examples/eventdev: add generic worker pipeline
authorPavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Wed, 10 Jan 2018 11:10:02 +0000 (16:40 +0530)
committerJerin Jacob <jerin.jacob@caviumnetworks.com>
Fri, 19 Jan 2018 15:09:56 +0000 (16:09 +0100)
Rename existing pipeline as generic worker pipeline.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Acked-by: Harry van Haaren <harry.van.haaren@intel.com>
examples/eventdev_pipeline_sw_pmd/Makefile
examples/eventdev_pipeline_sw_pmd/main.c
examples/eventdev_pipeline_sw_pmd/pipeline_common.h
examples/eventdev_pipeline_sw_pmd/pipeline_worker_generic.c [new file with mode: 0644]

index 241d0c1..f2578da 100644 (file)
@@ -15,6 +15,7 @@ APP = eventdev_pipeline_sw_pmd
 
 # all source are stored in SRCS-y
 SRCS-y := main.c
+SRCS-y += pipeline_worker_generic.c
 
 CFLAGS += -O3
 CFLAGS += $(WERROR_FLAGS)
index fc445a1..5635fc5 100644 (file)
@@ -40,179 +40,6 @@ eth_tx_buffer_retry(struct rte_mbuf **pkts, uint16_t unsent,
        } while (_sent != unsent);
 }
 
-static int
-consumer(void)
-{
-       const uint64_t freq_khz = rte_get_timer_hz() / 1000;
-       struct rte_event packets[BATCH_SIZE];
-
-       static uint64_t received;
-       static uint64_t last_pkts;
-       static uint64_t last_time;
-       static uint64_t start_time;
-       unsigned int i, j;
-       uint8_t dev_id = cons_data.dev_id;
-       uint8_t port_id = cons_data.port_id;
-
-       uint16_t n = rte_event_dequeue_burst(dev_id, port_id,
-                       packets, RTE_DIM(packets), 0);
-
-       if (n == 0) {
-               for (j = 0; j < rte_eth_dev_count(); j++)
-                       rte_eth_tx_buffer_flush(j, 0, fdata->tx_buf[j]);
-               return 0;
-       }
-       if (start_time == 0)
-               last_time = start_time = rte_get_timer_cycles();
-
-       received += n;
-       for (i = 0; i < n; i++) {
-               uint8_t outport = packets[i].mbuf->port;
-               rte_eth_tx_buffer(outport, 0, fdata->tx_buf[outport],
-                               packets[i].mbuf);
-
-               packets[i].op = RTE_EVENT_OP_RELEASE;
-       }
-
-       if (cons_data.release) {
-               uint16_t nb_tx;
-
-               nb_tx = rte_event_enqueue_burst(dev_id, port_id, packets, n);
-               while (nb_tx < n)
-                       nb_tx += rte_event_enqueue_burst(dev_id, port_id,
-                                                        packets + nb_tx,
-                                                        n - nb_tx);
-       }
-
-       /* Print out mpps every 1<22 packets */
-       if (!cdata.quiet && received >= last_pkts + (1<<22)) {
-               const uint64_t now = rte_get_timer_cycles();
-               const uint64_t total_ms = (now - start_time) / freq_khz;
-               const uint64_t delta_ms = (now - last_time) / freq_khz;
-               uint64_t delta_pkts = received - last_pkts;
-
-               printf("# consumer RX=%"PRIu64", time %"PRIu64 "ms, "
-                       "avg %.3f mpps [current %.3f mpps]\n",
-                               received,
-                               total_ms,
-                               received / (total_ms * 1000.0),
-                               delta_pkts / (delta_ms * 1000.0));
-               last_pkts = received;
-               last_time = now;
-       }
-
-       cdata.num_packets -= n;
-       if (cdata.num_packets <= 0)
-               fdata->done = 1;
-
-       return 0;
-}
-
-static inline void
-schedule_devices(unsigned int lcore_id)
-{
-       if (fdata->rx_core[lcore_id]) {
-               rte_service_run_iter_on_app_lcore(fdata->rxadptr_service_id,
-                               !fdata->rx_single);
-       }
-
-       if (fdata->sched_core[lcore_id]) {
-               rte_service_run_iter_on_app_lcore(fdata->evdev_service_id,
-                               !fdata->sched_single);
-               if (cdata.dump_dev_signal) {
-                       rte_event_dev_dump(0, stdout);
-                       cdata.dump_dev_signal = 0;
-               }
-       }
-
-       if (fdata->tx_core[lcore_id] && (fdata->tx_single ||
-           rte_atomic32_cmpset(&(fdata->tx_lock), 0, 1))) {
-               consumer();
-               rte_atomic32_clear((rte_atomic32_t *)&(fdata->tx_lock));
-       }
-}
-
-static inline void
-work(struct rte_mbuf *m)
-{
-       struct ether_hdr *eth;
-       struct ether_addr addr;
-
-       /* change mac addresses on packet (to use mbuf data) */
-       /*
-        * FIXME Swap mac address properly and also handle the
-        * case for both odd and even number of stages that the
-        * addresses end up the same at the end of the pipeline
-        */
-       eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
-       ether_addr_copy(&eth->d_addr, &addr);
-       ether_addr_copy(&addr, &eth->d_addr);
-
-       /* do a number of cycles of work per packet */
-       volatile uint64_t start_tsc = rte_rdtsc();
-       while (rte_rdtsc() < start_tsc + cdata.worker_cycles)
-               rte_pause();
-}
-
-static int
-worker(void *arg)
-{
-       struct rte_event events[BATCH_SIZE];
-
-       struct worker_data *data = (struct worker_data *)arg;
-       uint8_t dev_id = data->dev_id;
-       uint8_t port_id = data->port_id;
-       size_t sent = 0, received = 0;
-       unsigned int lcore_id = rte_lcore_id();
-
-       while (!fdata->done) {
-               uint16_t i;
-
-               schedule_devices(lcore_id);
-
-               if (!fdata->worker_core[lcore_id]) {
-                       rte_pause();
-                       continue;
-               }
-
-               const uint16_t nb_rx = rte_event_dequeue_burst(dev_id, port_id,
-                               events, RTE_DIM(events), 0);
-
-               if (nb_rx == 0) {
-                       rte_pause();
-                       continue;
-               }
-               received += nb_rx;
-
-               for (i = 0; i < nb_rx; i++) {
-
-                       /* The first worker stage does classification */
-                       if (events[i].queue_id == cdata.qid[0])
-                               events[i].flow_id = events[i].mbuf->hash.rss
-                                                       % cdata.num_fids;
-
-                       events[i].queue_id = cdata.next_qid[events[i].queue_id];
-                       events[i].op = RTE_EVENT_OP_FORWARD;
-                       events[i].sched_type = cdata.queue_type;
-
-                       work(events[i].mbuf);
-               }
-               uint16_t nb_tx = rte_event_enqueue_burst(dev_id, port_id,
-                               events, nb_rx);
-               while (nb_tx < nb_rx && !fdata->done)
-                       nb_tx += rte_event_enqueue_burst(dev_id, port_id,
-                                                       events + nb_tx,
-                                                       nb_rx - nb_tx);
-               sent += nb_tx;
-       }
-
-       if (!cdata.quiet)
-               printf("  worker %u thread done. RX=%zu TX=%zu\n",
-                               rte_lcore_id(), received, sent);
-
-       return 0;
-}
-
 /*
  * Parse the coremask given as argument (hexadecimal string) and fill
  * the global configuration (core role and core count) with the parsed
@@ -425,70 +252,6 @@ parse_app_args(int argc, char **argv)
        }
 }
 
