DPDK patches and discussions
 help / color / mirror / Atom feed
* [PATCH] examples: add eventdev_producer_consumer example
@ 2022-08-18 17:02 Timothy McDaniel
  2022-08-22 19:26 ` [PATCH v2] examples/eventdev_producer_consumer: fix 32-bit checkpatch issues Timothy McDaniel
                   ` (2 more replies)
  0 siblings, 3 replies; 5+ messages in thread
From: Timothy McDaniel @ 2022-08-18 17:02 UTC (permalink / raw)
  To: jerinj
  Cc: thomas, pbhagavatula, sthotton, sachin.saxena, hemant.agrawal,
	harry.van.haaren, mattias.ronnblom, liangma, peter.mccarthy, dev

The eventdev-producer-consumer application is a single-stage
producer-worker-consumer pipeline sample to mimic real-world applications.
It is useful in measuring performance impact when any eventdev
configuration is changed. Unlike test-eventdev, it allows configuring a
load balanced queue between the producer and workers and a single-link
queue between the workers and consumer. With test-eventdev, multiple worker
stages can be configured but there is no single consumer receiving events
from all the workers. It also does not require configuring TX/RX adapters
like in the case of eventdev_pipeline app.

Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>
---
 examples/eventdev_producer_consumer/Makefile  |  22 +
 examples/eventdev_producer_consumer/main.c    | 670 ++++++++++++++++++
 .../eventdev_producer_consumer/meson.build    |  13 +
 examples/meson.build                          |   1 +
 4 files changed, 706 insertions(+)
 create mode 100644 examples/eventdev_producer_consumer/Makefile
 create mode 100644 examples/eventdev_producer_consumer/main.c
 create mode 100644 examples/eventdev_producer_consumer/meson.build

diff --git a/examples/eventdev_producer_consumer/Makefile b/examples/eventdev_producer_consumer/Makefile
new file mode 100644
index 0000000000..761689eab7
--- /dev/null
+++ b/examples/eventdev_producer_consumer/Makefile
@@ -0,0 +1,22 @@
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(c) 2016-2017 Intel Corporation.
+
+ifeq ($(RTE_SDK),)
+$(error "Please define RTE_SDK environment variable")
+endif
+
+# Default target, can be overridden by command line or environment
+RTE_TARGET ?= x86_64-native-linuxapp-gcc
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+# binary name
+APP = eventdev_producer_consumer
+
+# all source are stored in SRCS-y
+SRCS-y := main.c
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+include $(RTE_SDK)/mk/rte.extapp.mk
diff --git a/examples/eventdev_producer_consumer/main.c b/examples/eventdev_producer_consumer/main.c
new file mode 100644
index 0000000000..54a550b459
--- /dev/null
+++ b/examples/eventdev_producer_consumer/main.c
@@ -0,0 +1,670 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2022 Intel Corporation
+ */
+
+#include <getopt.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+
+#include <rte_eal.h>
+#include <rte_mempool.h>
+#include <rte_mbuf.h>
+#include <rte_launch.h>
+#include <rte_malloc.h>
+#include <rte_cycles.h>
+#include <rte_ethdev.h>
+#include <rte_eventdev.h>
+
+#define BATCH_SIZE 32
+
+static unsigned int num_workers = 4;
+static bool g_is_mbuf;
+static unsigned long num_packets = (1L << 25); /* do ~32M packets */
+static int sched_type = RTE_SCHED_TYPE_ATOMIC;
+
+struct prod_data {
+	uint8_t event_dev_id;
+	uint8_t event_port_id;
+	int32_t qid;
+};
+
+struct cons_data {
+	uint8_t event_dev_id;
+	uint8_t event_port_id;
+};
+
+struct worker_data {
+	uint8_t event_dev_id;
+	int event_port_id;
+	int32_t qid;
+};
+
+static volatile int done;
+static int quiet;
+
+#define PORT_0 0
+#define QUEUE_0 0
+static struct rte_mempool *mp;
+
+static int
+worker(void *arg)
+{
+	struct rte_event rcv_events[BATCH_SIZE] = {0};
+
+	struct worker_data *data = (struct worker_data *)arg;
+	uint8_t event_dev_id = data->event_dev_id;
+	uint8_t event_port_id = data->event_port_id;
+	int32_t qid = data->qid;
+	size_t sent = 0, received = 0;
+	uint16_t n;
+
+	if (!quiet)
+		printf("Worker core %d started, portId=%d, sending to qid=%d\n",
+		       rte_lcore_id(), event_port_id, qid);
+
+	while (!done) {
+		uint16_t k;
+		int npkts_to_send, npkts_sent = 0;
+		struct rte_event *ev;
+		uint64_t delay_start;
+
+		/* Cannot wait for IRQ here due to the way that
+		 * we check for when we are done.
+		 */
+		n = rte_event_dequeue_burst(event_dev_id,
+					    event_port_id,
+					    rcv_events,
+					    RTE_DIM(rcv_events),
+					    0);
+
+		if (n == 0) {
+			rte_pause();
+			continue;
+		} else if (!quiet)
+			printf("Worker received %d events (%zu total)\n",
+			       n, received);
+
+		delay_start = rte_rdtsc();
+		while (delay_start > rte_rdtsc())
+			;
+
+		received += n;
+
+		ev = &rcv_events[0];
+		for (k = 0; k < n; k++) {
+			ev->queue_id = qid;
+			ev->op = RTE_EVENT_OP_FORWARD;
+			ev++;
+		}
+
+		ev = &rcv_events[0];
+		npkts_to_send = n, npkts_sent = 0;
+
+		while (npkts_sent < npkts_to_send)  {
+			int iter_sent = 0;
+			iter_sent = rte_event_enqueue_burst(event_dev_id,
+							    event_port_id,
+							    &ev[npkts_sent],
+							    n - npkts_sent);
+			npkts_sent += iter_sent;
+		}
+	} /* while (!done) */
+
+	if (!quiet)
+		printf("%s %d thread done. RX=%zu TX=%zu\n",
+			__func__, rte_lcore_id(), received, sent);
+
+	return 0;
+}
+
+static int
+consumer(void *arg)
+{
+	struct rte_event events[BATCH_SIZE] = {0};
+	struct cons_data *data = (struct cons_data *)arg;
+	uint8_t event_dev_id = data->event_dev_id;
+	uint8_t event_port_id = data->event_port_id;
+	int64_t npackets = num_packets;
+	uint64_t start_time = 0;
+	uint64_t freq_khz = rte_get_timer_hz() / 1000;
+	uint16_t n;
+	uint64_t deq_start, deq_end;
+
+	deq_start = rte_rdtsc();
+	while (npackets > 0) {
+		uint16_t i;
+		n = rte_event_dequeue_burst(event_dev_id,
+					    event_port_id,
+					    events,
+					    RTE_DIM(events),
+					    0);
+
+		if (g_is_mbuf) {
+			for (i = 0; i < n; i++) {
+				/* Could pack these up and do a bulk free */
+				if (!quiet)
+					printf("%s: mbuf[%d].seqno = %"
+						PRIu64"\n", __func__, i,
+						events[i].mbuf->tx_offload);
+				if (events[i].mbuf->tx_offload < 100000000000)
+					rte_pktmbuf_free(events[i].mbuf);
+				rte_cldemote(events[i].mbuf);
+			}
+		} /* if (g_is_mbuf) */
+		npackets -= n;
+	} /* while */
+
+	deq_end = rte_rdtsc();
+	printf("Consumer done in %"PRIu64" cycles (%f cycles/evt)"
+	       " (%f pkts/sec)\n", deq_end-deq_start,
+	       (float)(deq_end - deq_start)/(float)num_packets,
+	       (float) (num_packets * rte_get_timer_hz()) /
+	       (float) (deq_end - deq_start));
+	printf("deq_end = %"PRIu64", deq_start = %"PRIu64"\n",
+	       deq_end, deq_start);
+
+	printf("Consumer done! RX=%zu, time %"PRIu64"ms\n",
+	       num_packets,
+	       (rte_get_timer_cycles() - start_time) / freq_khz);
+	done = 1;
+	return 0;
+}
+
+
+static int
+producer(void *arg)
+{
+	struct prod_data *data = (struct prod_data *)arg;
+	int64_t npackets = num_packets;
+	uint64_t mbuf_seqno = 0;
+	uint8_t event_dev_id;
+	uint8_t event_port_id;
+	int fid_counter = 0;
+	int err;
+	int64_t retry_count = 0;
+	int32_t qid = data->qid;
+	uint64_t enq_start, enq_end;
+	int k = 0;
+	struct rte_mbuf *m;
+	struct rte_event producer_events[BATCH_SIZE] = {0};
+	struct rte_event *ev = &producer_events[0];
+	int l = 0;
+	struct rte_mbuf *mbufs[BATCH_SIZE];
+
+	event_dev_id = data->event_dev_id;
+	event_port_id = data->event_port_id;
+
+	for (k = 0; k < BATCH_SIZE; k++) {
+		if (!g_is_mbuf)
+			m = NULL;
+		ev->queue_id = qid;
+		ev->priority = 0;
+		ev->mbuf = m;
+		ev->sched_type = sched_type;
+		ev->op = RTE_EVENT_OP_NEW;
+		ev++;
+	}
+
+	enq_start = rte_rdtsc();
+	do {
+		int64_t npkt_start;
+		ev = &producer_events[0];
+		retry_count = 0;
+
+		if (g_is_mbuf) {
+			err = rte_pktmbuf_alloc_bulk(mp,
+						     &mbufs[0],
+						     BATCH_SIZE);
+			if (err) {
+				printf("mbuf alloc failed after sending %"
+				       PRIu64" with err=%d\n",
+				       num_packets - npackets, err);
+				return -1;
+			}
+
+			for (l = 0; l < BATCH_SIZE; l++) {
+				m = mbufs[l];
+				/* Using tx_offload field of rte_mbuf to store
+				 * seq nums as .udata64 has been removed
+				 */
+				m->tx_offload = mbuf_seqno++;
+				producer_events[l].mbuf = m;
+				producer_events[l].flow_id = fid_counter++;
+				if (!quiet)
+					printf("%s: mbuf[%d].seqno = %"PRIu64"\n",
+						__func__, l,
+						producer_events[l].mbuf->tx_offload);
+			} /* for l = 0 - BATCH_SIZE */
+		} /* if g_is_mbuf */
+		else {
+			for (l = 0; l < BATCH_SIZE; l++)
+				producer_events[l].flow_id = fid_counter++;
+		}
+		npkt_start = npackets;
+		while (npackets > npkt_start - BATCH_SIZE) {
+			int64_t num_sent = npkt_start - npackets;
+			npackets -= rte_event_enqueue_burst(event_dev_id,
+							    event_port_id,
+							    &ev[num_sent],
+							    BATCH_SIZE -
+							    num_sent);
+		}
+	} while ((npackets > 0) && retry_count++ < 100000000000);
+
+	enq_end = rte_rdtsc();
+
+	if (npackets > 0)
+		rte_panic("%s thread failed to enqueue events\n", __func__);
+
+	if (num_packets > 0 && npackets > 0)
+		printf("npackets not sent: %"PRIu64"\n", npackets);
+
+	printf("Producer done. %"PRIu64" packets sent in %"PRIu64" cycles"
+	       "(%f cycles/evt) (%f pkts/sec)\n",
+	       num_packets, enq_end-enq_start,
+	       (float)(enq_end - enq_start)/(float)num_packets,
+	       (float) (num_packets * rte_get_timer_hz()) /
+	       (float) (enq_end - enq_start));
+	printf("enq_enq = %"PRIu64", enq_start = %"PRIu64"\n",
+	       enq_end, enq_start);
+	return 0;
+}
+
+static struct option long_options[] = {
+	{"workers", required_argument, 0, 'w'},
+	{"packets", required_argument, 0, 'n'},
+	{"ordered", no_argument, 0, 'o'},
+	{"parallel", no_argument, 0, 'u'},
+	{"quiet", no_argument, 0, 'q'},
+	{"useMbufs", no_argument, 0, 'm'},
+	{0, 0, 0, 0}
+};
+
+static void
+usage(void)
+{
+	const char *usage_str =
+		"  Usage: eventdev_producer_consumer [options]\n"
+		"  Options:\n"
+		"  -w, --workers=N       Use N workers (default 4)\n"
+		"  -n, --packets=N       Send N packets (default ~32M),"
+					 " 0 implies no limit\n"
+		"  -o, --ordered         Use ordered scheduling\n"
+		"  -u, --parallel        Use parallel scheduling\n"
+		"  -q, --quiet           Minimize printed output\n"
+		"  -m, --use-mbufs       Use mbufs for enqueue\n"
+		"\n";
+
+	fprintf(stderr, "%s", usage_str);
+	exit(1);
+}
+
+static void
+parse_app_args(int argc, char **argv)
+{
+	/* Parse cli options*/
+	int option_index;
+	int c;
+	opterr = 0;
+
+	for (;;) {
+		c = getopt_long(argc, argv, "w:n:ouqm", long_options,
+				&option_index);
+		if (c == -1)
+			break;
+
+		switch (c) {
+		case 'w':
+			num_workers = (unsigned int)atoi(optarg);
+			break;
+		case 'n':
+			num_packets = (unsigned long)atol(optarg);
