X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;ds=sidebyside;f=app%2Ftest-eventdev%2Ftest_perf_common.c;h=ed2b81f0632d276f0a8096d92b7a30959d4969b7;hb=80da61198b6be66ab16964eb0944c79c0325c40e;hp=e7cf75a7d8912da711669637a72687b93c6928f8;hpb=99c25664149b9fe4b0d66e6897ab803552e649fb;p=dpdk.git

diff --git a/app/test-eventdev/test_perf_common.c b/app/test-eventdev/test_perf_common.c
index e7cf75a7d8..ed2b81f063 100644
--- a/app/test-eventdev/test_perf_common.c
+++ b/app/test-eventdev/test_perf_common.c
@@ -2,6 +2,8 @@
  * Copyright(c) 2017 Cavium, Inc
  */
 
+#include <math.h>
+
 #include "test_perf_common.h"
 
 int
@@ -17,7 +19,7 @@ perf_test_result(struct evt_test *test, struct evt_options *opt)
 		total += t->worker[i].processed_pkts;
 	for (i = 0; i < t->nb_workers; i++)
 		printf("Worker %d packets: "CLGRN"%"PRIx64" "CLNRM"percentage:"
-				CLGRN" %3.2f\n"CLNRM, i,
+				CLGRN" %3.2f"CLNRM"\n", i,
 				t->worker[i].processed_pkts,
 				(((double)t->worker[i].processed_pkts)/total)
 				* 100);
@@ -75,6 +77,71 @@ perf_producer(void *arg)
 	return 0;
 }
 
+static inline int
+perf_producer_burst(void *arg)
+{
+	uint32_t i;
+	uint64_t timestamp;
+	struct rte_event_dev_info dev_info;
+	struct prod_data *p  = arg;
+	struct test_perf *t = p->t;
+	struct evt_options *opt = t->opt;
+	const uint8_t dev_id = p->dev_id;
+	const uint8_t port = p->port_id;
+	struct rte_mempool *pool = t->pool;
+	const uint64_t nb_pkts = t->nb_pkts;
+	const uint32_t nb_flows = t->nb_flows;
+	uint32_t flow_counter = 0;
+	uint16_t enq = 0;
+	uint64_t count = 0;
+	struct perf_elt *m[MAX_PROD_ENQ_BURST_SIZE + 1];
+	struct rte_event ev[MAX_PROD_ENQ_BURST_SIZE + 1];
+	uint32_t burst_size = opt->prod_enq_burst_sz;
+
+	memset(m, 0, sizeof(*m) * (MAX_PROD_ENQ_BURST_SIZE + 1));
+	rte_event_dev_info_get(dev_id, &dev_info);
+	if (dev_info.max_event_port_enqueue_depth < burst_size)
+		burst_size = dev_info.max_event_port_enqueue_depth;
+
+	if (opt->verbose_level > 1)
+		printf("%s(): lcore %d dev_id %d port=%d queue %d\n", __func__,
+				rte_lcore_id(), dev_id, port, p->queue_id);
+
+	for (i = 0; i < burst_size; i++) {
+		ev[i].op = RTE_EVENT_OP_NEW;
+		ev[i].queue_id = p->queue_id;
+		ev[i].sched_type = t->opt->sched_type_list[0];
+		ev[i].priority = RTE_EVENT_DEV_PRIORITY_NORMAL;
+		ev[i].event_type =  RTE_EVENT_TYPE_CPU;
+		ev[i].sub_event_type = 0; /* stage 0 */
+	}
+
+	while (count < nb_pkts && t->done == false) {
+		if (rte_mempool_get_bulk(pool, (void **)m, burst_size) < 0)
+			continue;
+		timestamp = rte_get_timer_cycles();
+		for (i = 0; i < burst_size; i++) {
+			ev[i].flow_id = flow_counter++ % nb_flows;
+			ev[i].event_ptr = m[i];
+			m[i]->timestamp = timestamp;
+		}
+		enq = rte_event_enqueue_burst(dev_id, port, ev, burst_size);
+		while (enq < burst_size) {
+			enq += rte_event_enqueue_burst(dev_id, port,
+							ev + enq,
+							burst_size - enq);
+			if (t->done)
+				break;
+			rte_pause();
+			timestamp = rte_get_timer_cycles();
+			for (i = enq; i < burst_size; i++)
+				m[i]->timestamp = timestamp;
+		}
+		count += burst_size;
+	}
+	return 0;
+}
+
 static inline int
 perf_event_timer_producer(void *arg)
 {
@@ -95,11 +162,13 @@ perf_event_timer_producer(void *arg)
 	uint64_t timeout_ticks = opt->expiry_nsec / opt->timer_tick_nsec;
 