-static inline void
-init_rx_adapter(uint16_t nb_ports)
-{
-       int i;
-       int ret;
-       uint8_t evdev_id = 0;
-       struct rte_event_dev_info dev_info;
-
-       ret = rte_event_dev_info_get(evdev_id, &dev_info);
-
-       struct rte_event_port_conf rx_p_conf = {
-               .dequeue_depth = 8,
-               .enqueue_depth = 8,
-               .new_event_threshold = 1200,
-       };
-
-       if (rx_p_conf.dequeue_depth > dev_info.max_event_port_dequeue_depth)
-               rx_p_conf.dequeue_depth = dev_info.max_event_port_dequeue_depth;
-       if (rx_p_conf.enqueue_depth > dev_info.max_event_port_enqueue_depth)
-               rx_p_conf.enqueue_depth = dev_info.max_event_port_enqueue_depth;
-
-       ret = rte_event_eth_rx_adapter_create(cdata.rx_adapter_id, evdev_id,
-                       &rx_p_conf);
-       if (ret)
-               rte_exit(EXIT_FAILURE, "failed to create rx adapter[%d]",
-                               cdata.rx_adapter_id);
-
-       struct rte_event_eth_rx_adapter_queue_conf queue_conf = {
-               .ev.sched_type = cdata.queue_type,
-               .ev.queue_id = cdata.qid[0],
-       };
-
-       for (i = 0; i < nb_ports; i++) {
-               uint32_t cap;
-
-               ret = rte_event_eth_rx_adapter_caps_get(evdev_id, i, &cap);
-               if (ret)
-                       rte_exit(EXIT_FAILURE,
-                                       "failed to get event rx adapter "
-                                       "capabilities");
-
-               ret = rte_event_eth_rx_adapter_queue_add(cdata.rx_adapter_id, i,
-                               -1, &queue_conf);
-               if (ret)
-                       rte_exit(EXIT_FAILURE,
-                                       "Failed to add queues to Rx adapter");
-       }
-
-       ret = rte_event_eth_rx_adapter_service_id_get(cdata.rx_adapter_id,
-                               &fdata->rxadptr_service_id);
-       if (ret != -ESRCH && ret != 0) {
-               rte_exit(EXIT_FAILURE,
-                       "Error getting the service ID for Rx adapter\n");
-       }
-       rte_service_runstate_set(fdata->rxadptr_service_id, 1);
-       rte_service_set_runstate_mapped_check(fdata->rxadptr_service_id, 0);
-
-       ret = rte_event_eth_rx_adapter_start(cdata.rx_adapter_id);
-       if (ret)
-               rte_exit(EXIT_FAILURE, "Rx adapter[%d] start failed",
-                               cdata.rx_adapter_id);
-
-}
-
 /*
  * Initializes a given port using global settings and with the RX buffers
  * coming from the mbuf_pool passed as a parameter.
@@ -604,191 +367,12 @@ init_ports(unsigned int num_ports)
        return 0;
 }
 