+			break;
+		case 'o':
+			if (sched_type == RTE_SCHED_TYPE_PARALLEL)
+				rte_panic("Cannot specify both -o and -u\n");
+			sched_type = RTE_SCHED_TYPE_ORDERED;
+			break;
+		case 'u':
+			if (sched_type == RTE_SCHED_TYPE_ORDERED)
+				rte_panic("Cannot specify both -o and -u\n");
+			sched_type = RTE_SCHED_TYPE_PARALLEL;
+			break;
+		case 'q':
+			quiet = 1;
+			break;
+		case 'm':
+			g_is_mbuf = true;
+			break;
+		default:
+			usage();
+		}
+	}
+}
+
+static uint8_t
+setup_event_dev(struct prod_data *prod_data,
+		struct cons_data *cons_data,
+		struct worker_data *worker_data,
+		int id)
+{
+	struct rte_event_dev_info dev_info;
+	struct rte_event_dev_config config = {0};
+	struct rte_event_queue_conf queue_config = {0};
+	struct rte_event_port_conf port_config = {0};
+	uint8_t queue_id;
+	uint8_t priority;
+	int prod_port = 0;
+	int cons_port = 1;
+	int worker_port_base = 2;
+	int prod_qid = 0;
+	int cons_qid = 1;
+	int worker_qid = 2;
+	unsigned int i;
+	int ret;
+
+	/* Better yet, always use event dev 0 so the app can use either. You can
+	 * check that there's at least 1 eventdev with rte_event_dev_count().
+	 */
+
+	if (id < 0)
+		rte_panic("%s: invalid ev_dev ID %d\n", __func__, id);
+	else
+		printf("%s: ev_dev ID %d\n", __func__, id);
+
+	rte_event_dev_info_get(id, &dev_info);
+
+	if (num_workers)
+		config.nb_event_queues = 3;
+	else
+		config.nb_event_queues = 2;
+
+	config.nb_single_link_event_port_queues = 2;
+	config.nb_event_ports = num_workers +
+		config.nb_single_link_event_port_queues;
+	config.nb_events_limit = dev_info.max_num_events;
+	config.nb_event_queue_flows = dev_info.max_event_queue_flows;
+	config.nb_event_port_dequeue_depth =
+					dev_info.max_event_port_dequeue_depth;
+	config.nb_event_port_enqueue_depth =
+					dev_info.max_event_port_enqueue_depth;
+	config.dequeue_timeout_ns = 0;
+	config.event_dev_cfg = RTE_EVENT_DEV_CFG_PER_DEQUEUE_TIMEOUT;
+
+	ret = rte_event_dev_configure(id, &config);
+	if (ret)
+		rte_panic("Failed to configure the event dev\n");
+	else
+		printf("eventdev configured!\n");
+
+	/* Create queues */
+	queue_config.event_queue_cfg = 0;
+	queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+	queue_config.nb_atomic_order_sequences =
+		(sched_type == RTE_SCHED_TYPE_ORDERED) ? 1024 : 0;
+	queue_config.nb_atomic_flows = dev_info.max_event_queue_flows;
+	queue_config.schedule_type = sched_type;
+
+	if (num_workers) {
+		ret = rte_event_queue_setup(id, worker_qid, &queue_config);
+		if (ret < 0)
+			rte_panic("Failed to create the scheduled QID\n");
+		else
+			printf("rte_event_queue_setup success for worker_qid\n");
+	}
+
+	queue_config.event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK;
+	queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	cons_qid = 1;
+	ret = rte_event_queue_setup(id, cons_qid, &queue_config);
+	if (ret < 0)
+		rte_panic("Failed to create the cons directed QID\n");
+	else
+		printf("rte_event_queue_setup success for cons_qid\n");
+
+	queue_config.event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK;
+	queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	prod_qid = 0;
+	ret = rte_event_queue_setup(id, prod_qid, &queue_config);
+	if (ret < 0)
+		rte_panic("Failed to create the prod directed QID\n");
+	else
+		printf("rte_event_queue_setup success for prod_qid\n");
+
+	/* Create two directed ports */
+
+	port_config.dequeue_depth = dev_info.max_event_port_dequeue_depth;
+	port_config.enqueue_depth = dev_info.max_event_port_enqueue_depth;
+
+	/* Set producer new event threshold to 3/4 max */
+	port_config.new_event_threshold = 3 * (dev_info.max_num_events >> 2);
+	port_config.event_port_cfg = RTE_EVENT_PORT_CFG_SINGLE_LINK;
+	ret = rte_event_port_setup(id, prod_port, &port_config);
+	if (ret < 0)
+		rte_panic("Failed to create the producer port\n");
+	else
+		printf("rte_event_port_setup for prod_port ok\n");
+
+	/* Set consumer and worker new event threshold to max */
+	port_config.new_event_threshold = dev_info.max_num_events;
+	ret = rte_event_port_setup(id, cons_port, &port_config);
+	if (ret < 0)
+		rte_panic("Failed to create the consumer port\n");
+	else
+		printf("rte_event_port_setup for cons_port ok\n");
+
+	port_config.event_port_cfg = 0;
+
+	/* Create load-balanced worker ports */
+	for (i = 0; i < num_workers; i++) {
+		worker_data[i].event_port_id = i + worker_port_base;
+		ret = rte_event_port_setup(id, worker_data[i].event_port_id,
+					   &port_config);
+		if (ret < 0)
+			rte_panic("Failed to create worker port #%d\n", i);
+		else
+			printf("rte_event_port_setup for worker port %d ok\n",
+			       i);
+	}
+
+	printf("link worker queues\n");
+	/* Map ports/qids */
+	for (i = 0; i < num_workers; i++) {
+		queue_id = worker_qid;
+		priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+		ret = rte_event_port_link(id, worker_data[i].event_port_id,
+					  &queue_id, &priority, 1);
+		if (ret != 1)
+			rte_panic("Failed to map worker%d port to worker_qid\n",
+				  i);
+	}
+
+	printf("link consumer queue\n");
+	/* Link consumer port to its QID */
+	queue_id = cons_qid;
+	priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	ret = rte_event_port_link(id, cons_port, &queue_id, &priority, 1);
+	if (ret != 1)
+		rte_panic("Failed to map consumer port to cons_qid\n");
+
+	printf("link producer queue\n");
+	/* Link producer port to its QID */
+	queue_id = prod_qid;
+	priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	ret = rte_event_port_link(id, prod_port, &queue_id, &priority, 1);
+	if (ret != 1)
+		rte_panic("Failed to map producer port to prod_qid\n");
+
+	/* Dispatch to workers */
+	if (num_workers) {
+		*prod_data = (struct prod_data){.event_dev_id = id,
+						.event_port_id = prod_port,
+						.qid = worker_qid};
+		*cons_data = (struct cons_data){.event_dev_id = id,
+						.event_port_id = cons_port};
+		for (i = 0; i < num_workers; i++) {
+			struct worker_data *w = &worker_data[i];
+			w->event_dev_id = id;
+			w->qid = cons_qid;
+		}
+	} else {
+		*prod_data = (struct prod_data){.event_dev_id = id,
+						.event_port_id = prod_port,
+						.qid = cons_qid};
+		*cons_data = (struct cons_data){.event_dev_id = id,
+						.event_port_id = cons_port};
+	}
+
+	ret = rte_event_dev_start(id);
+	if (ret)
+		rte_panic("Failed to start the event dev\n");
+	if (g_is_mbuf) {
+		mp = rte_pktmbuf_pool_create("packet_pool",
+				/* mbufs */ dev_info.max_num_events,
+				/* cache_size */ 512,
+				/* priv_size*/ 0,
+				/* data_room_size */ 2048,
+				rte_socket_id());
+
+		if (mp == NULL) {
+			printf("mbuf pool create failed\n");
+			return -1;
+		}
+	}
+	return (uint8_t) id;
+}
+
+int
+main(int argc, char **argv)
+{
+	struct prod_data prod_data = {0};
+	struct cons_data cons_data = {0};
+	uint64_t start, end;
+	struct worker_data *worker_data = NULL;
+	unsigned int nworkers = 0;
+	int lcore_id;
+	int err;
+	int has_prod = 0;
+	int has_cons = 0;
+	int evdev_id = 0; /* TODO - allow app to override */
+
+	done = 0;
+	quiet = 0;
+	mp = NULL;
+	g_is_mbuf = false;
+
+	err = rte_eal_init(argc, argv);
+	if (err < 0)
+		rte_panic("Invalid EAL arguments\n");
+
+	argc -= err;
+	argv += err;
+
+	/* Parse cli options*/
+	parse_app_args(argc, argv);
+
+	if (!quiet) {
+		printf("  Config:\n");
+		printf("\tworkers: %d\n", num_workers);
+		printf("\tpackets: %"PRIu64"\n", num_packets);
+		if (sched_type == RTE_SCHED_TYPE_ORDERED)
+			printf("\tworker_qid type: ordered\n");
+		if (sched_type == RTE_SCHED_TYPE_ATOMIC)
+			printf("\tworker_qid type: atomic\n");
+		printf("\n");
+	}
+
+	const unsigned int cores_needed = num_workers +
+			/*main*/ 1 +
+			/*producer*/ 1 +
+			/*consumer*/ 1;
+
+	if (!quiet) {
+		printf("Number of cores available: %d\n", rte_lcore_count());
+		printf("Number of cores to be used: %d\n", cores_needed);
+	}
+
+	if (rte_lcore_count() < cores_needed)
+		rte_panic("Too few cores\n");
+
+	const uint8_t ndevs = rte_event_dev_count();
+	if (ndevs == 0)
+		rte_panic(
+			"No event devs found. Do you need"
+			" to pass in a --vdev flag?\n");
+	if (ndevs > 1)
+		fprintf(stderr,
+			"Warning: More than one event dev, but using idx 0");
+
+	if (num_workers) {
+		worker_data = rte_calloc(0, num_workers,
+					 sizeof(worker_data[0]), 0);
+		if (worker_data == NULL)
+			rte_panic("rte_calloc failed\n");
+	}
+
+	uint8_t id = setup_event_dev(&prod_data, &cons_data, worker_data,
+				    evdev_id);
+
+	printf("setup_event_dev returned eventdev_id = %d\n", id);
+
+	start = rte_rdtsc();
+
+	RTE_LCORE_FOREACH_WORKER(lcore_id) {
+		if (has_prod && has_cons && nworkers == num_workers)
+			break;
+
+		if (!has_prod) {
+			err = rte_eal_remote_launch(producer, &prod_data,
+						    lcore_id);
+			if (err)
+				rte_panic("Failed to launch producer\n");
+			printf("Launched producer\n");
+			has_prod = 1;
+			continue;
+		}
+
+		if (!has_cons) {
+			err = rte_eal_remote_launch(consumer, &cons_data,
+						    lcore_id);
+			if (err)
+				rte_panic("Failed to launch consumer\n");
+			printf("Launched consumer\n");
+			has_cons = 1;
+			continue;
+		}
+
+		if (nworkers < num_workers) {
+			err = rte_eal_remote_launch(worker,
+						    &worker_data[nworkers],
+						    lcore_id);
+			if (err)
+				rte_panic("Failed to launch worker%d\n",
+					  nworkers);
+			nworkers++;
+			printf("Launched worker %d\n", nworkers);
+			continue;
+		}
+	}
+
+	rte_eal_mp_wait_lcore();
+	end = rte_rdtsc();
+	printf("[%s()] DLB scheduled %"PRIu64" packets in %"PRIu64" cycles\n",
+	       __func__, num_packets, end - start);
+	printf("[%s()] \t %f packets/sec\n",
+	       __func__, (float) (num_packets * rte_get_timer_hz()) /
+	       (float) (end - start));
+
+	printf("Test Complete\n");
+
+	/* Cleanup done automatically by kernel on app exit */
+
+	return 0;
+}
diff --git a/examples/eventdev_producer_consumer/meson.build b/examples/eventdev_producer_consumer/meson.build
new file mode 100644
index 0000000000..e18739432c
--- /dev/null
+++ b/examples/eventdev_producer_consumer/meson.build
@@ -0,0 +1,13 @@
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(c) 2017 Intel Corporation
+
+# meson file, for building this example as part of a main DPDK build.
+#
+# To build this example as a standalone application with an already-installed
+# DPDK instance, use 'make'
+
+allow_experimental_apis = true
+deps += 'eventdev'
+sources = files(
+        'main.c',
+)
diff --git a/examples/meson.build b/examples/meson.build
index 81e93799f2..9d95ecc8e0 100644
--- a/examples/meson.build
+++ b/examples/meson.build
@@ -15,6 +15,7 @@ all_examples = [
         'dma',
         'ethtool',
         'eventdev_pipeline',
+	'eventdev_producer_consumer',
         'fips_validation',
         'flow_classify',
         'flow_filtering',
-- 
2.25.1


