void
perf_opt_dump(struct evt_options *opt, uint8_t nb_queues)
{
evt_dump("nb_prod_lcores", "%d", evt_nr_active_lcores(opt->plcores));
evt_dump_producer_lcores(opt);
evt_dump("nb_worker_lcores", "%d", evt_nr_active_lcores(opt->wlcores));
evt_dump_worker_lcores(opt);
evt_dump_nb_stages(opt);
evt_dump("nb_evdev_ports", "%d", perf_nb_event_ports(opt));
evt_dump("nb_evdev_queues", "%d", nb_queues);
evt_dump_queue_priority(opt);
evt_dump_sched_type_list(opt);
evt_dump_producer_type(opt);
}
static inline int
atq_nb_event_queues(struct evt_options *opt)
{
/* nb_queues = number of producers */
return opt->prod_type == EVT_PROD_TYPE_ETH_RX_ADPTR ?
rte_eth_dev_count_avail() : evt_nr_active_lcores(opt->plcores);
}
static int
order_queue_eventdev_setup(struct evt_test *test, struct evt_options *opt)
{
int ret;
const uint8_t nb_workers = evt_nr_active_lcores(opt->wlcores);
/* number of active worker cores + 1 producer */
const uint8_t nb_ports = nb_workers + 1;
const struct rte_event_dev_config config = {
.nb_event_queues = NB_QUEUES,/* q0 ordered, q1 atomic */
.nb_event_ports = nb_ports,
.nb_events_limit = 4096,
.nb_event_queue_flows = opt->nb_flows,
.nb_event_port_dequeue_depth = 128,
.nb_event_port_enqueue_depth = 128,
};
ret = rte_event_dev_configure(opt->dev_id, &config);
if (ret) {
evt_err("failed to configure eventdev %d", opt->dev_id);
return ret;
}
/* q0 (ordered queue) configuration */
struct rte_event_queue_conf q0_ordered_conf = {
.priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
.schedule_type = RTE_SCHED_TYPE_ORDERED,
.nb_atomic_flows = opt->nb_flows,
.nb_atomic_order_sequences = opt->nb_flows,
};
ret = rte_event_queue_setup(opt->dev_id, 0, &q0_ordered_conf);
if (ret) {
evt_err("failed to setup queue0 eventdev %d", opt->dev_id);
return ret;
}
/* q1 (atomic queue) configuration */
struct rte_event_queue_conf q1_atomic_conf = {
.priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
.schedule_type = RTE_SCHED_TYPE_ATOMIC,
.nb_atomic_flows = opt->nb_flows,
.nb_atomic_order_sequences = opt->nb_flows,
};
ret = rte_event_queue_setup(opt->dev_id, 1, &q1_atomic_conf);
if (ret) {
evt_err("failed to setup queue1 eventdev %d", opt->dev_id);
return ret;
}
/* setup one port per worker, linking to all queues */
ret = order_event_dev_port_setup(test, opt, nb_workers, NB_QUEUES);
if (ret)
return ret;
ret = evt_service_setup(opt->dev_id);
if (ret) {
evt_err("No service lcore found to run event dev.");
return ret;
}
ret = rte_event_dev_start(opt->dev_id);
if (ret) {
evt_err("failed to start eventdev %d", opt->dev_id);
return ret;
}
return 0;
}
static void
order_queue_opt_dump(struct evt_options *opt)
{
order_opt_dump(opt);
evt_dump("nb_evdev_queues", "%d", NB_QUEUES);
}
static bool
order_queue_capability_check(struct evt_options *opt)
{
struct rte_event_dev_info dev_info;
rte_event_dev_info_get(opt->dev_id, &dev_info);
if (dev_info.max_event_queues < NB_QUEUES || dev_info.max_event_ports <
order_nb_event_ports(opt)) {
evt_err("not enough eventdev queues=%d/%d or ports=%d/%d",
NB_QUEUES, dev_info.max_event_queues,
order_nb_event_ports(opt), dev_info.max_event_ports);
return false;
}
return true;
}
static const struct evt_test_ops order_queue = {