-struct port_link {
-       uint8_t queue_id;
-       uint8_t priority;
-};
-
-static int
-setup_eventdev(struct cons_data *cons_data,
-               struct worker_data *worker_data)
+static void
+do_capability_setup(uint16_t nb_ethdev, uint8_t eventdev_id)
 {
-       const uint8_t dev_id = 0;
-       /* +1 stages is for a SINGLE_LINK TX stage */
-       const uint8_t nb_queues = cdata.num_stages + 1;
-       /* + 1 for consumer */
-       const uint8_t nb_ports = cdata.num_workers + 1;
-       struct rte_event_dev_config config = {
-                       .nb_event_queues = nb_queues,
-                       .nb_event_ports = nb_ports,
-                       .nb_events_limit  = 4096,
-                       .nb_event_queue_flows = 1024,
-                       .nb_event_port_dequeue_depth = 128,
-                       .nb_event_port_enqueue_depth = 128,
-       };
-       struct rte_event_port_conf wkr_p_conf = {
-                       .dequeue_depth = cdata.worker_cq_depth,
-                       .enqueue_depth = 64,
-                       .new_event_threshold = 4096,
-       };
-       struct rte_event_queue_conf wkr_q_conf = {
-                       .schedule_type = cdata.queue_type,
-                       .priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
-                       .nb_atomic_flows = 1024,
-                       .nb_atomic_order_sequences = 1024,
-       };
-       struct rte_event_port_conf tx_p_conf = {
-                       .dequeue_depth = 128,
-                       .enqueue_depth = 128,
-                       .new_event_threshold = 4096,
-       };
-       const struct rte_event_queue_conf tx_q_conf = {
-                       .priority = RTE_EVENT_DEV_PRIORITY_HIGHEST,
-                       .event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK,
-       };
-
-       struct port_link worker_queues[MAX_NUM_STAGES];
-       uint8_t disable_implicit_release;
-       struct port_link tx_queue;
-       unsigned int i;
-
-       int ret, ndev = rte_event_dev_count();
-       if (ndev < 1) {
-               printf("%d: No Eventdev Devices Found\n", __LINE__);
-               return -1;
-       }
-
-       struct rte_event_dev_info dev_info;
-       ret = rte_event_dev_info_get(dev_id, &dev_info);
-       printf("\tEventdev %d: %s\n", dev_id, dev_info.driver_name);
-
-       disable_implicit_release = (dev_info.event_dev_cap &
-                                   RTE_EVENT_DEV_CAP_IMPLICIT_RELEASE_DISABLE);
-
-       wkr_p_conf.disable_implicit_release = disable_implicit_release;
-       tx_p_conf.disable_implicit_release = disable_implicit_release;
-
-       if (dev_info.max_event_port_dequeue_depth <
-                       config.nb_event_port_dequeue_depth)
-               config.nb_event_port_dequeue_depth =
-                               dev_info.max_event_port_dequeue_depth;
-       if (dev_info.max_event_port_enqueue_depth <
-                       config.nb_event_port_enqueue_depth)
-               config.nb_event_port_enqueue_depth =
-                               dev_info.max_event_port_enqueue_depth;
-
-       ret = rte_event_dev_configure(dev_id, &config);
-       if (ret < 0) {
-               printf("%d: Error configuring device\n", __LINE__);
-               return -1;
-       }
-
-       /* Q creation - one load balanced per pipeline stage*/
-       printf("  Stages:\n");
-       for (i = 0; i < cdata.num_stages; i++) {
-               if (rte_event_queue_setup(dev_id, i, &wkr_q_conf) < 0) {
-                       printf("%d: error creating qid %d\n", __LINE__, i);
-                       return -1;
-               }
-               cdata.qid[i] = i;
-               cdata.next_qid[i] = i+1;
-               worker_queues[i].queue_id = i;
-               if (cdata.enable_queue_priorities) {
-                       /* calculate priority stepping for each stage, leaving
-                        * headroom of 1 for the SINGLE_LINK TX below
-                        */
-                       const uint32_t prio_delta =
-                               (RTE_EVENT_DEV_PRIORITY_LOWEST-1) /  nb_queues;
-
-                       /* higher priority for queues closer to tx */
-                       wkr_q_conf.priority =
-                               RTE_EVENT_DEV_PRIORITY_LOWEST - prio_delta * i;
-               }
-
-               const char *type_str = "Atomic";
-               switch (wkr_q_conf.schedule_type) {
-               case RTE_SCHED_TYPE_ORDERED:
-                       type_str = "Ordered";
-                       break;
-               case RTE_SCHED_TYPE_PARALLEL:
-                       type_str = "Parallel";
-                       break;
-               }
-               printf("\tStage %d, Type %s\tPriority = %d\n", i, type_str,
-                               wkr_q_conf.priority);
-       }
-       printf("\n");
-
-       /* final queue for sending to TX core */
-       if (rte_event_queue_setup(dev_id, i, &tx_q_conf) < 0) {
-               printf("%d: error creating qid %d\n", __LINE__, i);
-               return -1;
-       }
-       tx_queue.queue_id = i;
-       tx_queue.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
-
-       if (wkr_p_conf.dequeue_depth > config.nb_event_port_dequeue_depth)
-               wkr_p_conf.dequeue_depth = config.nb_event_port_dequeue_depth;
-       if (wkr_p_conf.enqueue_depth > config.nb_event_port_enqueue_depth)
-               wkr_p_conf.enqueue_depth = config.nb_event_port_enqueue_depth;
-
-       /* set up one port per worker, linking to all stage queues */
-       for (i = 0; i < cdata.num_workers; i++) {
-               struct worker_data *w = &worker_data[i];
-               w->dev_id = dev_id;
-               if (rte_event_port_setup(dev_id, i, &wkr_p_conf) < 0) {
-                       printf("Error setting up port %d\n", i);
-                       return -1;
-               }
-
-               uint32_t s;
-               for (s = 0; s < cdata.num_stages; s++) {
-                       if (rte_event_port_link(dev_id, i,
-                                               &worker_queues[s].queue_id,
-                                               &worker_queues[s].priority,
-                                               1) != 1) {
-                               printf("%d: error creating link for port %d\n",
-                                               __LINE__, i);
-                               return -1;
-                       }
-               }
-               w->port_id = i;
-       }
-
-       if (tx_p_conf.dequeue_depth > config.nb_event_port_dequeue_depth)
-               tx_p_conf.dequeue_depth = config.nb_event_port_dequeue_depth;
-       if (tx_p_conf.enqueue_depth > config.nb_event_port_enqueue_depth)
-               tx_p_conf.enqueue_depth = config.nb_event_port_enqueue_depth;
-
-       /* port for consumer, linked to TX queue */
-       if (rte_event_port_setup(dev_id, i, &tx_p_conf) < 0) {
-               printf("Error setting up port %d\n", i);
-               return -1;
-       }
-       if (rte_event_port_link(dev_id, i, &tx_queue.queue_id,
-                               &tx_queue.priority, 1) != 1) {
-               printf("%d: error creating link for port %d\n",
-                               __LINE__, i);
-               return -1;
-       }
-       *cons_data = (struct cons_data){.dev_id = dev_id,
-                                       .port_id = i,
-                                       .release = disable_implicit_release };
-
-       ret = rte_event_dev_service_id_get(dev_id,
-                               &fdata->evdev_service_id);
-       if (ret != -ESRCH && ret != 0) {
-               printf("Error getting the service ID for sw eventdev\n");
-               return -1;
-       }
-       rte_service_runstate_set(fdata->evdev_service_id, 1);
-       rte_service_set_runstate_mapped_check(fdata->evdev_service_id, 0);
-       if (rte_event_dev_start(dev_id) < 0) {
-               printf("Error starting eventdev\n");
-               return -1;
-       }
-
-       return dev_id;
+       RTE_SET_USED(nb_ethdev);
+       RTE_SET_USED(eventdev_id);
+       set_worker_generic_setup_data(&fdata->cap, 1);
 }
 