^ permalink raw reply	[flat|nested] 5+ messages in thread

* [PATCH v2] examples/eventdev_producer_consumer: fix 32-bit checkpatch issues
  2022-08-18 17:02 [PATCH] examples: add eventdev_producer_consumer example Timothy McDaniel
@ 2022-08-22 19:26 ` Timothy McDaniel
  2022-08-22 20:04 ` [PATCH v3] examples: add eventdev_producer_consumer example Timothy McDaniel
  2022-08-22 20:51 ` [PATCH v4] " Timothy McDaniel
  2 siblings, 0 replies; 5+ messages in thread
From: Timothy McDaniel @ 2022-08-22 19:26 UTC (permalink / raw)
  To: jerinj; +Cc: mdr, thomas, dev

Fixed style and format issues, primarily those involving data types
whose size varies depending on whether we are building for 32 or
64 bit platforms.

Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>
---
 examples/eventdev_producer_consumer/main.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/examples/eventdev_producer_consumer/main.c b/examples/eventdev_producer_consumer/main.c
index 54a550b459..164bdf6f74 100644
--- a/examples/eventdev_producer_consumer/main.c
+++ b/examples/eventdev_producer_consumer/main.c
@@ -21,7 +21,7 @@
 
 static unsigned int num_workers = 4;
 static bool g_is_mbuf;
-static unsigned long num_packets = (1L << 25); /* do ~32M packets */
+static uint64_t num_packets = (1L << 25); /* do ~32M packets */
 static int sched_type = RTE_SCHED_TYPE_ATOMIC;
 
 struct prod_data {
@@ -51,13 +51,13 @@ static struct rte_mempool *mp;
 static int
 worker(void *arg)
 {
-	struct rte_event rcv_events[BATCH_SIZE] = {0};
+	struct rte_event rcv_events[BATCH_SIZE];
 
 	struct worker_data *data = (struct worker_data *)arg;
 	uint8_t event_dev_id = data->event_dev_id;
 	uint8_t event_port_id = data->event_port_id;
 	int32_t qid = data->qid;
-	size_t sent = 0, received = 0;
+	uint64_t sent = 0, received = 0;
 	uint16_t n;
 
 	if (!quiet)
@@ -83,7 +83,7 @@ worker(void *arg)
 			rte_pause();
 			continue;
 		} else if (!quiet)
-			printf("Worker received %d events (%zu total)\n",
+			printf("Worker received %d events(%"PRIu64" total)\n",
 			       n, received);
 
 		delay_start = rte_rdtsc();
@@ -113,7 +113,7 @@ worker(void *arg)
 	} /* while (!done) */
 
 	if (!quiet)
-		printf("%s %d thread done. RX=%zu TX=%zu\n",
+		printf("%s %d thread done. RX= %"PRIu64" TX= %"PRIu64"\n",
 			__func__, rte_lcore_id(), received, sent);
 
 	return 0;
@@ -122,7 +122,7 @@ worker(void *arg)
 static int
 consumer(void *arg)
 {
-	struct rte_event events[BATCH_SIZE] = {0};
+	struct rte_event events[BATCH_SIZE];
 	struct cons_data *data = (struct cons_data *)arg;
 	uint8_t event_dev_id = data->event_dev_id;
 	uint8_t event_port_id = data->event_port_id;
@@ -165,7 +165,7 @@ consumer(void *arg)
 	printf("deq_end = %"PRIu64", deq_start = %"PRIu64"\n",
 	       deq_end, deq_start);
 
-	printf("Consumer done! RX=%zu, time %"PRIu64"ms\n",
+	printf("Consumer done! RX=%"PRIu64", time %"PRIu64"ms\n",
 	       num_packets,
 	       (rte_get_timer_cycles() - start_time) / freq_khz);
 	done = 1;
@@ -188,7 +188,7 @@ producer(void *arg)
 	uint64_t enq_start, enq_end;
 	int k = 0;
 	struct rte_mbuf *m;
-	struct rte_event producer_events[BATCH_SIZE] = {0};
+	struct rte_event producer_events[BATCH_SIZE];
 	struct rte_event *ev = &producer_events[0];
 	int l = 0;
 	struct rte_mbuf *mbufs[BATCH_SIZE];
@@ -263,7 +263,7 @@ producer(void *arg)
 
 	printf("Producer done. %"PRIu64" packets sent in %"PRIu64" cycles"
 	       "(%f cycles/evt) (%f pkts/sec)\n",
-	       num_packets, enq_end-enq_start,
+	       num_packets, enq_end - enq_start,
 	       (float)(enq_end - enq_start)/(float)num_packets,
 	       (float) (num_packets * rte_get_timer_hz()) /
 	       (float) (enq_end - enq_start));
-- 
2.25.1


^ permalink raw reply	[flat|nested] 5+ messages in thread

* [PATCH v3] examples: add eventdev_producer_consumer example
  2022-08-18 17:02 [PATCH] examples: add eventdev_producer_consumer example Timothy McDaniel
  2022-08-22 19:26 ` [PATCH v2] examples/eventdev_producer_consumer: fix 32-bit checkpatch issues Timothy McDaniel
@ 2022-08-22 20:04 ` Timothy McDaniel
  2022-08-22 20:51 ` [PATCH v4] " Timothy McDaniel
  2 siblings, 0 replies; 5+ messages in thread
From: Timothy McDaniel @ 2022-08-22 20:04 UTC (permalink / raw)
  To: jerinj; +Cc: mdr, thomas, dev

The eventdev-producer-consumer application is a single-stage
producer-worker-consumer pipeline sample to mimic real-world applications.
It is useful in measuring performance impact when any eventdev
configuration is changed. Unlike test-eventdev, it allows configuring a
load balanced queue between the producer and workers and a single-link
queue between the workers and consumer. With test-eventdev, multiple worker
stages can be configured but there is no single consumer receiving events
from all the workers. It also does not require configuring TX/RX adapters
like in the case of eventdev_pipeline app.

Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>

---

V3: Fixed style and format issues, primarily those involving data types
whose size varies depending on whether we are building for 32 or
64 bit platforms.

V2: Disregard - forgot to resubmit entire patch

---
---
 examples/eventdev_producer_consumer/Makefile  |  22 +
 examples/eventdev_producer_consumer/main.c    | 670 ++++++++++++++++++
 .../eventdev_producer_consumer/meson.build    |  13 +
 examples/meson.build                          |   1 +
 4 files changed, 706 insertions(+)
 create mode 100644 examples/eventdev_producer_consumer/Makefile
 create mode 100644 examples/eventdev_producer_consumer/main.c
 create mode 100644 examples/eventdev_producer_consumer/meson.build

