46 ` Hunt, David
2017-06-27 9:35 ` Jerin Jacob
2017-06-27 13:12 ` Hunt, David
2017-06-29 7:17 ` Jerin Jacob
2017-06-29 12:51 ` Hunt, David
2017-05-17 18:03 ` Jerin Jacob
2017-05-18 10:13 ` Bruce Richardson
2017-06-26 14:41 ` [dpdk-dev] [PATCH v2 0/3] next-eventdev: evendev pipeline " David Hunt
2017-06-26 14:41 ` [dpdk-dev] [PATCH v2 1/3] examples/eventdev_pipeline: added " David Hunt
2017-06-27 12:54 ` [dpdk-dev] [PATCH v3 0/3] next-eventdev: evendev pipeline " David Hunt
2017-06-27 12:54 ` [dpdk-dev] [PATCH v3 1/3] examples/eventdev_pipeline: added " David Hunt
2017-06-29 15:49 ` [dpdk-dev] [PATCH v4 0/3] next-eventdev: evendev pipeline " David Hunt
2017-06-29 15:49 ` [dpdk-dev] [PATCH v4 1/3] examples/eventdev_pipeline: added " David Hunt
2017-06-30 13:51 ` [dpdk-dev] [PATCH v5 0/3] next-eventdev: evendev pipeline " David Hunt
2017-06-30 13:51 ` [dpdk-dev] [PATCH v5 1/3] examples/eventdev_pipeline: added " David Hunt
2017-07-03 3:57 ` Jerin Jacob [this message]
2017-07-04 7:55 ` Hunt, David
2017-07-05 5:30 ` Jerin Jacob
2017-07-05 11:15 ` Hunt, David
2017-07-06 3:31 ` Jerin Jacob
2017-07-06 10:04 ` Hunt, David
2017-07-06 10:39 ` Hunt, David
2017-07-06 13:26 ` Hunt, David
2017-07-06 13:38 ` Jerin Jacob
2017-07-04 8:14 ` [dpdk-dev] [PATCH v6 0/3] next-eventdev: evendev pipeline " David Hunt
2017-07-04 8:14 ` [dpdk-dev] [PATCH v6 1/3] examples/eventdev_pipeline: added " David Hunt
2017-07-05 12:52 ` [dpdk-dev] [PATCH v7 0/3] next-eventdev: evendev pipeline " David Hunt
2017-07-05 12:52 ` [dpdk-dev] [PATCH v7 1/3] examples/eventdev_pipeline: added " David Hunt
2017-07-06 14:35 ` [dpdk-dev] [PATCH v8 0/3] next-eventdev: evendev pipeline " David Hunt
2017-07-06 14:35 ` [dpdk-dev] [PATCH v8 1/3] examples/eventdev_pipeline_sw_pmd: add " David Hunt
2017-07-06 14:35 ` [dpdk-dev] [PATCH v8 2/3] doc: add SW eventdev pipeline to sample app ug David Hunt
2017-07-06 14:35 ` [dpdk-dev] [PATCH v8 3/3] doc: add eventdev library to programmers guide David Hunt
2017-07-07 4:50 ` [dpdk-dev] [PATCH v8 0/3] next-eventdev: evendev pipeline sample app Jerin Jacob
2017-07-05 12:52 ` [dpdk-dev] [PATCH v7 2/3] doc: add sw eventdev pipeline to sample app ug David Hunt
2017-07-05 12:52 ` [dpdk-dev] [PATCH v7 3/3] doc: add eventdev library to programmers guide David Hunt
2017-07-04 8:14 ` [dpdk-dev] [PATCH v6 2/3] doc: add sw eventdev pipeline to sample app ug David Hunt
2017-07-05 4:30 ` Jerin Jacob
2017-07-04 8:14 ` [dpdk-dev] [PATCH v6 3/3] doc: add eventdev library to programmers guide David Hunt
2017-06-30 13:51 ` [dpdk-dev] [PATCH v5 2/3] doc: add sw eventdev pipeline to sample app ug David Hunt
2017-06-30 14:37 ` Mcnamara, John
2017-07-03 5:37 ` Jerin Jacob
2017-07-03 9:25 ` Hunt, David
2017-07-03 9:32 ` Jerin Jacob
2017-07-04 8:20 ` Hunt, David
2017-06-30 13:51 ` [dpdk-dev] [PATCH v5 3/3] doc: add eventdev library to programmers guide David Hunt
2017-06-30 14:38 ` Mcnamara, John
2017-07-02 12:08 ` Jerin Jacob
2017-06-29 15:49 ` [dpdk-dev] [PATCH v4 2/3] doc: add sw eventdev pipeline to sample app ug David Hunt
2017-06-30 12:25 ` Mcnamara, John
2017-06-29 15:49 ` [dpdk-dev] [PATCH v4 3/3] doc: add eventdev library to programmers guide David Hunt
2017-06-30 12:26 ` Mcnamara, John
2017-06-27 12:54 ` [dpdk-dev] [PATCH v3 2/3] doc: add eventdev pipeline to sample app ug David Hunt
2017-06-27 12:54 ` [dpdk-dev] [PATCH v3 3/3] doc: add eventdev library to programmers guide David Hunt
2017-06-26 14:41 ` [dpdk-dev] [PATCH v2 2/3] doc: add eventdev pipeline to sample app ug David Hunt
2017-06-26 14:41 ` [dpdk-dev] [PATCH v2 3/3] doc: add eventdev library to programmers guide David Hunt
2017-04-21 9:51 ` [dpdk-dev] [PATCH 2/3] doc: add eventdev pipeline to sample app ug Harry van Haaren
2017-04-21 9:51 ` [dpdk-dev] [PATCH 3/3] doc: add eventdev library to programmers guide Harry van Haaren
2017-04-21 11:14 ` Bruce Richardson
2017-04-21 14:00 ` 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=20170703035755.GA6275@jerin \
--to=jerin.jacob@caviumnetworks.com \
--cc=bruce.richardson@intel.com \
--cc=david.hunt@intel.com \
--cc=dev@dpdk.org \
--cc=gage.eads@intel.com \
--cc=harry.van.haaren@intel.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).