 static void
@@ -869,17 +453,21 @@ main(int argc, char **argv)
        if (ndevs > 1)
                fprintf(stderr, "Warning: More than one eventdev, using idx 0");
 
+
+       do_capability_setup(num_ports, 0);
+       fdata->cap.check_opt();
+
        worker_data = rte_calloc(0, cdata.num_workers,
                        sizeof(worker_data[0]), 0);
        if (worker_data == NULL)
                rte_panic("rte_calloc failed\n");
 
-       int dev_id = setup_eventdev(&cons_data, worker_data);
+       int dev_id = fdata->cap.evdev_setup(&cons_data, worker_data);
        if (dev_id < 0)
                rte_exit(EXIT_FAILURE, "Error setting up eventdev\n");
 
        init_ports(num_ports);
-       init_rx_adapter(num_ports);
+       fdata->cap.adptr_setup(num_ports);
 
        int worker_idx = 0;
        RTE_LCORE_FOREACH_SLAVE(lcore_id) {
@@ -912,8 +500,8 @@ main(int argc, char **argv)
                                __func__, lcore_id,
                                worker_data[worker_idx].port_id);
 
-               err = rte_eal_remote_launch(worker, &worker_data[worker_idx],
-                                           lcore_id);
+               err = rte_eal_remote_launch(fdata->cap.worker,
+                               &worker_data[worker_idx], lcore_id);
                if (err) {
                        rte_panic("Failed to launch worker on core %d\n",
                                        lcore_id);
@@ -926,7 +514,7 @@ main(int argc, char **argv)
        lcore_id = rte_lcore_id();
 
        if (core_in_use(lcore_id))
-               worker(&worker_data[worker_idx++]);
+               fdata->cap.worker(&worker_data[worker_idx++]);
 
        rte_eal_mp_wait_lcore();
 
index 00721ea..379ba9d 100644 (file)
@@ -84,7 +84,60 @@ struct config_data {
        uint8_t rx_adapter_id;
 };
 
+struct port_link {
+       uint8_t queue_id;
+       uint8_t priority;
+};
+
 struct cons_data cons_data;
 