diff --git a/examples/eventdev_producer_consumer/Makefile b/examples/eventdev_producer_consumer/Makefile
new file mode 100644
index 0000000000..761689eab7
--- /dev/null
+++ b/examples/eventdev_producer_consumer/Makefile
@@ -0,0 +1,22 @@
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(c) 2016-2017 Intel Corporation.
+
+ifeq ($(RTE_SDK),)
+$(error "Please define RTE_SDK environment variable")
+endif
+
+# Default target, can be overridden by command line or environment
+RTE_TARGET ?= x86_64-native-linuxapp-gcc
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+# binary name
+APP = eventdev_producer_consumer
+
+# all source are stored in SRCS-y
+SRCS-y := main.c
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+include $(RTE_SDK)/mk/rte.extapp.mk
diff --git a/examples/eventdev_producer_consumer/main.c b/examples/eventdev_producer_consumer/main.c
new file mode 100644
index 0000000000..164bdf6f74
--- /dev/null
+++ b/examples/eventdev_producer_consumer/main.c
@@ -0,0 +1,670 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2022 Intel Corporation
+ */
+
+#include <getopt.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+
+#include <rte_eal.h>
+#include <rte_mempool.h>
+#include <rte_mbuf.h>
+#include <rte_launch.h>
+#include <rte_malloc.h>
+#include <rte_cycles.h>
+#include <rte_ethdev.h>
+#include <rte_eventdev.h>
+
+#define BATCH_SIZE 32
+
+static unsigned int num_workers = 4;
+static bool g_is_mbuf;
+static uint64_t num_packets = (1L << 25); /* do ~32M packets */
+static int sched_type = RTE_SCHED_TYPE_ATOMIC;
+
+struct prod_data {
+	uint8_t event_dev_id;
+	uint8_t event_port_id;
+	int32_t qid;
+};
+
+struct cons_data {
+	uint8_t event_dev_id;
+	uint8_t event_port_id;
+};
+
+struct worker_data {
+	uint8_t event_dev_id;
+	int event_port_id;
+	int32_t qid;
+};
+
+static volatile int done;
+static int quiet;
+
+#define PORT_0 0
+#define QUEUE_0 0
+static struct rte_mempool *mp;
+
+static int
+worker(void *arg)
+{
+	struct rte_event rcv_events[BATCH_SIZE];
+
+	struct worker_data *data = (struct worker_data *)arg;
+	uint8_t event_dev_id = data->event_dev_id;
+	uint8_t event_port_id = data->event_port_id;
+	int32_t qid = data->qid;
+	uint64_t sent = 0, received = 0;
+	uint16_t n;
+
+	if (!quiet)
+		printf("Worker core %d started, portId=%d, sending to qid=%d\n",
+		       rte_lcore_id(), event_port_id, qid);
+
+	while (!done) {
+		uint16_t k;
+		int npkts_to_send, npkts_sent = 0;
+		struct rte_event *ev;
+		uint64_t delay_start;
+
+		/* Cannot wait for IRQ here due to the way that
+		 * we check for when we are done.
+		 */
+		n = rte_event_dequeue_burst(event_dev_id,
+					    event_port_id,
+					    rcv_events,
+					    RTE_DIM(rcv_events),
+					    0);
+
+		if (n == 0) {
+			rte_pause();
+			continue;
+		} else if (!quiet)
+			printf("Worker received %d events(%"PRIu64" total)\n",
+			       n, received);
+
+		delay_start = rte_rdtsc();
+		while (delay_start > rte_rdtsc())
+			;
+
+		received += n;
+
+		ev = &rcv_events[0];
+		for (k = 0; k < n; k++) {
+			ev->queue_id = qid;
+			ev->op = RTE_EVENT_OP_FORWARD;
+			ev++;
+		}
+
+		ev = &rcv_events[0];
+		npkts_to_send = n, npkts_sent = 0;
+
+		while (npkts_sent < npkts_to_send)  {
+			int iter_sent = 0;
+			iter_sent = rte_event_enqueue_burst(event_dev_id,
+							    event_port_id,
+							    &ev[npkts_sent],
+							    n - npkts_sent);
+			npkts_sent += iter_sent;
+		}
+	} /* while (!done) */
+
+	if (!quiet)
+		printf("%s %d thread done. RX= %"PRIu64" TX= %"PRIu64"\n",
+			__func__, rte_lcore_id(), received, sent);
+
+	return 0;
+}
+
+static int
+consumer(void *arg)
+{
+	struct rte_event events[BATCH_SIZE];
+	struct cons_data *data = (struct cons_data *)arg;
+	uint8_t event_dev_id = data->event_dev_id;
+	uint8_t event_port_id = data->event_port_id;
+	int64_t npackets = num_packets;
+	uint64_t start_time = 0;
+	uint64_t freq_khz = rte_get_timer_hz() / 1000;
+	uint16_t n;
+	uint64_t deq_start, deq_end;
+
+	deq_start = rte_rdtsc();
+	while (npackets > 0) {
+		uint16_t i;
+		n = rte_event_dequeue_burst(event_dev_id,
+					    event_port_id,
+					    events,
+					    RTE_DIM(events),
+					    0);
+
+		if (g_is_mbuf) {
+			for (i = 0; i < n; i++) {
+				/* Could pack these up and do a bulk free */
+				if (!quiet)
+					printf("%s: mbuf[%d].seqno = %"
+						PRIu64"\n", __func__, i,
+						events[i].mbuf->tx_offload);
+				if (events[i].mbuf->tx_offload < 100000000000)
+					rte_pktmbuf_free(events[i].mbuf);
+				rte_cldemote(events[i].mbuf);
+			}
+		} /* if (g_is_mbuf) */
+		npackets -= n;
+	} /* while */
+
+	deq_end = rte_rdtsc();
+	printf("Consumer done in %"PRIu64" cycles (%f cycles/evt)"
+	       " (%f pkts/sec)\n", deq_end-deq_start,
+	       (float)(deq_end - deq_start)/(float)num_packets,
+	       (float) (num_packets * rte_get_timer_hz()) /
+	       (float) (deq_end - deq_start));
+	printf("deq_end = %"PRIu64", deq_start = %"PRIu64"\n",
+	       deq_end, deq_start);
+
+	printf("Consumer done! RX=%"PRIu64", time %"PRIu64"ms\n",
+	       num_packets,
+	       (rte_get_timer_cycles() - start_time) / freq_khz);
+	done = 1;
+	return 0;
+}
+
+
+static int
+producer(void *arg)
+{
+	struct prod_data *data = (struct prod_data *)arg;
+	int64_t npackets = num_packets;
+	uint64_t mbuf_seqno = 0;
+	uint8_t event_dev_id;
+	uint8_t event_port_id;
+	int fid_counter = 0;
+	int err;
+	int64_t retry_count = 0;
+	int32_t qid = data->qid;
+	uint64_t enq_start, enq_end;
+	int k = 0;
+	struct rte_mbuf *m;
+	struct rte_event producer_events[BATCH_SIZE];
+	struct rte_event *ev = &producer_events[0];
+	int l = 0;
+	struct rte_mbuf *mbufs[BATCH_SIZE];
+
+	event_dev_id = data->event_dev_id;
+	event_port_id = data->event_port_id;
+
+	for (k = 0; k < BATCH_SIZE; k++) {
+		if (!g_is_mbuf)
+			m = NULL;
+		ev->queue_id = qid;
+		ev->priority = 0;
+		ev->mbuf = m;
+		ev->sched_type = sched_type;
+		ev->op = RTE_EVENT_OP_NEW;
+		ev++;
+	}
+
+	enq_start = rte_rdtsc();
+	do {
+		int64_t npkt_start;
+		ev = &producer_events[0];
+		retry_count = 0;
+
+		if (g_is_mbuf) {
+			err = rte_pktmbuf_alloc_bulk(mp,
+						     &mbufs[0],
+						     BATCH_SIZE);
+			if (err) {
+				printf("mbuf alloc failed after sending %"
+				       PRIu64" with err=%d\n",
+				       num_packets - npackets, err);
+				return -1;
+			}
+
+			for (l = 0; l < BATCH_SIZE; l++) {
+				m = mbufs[l];
+				/* Using tx_offload field of rte_mbuf to store
+				 * seq nums as .udata64 has been removed
+				 */
+				m->tx_offload = mbuf_seqno++;
+				producer_events[l].mbuf = m;
+				producer_events[l].flow_id = fid_counter++;
+				if (!quiet)
+					printf("%s: mbuf[%d].seqno = %"PRIu64"\n",
+						__func__, l,
+						producer_events[l].mbuf->tx_offload);
+			} /* for l = 0 - BATCH_SIZE */
+		} /* if g_is_mbuf */
+		else {
+			for (l = 0; l < BATCH_SIZE; l++)
+				producer_events[l].flow_id = fid_counter++;
+		}
+		npkt_start = npackets;
+		while (npackets > npkt_start - BATCH_SIZE) {
+			int64_t num_sent = npkt_start - npackets;
+			npackets -= rte_event_enqueue_burst(event_dev_id,
+							    event_port_id,
+							    &ev[num_sent],
+							    BATCH_SIZE -
+							    num_sent);
+		}
+	} while ((npackets > 0) && retry_count++ < 100000000000);
+
+	enq_end = rte_rdtsc();
+
+	if (npackets > 0)
+		rte_panic("%s thread failed to enqueue events\n", __func__);
+
+	if (num_packets > 0 && npackets > 0)
+		printf("npackets not sent: %"PRIu64"\n", npackets);
+
+	printf("Producer done. %"PRIu64" packets sent in %"PRIu64" cycles"
+	       "(%f cycles/evt) (%f pkts/sec)\n",
+	       num_packets, enq_end - enq_start,
+	       (float)(enq_end - enq_start)/(float)num_packets,
+	       (float) (num_packets * rte_get_timer_hz()) /
+	       (float) (enq_end - enq_start));
+	printf("enq_enq = %"PRIu64", enq_start = %"PRIu64"\n",
+	       enq_end, enq_start);
+	return 0;
+}
+
+static struct option long_options[] = {
+	{"workers", required_argument, 0, 'w'},
+	{"packets", required_argument, 0, 'n'},
+	{"ordered", no_argument, 0, 'o'},
+	{"parallel", no_argument, 0, 'u'},
+	{"quiet", no_argument, 0, 'q'},
+	{"useMbufs", no_argument, 0, 'm'},
+	{0, 0, 0, 0}
+};
+
+static void
+usage(void)
+{
+	const char *usage_str =
+		"  Usage: eventdev_producer_consumer [options]\n"
+		"  Options:\n"
+		"  -w, --workers=N       Use N workers (default 4)\n"
+		"  -n, --packets=N       Send N packets (default ~32M),"
+					 " 0 implies no limit\n"
+		"  -o, --ordered         Use ordered scheduling\n"
+		"  -u, --parallel        Use parallel scheduling\n"
+		"  -q, --quiet           Minimize printed output\n"
+		"  -m, --use-mbufs       Use mbufs for enqueue\n"
+		"\n";
+
+	fprintf(stderr, "%s", usage_str);
+	exit(1);
+}
+
+static void
+parse_app_args(int argc, char **argv)
+{
+	/* Parse cli options*/
+	int option_index;
+	int c;
+	opterr = 0;
+
+	for (;;) {
+		c = getopt_long(argc, argv, "w:n:ouqm", long_options,
+				&option_index);
+		if (c == -1)
+			break;
+
+		switch (c) {
+		case 'w':
+			num_workers = (unsigned int)atoi(optarg);
+			break;
+		case 'n':
+			num_packets = (unsigned long)atol(optarg);
+			break;
+		case 'o':
+			if (sched_type == RTE_SCHED_TYPE_PARALLEL)
+				rte_panic("Cannot specify both -o and -u\n");
+			sched_type = RTE_SCHED_TYPE_ORDERED;
+			break;
+		case 'u':
+			if (sched_type == RTE_SCHED_TYPE_ORDERED)
+				rte_panic("Cannot specify both -o and -u\n");
+			sched_type = RTE_SCHED_TYPE_PARALLEL;
+			break;
+		case 'q':
+			quiet = 1;
+			break;
+		case 'm':
+			g_is_mbuf = true;
+			break;
+		default:
+			usage();
+		}
+	}
+}
+
+static uint8_t
+setup_event_dev(struct prod_data *prod_data,
+		struct cons_data *cons_data,
+		struct worker_data *worker_data,
+		int id)
+{
+	struct rte_event_dev_info dev_info;
+	struct rte_event_dev_config config = {0};
+	struct rte_event_queue_conf queue_config = {0};
+	struct rte_event_port_conf port_config = {0};
+	uint8_t queue_id;
+	uint8_t priority;
+	int prod_port = 0;
+	int cons_port = 1;
+	int worker_port_base = 2;
+	int prod_qid = 0;
+	int cons_qid = 1;
+	int worker_qid = 2;
+	unsigned int i;
+	int ret;
+
+	/* Better yet, always use event dev 0 so the app can use either. You can
+	 * check that there's at least 1 eventdev with rte_event_dev_count().
+	 */
+
+	if (id < 0)
+		rte_panic("%s: invalid ev_dev ID %d\n", __func__, id);
+	else
+		printf("%s: ev_dev ID %d\n", __func__, id);
+
+	rte_event_dev_info_get(id, &dev_info);
+
+	if (num_workers)
+		config.nb_event_queues = 3;
+	else
+		config.nb_event_queues = 2;
+
+	config.nb_single_link_event_port_queues = 2;
+	config.nb_event_ports = num_workers +
+		config.nb_single_link_event_port_queues;
+	config.nb_events_limit = dev_info.max_num_events;
+	config.nb_event_queue_flows = dev_info.max_event_queue_flows;
+	config.nb_event_port_dequeue_depth =
+					dev_info.max_event_port_dequeue_depth;
+	config.nb_event_port_enqueue_depth =
+					dev_info.max_event_port_enqueue_depth;
+	config.dequeue_timeout_ns = 0;
+	config.event_dev_cfg = RTE_EVENT_DEV_CFG_PER_DEQUEUE_TIMEOUT;
+
+	ret = rte_event_dev_configure(id, &config);
+	if (ret)
+		rte_panic("Failed to configure the event dev\n");
+	else
+		printf("eventdev configured!\n");
+
+	/* Create queues */
+	queue_config.event_queue_cfg = 0;
+	queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+	queue_config.nb_atomic_order_sequences =
+		(sched_type == RTE_SCHED_TYPE_ORDERED) ? 1024 : 0;
+	queue_config.nb_atomic_flows = dev_info.max_event_queue_flows;
+	queue_config.schedule_type = sched_type;
+
+	if (num_workers) {
+		ret = rte_event_queue_setup(id, worker_qid, &queue_config);
+		if (ret < 0)
+			rte_panic("Failed to create the scheduled QID\n");
+		else
+			printf("rte_event_queue_setup success for worker_qid\n");
+	}
+
+	queue_config.event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK;
+	queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	cons_qid = 1;
+	ret = rte_event_queue_setup(id, cons_qid, &queue_config);
+	if (ret < 0)
+		rte_panic("Failed to create the cons directed QID\n");
+	else
+		printf("rte_event_queue_setup success for cons_qid\n");
+
+	queue_config.event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK;
+	queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	prod_qid = 0;
+	ret = rte_event_queue_setup(id, prod_qid, &queue_config);
+	if (ret < 0)
+		rte_panic("Failed to create the prod directed QID\n");
+	else
+		printf("rte_event_queue_setup success for prod_qid\n");
+
+	/* Create two directed ports */
+
+	port_config.dequeue_depth = dev_info.max_event_port_dequeue_depth;
+	port_config.enqueue_depth = dev_info.max_event_port_enqueue_depth;
+
+	/* Set producer new event threshold to 3/4 max */
+	port_config.new_event_threshold = 3 * (dev_info.max_num_events >> 2);
+	port_config.event_port_cfg = RTE_EVENT_PORT_CFG_SINGLE_LINK;
+	ret = rte_event_port_setup(id, prod_port, &port_config);
+	if (ret < 0)
+		rte_panic("Failed to create the producer port\n");
+	else
+		printf("rte_event_port_setup for prod_port ok\n");
+
+	/* Set consumer and worker new event threshold to max */
+	port_config.new_event_threshold = dev_info.max_num_events;
+	ret = rte_event_port_setup(id, cons_port, &port_config);
+	if (ret < 0)
+		rte_panic("Failed to create the consumer port\n");
+	else
+		printf("rte_event_port_setup for cons_port ok\n");
+
+	port_config.event_port_cfg = 0;
+
+	/* Create load-balanced worker ports */
+	for (i = 0; i < num_workers; i++) {
+		worker_data[i].event_port_id = i + worker_port_base;
+		ret = rte_event_port_setup(id, worker_data[i].event_port_id,
+					   &port_config);
+		if (ret < 0)
+			rte_panic("Failed to create worker port #%d\n", i);
+		else
+			printf("rte_event_port_setup for worker port %d ok\n",
+			       i);
+	}
+
+	printf("link worker queues\n");
+	/* Map ports/qids */
+	for (i = 0; i < num_workers; i++) {
+		queue_id = worker_qid;
+		priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+		ret = rte_event_port_link(id, worker_data[i].event_port_id,
+					  &queue_id, &priority, 1);
+		if (ret != 1)
+			rte_panic("Failed to map worker%d port to worker_qid\n",
+				  i);
+	}
+
+	printf("link consumer queue\n");
+	/* Link consumer port to its QID */
+	queue_id = cons_qid;
+	priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	ret = rte_event_port_link(id, cons_port, &queue_id, &priority, 1);
+	if (ret != 1)
+		rte_panic("Failed to map consumer port to cons_qid\n");
+
+	printf("link producer queue\n");
+	/* Link producer port to its QID */
+	queue_id = prod_qid;
+	priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	ret = rte_event_port_link(id, prod_port, &queue_id, &priority, 1);
+	if (ret != 1)
+		rte_panic("Failed to map producer port to prod_qid\n");
+
+	/* Dispatch to workers */
+	if (num_workers) {
+		*prod_data = (struct prod_data){.event_dev_id = id,
+						.event_port_id = prod_port,
+						.qid = worker_qid};
+		*cons_data = (struct cons_data){.event_dev_id = id,
+						.event_port_id = cons_port};
+		for (i = 0; i < num_workers; i++) {
+			struct worker_data *w = &worker_data[i];
+			w->event_dev_id = id;
+			w->qid = cons_qid;
+		}
+	} else {
+		*prod_data = (struct prod_data){.event_dev_id = id,
+						.event_port_id = prod_port,
+						.qid = cons_qid};
+		*cons_data = (struct cons_data){.event_dev_id = id,
+						.event_port_id = cons_port};
+	}
+
+	ret = rte_event_dev_start(id);
+	if (ret)
+		rte_panic("Failed to start the event dev\n");
+	if (g_is_mbuf) {
+		mp = rte_pktmbuf_pool_create("packet_pool",
+				/* mbufs */ dev_info.max_num_events,
+				/* cache_size */ 512,
+				/* priv_size*/ 0,
+				/* data_room_size */ 2048,
+				rte_socket_id());
+
+		if (mp == NULL) {
+			printf("mbuf pool create failed\n");
+			return -1;
+		}
+	}
+	return (uint8_t) id;
+}
+
+int
+main(int argc, char **argv)
+{
+	struct prod_data prod_data = {0};
+	struct cons_data cons_data = {0};
+	uint64_t start, end;
+	struct worker_data *worker_data = NULL;
+	unsigned int nworkers = 0;
+	int lcore_id;
+	int err;
+	int has_prod = 0;
+	int has_cons = 0;
+	int evdev_id = 0; /* TODO - allow app to override */
+
+	done = 0;
+	quiet = 0;
+	mp = NULL;
+	g_is_mbuf = false;
+
+	err = rte_eal_init(argc, argv);
+	if (err < 0)
+		rte_panic("Invalid EAL arguments\n");
+
+	argc -= err;
+	argv += err;
+
+	/* Parse cli options*/
+	parse_app_args(argc, argv);
+
+	if (!quiet) {
+		printf("  Config:\n");
+		printf("\tworkers: %d\n", num_workers);
+		printf("\tpackets: %"PRIu64"\n", num_packets);
+		if (sched_type == RTE_SCHED_TYPE_ORDERED)
+			printf("\tworker_qid type: ordered\n");
+		if (sched_type == RTE_SCHED_TYPE_ATOMIC)
+			printf("\tworker_qid type: atomic\n");
+		printf("\n");
+	}
+
+	const unsigned int cores_needed = num_workers +
+			/*main*/ 1 +
+			/*producer*/ 1 +
+			/*consumer*/ 1;
+
+	if (!quiet) {
+		printf("Number of cores available: %d\n", rte_lcore_count());
+		printf("Number of cores to be used: %d\n", cores_needed);
+	}
+
+	if (rte_lcore_count() < cores_needed)
+		rte_panic("Too few cores\n");
+
+	const uint8_t ndevs = rte_event_dev_count();
+	if (ndevs == 0)
+		rte_panic(
+			"No event devs found. Do you need"
+			" to pass in a --vdev flag?\n");
+	if (ndevs > 1)
+		fprintf(stderr,
+			"Warning: More than one event dev, but using idx 0");
+
+	if (num_workers) {
+		worker_data = rte_calloc(0, num_workers,
+					 sizeof(worker_data[0]), 0);
+		if (worker_data == NULL)
+			rte_panic("rte_calloc failed\n");
+	}
+
+	uint8_t id = setup_event_dev(&prod_data, &cons_data, worker_data,
+				    evdev_id);
+
+	printf("setup_event_dev returned eventdev_id = %d\n", id);
+
+	start = rte_rdtsc();
+
+	RTE_LCORE_FOREACH_WORKER(lcore_id) {
+		if (has_prod && has_cons && nworkers == num_workers)
+			break;
+
+		if (!has_prod) {
+			err = rte_eal_remote_launch(producer, &prod_data,
+						    lcore_id);
+			if (err)
+				rte_panic("Failed to launch producer\n");
+			printf("Launched producer\n");
+			has_prod = 1;
+			continue;
+		}
+
+		if (!has_cons) {
+			err = rte_eal_remote_launch(consumer, &cons_data,
+						    lcore_id);
+			if (err)
+				rte_panic("Failed to launch consumer\n");
+			printf("Launched consumer\n");
+			has_cons = 1;
+			continue;
+		}
+
+		if (nworkers < num_workers) {
+			err = rte_eal_remote_launch(worker,
+						    &worker_data[nworkers],
+						    lcore_id);
+			if (err)
+				rte_panic("Failed to launch worker%d\n",
+					  nworkers);
+			nworkers++;
+			printf("Launched worker %d\n", nworkers);
+			continue;
+		}
+	}
+
+	rte_eal_mp_wait_lcore();
+	end = rte_rdtsc();
+	printf("[%s()] DLB scheduled %"PRIu64" packets in %"PRIu64" cycles\n",
+	       __func__, num_packets, end - start);
+	printf("[%s()] \t %f packets/sec\n",
+	       __func__, (float) (num_packets * rte_get_timer_hz()) /
+	       (float) (end - start));
+
+	printf("Test Complete\n");
+
+	/* Cleanup done automatically by kernel on app exit */
+
+	return 0;
+}
diff --git a/examples/eventdev_producer_consumer/meson.build b/examples/eventdev_producer_consumer/meson.build
new file mode 100644
index 0000000000..e18739432c
--- /dev/null
+++ b/examples/eventdev_producer_consumer/meson.build
@@ -0,0 +1,13 @@
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(c) 2017 Intel Corporation
+
+# meson file, for building this example as part of a main DPDK build.
+#
+# To build this example as a standalone application with an already-installed
+# DPDK instance, use 'make'
+
+allow_experimental_apis = true
+deps += 'eventdev'
+sources = files(
+        'main.c',
+)
diff --git a/examples/meson.build b/examples/meson.build
index 81e93799f2..9d95ecc8e0 100644
--- a/examples/meson.build
+++ b/examples/meson.build
@@ -15,6 +15,7 @@ all_examples = [
         'dma',
         'ethtool',
         'eventdev_pipeline',
+	'eventdev_producer_consumer',
         'fips_validation',
         'flow_classify',
         'flow_filtering',
-- 
2.23.0


^ permalink raw reply	[flat|nested] 5+ messages in thread

* [PATCH v4] examples: add eventdev_producer_consumer example
  2022-08-18 17:02 [PATCH] examples: add eventdev_producer_consumer example Timothy McDaniel
  2022-08-22 19:26 ` [PATCH v2] examples/eventdev_producer_consumer: fix 32-bit checkpatch issues Timothy McDaniel
  2022-08-22 20:04 ` [PATCH v3] examples: add eventdev_producer_consumer example Timothy McDaniel
@ 2022-08-22 20:51 ` Timothy McDaniel
  2022-08-27 11:57   ` Jerin Jacob
  2 siblings, 1 reply; 5+ messages in thread
From: Timothy McDaniel @ 2022-08-22 20:51 UTC (permalink / raw)
  To: jerinj; +Cc: mdr, thomas, dev

The eventdev-producer-consumer application is a single-stage
producer-worker-consumer pipeline sample to mimic real-world applications.
It is useful in measuring performance impact when any eventdev
configuration is changed. Unlike test-eventdev, it allows configuring a
load balanced queue between the producer and workers and a single-link
queue between the workers and consumer. With test-eventdev, multiple worker
stages can be configured but there is no single consumer receiving events
from all the workers. It also does not require configuring TX/RX adapters
like in the case of eventdev_pipeline app.

Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>

---

V4: Fixed a coding style issue

V3: Fixed style and format issues, primarily those involving data types
whose size varies depending on whether we are building for 32 or
64 bit platforms.

V2: Disregard - forgot to resubmit entire patch

---
---
 examples/eventdev_producer_consumer/Makefile  |  22 +
 examples/eventdev_producer_consumer/main.c    | 671 ++++++++++++++++++
 .../eventdev_producer_consumer/meson.build    |  13 +
 examples/meson.build                          |   1 +
 4 files changed, 707 insertions(+)
 create mode 100644 examples/eventdev_producer_consumer/Makefile
 create mode 100644 examples/eventdev_producer_consumer/main.c
 create mode 100644 examples/eventdev_producer_consumer/meson.build

diff --git a/examples/eventdev_producer_consumer/Makefile b/examples/eventdev_producer_consumer/Makefile
new file mode 100644
index 0000000000..761689eab7
--- /dev/null
+++ b/examples/eventdev_producer_consumer/Makefile
@@ -0,0 +1,22 @@
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(c) 2016-2017 Intel Corporation.
+
+ifeq ($(RTE_SDK),)
+$(error "Please define RTE_SDK environment variable")
+endif
+
+# Default target, can be overridden by command line or environment
+RTE_TARGET ?= x86_64-native-linuxapp-gcc
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+# binary name
+APP = eventdev_producer_consumer
+
+# all source are stored in SRCS-y
+SRCS-y := main.c
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+include $(RTE_SDK)/mk/rte.extapp.mk
diff --git a/examples/eventdev_producer_consumer/main.c b/examples/eventdev_producer_consumer/main.c
new file mode 100644
index 0000000000..4c9f51d8c2
--- /dev/null
+++ b/examples/eventdev_producer_consumer/main.c
@@ -0,0 +1,671 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2022 Intel Corporation
+ */
+
+#include <getopt.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+
+#include <rte_eal.h>
+#include <rte_mempool.h>
+#include <rte_mbuf.h>
+#include <rte_launch.h>
+#include <rte_malloc.h>
+#include <rte_cycles.h>
+#include <rte_ethdev.h>
+#include <rte_eventdev.h>
+
+#define BATCH_SIZE 32
+
+static unsigned int num_workers = 4;
+static bool g_is_mbuf;
+static uint64_t num_packets = (1L << 25); /* do ~32M packets */
+static int sched_type = RTE_SCHED_TYPE_ATOMIC;
+
+struct prod_data {
+	uint8_t event_dev_id;
+	uint8_t event_port_id;
+	int32_t qid;
+};
+
+struct cons_data {
+	uint8_t event_dev_id;
+	uint8_t event_port_id;
+};
+
+struct worker_data {
+	uint8_t event_dev_id;
+	int event_port_id;
+	int32_t qid;
+};
+
+static volatile int done;
+static int quiet;
+
+#define PORT_0 0
+#define QUEUE_0 0
+static struct rte_mempool *mp;
+
+static int
+worker(void *arg)
+{
+	struct rte_event rcv_events[BATCH_SIZE];
+
+	struct worker_data *data = (struct worker_data *)arg;
+	uint8_t event_dev_id = data->event_dev_id;
+	uint8_t event_port_id = data->event_port_id;
+	int32_t qid = data->qid;
+	uint64_t sent = 0, received = 0;
+	uint16_t n;
+
+	if (!quiet)
+		printf("Worker core %d started, portId=%d, sending to qid=%d\n",
+		       rte_lcore_id(), event_port_id, qid);
+
+	while (!done) {
+		uint16_t k;
+		int npkts_to_send, npkts_sent = 0;
+		struct rte_event *ev;
+		uint64_t delay_start;
+
+		/* Cannot wait for IRQ here due to the way that
+		 * we check for when we are done.
+		 */
+		n = rte_event_dequeue_burst(event_dev_id,
+					    event_port_id,
+					    rcv_events,
+					    RTE_DIM(rcv_events),
+					    0);
+
+		if (n == 0) {
+			rte_pause();
+			continue;
+		} else if (!quiet)
+			printf("Worker received %d events(%"PRIu64" total)\n",
+			       n, received);
+
+		delay_start = rte_rdtsc();
+		while (delay_start > rte_rdtsc())
+			;
+
+		received += n;
+
+		ev = &rcv_events[0];
+		for (k = 0; k < n; k++) {
+			ev->queue_id = qid;
+			ev->op = RTE_EVENT_OP_FORWARD;
+			ev++;
+		}
+
+		ev = &rcv_events[0];
+		npkts_to_send = n;
+		npkts_sent = 0;
+
+		while (npkts_sent < npkts_to_send)  {
+			int iter_sent = 0;
+			iter_sent = rte_event_enqueue_burst(event_dev_id,
+							    event_port_id,
+							    &ev[npkts_sent],
+							    n - npkts_sent);
+			npkts_sent += iter_sent;
+		}
+	} /* while (!done) */
+
+	if (!quiet)
+		printf("%s %d thread done. RX= %"PRIu64" TX= %"PRIu64"\n",
+			__func__, rte_lcore_id(), received, sent);
+
+	return 0;
+}
+
+static int
+consumer(void *arg)
+{
+	struct rte_event events[BATCH_SIZE];
+	struct cons_data *data = (struct cons_data *)arg;
+	uint8_t event_dev_id = data->event_dev_id;
+	uint8_t event_port_id = data->event_port_id;
+	int64_t npackets = num_packets;
+	uint64_t start_time = 0;
+	uint64_t freq_khz = rte_get_timer_hz() / 1000;
+	uint16_t n;
+	uint64_t deq_start, deq_end;
+
+	deq_start = rte_rdtsc();
+	while (npackets > 0) {
+		uint16_t i;
+		n = rte_event_dequeue_burst(event_dev_id,
+					    event_port_id,
+					    events,
+					    RTE_DIM(events),
+					    0);
+
+		if (g_is_mbuf) {
+			for (i = 0; i < n; i++) {
+				/* Could pack these up and do a bulk free */
+				if (!quiet)
+					printf("%s: mbuf[%d].seqno = %"
+						PRIu64"\n", __func__, i,
+						events[i].mbuf->tx_offload);
+				if (events[i].mbuf->tx_offload < 100000000000)
+					rte_pktmbuf_free(events[i].mbuf);
+				rte_cldemote(events[i].mbuf);
+			}
+		} /* if (g_is_mbuf) */
+		npackets -= n;
+	} /* while */
+
+	deq_end = rte_rdtsc();
+	printf("Consumer done in %"PRIu64" cycles (%f cycles/evt)"
+	       " (%f pkts/sec)\n", deq_end-deq_start,
+	       (float)(deq_end - deq_start)/(float)num_packets,
+	       (float) (num_packets * rte_get_timer_hz()) /
+	       (float) (deq_end - deq_start));
+	printf("deq_end = %"PRIu64", deq_start = %"PRIu64"\n",
+	       deq_end, deq_start);
+
+	printf("Consumer done! RX=%"PRIu64", time %"PRIu64"ms\n",
+	       num_packets,
+	       (rte_get_timer_cycles() - start_time) / freq_khz);
+	done = 1;
+	return 0;
+}
+
+
+static int
+producer(void *arg)
+{
+	struct prod_data *data = (struct prod_data *)arg;
+	int64_t npackets = num_packets;
+	uint64_t mbuf_seqno = 0;
+	uint8_t event_dev_id;
+	uint8_t event_port_id;
+	int fid_counter = 0;
+	int err;
+	int64_t retry_count = 0;
+	int32_t qid = data->qid;
+	uint64_t enq_start, enq_end;
+	int k = 0;
+	struct rte_mbuf *m;
+	struct rte_event producer_events[BATCH_SIZE];
+	struct rte_event *ev = &producer_events[0];
+	int l = 0;
+	struct rte_mbuf *mbufs[BATCH_SIZE];
+
+	event_dev_id = data->event_dev_id;
+	event_port_id = data->event_port_id;
+
+	for (k = 0; k < BATCH_SIZE; k++) {
+		if (!g_is_mbuf)
+			m = NULL;
+		ev->queue_id = qid;
+		ev->priority = 0;
+		ev->mbuf = m;
+		ev->sched_type = sched_type;
+		ev->op = RTE_EVENT_OP_NEW;
+		ev++;
+	}
+
+	enq_start = rte_rdtsc();
+	do {
+		int64_t npkt_start;
+		ev = &producer_events[0];
+		retry_count = 0;
+
+		if (g_is_mbuf) {
+			err = rte_pktmbuf_alloc_bulk(mp,
+						     &mbufs[0],
+						     BATCH_SIZE);
+			if (err) {
+				printf("mbuf alloc failed after sending %"
+				       PRIu64" with err=%d\n",
+				       num_packets - npackets, err);
+				return -1;
+			}
+
+			for (l = 0; l < BATCH_SIZE; l++) {
+				m = mbufs[l];
+				/* Using tx_offload field of rte_mbuf to store
+				 * seq nums as .udata64 has been removed
+				 */
+				m->tx_offload = mbuf_seqno++;
+				producer_events[l].mbuf = m;
+				producer_events[l].flow_id = fid_counter++;
+				if (!quiet)
+					printf("%s: mbuf[%d].seqno = %"PRIu64"\n",
+						__func__, l,
+						producer_events[l].mbuf->tx_offload);
+			} /* for l = 0 - BATCH_SIZE */
+		} /* if g_is_mbuf */
+		else {
+			for (l = 0; l < BATCH_SIZE; l++)
+				producer_events[l].flow_id = fid_counter++;
+		}
+		npkt_start = npackets;
+		while (npackets > npkt_start - BATCH_SIZE) {
+			int64_t num_sent = npkt_start - npackets;
+			npackets -= rte_event_enqueue_burst(event_dev_id,
+							    event_port_id,
+							    &ev[num_sent],
+							    BATCH_SIZE -
+							    num_sent);
+		}
+	} while ((npackets > 0) && retry_count++ < 100000000000);
+
+	enq_end = rte_rdtsc();
+
+	if (npackets > 0)
+		rte_panic("%s thread failed to enqueue events\n", __func__);
+
+	if (num_packets > 0 && npackets > 0)
+		printf("npackets not sent: %"PRIu64"\n", npackets);
+
+	printf("Producer done. %"PRIu64" packets sent in %"PRIu64" cycles"
+	       "(%f cycles/evt) (%f pkts/sec)\n",
+	       num_packets, enq_end - enq_start,
+	       (float)(enq_end - enq_start)/(float)num_packets,
+	       (float) (num_packets * rte_get_timer_hz()) /
+	       (float) (enq_end - enq_start));
+	printf("enq_enq = %"PRIu64", enq_start = %"PRIu64"\n",
+	       enq_end, enq_start);
+	return 0;
+}
+
+static struct option long_options[] = {
+	{"workers", required_argument, 0, 'w'},
+	{"packets", required_argument, 0, 'n'},
+	{"ordered", no_argument, 0, 'o'},
+	{"parallel", no_argument, 0, 'u'},
+	{"quiet", no_argument, 0, 'q'},
+	{"useMbufs", no_argument, 0, 'm'},
+	{0, 0, 0, 0}
+};
+
+static void
+usage(void)
+{
+	const char *usage_str =
+		"  Usage: eventdev_producer_consumer [options]\n"
+		"  Options:\n"
+		"  -w, --workers=N       Use N workers (default 4)\n"
+		"  -n, --packets=N       Send N packets (default ~32M),"
+					 " 0 implies no limit\n"
+		"  -o, --ordered         Use ordered scheduling\n"
+		"  -u, --parallel        Use parallel scheduling\n"
+		"  -q, --quiet           Minimize printed output\n"
+		"  -m, --use-mbufs       Use mbufs for enqueue\n"
+		"\n";
+
+	fprintf(stderr, "%s", usage_str);
+	exit(1);
+}
+
+static void
+parse_app_args(int argc, char **argv)
+{
+	/* Parse cli options*/
+	int option_index;
+	int c;
+	opterr = 0;
+
+	for (;;) {
+		c = getopt_long(argc, argv, "w:n:ouqm", long_options,
+				&option_index);
+		if (c == -1)
+			break;
+
+		switch (c) {
+		case 'w':
+			num_workers = (unsigned int)atoi(optarg);
+			break;
+		case 'n':
+			num_packets = (unsigned long)atol(optarg);
+			break;
+		case 'o':
+			if (sched_type == RTE_SCHED_TYPE_PARALLEL)
+				rte_panic("Cannot specify both -o and -u\n");
+			sched_type = RTE_SCHED_TYPE_ORDERED;
+			break;
+		case 'u':
+			if (sched_type == RTE_SCHED_TYPE_ORDERED)
+				rte_panic("Cannot specify both -o and -u\n");
+			sched_type = RTE_SCHED_TYPE_PARALLEL;
+			break;
+		case 'q':
+			quiet = 1;
+			break;
+		case 'm':
+			g_is_mbuf = true;
+			break;
+		default:
+			usage();
+		}
+	}
+}
+
+static uint8_t
+setup_event_dev(struct prod_data *prod_data,
+		struct cons_data *cons_data,
+		struct worker_data *worker_data,
+		int id)
+{
+	struct rte_event_dev_info dev_info;
+	struct rte_event_dev_config config = {0};
+	struct rte_event_queue_conf queue_config = {0};
+	struct rte_event_port_conf port_config = {0};
+	uint8_t queue_id;
+	uint8_t priority;
+	int prod_port = 0;
+	int cons_port = 1;
+	int worker_port_base = 2;
+	int prod_qid = 0;
+	int cons_qid = 1;
+	int worker_qid = 2;
+	unsigned int i;
+	int ret;
+
+	/* Better yet, always use event dev 0 so the app can use either. You can
+	 * check that there's at least 1 eventdev with rte_event_dev_count().
+	 */
+
+	if (id < 0)