.cap_check = order_queue_capability_check,
.opt_check = order_opt_check,
.opt_dump = order_queue_opt_dump,
.test_setup = order_test_setup,
.mempool_setup = order_mempool_setup,
.eventdev_setup = order_queue_eventdev_setup,
.launch_lcores = order_queue_launch_lcores,
.eventdev_destroy = order_eventdev_destroy,
.mempool_destroy = order_mempool_destroy,
.test_result = order_test_result,
.test_destroy = order_test_destroy,
};
EVT_TEST_REGISTER(order_queue);
int
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
*/
lcores = opt->prod_type == EVT_PROD_TYPE_SYNT ? 3 : 2;
if (rte_lcore_count() < lcores) {
evt_err("test need minimum %d lcores", lcores);
return -1;
}
/* Validate worker lcores */
if (evt_lcores_has_overlap(opt->wlcores, rte_get_master_lcore())) {
evt_err("worker lcores overlaps with master lcore");
return -1;
}
if (evt_lcores_has_overlap_multi(opt->wlcores, opt->plcores)) {
evt_err("worker lcores overlaps producer lcores");
return -1;
}
if (evt_has_disabled_lcore(opt->wlcores)) {
evt_err("one or more workers lcores are not enabled");
return -1;
}
if (!evt_has_active_lcore(opt->wlcores)) {
evt_err("minimum one worker is required");
return -1;
}
if (opt->prod_type == EVT_PROD_TYPE_SYNT) {
/* Validate producer lcores */
if (evt_lcores_has_overlap(opt->plcores,
rte_get_master_lcore())) {
evt_err("producer lcores overlaps with master lcore");
return -1;
}
if (evt_has_disabled_lcore(opt->plcores)) {
evt_err("one or more producer lcores are not enabled");
return -1;
}
if (!evt_has_active_lcore(opt->plcores)) {
evt_err("minimum one producer is required");
return -1;
}
}
if (evt_has_invalid_stage(opt))
return -1;
if (evt_has_invalid_sched_type(opt))
return -1;
if (nb_queues > EVT_MAX_QUEUES) {
evt_err("number of queues exceeds %d", EVT_MAX_QUEUES);
return -1;
}
if (perf_nb_event_ports(opt) > EVT_MAX_PORTS) {
evt_err("number of ports exceeds %d", EVT_MAX_PORTS);
return -1;
}
/* Fixups */
if (opt->nb_stages == 1 && opt->fwd_latency) {
evt_info("fwd_latency is valid when nb_stages > 1, disabling");
opt->fwd_latency = 0;
}
if (opt->fwd_latency && !opt->q_priority) {
evt_info("enabled queue priority for latency measurement");
opt->q_priority = 1;
}
if (opt->nb_pkts == 0)
opt->nb_pkts = INT64_MAX/evt_nr_active_lcores(opt->plcores);
return 0;
}
int
perf_event_dev_port_setup(struct evt_test *test, struct evt_options *opt,
uint8_t stride, uint8_t nb_queues,
const struct rte_event_port_conf *port_conf)
{
struct test_perf *t = evt_test_priv(test);
uint16_t port, prod;
int ret = -1;
/* setup one port per worker, linking to all queues */
for (port = 0; port < evt_nr_active_lcores(opt->wlcores);
port++) {
struct worker_data *w = &t->worker[port];
w->dev_id = opt->dev_id;
w->port_id = port;
w->t = t;
w->processed_pkts = 0;
w->latency = 0;
ret = rte_event_port_setup(opt->dev_id, port, port_conf);
if (ret) {
evt_err("failed to setup port %d", port);
return ret;
}
ret = rte_event_port_link(opt->dev_id, port, NULL, NULL, 0);
if (ret != nb_queues) {
evt_err("failed to link all queues to port %d", port);
return -EINVAL;
}
}
/* port for producers, no links */
if (opt->prod_type == EVT_PROD_TYPE_ETH_RX_ADPTR) {
for ( ; port < perf_nb_event_ports(opt); port++) {