 struct fastpath_data *fdata;
 struct config_data cdata;
+
+static __rte_always_inline void
+work(struct rte_mbuf *m)
+{
+       struct ether_hdr *eth;
+       struct ether_addr addr;
+
+       /* change mac addresses on packet (to use mbuf data) */
+       /*
+        * FIXME Swap mac address properly and also handle the
+        * case for both odd and even number of stages that the
+        * addresses end up the same at the end of the pipeline
+        */
+       eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
+       ether_addr_copy(&eth->d_addr, &addr);
+       ether_addr_copy(&addr, &eth->d_addr);
+
+       /* do a number of cycles of work per packet */
+       volatile uint64_t start_tsc = rte_rdtsc();
+       while (rte_rdtsc() < start_tsc + cdata.worker_cycles)
+               rte_pause();
+}
+
+static __rte_always_inline void
+schedule_devices(unsigned int lcore_id)
+{
+       if (fdata->rx_core[lcore_id]) {
+               rte_service_run_iter_on_app_lcore(fdata->rxadptr_service_id,
+                               !fdata->rx_single);
+       }
+
+       if (fdata->sched_core[lcore_id]) {
+               rte_service_run_iter_on_app_lcore(fdata->evdev_service_id,
+                               !fdata->sched_single);
+               if (cdata.dump_dev_signal) {
+                       rte_event_dev_dump(0, stdout);
+                       cdata.dump_dev_signal = 0;
+               }
+       }
+
+       if (fdata->tx_core[lcore_id] && (fdata->tx_single ||
+                        rte_atomic32_cmpset(&(fdata->tx_lock), 0, 1))) {
+               fdata->cap.consumer();
+               rte_atomic32_clear((rte_atomic32_t *)&(fdata->tx_lock));
+       }
+}
+
+void set_worker_generic_setup_data(struct setup_data *caps, bool burst);
diff --git a/examples/eventdev_pipeline_sw_pmd/pipeline_worker_generic.c b/examples/eventdev_pipeline_sw_pmd/pipeline_worker_generic.c
new file mode 100644 (file)
index 0000000..d2bc6d3
--- /dev/null
@@ -0,0 +1,398 @@
+/*
+ * SPDX-License-Identifier: BSD-3-Clause
+ * Copyright 2016 Intel Corporation.
+ * Copyright 2017 Cavium, Inc.
+ */
+
+#include "pipeline_common.h"
+
+static int
+worker_generic_burst(void *arg)
+{
+       struct rte_event events[BATCH_SIZE];
+
+       struct worker_data *data = (struct worker_data *)arg;
+       uint8_t dev_id = data->dev_id;
+       uint8_t port_id = data->port_id;
+       size_t sent = 0, received = 0;
+       unsigned int lcore_id = rte_lcore_id();
+
+       while (!fdata->done) {
+               uint16_t i;
+
+               if (fdata->cap.scheduler)
+                       fdata->cap.scheduler(lcore_id);
+
+               if (!fdata->worker_core[lcore_id]) {
+                       rte_pause();
+                       continue;
+               }
+
+               const uint16_t nb_rx = rte_event_dequeue_burst(dev_id, port_id,
+                               events, RTE_DIM(events), 0);
+
+               if (nb_rx == 0) {
+                       rte_pause();
+                       continue;
+               }
+               received += nb_rx;
+
+               for (i = 0; i < nb_rx; i++) {
+
+                       /* The first worker stage does classification */
+                       if (events[i].queue_id == cdata.qid[0])
+                               events[i].flow_id = events[i].mbuf->hash.rss
+                                                       % cdata.num_fids;
+
+                       events[i].queue_id = cdata.next_qid[events[i].queue_id];
+                       events[i].op = RTE_EVENT_OP_FORWARD;
+                       events[i].sched_type = cdata.queue_type;
+
+                       work(events[i].mbuf);
+               }
+               uint16_t nb_tx = rte_event_enqueue_burst(dev_id, port_id,
+                               events, nb_rx);
+               while (nb_tx < nb_rx && !fdata->done)
+                       nb_tx += rte_event_enqueue_burst(dev_id, port_id,
+                                                       events + nb_tx,
+                                                       nb_rx - nb_tx);
+               sent += nb_tx;
+       }
+
+       if (!cdata.quiet)
+               printf("  worker %u thread done. RX=%zu TX=%zu\n",
+                               rte_lcore_id(), received, sent);
+
+       return 0;
+}
+
+static __rte_always_inline int
+consumer_burst(void)
+{
+       const uint64_t freq_khz = rte_get_timer_hz() / 1000;
+       struct rte_event packets[BATCH_SIZE];
+
+       static uint64_t received;