+		rte_panic("%s: invalid ev_dev ID %d\n", __func__, id);
+	else
+		printf("%s: ev_dev ID %d\n", __func__, id);
+
+	rte_event_dev_info_get(id, &dev_info);
+
+	if (num_workers)
+		config.nb_event_queues = 3;
+	else
+		config.nb_event_queues = 2;
+
+	config.nb_single_link_event_port_queues = 2;
+	config.nb_event_ports = num_workers +
+		config.nb_single_link_event_port_queues;
+	config.nb_events_limit = dev_info.max_num_events;
+	config.nb_event_queue_flows = dev_info.max_event_queue_flows;
+	config.nb_event_port_dequeue_depth =
+					dev_info.max_event_port_dequeue_depth;
+	config.nb_event_port_enqueue_depth =
+					dev_info.max_event_port_enqueue_depth;
+	config.dequeue_timeout_ns = 0;
+	config.event_dev_cfg = RTE_EVENT_DEV_CFG_PER_DEQUEUE_TIMEOUT;
+
+	ret = rte_event_dev_configure(id, &config);
+	if (ret)
+		rte_panic("Failed to configure the event dev\n");
+	else
+		printf("eventdev configured!\n");
+
+	/* Create queues */
+	queue_config.event_queue_cfg = 0;
+	queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+	queue_config.nb_atomic_order_sequences =
+		(sched_type == RTE_SCHED_TYPE_ORDERED) ? 1024 : 0;
+	queue_config.nb_atomic_flows = dev_info.max_event_queue_flows;
+	queue_config.schedule_type = sched_type;
+
+	if (num_workers) {
+		ret = rte_event_queue_setup(id, worker_qid, &queue_config);
+		if (ret < 0)
+			rte_panic("Failed to create the scheduled QID\n");
+		else
+			printf("rte_event_queue_setup success for worker_qid\n");
+	}
+
+	queue_config.event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK;
+	queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	cons_qid = 1;
+	ret = rte_event_queue_setup(id, cons_qid, &queue_config);
+	if (ret < 0)
+		rte_panic("Failed to create the cons directed QID\n");
+	else
+		printf("rte_event_queue_setup success for cons_qid\n");
+
+	queue_config.event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK;
+	queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	prod_qid = 0;
+	ret = rte_event_queue_setup(id, prod_qid, &queue_config);
+	if (ret < 0)
+		rte_panic("Failed to create the prod directed QID\n");
+	else
+		printf("rte_event_queue_setup success for prod_qid\n");
+
+	/* Create two directed ports */
+
+	port_config.dequeue_depth = dev_info.max_event_port_dequeue_depth;
+	port_config.enqueue_depth = dev_info.max_event_port_enqueue_depth;
+
+	/* Set producer new event threshold to 3/4 max */
+	port_config.new_event_threshold = 3 * (dev_info.max_num_events >> 2);
+	port_config.event_port_cfg = RTE_EVENT_PORT_CFG_SINGLE_LINK;
+	ret = rte_event_port_setup(id, prod_port, &port_config);
+	if (ret < 0)
+		rte_panic("Failed to create the producer port\n");
+	else
+		printf("rte_event_port_setup for prod_port ok\n");
+
+	/* Set consumer and worker new event threshold to max */
+	port_config.new_event_threshold = dev_info.max_num_events;
+	ret = rte_event_port_setup(id, cons_port, &port_config);
+	if (ret < 0)
+		rte_panic("Failed to create the consumer port\n");
+	else
+		printf("rte_event_port_setup for cons_port ok\n");
+
+	port_config.event_port_cfg = 0;
+
+	/* Create load-balanced worker ports */
+	for (i = 0; i < num_workers; i++) {
+		worker_data[i].event_port_id = i + worker_port_base;
+		ret = rte_event_port_setup(id, worker_data[i].event_port_id,
+					   &port_config);
+		if (ret < 0)
+			rte_panic("Failed to create worker port #%d\n", i);
+		else
+			printf("rte_event_port_setup for worker port %d ok\n",
+			       i);
+	}
+
+	printf("link worker queues\n");
+	/* Map ports/qids */
+	for (i = 0; i < num_workers; i++) {
+		queue_id = worker_qid;
+		priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+		ret = rte_event_port_link(id, worker_data[i].event_port_id,
+					  &queue_id, &priority, 1);
+		if (ret != 1)
+			rte_panic("Failed to map worker%d port to worker_qid\n",
+				  i);
+	}
+
+	printf("link consumer queue\n");
+	/* Link consumer port to its QID */
+	queue_id = cons_qid;
+	priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	ret = rte_event_port_link(id, cons_port, &queue_id, &priority, 1);
+	if (ret != 1)
+		rte_panic("Failed to map consumer port to cons_qid\n");
+
+	printf("link producer queue\n");
+	/* Link producer port to its QID */
+	queue_id = prod_qid;
+	priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+	ret = rte_event_port_link(id, prod_port, &queue_id, &priority, 1);
+	if (ret != 1)
+		rte_panic("Failed to map producer port to prod_qid\n");
+
+	/* Dispatch to workers */
+	if (num_workers) {
+		*prod_data = (struct prod_data){.event_dev_id = id,
+						.event_port_id = prod_port,
+						.qid = worker_qid};
+		*cons_data = (struct cons_data){.event_dev_id = id,
+						.event_port_id = cons_port};
+		for (i = 0; i < num_workers; i++) {
+			struct worker_data *w = &worker_data[i];
+			w->event_dev_id = id;
+			w->qid = cons_qid;
+		}
+	} else {
+		*prod_data = (struct prod_data){.event_dev_id = id,
+						.event_port_id = prod_port,
+						.qid = cons_qid};
+		*cons_data = (struct cons_data){.event_dev_id = id,
+						.event_port_id = cons_port};
+	}
+
+	ret = rte_event_dev_start(id);
+	if (ret)
+		rte_panic("Failed to start the event dev\n");
+	if (g_is_mbuf) {
+		mp = rte_pktmbuf_pool_create("packet_pool",
+				/* mbufs */ dev_info.max_num_events,
+				/* cache_size */ 512,
+				/* priv_size*/ 0,
+				/* data_room_size */ 2048,
+				rte_socket_id());
+
+		if (mp == NULL) {
+			printf("mbuf pool create failed\n");
+			return -1;
+		}
+	}
+	return (uint8_t) id;
+}
+
+int
+main(int argc, char **argv)
+{
+	struct prod_data prod_data = {0};
+	struct cons_data cons_data = {0};
+	uint64_t start, end;
+	struct worker_data *worker_data = NULL;
+	unsigned int nworkers = 0;
+	int lcore_id;
+	int err;
+	int has_prod = 0;
+	int has_cons = 0;
+	int evdev_id = 0; /* TODO - allow app to override */
+
+	done = 0;
+	quiet = 0;
+	mp = NULL;
+	g_is_mbuf = false;
+
+	err = rte_eal_init(argc, argv);
+	if (err < 0)
+		rte_panic("Invalid EAL arguments\n");
+
+	argc -= err;
+	argv += err;
+
+	/* Parse cli options*/
+	parse_app_args(argc, argv);
+
+	if (!quiet) {
+		printf("  Config:\n");
+		printf("\tworkers: %d\n", num_workers);
+		printf("\tpackets: %"PRIu64"\n", num_packets);
+		if (sched_type == RTE_SCHED_TYPE_ORDERED)
+			printf("\tworker_qid type: ordered\n");
+		if (sched_type == RTE_SCHED_TYPE_ATOMIC)
+			printf("\tworker_qid type: atomic\n");
+		printf("\n");
+	}
+
+	const unsigned int cores_needed = num_workers +
+			/*main*/ 1 +
+			/*producer*/ 1 +
+			/*consumer*/ 1;
+
+	if (!quiet) {
+		printf("Number of cores available: %d\n", rte_lcore_count());
+		printf("Number of cores to be used: %d\n", cores_needed);
+	}
+
+	if (rte_lcore_count() < cores_needed)
+		rte_panic("Too few cores\n");
+
+	const uint8_t ndevs = rte_event_dev_count();
+	if (ndevs == 0)
+		rte_panic(
+			"No event devs found. Do you need"
+			" to pass in a --vdev flag?\n");
+	if (ndevs > 1)
+		fprintf(stderr,
+			"Warning: More than one event dev, but using idx 0");
+
+	if (num_workers) {
+		worker_data = rte_calloc(0, num_workers,
+					 sizeof(worker_data[0]), 0);
+		if (worker_data == NULL)
+			rte_panic("rte_calloc failed\n");
+	}
+
+	uint8_t id = setup_event_dev(&prod_data, &cons_data, worker_data,
+				    evdev_id);
+
+	printf("setup_event_dev returned eventdev_id = %d\n", id);
+
+	start = rte_rdtsc();
+
+	RTE_LCORE_FOREACH_WORKER(lcore_id) {
+		if (has_prod && has_cons && nworkers == num_workers)
+			break;
+
+		if (!has_prod) {
+			err = rte_eal_remote_launch(producer, &prod_data,
+						    lcore_id);
+			if (err)
+				rte_panic("Failed to launch producer\n");
+			printf("Launched producer\n");
+			has_prod = 1;
+			continue;
+		}
+
+		if (!has_cons) {
+			err = rte_eal_remote_launch(consumer, &cons_data,
+						    lcore_id);
+			if (err)
+				rte_panic("Failed to launch consumer\n");
+			printf("Launched consumer\n");
+			has_cons = 1;
+			continue;
+		}
+
+		if (nworkers < num_workers) {
+			err = rte_eal_remote_launch(worker,
+						    &worker_data[nworkers],
+						    lcore_id);
+			if (err)
+				rte_panic("Failed to launch worker%d\n",
+					  nworkers);
+			nworkers++;
+			printf("Launched worker %d\n", nworkers);
+			continue;
+		}
+	}
+
+	rte_eal_mp_wait_lcore();
+	end = rte_rdtsc();
+	printf("[%s()] DLB scheduled %"PRIu64" packets in %"PRIu64" cycles\n",
+	       __func__, num_packets, end - start);
+	printf("[%s()] \t %f packets/sec\n",
+	       __func__, (float) (num_packets * rte_get_timer_hz()) /
+	       (float) (end - start));
+
+	printf("Test Complete\n");
+
+	/* Cleanup done automatically by kernel on app exit */
+
+	return 0;
+}
diff --git a/examples/eventdev_producer_consumer/meson.build b/examples/eventdev_producer_consumer/meson.build
new file mode 100644
index 0000000000..e18739432c
--- /dev/null
+++ b/examples/eventdev_producer_consumer/meson.build
@@ -0,0 +1,13 @@
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(c) 2017 Intel Corporation
+
+# meson file, for building this example as part of a main DPDK build.
+#
+# To build this example as a standalone application with an already-installed
+# DPDK instance, use 'make'
+
+allow_experimental_apis = true
+deps += 'eventdev'
+sources = files(
+        'main.c',
+)
diff --git a/examples/meson.build b/examples/meson.build
index 81e93799f2..9d95ecc8e0 100644
--- a/examples/meson.build
+++ b/examples/meson.build
@@ -15,6 +15,7 @@ all_examples = [
         'dma',
         'ethtool',
         'eventdev_pipeline',
+	'eventdev_producer_consumer',
         'fips_validation',
         'flow_classify',
         'flow_filtering',
-- 
2.23.0


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH v4] examples: add eventdev_producer_consumer example
  2022-08-22 20:51 ` [PATCH v4] " Timothy McDaniel
@ 2022-08-27 11:57   ` Jerin Jacob
  0 siblings, 0 replies; 5+ messages in thread
From: Jerin Jacob @ 2022-08-27 11:57 UTC (permalink / raw)
  To: Timothy McDaniel; +Cc: Jerin Jacob, Ray Kinsella, Thomas Monjalon, dpdk-dev

On Tue, Aug 23, 2022 at 2:21 AM Timothy McDaniel
<timothy.mcdaniel@intel.com> wrote:
>
> The eventdev-producer-consumer application is a single-stage
> producer-worker-consumer pipeline sample to mimic real-world applications.
> It is useful in measuring performance impact when any eventdev
> configuration is changed. Unlike test-eventdev, it allows configuring a
> load balanced queue between the producer and workers and a single-link
> queue between the workers and consumer. With test-eventdev, multiple worker
> stages can be configured but there is no single consumer receiving events
> from all the workers. It also does not require configuring TX/RX adapters
> like in the case of eventdev_pipeline app.
>
> Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>


We are in the process of reducing/converging the example apps and it
needs TB approval to add a new example app.
IMO, It is not worth adding a new example app for this.

You can add a new test and integrate to app/test-eventdev. Please see
"11.2. Eventdev Tests" in
https://doc.dpdk.org/guides/tools/testeventdev.html.