struct prod_data *p = &t->prod[port];
p->t = t;
}
ret = perf_event_rx_adapter_setup(opt, stride, *port_conf);
if (ret)
return ret;
} else {
prod = 0;
for ( ; port < perf_nb_event_ports(opt); port++) {
struct prod_data *p = &t->prod[port];
p->dev_id = opt->dev_id;
p->port_id = port;
p->queue_id = prod * stride;
p->t = t;
ret = rte_event_port_setup(opt->dev_id, port,
port_conf);
if (ret) {
evt_err("failed to setup port %d", port);
return ret;
}
prod++;
}
}
return ret;
}
int
perf_launch_lcores(struct evt_test *test, struct evt_options *opt,
int (*worker)(void *))
{
int ret, lcore_id;
struct test_perf *t = evt_test_priv(test);
int port_idx = 0;
/* launch workers */
RTE_LCORE_FOREACH_SLAVE(lcore_id) {
if (!(opt->wlcores[lcore_id]))
continue;
ret = rte_eal_remote_launch(worker,
&t->worker[port_idx], lcore_id);
if (ret) {
evt_err("failed to launch worker %d", lcore_id);
return ret;
}
port_idx++;
}
/* launch producers */
RTE_LCORE_FOREACH_SLAVE(lcore_id) {
if (!(opt->plcores[lcore_id]))
continue;
ret = rte_eal_remote_launch(perf_producer_wrapper,
&t->prod[port_idx], lcore_id);
if (ret) {
evt_err("failed to launch perf_producer %d", lcore_id);
return ret;
}
port_idx++;
}
const uint64_t total_pkts = opt->nb_pkts *
evt_nr_active_lcores(opt->plcores);
uint64_t dead_lock_cycles = rte_get_timer_cycles();
int64_t dead_lock_remaining = total_pkts;
const uint64_t dead_lock_sample = rte_get_timer_hz() * 5;
uint64_t perf_cycles = rte_get_timer_cycles();
int64_t perf_remaining = total_pkts;
const uint64_t perf_sample = rte_get_timer_hz();
static float total_mpps;
static uint64_t samples;
const uint64_t freq_mhz = rte_get_timer_hz() / 1000000;
int64_t remaining = t->outstand_pkts - processed_pkts(t);
while (t->done == false) {
const uint64_t new_cycles = rte_get_timer_cycles();
if ((new_cycles - perf_cycles) > perf_sample) {
const uint64_t latency = total_latency(t);
const uint64_t pkts = processed_pkts(t);
remaining = t->outstand_pkts - pkts;
float mpps = (float)(perf_remaining-remaining)/1000000;
perf_remaining = remaining;
perf_cycles = new_cycles;
total_mpps += mpps;
++samples;
if (opt->fwd_latency && pkts > 0) {
printf(CLGRN"\r%.3f mpps avg %.3f mpps [avg fwd latency %.3f us] "CLNRM,
mpps, total_mpps/samples,
(float)(latency/pkts)/freq_mhz);
} else {
printf(CLGRN"\r%.3f mpps avg %.3f mpps"CLNRM,
mpps, total_mpps/samples);
}
fflush(stdout);
if (remaining <= 0) {
t->result = EVT_TEST_SUCCESS;
if (opt->prod_type == EVT_PROD_TYPE_SYNT) {
t->done = true;
rte_smp_wmb();
break;
}
}
}
if (new_cycles - dead_lock_cycles > dead_lock_sample &&
opt->prod_type == EVT_PROD_TYPE_SYNT) {
remaining = t->outstand_pkts - processed_pkts(t);
if (dead_lock_remaining == remaining) {
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;
dead_lock_cycles = new_cycles;
}
}
printf("\n");
return 0;
}
static int
perf_atq_eventdev_setup(struct evt_test *test, struct evt_options *opt)
{
int ret;
uint8_t queue;
uint8_t nb_queues;
uint8_t nb_ports;
struct rte_event_dev_info dev_info;
nb_ports = evt_nr_active_lcores(opt->wlcores);
nb_ports += (opt->prod_type == EVT_PROD_TYPE_ETH_RX_ADPTR ||
opt->prod_type == EVT_PROD_TYPE_EVENT_TIMER_ADPTR) ? 0 :
evt_nr_active_lcores(opt->plcores);