 	memset(&tim, 0, sizeof(struct rte_event_timer));
-	timeout_ticks = opt->optm_timer_tick_nsec ?
-			(timeout_ticks * opt->timer_tick_nsec)
-			/ opt->optm_timer_tick_nsec : timeout_ticks;
+	timeout_ticks =
+		opt->optm_timer_tick_nsec
+			? ceil((double)(timeout_ticks * opt->timer_tick_nsec) /
+			       opt->optm_timer_tick_nsec)
+			: timeout_ticks;
 	timeout_ticks += timeout_ticks ? 0 : 1;
-	tim.ev.event_type =  RTE_EVENT_TYPE_TIMER;
+	tim.ev.event_type = RTE_EVENT_TYPE_TIMER;
 	tim.ev.op = RTE_EVENT_OP_NEW;
 	tim.ev.sched_type = t->opt->sched_type_list[0];
 	tim.ev.queue_id = p->queue_id;
@@ -133,8 +202,9 @@ perf_event_timer_producer(void *arg)
 	fflush(stdout);
 	rte_delay_ms(1000);
 	printf("%s(): lcore %d Average event timer arm latency = %.3f us\n",
-			__func__, rte_lcore_id(), (float)(arm_latency / count) /
-			(rte_get_timer_hz() / 1000000));
+			__func__, rte_lcore_id(),
+			count ? (float)(arm_latency / count) /
+			(rte_get_timer_hz() / 1000000) : 0);
 	return 0;
 }
 
@@ -158,11 +228,13 @@ perf_event_timer_producer_burst(void *arg)
 	uint64_t timeout_ticks = opt->expiry_nsec / opt->timer_tick_nsec;
 
 	memset(&tim, 0, sizeof(struct rte_event_timer));
-	timeout_ticks = opt->optm_timer_tick_nsec ?
-			(timeout_ticks * opt->timer_tick_nsec)
-			/ opt->optm_timer_tick_nsec : timeout_ticks;
+	timeout_ticks =
+		opt->optm_timer_tick_nsec
+			? ceil((double)(timeout_ticks * opt->timer_tick_nsec) /
+			       opt->optm_timer_tick_nsec)
+			: timeout_ticks;
 	timeout_ticks += timeout_ticks ? 0 : 1;
-	tim.ev.event_type =  RTE_EVENT_TYPE_TIMER;
+	tim.ev.event_type = RTE_EVENT_TYPE_TIMER;
 	tim.ev.op = RTE_EVENT_OP_NEW;
 	tim.ev.sched_type = t->opt->sched_type_list[0];
 	tim.ev.queue_id = p->queue_id;
@@ -194,8 +266,9 @@ perf_event_timer_producer_burst(void *arg)
 	fflush(stdout);
 	rte_delay_ms(1000);
 	printf("%s(): lcore %d Average event timer arm latency = %.3f us\n",
-			__func__, rte_lcore_id(), (float)(arm_latency / count) /
-			(rte_get_timer_hz() / 1000000));
+			__func__, rte_lcore_id(),
+			count ? (float)(arm_latency / count) /
+			(rte_get_timer_hz() / 1000000) : 0);
 	return 0;
 }
 