>
> ---
>
> V4: Fixed a coding style issue
>
> V3: Fixed style and format issues, primarily those involving data types
> whose size varies depending on whether we are building for 32 or
> 64 bit platforms.
>
> V2: Disregard - forgot to resubmit entire patch
>
> ---
> ---
>  examples/eventdev_producer_consumer/Makefile  |  22 +
>  examples/eventdev_producer_consumer/main.c    | 671 ++++++++++++++++++
>  .../eventdev_producer_consumer/meson.build    |  13 +
>  examples/meson.build                          |   1 +
>  4 files changed, 707 insertions(+)
>  create mode 100644 examples/eventdev_producer_consumer/Makefile
>  create mode 100644 examples/eventdev_producer_consumer/main.c
>  create mode 100644 examples/eventdev_producer_consumer/meson.build
>
> diff --git a/examples/eventdev_producer_consumer/Makefile b/examples/eventdev_producer_consumer/Makefile
> new file mode 100644
> index 0000000000..761689eab7
> --- /dev/null
> +++ b/examples/eventdev_producer_consumer/Makefile
> @@ -0,0 +1,22 @@
> +# SPDX-License-Identifier: BSD-3-Clause
> +# Copyright(c) 2016-2017 Intel Corporation.
> +
> +ifeq ($(RTE_SDK),)
> +$(error "Please define RTE_SDK environment variable")
> +endif
> +
> +# Default target, can be overridden by command line or environment
> +RTE_TARGET ?= x86_64-native-linuxapp-gcc
> +
> +include $(RTE_SDK)/mk/rte.vars.mk
> +
> +# binary name
> +APP = eventdev_producer_consumer
> +
> +# all source are stored in SRCS-y
> +SRCS-y := main.c
> +
> +CFLAGS += -O3
> +CFLAGS += $(WERROR_FLAGS)
> +
> +include $(RTE_SDK)/mk/rte.extapp.mk
> diff --git a/examples/eventdev_producer_consumer/main.c b/examples/eventdev_producer_consumer/main.c
> new file mode 100644
> index 0000000000..4c9f51d8c2
> --- /dev/null
> +++ b/examples/eventdev_producer_consumer/main.c
> @@ -0,0 +1,671 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(c) 2022 Intel Corporation
> + */
> +
> +#include <getopt.h>
> +#include <stdint.h>
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <unistd.h>
> +
> +#include <rte_eal.h>
> +#include <rte_mempool.h>
> +#include <rte_mbuf.h>
> +#include <rte_launch.h>
> +#include <rte_malloc.h>
> +#include <rte_cycles.h>
> +#include <rte_ethdev.h>
> +#include <rte_eventdev.h>
> +
> +#define BATCH_SIZE 32
> +
> +static unsigned int num_workers = 4;
> +static bool g_is_mbuf;
> +static uint64_t num_packets = (1L << 25); /* do ~32M packets */
> +static int sched_type = RTE_SCHED_TYPE_ATOMIC;
> +
> +struct prod_data {
> +       uint8_t event_dev_id;
> +       uint8_t event_port_id;
> +       int32_t qid;
> +};
> +
> +struct cons_data {
> +       uint8_t event_dev_id;
> +       uint8_t event_port_id;
> +};
> +
> +struct worker_data {
> +       uint8_t event_dev_id;
> +       int event_port_id;
> +       int32_t qid;
> +};
> +
> +static volatile int done;
> +static int quiet;
> +
> +#define PORT_0 0
> +#define QUEUE_0 0
> +static struct rte_mempool *mp;
> +
> +static int
> +worker(void *arg)
> +{
> +       struct rte_event rcv_events[BATCH_SIZE];
> +
> +       struct worker_data *data = (struct worker_data *)arg;
> +       uint8_t event_dev_id = data->event_dev_id;
> +       uint8_t event_port_id = data->event_port_id;
> +       int32_t qid = data->qid;
> +       uint64_t sent = 0, received = 0;
> +       uint16_t n;
> +
> +       if (!quiet)
> +               printf("Worker core %d started, portId=%d, sending to qid=%d\n",
> +                      rte_lcore_id(), event_port_id, qid);
> +
> +       while (!done) {
> +               uint16_t k;
> +               int npkts_to_send, npkts_sent = 0;
> +               struct rte_event *ev;
> +               uint64_t delay_start;
> +
> +               /* Cannot wait for IRQ here due to the way that
> +                * we check for when we are done.
> +                */
> +               n = rte_event_dequeue_burst(event_dev_id,
> +                                           event_port_id,
> +                                           rcv_events,
> +                                           RTE_DIM(rcv_events),
> +                                           0);
> +
> +               if (n == 0) {
> +                       rte_pause();
> +                       continue;
> +               } else if (!quiet)
> +                       printf("Worker received %d events(%"PRIu64" total)\n",
> +                              n, received);
> +
> +               delay_start = rte_rdtsc();
> +               while (delay_start > rte_rdtsc())
> +                       ;
> +
> +               received += n;
> +
> +               ev = &rcv_events[0];
> +               for (k = 0; k < n; k++) {
> +                       ev->queue_id = qid;
> +                       ev->op = RTE_EVENT_OP_FORWARD;
> +                       ev++;
> +               }
> +
> +               ev = &rcv_events[0];
> +               npkts_to_send = n;
> +               npkts_sent = 0;
> +
> +               while (npkts_sent < npkts_to_send)  {
> +                       int iter_sent = 0;
> +                       iter_sent = rte_event_enqueue_burst(event_dev_id,
> +                                                           event_port_id,
> +                                                           &ev[npkts_sent],
> +                                                           n - npkts_sent);
> +                       npkts_sent += iter_sent;
> +               }
> +       } /* while (!done) */
> +
> +       if (!quiet)
> +               printf("%s %d thread done. RX= %"PRIu64" TX= %"PRIu64"\n",
> +                       __func__, rte_lcore_id(), received, sent);
> +
> +       return 0;
> +}
> +
> +static int
> +consumer(void *arg)
> +{
> +       struct rte_event events[BATCH_SIZE];
> +       struct cons_data *data = (struct cons_data *)arg;
> +       uint8_t event_dev_id = data->event_dev_id;
> +       uint8_t event_port_id = data->event_port_id;
> +       int64_t npackets = num_packets;
> +       uint64_t start_time = 0;
> +       uint64_t freq_khz = rte_get_timer_hz() / 1000;
> +       uint16_t n;
> +       uint64_t deq_start, deq_end;
> +
> +       deq_start = rte_rdtsc();
> +       while (npackets > 0) {
> +               uint16_t i;
> +               n = rte_event_dequeue_burst(event_dev_id,
> +                                           event_port_id,
> +                                           events,
> +                                           RTE_DIM(events),
> +                                           0);
> +
> +               if (g_is_mbuf) {
> +                       for (i = 0; i < n; i++) {
> +                               /* Could pack these up and do a bulk free */
> +                               if (!quiet)
> +                                       printf("%s: mbuf[%d].seqno = %"
> +                                               PRIu64"\n", __func__, i,
> +                                               events[i].mbuf->tx_offload);
> +                               if (events[i].mbuf->tx_offload < 100000000000)
> +                                       rte_pktmbuf_free(events[i].mbuf);
> +                               rte_cldemote(events[i].mbuf);
> +                       }
> +               } /* if (g_is_mbuf) */
> +               npackets -= n;
> +       } /* while */
> +
> +       deq_end = rte_rdtsc();
> +       printf("Consumer done in %"PRIu64" cycles (%f cycles/evt)"
> +              " (%f pkts/sec)\n", deq_end-deq_start,
> +              (float)(deq_end - deq_start)/(float)num_packets,
> +              (float) (num_packets * rte_get_timer_hz()) /
> +              (float) (deq_end - deq_start));
> +       printf("deq_end = %"PRIu64", deq_start = %"PRIu64"\n",
> +              deq_end, deq_start);
> +
> +       printf("Consumer done! RX=%"PRIu64", time %"PRIu64"ms\n",
> +              num_packets,
> +              (rte_get_timer_cycles() - start_time) / freq_khz);
> +       done = 1;
> +       return 0;
> +}
> +
> +
> +static int
> +producer(void *arg)
> +{
> +       struct prod_data *data = (struct prod_data *)arg;
> +       int64_t npackets = num_packets;
> +       uint64_t mbuf_seqno = 0;
> +       uint8_t event_dev_id;
> +       uint8_t event_port_id;
> +       int fid_counter = 0;
> +       int err;
> +       int64_t retry_count = 0;
> +       int32_t qid = data->qid;
> +       uint64_t enq_start, enq_end;
> +       int k = 0;
> +       struct rte_mbuf *m;
> +       struct rte_event producer_events[BATCH_SIZE];
> +       struct rte_event *ev = &producer_events[0];
> +       int l = 0;
> +       struct rte_mbuf *mbufs[BATCH_SIZE];
> +
> +       event_dev_id = data->event_dev_id;
> +       event_port_id = data->event_port_id;
> +
> +       for (k = 0; k < BATCH_SIZE; k++) {
> +               if (!g_is_mbuf)
> +                       m = NULL;
> +               ev->queue_id = qid;
> +               ev->priority = 0;
> +               ev->mbuf = m;
> +               ev->sched_type = sched_type;
> +               ev->op = RTE_EVENT_OP_NEW;
> +               ev++;
> +       }
> +
> +       enq_start = rte_rdtsc();
> +       do {
> +               int64_t npkt_start;
> +               ev = &producer_events[0];
> +               retry_count = 0;
> +
> +               if (g_is_mbuf) {
> +                       err = rte_pktmbuf_alloc_bulk(mp,
> +                                                    &mbufs[0],
> +                                                    BATCH_SIZE);
> +                       if (err) {
> +                               printf("mbuf alloc failed after sending %"
> +                                      PRIu64" with err=%d\n",
> +                                      num_packets - npackets, err);
> +                               return -1;
> +                       }
> +
> +                       for (l = 0; l < BATCH_SIZE; l++) {
> +                               m = mbufs[l];
> +                               /* Using tx_offload field of rte_mbuf to store
> +                                * seq nums as .udata64 has been removed
> +                                */
> +                               m->tx_offload = mbuf_seqno++;
> +                               producer_events[l].mbuf = m;
> +                               producer_events[l].flow_id = fid_counter++;
> +                               if (!quiet)
> +                                       printf("%s: mbuf[%d].seqno = %"PRIu64"\n",
> +                                               __func__, l,
> +                                               producer_events[l].mbuf->tx_offload);
> +                       } /* for l = 0 - BATCH_SIZE */
> +               } /* if g_is_mbuf */
> +               else {
> +                       for (l = 0; l < BATCH_SIZE; l++)
> +                               producer_events[l].flow_id = fid_counter++;
> +               }
> +               npkt_start = npackets;
> +               while (npackets > npkt_start - BATCH_SIZE) {
> +                       int64_t num_sent = npkt_start - npackets;
> +                       npackets -= rte_event_enqueue_burst(event_dev_id,
> +                                                           event_port_id,
> +                                                           &ev[num_sent],
> +                                                           BATCH_SIZE -
> +                                                           num_sent);
> +               }
> +       } while ((npackets > 0) && retry_count++ < 100000000000);
> +
> +       enq_end = rte_rdtsc();
> +
> +       if (npackets > 0)
> +               rte_panic("%s thread failed to enqueue events\n", __func__);
> +
> +       if (num_packets > 0 && npackets > 0)
> +               printf("npackets not sent: %"PRIu64"\n", npackets);
> +
> +       printf("Producer done. %"PRIu64" packets sent in %"PRIu64" cycles"
> +              "(%f cycles/evt) (%f pkts/sec)\n",
> +              num_packets, enq_end - enq_start,
> +              (float)(enq_end - enq_start)/(float)num_packets,
> +              (float) (num_packets * rte_get_timer_hz()) /
> +              (float) (enq_end - enq_start));
> +       printf("enq_enq = %"PRIu64", enq_start = %"PRIu64"\n",
> +              enq_end, enq_start);
> +       return 0;
> +}
> +
> +static struct option long_options[] = {
> +       {"workers", required_argument, 0, 'w'},
> +       {"packets", required_argument, 0, 'n'},
> +       {"ordered", no_argument, 0, 'o'},
> +       {"parallel", no_argument, 0, 'u'},
> +       {"quiet", no_argument, 0, 'q'},
> +       {"useMbufs", no_argument, 0, 'm'},
> +       {0, 0, 0, 0}
> +};
> +
> +static void
> +usage(void)
> +{
> +       const char *usage_str =
> +               "  Usage: eventdev_producer_consumer [options]\n"
> +               "  Options:\n"
> +               "  -w, --workers=N       Use N workers (default 4)\n"
> +               "  -n, --packets=N       Send N packets (default ~32M),"
> +                                        " 0 implies no limit\n"
> +               "  -o, --ordered         Use ordered scheduling\n"
> +               "  -u, --parallel        Use parallel scheduling\n"
> +               "  -q, --quiet           Minimize printed output\n"
> +               "  -m, --use-mbufs       Use mbufs for enqueue\n"
> +               "\n";
> +
> +       fprintf(stderr, "%s", usage_str);
> +       exit(1);
> +}
> +
> +static void
> +parse_app_args(int argc, char **argv)
> +{
> +       /* Parse cli options*/
> +       int option_index;
> +       int c;
> +       opterr = 0;
> +
> +       for (;;) {
> +               c = getopt_long(argc, argv, "w:n:ouqm", long_options,
> +                               &option_index);
> +               if (c == -1)
> +                       break;
> +
> +               switch (c) {
> +               case 'w':
> +                       num_workers = (unsigned int)atoi(optarg);
> +                       break;
> +               case 'n':
> +                       num_packets = (unsigned long)atol(optarg);
> +                       break;
> +               case 'o':
> +                       if (sched_type == RTE_SCHED_TYPE_PARALLEL)
> +                               rte_panic("Cannot specify both -o and -u\n");
> +                       sched_type = RTE_SCHED_TYPE_ORDERED;
> +                       break;
> +               case 'u':
> +                       if (sched_type == RTE_SCHED_TYPE_ORDERED)
> +                               rte_panic("Cannot specify both -o and -u\n");
> +                       sched_type = RTE_SCHED_TYPE_PARALLEL;
> +                       break;
> +               case 'q':
> +                       quiet = 1;
> +                       break;
> +               case 'm':
> +                       g_is_mbuf = true;
> +                       break;
> +               default:
> +                       usage();
> +               }
> +       }
> +}
> +
> +static uint8_t