nb_queues = atq_nb_event_queues(opt);
memset(&dev_info, 0, sizeof(struct rte_event_dev_info));
ret = rte_event_dev_info_get(opt->dev_id, &dev_info);
if (ret) {
evt_err("failed to get eventdev info %d", opt->dev_id);
return ret;
}
const struct rte_event_dev_config config = {
.nb_event_queues = nb_queues,
.nb_event_ports = nb_ports,
.nb_events_limit = dev_info.max_num_events,
.nb_event_queue_flows = opt->nb_flows,
.nb_event_port_dequeue_depth =
dev_info.max_event_port_dequeue_depth,
.nb_event_port_enqueue_depth =
dev_info.max_event_port_enqueue_depth,
};
ret = rte_event_dev_configure(opt->dev_id, &config);
if (ret) {
evt_err("failed to configure eventdev %d", opt->dev_id);
return ret;
}
struct rte_event_queue_conf q_conf = {
.priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
.event_queue_cfg = RTE_EVENT_QUEUE_CFG_ALL_TYPES,
.nb_atomic_flows = opt->nb_flows,
.nb_atomic_order_sequences = opt->nb_flows,
};
/* queue configurations */
for (queue = 0; queue < nb_queues; queue++) {
ret = rte_event_queue_setup(opt->dev_id, queue, &q_conf);
if (ret) {
evt_err("failed to setup queue=%d", queue);
return ret;
}
}
if (opt->wkr_deq_dep > dev_info.max_event_port_dequeue_depth)
opt->wkr_deq_dep = dev_info.max_event_port_dequeue_depth;
/* port configuration */
const struct rte_event_port_conf p_conf = {
.dequeue_depth = opt->wkr_deq_dep,
.enqueue_depth = dev_info.max_event_port_dequeue_depth,
.new_event_threshold = dev_info.max_num_events,
};
ret = perf_event_dev_port_setup(test, opt, 1 /* stride */, nb_queues,
&p_conf);
if (ret)
return ret;
if (!evt_has_distributed_sched(opt->dev_id)) {
uint32_t service_id;
rte_event_dev_service_id_get(opt->dev_id, &service_id);
ret = evt_service_setup(service_id);
if (ret) {
evt_err("No service lcore found to run event dev.");
return ret;
}
}
ret = rte_event_dev_start(opt->dev_id);
if (ret) {
evt_err("failed to start eventdev %d", opt->dev_id);
return ret;
}
return 0;
}
static void
perf_atq_opt_dump(struct evt_options *opt)
{
perf_opt_dump(opt, atq_nb_event_queues(opt));
}
static int
perf_atq_opt_check(struct evt_options *opt)
{
return perf_opt_check(opt, atq_nb_event_queues(opt));
}
static bool
perf_atq_capability_check(struct evt_options *opt)
{
struct rte_event_dev_info dev_info;
rte_event_dev_info_get(opt->dev_id, &dev_info);
if (dev_info.max_event_queues < atq_nb_event_queues(opt) ||
dev_info.max_event_ports < perf_nb_event_ports(opt)) {
evt_err("not enough eventdev queues=%d/%d or ports=%d/%d",
atq_nb_event_queues(opt), dev_info.max_event_queues,
perf_nb_event_ports(opt), dev_info.max_event_ports);
}
if (!evt_has_all_types_queue(opt->dev_id))
return false;
return true;
}
static const struct evt_test_ops perf_atq = {
.cap_check = perf_atq_capability_check,
.opt_check = perf_atq_opt_check,
.opt_dump = perf_atq_opt_dump,
.test_setup = perf_test_setup,
.ethdev_setup = perf_ethdev_setup,
.mempool_setup = perf_mempool_setup,
.eventdev_setup = perf_atq_eventdev_setup,
.launch_lcores = perf_atq_launch_lcores,
.eventdev_destroy = perf_eventdev_destroy,
.mempool_destroy = perf_mempool_destroy,
.ethdev_destroy = perf_ethdev_destroy,
.test_result = perf_test_result,
.test_destroy = perf_test_destroy,
};
EVT_TEST_REGISTER(perf_atq);