@@ -204,9 +277,21 @@ perf_producer_wrapper(void *arg)
 {
 	struct prod_data *p  = arg;
 	struct test_perf *t = p->t;
-	/* Launch the producer function only in case of synthetic producer. */
-	if (t->opt->prod_type == EVT_PROD_TYPE_SYNT)
+	bool burst = evt_has_burst_mode(p->dev_id);
+
+	/* In case of synthetic producer, launch perf_producer or
+	 * perf_producer_burst depending on producer enqueue burst size
+	 */
+	if (t->opt->prod_type == EVT_PROD_TYPE_SYNT &&
+			t->opt->prod_enq_burst_sz == 1)
 		return perf_producer(arg);
+	else if (t->opt->prod_type == EVT_PROD_TYPE_SYNT &&
+			t->opt->prod_enq_burst_sz > 1) {
+		if (!burst)
+			evt_err("This event device does not support burst mode");
+		else
+			return perf_producer_burst(arg);
+	}
 	else if (t->opt->prod_type == EVT_PROD_TYPE_EVENT_TIMER_ADPTR &&
 			!t->opt->timdev_use_burst)
 		return perf_event_timer_producer(arg);
@@ -222,7 +307,6 @@ processed_pkts(struct test_perf *t)
 	uint8_t i;
 	uint64_t total = 0;
 
-	rte_smp_rmb();
 	for (i = 0; i < t->nb_workers; i++)
 		total += t->worker[i].processed_pkts;
 
@@ -235,7 +319,6 @@ total_latency(struct test_perf *t)
 	uint8_t i;
 	uint64_t total = 0;
 
-	rte_smp_rmb();
 	for (i = 0; i < t->nb_workers; i++)
 		total += t->worker[i].latency;
 
@@ -252,7 +335,7 @@ perf_launch_lcores(struct evt_test *test, struct evt_options *opt,
 
 	int port_idx = 0;
 	/* launch workers */
-	RTE_LCORE_FOREACH_SLAVE(lcore_id) {
+	RTE_LCORE_FOREACH_WORKER(lcore_id) {
 		if (!(opt->wlcores[lcore_id]))
 			continue;
 
@@ -266,7 +349,7 @@ perf_launch_lcores(struct evt_test *test, struct evt_options *opt,
 	}
 
 	/* launch producers */
-	RTE_LCORE_FOREACH_SLAVE(lcore_id) {
+	RTE_LCORE_FOREACH_WORKER(lcore_id) {
 		if (!(opt->plcores[lcore_id]))
 			continue;
 
@@ -325,7 +408,6 @@ perf_launch_lcores(struct evt_test *test, struct evt_options *opt,
 					opt->prod_type ==
 					EVT_PROD_TYPE_EVENT_TIMER_ADPTR) {
 					t->done = true;
-					rte_smp_wmb();
 					break;
 				}
 			}
@@ -339,7 +421,6 @@ perf_launch_lcores(struct evt_test *test, struct evt_options *opt,
 				rte_event_dev_dump(opt->dev_id, stdout);
 				evt_err("No schedules for seconds, deadlock");
 				t->done = true;
-				rte_smp_wmb();
 				break;
 			}
 			dead_lock_remaining = remaining;
@@ -476,7 +557,10 @@ perf_event_dev_port_setup(struct evt_test *test, struct evt_options *opt,
 		w->processed_pkts = 0;
 		w->latency = 0;
 
-		ret = rte_event_port_setup(opt->dev_id, port, port_conf);
+		struct rte_event_port_conf conf = *port_conf;
+		conf.event_port_cfg |= RTE_EVENT_PORT_CFG_HINT_WORKER;
+
+		ret = rte_event_port_setup(opt->dev_id, port, &conf);
 		if (ret) {
 			evt_err("failed to setup port %d", port);
 			return ret;
@@ -496,7 +580,10 @@ perf_event_dev_port_setup(struct evt_test *test, struct evt_options *opt,
 			p->t = t;
 		}
 