> +setup_event_dev(struct prod_data *prod_data,
> +               struct cons_data *cons_data,
> +               struct worker_data *worker_data,
> +               int id)
> +{
> +       struct rte_event_dev_info dev_info;
> +       struct rte_event_dev_config config = {0};
> +       struct rte_event_queue_conf queue_config = {0};
> +       struct rte_event_port_conf port_config = {0};
> +       uint8_t queue_id;
> +       uint8_t priority;
> +       int prod_port = 0;
> +       int cons_port = 1;
> +       int worker_port_base = 2;
> +       int prod_qid = 0;
> +       int cons_qid = 1;
> +       int worker_qid = 2;
> +       unsigned int i;
> +       int ret;
> +
> +       /* Better yet, always use event dev 0 so the app can use either. You can
> +        * check that there's at least 1 eventdev with rte_event_dev_count().
> +        */
> +
> +       if (id < 0)
> +               rte_panic("%s: invalid ev_dev ID %d\n", __func__, id);
> +       else
> +               printf("%s: ev_dev ID %d\n", __func__, id);
> +
> +       rte_event_dev_info_get(id, &dev_info);
> +
> +       if (num_workers)
> +               config.nb_event_queues = 3;
> +       else
> +               config.nb_event_queues = 2;
> +
> +       config.nb_single_link_event_port_queues = 2;
> +       config.nb_event_ports = num_workers +
> +               config.nb_single_link_event_port_queues;
> +       config.nb_events_limit = dev_info.max_num_events;
> +       config.nb_event_queue_flows = dev_info.max_event_queue_flows;
> +       config.nb_event_port_dequeue_depth =
> +                                       dev_info.max_event_port_dequeue_depth;
> +       config.nb_event_port_enqueue_depth =
> +                                       dev_info.max_event_port_enqueue_depth;
> +       config.dequeue_timeout_ns = 0;
> +       config.event_dev_cfg = RTE_EVENT_DEV_CFG_PER_DEQUEUE_TIMEOUT;
> +
> +       ret = rte_event_dev_configure(id, &config);
> +       if (ret)
> +               rte_panic("Failed to configure the event dev\n");
> +       else
> +               printf("eventdev configured!\n");
> +
> +       /* Create queues */
> +       queue_config.event_queue_cfg = 0;
> +       queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
> +       queue_config.nb_atomic_order_sequences =
> +               (sched_type == RTE_SCHED_TYPE_ORDERED) ? 1024 : 0;
> +       queue_config.nb_atomic_flows = dev_info.max_event_queue_flows;
> +       queue_config.schedule_type = sched_type;
> +
> +       if (num_workers) {
> +               ret = rte_event_queue_setup(id, worker_qid, &queue_config);
> +               if (ret < 0)
> +                       rte_panic("Failed to create the scheduled QID\n");
> +               else
> +                       printf("rte_event_queue_setup success for worker_qid\n");
> +       }
> +
> +       queue_config.event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK;
> +       queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
> +
> +       cons_qid = 1;
> +       ret = rte_event_queue_setup(id, cons_qid, &queue_config);
> +       if (ret < 0)
> +               rte_panic("Failed to create the cons directed QID\n");
> +       else
> +               printf("rte_event_queue_setup success for cons_qid\n");
> +
> +       queue_config.event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK;
> +       queue_config.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
> +
> +       prod_qid = 0;
> +       ret = rte_event_queue_setup(id, prod_qid, &queue_config);
> +       if (ret < 0)
> +               rte_panic("Failed to create the prod directed QID\n");
> +       else
> +               printf("rte_event_queue_setup success for prod_qid\n");
> +
> +       /* Create two directed ports */
> +
> +       port_config.dequeue_depth = dev_info.max_event_port_dequeue_depth;
> +       port_config.enqueue_depth = dev_info.max_event_port_enqueue_depth;
> +
> +       /* Set producer new event threshold to 3/4 max */
> +       port_config.new_event_threshold = 3 * (dev_info.max_num_events >> 2);
> +       port_config.event_port_cfg = RTE_EVENT_PORT_CFG_SINGLE_LINK;
> +       ret = rte_event_port_setup(id, prod_port, &port_config);
> +       if (ret < 0)
> +               rte_panic("Failed to create the producer port\n");
> +       else
> +               printf("rte_event_port_setup for prod_port ok\n");
> +
> +       /* Set consumer and worker new event threshold to max */
> +       port_config.new_event_threshold = dev_info.max_num_events;
> +       ret = rte_event_port_setup(id, cons_port, &port_config);
> +       if (ret < 0)
> +               rte_panic("Failed to create the consumer port\n");
> +       else
> +               printf("rte_event_port_setup for cons_port ok\n");
> +
> +       port_config.event_port_cfg = 0;
> +
> +       /* Create load-balanced worker ports */
> +       for (i = 0; i < num_workers; i++) {
> +               worker_data[i].event_port_id = i + worker_port_base;
> +               ret = rte_event_port_setup(id, worker_data[i].event_port_id,
> +                                          &port_config);
> +               if (ret < 0)
> +                       rte_panic("Failed to create worker port #%d\n", i);
> +               else
> +                       printf("rte_event_port_setup for worker port %d ok\n",
> +                              i);
> +       }
> +
> +       printf("link worker queues\n");
> +       /* Map ports/qids */
> +       for (i = 0; i < num_workers; i++) {
> +               queue_id = worker_qid;
> +               priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
> +
> +               ret = rte_event_port_link(id, worker_data[i].event_port_id,
> +                                         &queue_id, &priority, 1);
> +               if (ret != 1)
> +                       rte_panic("Failed to map worker%d port to worker_qid\n",
> +                                 i);
> +       }
> +
> +       printf("link consumer queue\n");
> +       /* Link consumer port to its QID */
> +       queue_id = cons_qid;
> +       priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
> +
> +       ret = rte_event_port_link(id, cons_port, &queue_id, &priority, 1);
> +       if (ret != 1)
> +               rte_panic("Failed to map consumer port to cons_qid\n");
> +
> +       printf("link producer queue\n");
> +       /* Link producer port to its QID */
> +       queue_id = prod_qid;
> +       priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
> +
> +       ret = rte_event_port_link(id, prod_port, &queue_id, &priority, 1);
> +       if (ret != 1)
> +               rte_panic("Failed to map producer port to prod_qid\n");
> +
> +       /* Dispatch to workers */
> +       if (num_workers) {
> +               *prod_data = (struct prod_data){.event_dev_id = id,
> +                                               .event_port_id = prod_port,
> +                                               .qid = worker_qid};
> +               *cons_data = (struct cons_data){.event_dev_id = id,
> +                                               .event_port_id = cons_port};
> +               for (i = 0; i < num_workers; i++) {
> +                       struct worker_data *w = &worker_data[i];
> +                       w->event_dev_id = id;
> +                       w->qid = cons_qid;
> +               }
> +       } else {
> +               *prod_data = (struct prod_data){.event_dev_id = id,
> +                                               .event_port_id = prod_port,
> +                                               .qid = cons_qid};
> +               *cons_data = (struct cons_data){.event_dev_id = id,
> +                                               .event_port_id = cons_port};
> +       }
> +
> +       ret = rte_event_dev_start(id);
> +       if (ret)
> +               rte_panic("Failed to start the event dev\n");
> +       if (g_is_mbuf) {
> +               mp = rte_pktmbuf_pool_create("packet_pool",
> +                               /* mbufs */ dev_info.max_num_events,
> +                               /* cache_size */ 512,
> +                               /* priv_size*/ 0,
> +                               /* data_room_size */ 2048,
> +                               rte_socket_id());
> +
> +               if (mp == NULL) {
> +                       printf("mbuf pool create failed\n");
> +                       return -1;
> +               }
> +       }
> +       return (uint8_t) id;
> +}
> +
> +int
> +main(int argc, char **argv)
> +{
> +       struct prod_data prod_data = {0};
> +       struct cons_data cons_data = {0};
> +       uint64_t start, end;
> +       struct worker_data *worker_data = NULL;
> +       unsigned int nworkers = 0;
> +       int lcore_id;
> +       int err;
> +       int has_prod = 0;
> +       int has_cons = 0;
> +       int evdev_id = 0; /* TODO - allow app to override */
> +
> +       done = 0;
> +       quiet = 0;
> +       mp = NULL;
> +       g_is_mbuf = false;
> +
> +       err = rte_eal_init(argc, argv);
> +       if (err < 0)
> +               rte_panic("Invalid EAL arguments\n");
> +
> +       argc -= err;
> +       argv += err;
> +
> +       /* Parse cli options*/
> +       parse_app_args(argc, argv);
> +
> +       if (!quiet) {
> +               printf("  Config:\n");
> +               printf("\tworkers: %d\n", num_workers);
> +               printf("\tpackets: %"PRIu64"\n", num_packets);
> +               if (sched_type == RTE_SCHED_TYPE_ORDERED)
> +                       printf("\tworker_qid type: ordered\n");
> +               if (sched_type == RTE_SCHED_TYPE_ATOMIC)
> +                       printf("\tworker_qid type: atomic\n");
> +               printf("\n");
> +       }
> +
> +       const unsigned int cores_needed = num_workers +
> +                       /*main*/ 1 +
> +                       /*producer*/ 1 +
> +                       /*consumer*/ 1;
> +
> +       if (!quiet) {
> +               printf("Number of cores available: %d\n", rte_lcore_count());
> +               printf("Number of cores to be used: %d\n", cores_needed);
> +       }
> +
> +       if (rte_lcore_count() < cores_needed)
> +               rte_panic("Too few cores\n");
> +
> +       const uint8_t ndevs = rte_event_dev_count();
> +       if (ndevs == 0)
> +               rte_panic(
> +                       "No event devs found. Do you need"
> +                       " to pass in a --vdev flag?\n");
> +       if (ndevs > 1)
> +               fprintf(stderr,
> +                       "Warning: More than one event dev, but using idx 0");
> +
> +       if (num_workers) {
> +               worker_data = rte_calloc(0, num_workers,
> +                                        sizeof(worker_data[0]), 0);
> +               if (worker_data == NULL)
> +                       rte_panic("rte_calloc failed\n");
> +       }
> +
> +       uint8_t id = setup_event_dev(&prod_data, &cons_data, worker_data,
> +                                   evdev_id);
> +
> +       printf("setup_event_dev returned eventdev_id = %d\n", id);
> +
> +       start = rte_rdtsc();
> +
> +       RTE_LCORE_FOREACH_WORKER(lcore_id) {
> +               if (has_prod && has_cons && nworkers == num_workers)
> +                       break;
> +
> +               if (!has_prod) {
> +                       err = rte_eal_remote_launch(producer, &prod_data,
> +                                                   lcore_id);
> +                       if (err)
> +                               rte_panic("Failed to launch producer\n");
> +                       printf("Launched producer\n");
> +                       has_prod = 1;
> +                       continue;
> +               }
> +
> +               if (!has_cons) {
> +                       err = rte_eal_remote_launch(consumer, &cons_data,
> +                                                   lcore_id);
> +                       if (err)
> +                               rte_panic("Failed to launch consumer\n");
> +                       printf("Launched consumer\n");
> +                       has_cons = 1;
> +                       continue;
> +               }
> +
> +               if (nworkers < num_workers) {
> +                       err = rte_eal_remote_launch(worker,
> +                                                   &worker_data[nworkers],
> +                                                   lcore_id);
> +                       if (err)
> +                               rte_panic("Failed to launch worker%d\n",
> +                                         nworkers);
> +                       nworkers++;
> +                       printf("Launched worker %d\n", nworkers);
> +                       continue;
> +               }
> +       }
> +
> +       rte_eal_mp_wait_lcore();
> +       end = rte_rdtsc();
> +       printf("[%s()] DLB scheduled %"PRIu64" packets in %"PRIu64" cycles\n",
> +              __func__, num_packets, end - start);
> +       printf("[%s()] \t %f packets/sec\n",
> +              __func__, (float) (num_packets * rte_get_timer_hz()) /
> +              (float) (end - start));
> +
> +       printf("Test Complete\n");
> +
> +       /* Cleanup done automatically by kernel on app exit */
> +
> +       return 0;
> +}
> diff --git a/examples/eventdev_producer_consumer/meson.build b/examples/eventdev_producer_consumer/meson.build
> new file mode 100644
> index 0000000000..e18739432c
> --- /dev/null
> +++ b/examples/eventdev_producer_consumer/meson.build
> @@ -0,0 +1,13 @@
> +# SPDX-License-Identifier: BSD-3-Clause
> +# Copyright(c) 2017 Intel Corporation
> +
> +# meson file, for building this example as part of a main DPDK build.
> +#
> +# To build this example as a standalone application with an already-installed
> +# DPDK instance, use 'make'
> +
> +allow_experimental_apis = true
> +deps += 'eventdev'
> +sources = files(
> +        'main.c',
> +)
> diff --git a/examples/meson.build b/examples/meson.build
> index 81e93799f2..9d95ecc8e0 100644
> --- a/examples/meson.build
> +++ b/examples/meson.build
> @@ -15,6 +15,7 @@ all_examples = [
>          'dma',
>          'ethtool',
>          'eventdev_pipeline',
> +       'eventdev_producer_consumer',
>          'fips_validation',
>          'flow_classify',
>          'flow_filtering',
> --
> 2.23.0
>

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2022-08-27 11:58 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-08-18 17:02 [PATCH] examples: add eventdev_producer_consumer example Timothy McDaniel
2022-08-22 19:26 ` [PATCH v2] examples/eventdev_producer_consumer: fix 32-bit checkpatch issues Timothy McDaniel
2022-08-22 20:04 ` [PATCH v3] examples: add eventdev_producer_consumer example Timothy McDaniel
2022-08-22 20:51 ` [PATCH v4] " Timothy McDaniel
2022-08-27 11:57   ` Jerin Jacob

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).