+       static uint64_t last_pkts;
+       static uint64_t last_time;
+       static uint64_t start_time;
+       unsigned int i, j;
+       uint8_t dev_id = cons_data.dev_id;
+       uint8_t port_id = cons_data.port_id;
+       uint16_t nb_ports = rte_eth_dev_count();
+
+       do {
+               uint16_t n = rte_event_dequeue_burst(dev_id, port_id,
+                               packets, RTE_DIM(packets), 0);
+
+               if (n == 0) {
+                       for (j = 0; j < nb_ports; j++)
+                               rte_eth_tx_buffer_flush(j, 0, fdata->tx_buf[j]);
+                       return 0;
+               }
+               if (start_time == 0)
+                       last_time = start_time = rte_get_timer_cycles();
+
+               received += n;
+               for (i = 0; i < n; i++) {
+                       uint8_t outport = packets[i].mbuf->port;
+                       rte_eth_tx_buffer(outport, 0, fdata->tx_buf[outport],
+                                       packets[i].mbuf);
+
+                       packets[i].op = RTE_EVENT_OP_RELEASE;
+               }
+
+               if (cons_data.release) {
+                       uint16_t nb_tx;
+
+                       nb_tx = rte_event_enqueue_burst(dev_id, port_id,
+                                                               packets, n);
+                       while (nb_tx < n)
+                               nb_tx += rte_event_enqueue_burst(dev_id,
+                                               port_id, packets + nb_tx,
+                                               n - nb_tx);
+               }
+
+               /* Print out mpps every 1<22 packets */
+               if (!cdata.quiet && received >= last_pkts + (1<<22)) {
+                       const uint64_t now = rte_get_timer_cycles();
+                       const uint64_t total_ms = (now - start_time) / freq_khz;
+                       const uint64_t delta_ms = (now - last_time) / freq_khz;
+                       uint64_t delta_pkts = received - last_pkts;
+
+                       printf("# consumer RX=%"PRIu64", time %"PRIu64 "ms, "
+                                       "avg %.3f mpps [current %.3f mpps]\n",
+                                       received,
+                                       total_ms,
+                                       received / (total_ms * 1000.0),
+                                       delta_pkts / (delta_ms * 1000.0));
+                       last_pkts = received;
+                       last_time = now;
+               }
+
+               cdata.num_packets -= n;
+               if (cdata.num_packets <= 0)
+                       fdata->done = 1;
+       /* Be stuck in this loop if single. */
+       } while (!fdata->done && fdata->tx_single);
+
+       return 0;
+}
+
+static int
+setup_eventdev_generic(struct cons_data *cons_data,
+               struct worker_data *worker_data)
+{
+       const uint8_t dev_id = 0;
+       /* +1 stages is for a SINGLE_LINK TX stage */
+       const uint8_t nb_queues = cdata.num_stages + 1;
+       /* + 1 is one port for consumer */
+       const uint8_t nb_ports = cdata.num_workers + 1;
+       struct rte_event_dev_config config = {
+                       .nb_event_queues = nb_queues,
+                       .nb_event_ports = nb_ports,
+                       .nb_events_limit  = 4096,
+                       .nb_event_queue_flows = 1024,
+                       .nb_event_port_dequeue_depth = 128,
+                       .nb_event_port_enqueue_depth = 128,
+       };
+       struct rte_event_port_conf wkr_p_conf = {
+                       .dequeue_depth = cdata.worker_cq_depth,
+                       .enqueue_depth = 64,
+                       .new_event_threshold = 4096,
+       };
+       struct rte_event_queue_conf wkr_q_conf = {
+                       .schedule_type = cdata.queue_type,
+                       .priority = RTE_EVENT_DEV_PRIORITY_NORMAL,
+                       .nb_atomic_flows = 1024,
+               .nb_atomic_order_sequences = 1024,
+       };
+       struct rte_event_port_conf tx_p_conf = {
+                       .dequeue_depth = 128,
+                       .enqueue_depth = 128,
+                       .new_event_threshold = 4096,
+       };
+       struct rte_event_queue_conf tx_q_conf = {
+                       .priority = RTE_EVENT_DEV_PRIORITY_HIGHEST,
+                       .event_queue_cfg = RTE_EVENT_QUEUE_CFG_SINGLE_LINK,
+       };
+
+       struct port_link worker_queues[MAX_NUM_STAGES];