-		ret = perf_event_rx_adapter_setup(opt, stride, *port_conf);
+		struct rte_event_port_conf conf = *port_conf;
+		conf.event_port_cfg |= RTE_EVENT_PORT_CFG_HINT_PRODUCER;
+
+		ret = perf_event_rx_adapter_setup(opt, stride, conf);
 		if (ret)
 			return ret;
 	} else if (opt->prod_type == EVT_PROD_TYPE_EVENT_TIMER_ADPTR) {
@@ -521,8 +608,12 @@ perf_event_dev_port_setup(struct evt_test *test, struct evt_options *opt,
 			p->queue_id = prod * stride;
 			p->t = t;
 
-			ret = rte_event_port_setup(opt->dev_id, port,
-					port_conf);
+			struct rte_event_port_conf conf = *port_conf;
+			conf.event_port_cfg |=
+				RTE_EVENT_PORT_CFG_HINT_PRODUCER |
+				RTE_EVENT_PORT_CFG_HINT_CONSUMER;
+
+			ret = rte_event_port_setup(opt->dev_id, port, &conf);
 			if (ret) {
 				evt_err("failed to setup port %d", port);
 				return ret;
@@ -539,8 +630,8 @@ perf_opt_check(struct evt_options *opt, uint64_t nb_queues)
 {
 	unsigned int lcores;
 
-	/* N producer + N worker + 1 master when producer cores are used
-	 * Else N worker + 1 master when Rx adapter is used
+	/* N producer + N worker + main when producer cores are used
+	 * Else N worker + main when Rx adapter is used
 	 */
 	lcores = opt->prod_type == EVT_PROD_TYPE_SYNT ? 3 : 2;
 
@@ -550,8 +641,8 @@ perf_opt_check(struct evt_options *opt, uint64_t nb_queues)
 	}
 
 	/* Validate worker lcores */
-	if (evt_lcores_has_overlap(opt->wlcores, rte_get_master_lcore())) {
-		evt_err("worker lcores overlaps with master lcore");
+	if (evt_lcores_has_overlap(opt->wlcores, rte_get_main_lcore())) {
+		evt_err("worker lcores overlaps with main lcore");
 		return -1;
 	}
 	if (evt_lcores_has_overlap_multi(opt->wlcores, opt->plcores)) {
@@ -571,8 +662,8 @@ perf_opt_check(struct evt_options *opt, uint64_t nb_queues)
 			opt->prod_type == EVT_PROD_TYPE_EVENT_TIMER_ADPTR) {
 		/* Validate producer lcores */
 		if (evt_lcores_has_overlap(opt->plcores,
-					rte_get_master_lcore())) {
-			evt_err("producer lcores overlaps with master lcore");
+					rte_get_main_lcore())) {
+			evt_err("producer lcores overlaps with main lcore");
 			return -1;
 		}
 		if (evt_has_disabled_lcore(opt->plcores)) {
@@ -631,6 +722,7 @@ perf_opt_dump(struct evt_options *opt, uint8_t nb_queues)
 	evt_dump_queue_priority(opt);
 	evt_dump_sched_type_list(opt);
 	evt_dump_producer_type(opt);
+	evt_dump("prod_enq_burst_sz", "%d", opt->prod_enq_burst_sz);
 }
 
 void
@@ -665,7 +757,6 @@ perf_ethdev_setup(struct evt_test *test, struct evt_options *opt)
 	struct rte_eth_conf port_conf = {
 		.rxmode = {
 			.mq_mode = ETH_MQ_RX_RSS,
-			.max_rx_pkt_len = RTE_ETHER_MAX_LEN,
 			.split_hdr_size = 0,
 		},
 		.rx_adv_conf = {