+       uint8_t disable_implicit_release;
+       struct port_link tx_queue;
+       unsigned int i;
+
+       int ret, ndev = rte_event_dev_count();
+       if (ndev < 1) {
+               printf("%d: No Eventdev Devices Found\n", __LINE__);
+               return -1;
+       }
+
+       struct rte_event_dev_info dev_info;
+       ret = rte_event_dev_info_get(dev_id, &dev_info);
+       printf("\tEventdev %d: %s\n", dev_id, dev_info.driver_name);
+
+       disable_implicit_release = (dev_info.event_dev_cap &
+                       RTE_EVENT_DEV_CAP_IMPLICIT_RELEASE_DISABLE);
+
+       wkr_p_conf.disable_implicit_release = disable_implicit_release;
+       tx_p_conf.disable_implicit_release = disable_implicit_release;
+
+       if (dev_info.max_event_port_dequeue_depth <
+                       config.nb_event_port_dequeue_depth)
+               config.nb_event_port_dequeue_depth =
+                               dev_info.max_event_port_dequeue_depth;
+       if (dev_info.max_event_port_enqueue_depth <
+                       config.nb_event_port_enqueue_depth)
+               config.nb_event_port_enqueue_depth =
+                               dev_info.max_event_port_enqueue_depth;
+
+       ret = rte_event_dev_configure(dev_id, &config);
+       if (ret < 0) {
+               printf("%d: Error configuring device\n", __LINE__);
+               return -1;
+       }
+
+       /* Q creation - one load balanced per pipeline stage*/
+       printf("  Stages:\n");
+       for (i = 0; i < cdata.num_stages; i++) {
+               if (rte_event_queue_setup(dev_id, i, &wkr_q_conf) < 0) {
+                       printf("%d: error creating qid %d\n", __LINE__, i);
+                       return -1;
+               }
+               cdata.qid[i] = i;
+               cdata.next_qid[i] = i+1;
+               worker_queues[i].queue_id = i;
+               if (cdata.enable_queue_priorities) {
+                       /* calculate priority stepping for each stage, leaving
+                        * headroom of 1 for the SINGLE_LINK TX below
+                        */
+                       const uint32_t prio_delta =
+                               (RTE_EVENT_DEV_PRIORITY_LOWEST-1) /  nb_queues;
+
+                       /* higher priority for queues closer to tx */
+                       wkr_q_conf.priority =
+                               RTE_EVENT_DEV_PRIORITY_LOWEST - prio_delta * i;
+               }
+
+               const char *type_str = "Atomic";
+               switch (wkr_q_conf.schedule_type) {
+               case RTE_SCHED_TYPE_ORDERED:
+                       type_str = "Ordered";
+                       break;
+               case RTE_SCHED_TYPE_PARALLEL:
+                       type_str = "Parallel";
+                       break;
+               }
+               printf("\tStage %d, Type %s\tPriority = %d\n", i, type_str,
+                               wkr_q_conf.priority);
+       }
+       printf("\n");
+
+       /* final queue for sending to TX core */
+       if (rte_event_queue_setup(dev_id, i, &tx_q_conf) < 0) {
+               printf("%d: error creating qid %d\n", __LINE__, i);
+               return -1;
+       }
+       tx_queue.queue_id = i;
+       tx_queue.priority = RTE_EVENT_DEV_PRIORITY_HIGHEST;
+
+       if (wkr_p_conf.dequeue_depth > config.nb_event_port_dequeue_depth)
+               wkr_p_conf.dequeue_depth = config.nb_event_port_dequeue_depth;
+       if (wkr_p_conf.enqueue_depth > config.nb_event_port_enqueue_depth)
+               wkr_p_conf.enqueue_depth = config.nb_event_port_enqueue_depth;
+
+       /* set up one port per worker, linking to all stage queues */
+       for (i = 0; i < cdata.num_workers; i++) {
+               struct worker_data *w = &worker_data[i];
+               w->dev_id = dev_id;
+               if (rte_event_port_setup(dev_id, i, &wkr_p_conf) < 0) {
+                       printf("Error setting up port %d\n", i);
+                       return -1;
+               }
+
+               uint32_t s;
+               for (s = 0; s < cdata.num_stages; s++) {
+                       if (rte_event_port_link(dev_id, i,
+                                               &worker_queues[s].queue_id,
+                                               &worker_queues[s].priority,
+                                               1) != 1) {
+                               printf("%d: error creating link for port %d\n",
+                                               __LINE__, i);
+                               return -1;
+                       }
+               }
+               w->port_id = i;
+       }
+
+       if (tx_p_conf.dequeue_depth > config.nb_event_port_dequeue_depth)
+               tx_p_conf.dequeue_depth = config.nb_event_port_dequeue_depth;
+       if (tx_p_conf.enqueue_depth > config.nb_event_port_enqueue_depth)
+               tx_p_conf.enqueue_depth = config.nb_event_port_enqueue_depth;
+
+       /* port for consumer, linked to TX queue */
+       if (rte_event_port_setup(dev_id, i, &tx_p_conf) < 0) {
+               printf("Error setting up port %d\n", i);
+               return -1;
+       }
+       if (rte_event_port_link(dev_id, i, &tx_queue.queue_id,
+                               &tx_queue.priority, 1) != 1) {
+               printf("%d: error creating link for port %d\n",
+                               __LINE__, i);
+               return -1;
+       }
+       *cons_data = (struct cons_data){.dev_id = dev_id,
+                                       .port_id = i,
+                                       .release = disable_implicit_release };
+
+       ret = rte_event_dev_service_id_get(dev_id,
+                               &fdata->evdev_service_id);
+       if (ret != -ESRCH && ret != 0) {
+               printf("Error getting the service ID for sw eventdev\n");
+               return -1;
+       }
+       rte_service_runstate_set(fdata->evdev_service_id, 1);
+       rte_service_set_runstate_mapped_check(fdata->evdev_service_id, 0);
+       if (rte_event_dev_start(dev_id) < 0) {
+               printf("Error starting eventdev\n");
+               return -1;
+       }
+
+       return dev_id;
+}
+
+static void
+init_rx_adapter(uint16_t nb_ports)
+{
+       int i;
+       int ret;
+       uint8_t evdev_id = 0;
+       struct rte_event_dev_info dev_info;
+
+       ret = rte_event_dev_info_get(evdev_id, &dev_info);
+
+       struct rte_event_port_conf rx_p_conf = {
+               .dequeue_depth = 8,
+               .enqueue_depth = 8,
+               .new_event_threshold = 1200,
+       };
+
+       if (rx_p_conf.dequeue_depth > dev_info.max_event_port_dequeue_depth)
+               rx_p_conf.dequeue_depth = dev_info.max_event_port_dequeue_depth;
+       if (rx_p_conf.enqueue_depth > dev_info.max_event_port_enqueue_depth)
+               rx_p_conf.enqueue_depth = dev_info.max_event_port_enqueue_depth;
+
+       /* Create one adapter for all the ethernet ports. */
+       ret = rte_event_eth_rx_adapter_create(cdata.rx_adapter_id, evdev_id,
+                       &rx_p_conf);
+       if (ret)
+               rte_exit(EXIT_FAILURE, "failed to create rx adapter[%d]",
+                               cdata.rx_adapter_id);
+
+       struct rte_event_eth_rx_adapter_queue_conf queue_conf = {
+               .ev.sched_type = cdata.queue_type,
+               .ev.queue_id = cdata.qid[0],
+       };
+
+       for (i = 0; i < nb_ports; i++) {
+               uint32_t cap;
+
+               ret = rte_event_eth_rx_adapter_caps_get(evdev_id, i, &cap);
+               if (ret)
+                       rte_exit(EXIT_FAILURE,
+                                       "failed to get event rx adapter "
+                                       "capabilities");
+
+               ret = rte_event_eth_rx_adapter_queue_add(cdata.rx_adapter_id, i,
+                               -1, &queue_conf);
+               if (ret)
+                       rte_exit(EXIT_FAILURE,
+                                       "Failed to add queues to Rx adapter");
+       }
+
+       ret = rte_event_eth_rx_adapter_service_id_get(cdata.rx_adapter_id,
+                               &fdata->rxadptr_service_id);
+       if (ret != -ESRCH && ret != 0) {
+               rte_exit(EXIT_FAILURE,
+                       "Error getting the service ID for sw eventdev\n");
+       }
+       rte_service_runstate_set(fdata->rxadptr_service_id, 1);
+       rte_service_set_runstate_mapped_check(fdata->rxadptr_service_id, 0);
+
+       ret = rte_event_eth_rx_adapter_start(cdata.rx_adapter_id);
+       if (ret)
+               rte_exit(EXIT_FAILURE, "Rx adapter[%d] start failed",
+                               cdata.rx_adapter_id);
+}
+
+void
+set_worker_generic_setup_data(struct setup_data *caps, bool burst)
+{
+       RTE_SET_USED(burst);
+       caps->consumer = consumer_burst;
+       caps->worker = worker_generic_burst;
+
+       caps->adptr_setup = init_rx_adapter;
+       caps->scheduler = schedule_devices;
+       caps->evdev_setup = setup_eventdev_generic;
+}