1 /* 2 * BSD LICENSE 3 * 4 * Copyright (C) Cavium, Inc 2017. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 10 * * Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * * Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in 14 * the documentation and/or other materials provided with the 15 * distribution. 16 * * Neither the name of Cavium, Inc nor the names of its 17 * contributors may be used to endorse or promote products derived 18 * from this software without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 23 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 24 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 25 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 26 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 27 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 28 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 29 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 30 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 #ifndef _TEST_PERF_COMMON_ 34 #define _TEST_PERF_COMMON_ 35 36 #include <stdio.h> 37 #include <stdbool.h> 38 #include <unistd.h> 39 40 #include <rte_cycles.h> 41 #include <rte_eventdev.h> 42 #include <rte_lcore.h> 43 #include <rte_malloc.h> 44 #include <rte_mempool.h> 45 #include <rte_prefetch.h> 46 47 #include "evt_common.h" 48 #include "evt_options.h" 49 #include "evt_test.h" 50 51 struct test_perf; 52 53 struct worker_data { 54 uint64_t processed_pkts; 55 uint64_t latency; 56 uint8_t dev_id; 57 uint8_t port_id; 58 struct test_perf *t; 59 } __rte_cache_aligned; 60 61 struct prod_data { 62 uint8_t dev_id; 63 uint8_t port_id; 64 uint8_t queue_id; 65 struct test_perf *t; 66 } __rte_cache_aligned; 67 68 struct test_perf { 69 /* Don't change the offset of "done". Signal handler use this memory 70 * to terminate all lcores work. 71 */ 72 int done; 73 uint64_t outstand_pkts; 74 uint8_t nb_workers; 75 enum evt_test_result result; 76 uint32_t nb_flows; 77 uint64_t nb_pkts; 78 struct rte_mempool *pool; 79 struct prod_data prod[EVT_MAX_PORTS]; 80 struct worker_data worker[EVT_MAX_PORTS]; 81 struct evt_options *opt; 82 uint8_t sched_type_list[EVT_MAX_STAGES] __rte_cache_aligned; 83 } __rte_cache_aligned; 84 85 struct perf_elt { 86 uint64_t timestamp; 87 } __rte_cache_aligned; 88 89 #define BURST_SIZE 16 90 91 #define PERF_WORKER_INIT\ 92 struct worker_data *w = arg;\ 93 struct test_perf *t = w->t;\ 94 struct evt_options *opt = t->opt;\ 95 const uint8_t dev = w->dev_id;\ 96 const uint8_t port = w->port_id;\ 97 uint8_t *const sched_type_list = &t->sched_type_list[0];\ 98 struct rte_mempool *const pool = t->pool;\ 99 const uint8_t nb_stages = t->opt->nb_stages;\ 100 const uint8_t laststage = nb_stages - 1;\ 101 uint8_t cnt = 0;\ 102 void *bufs[16] __rte_cache_aligned;\ 103 int const sz = RTE_DIM(bufs);\ 104 if (opt->verbose_level > 1)\ 105 printf("%s(): lcore %d dev_id %d port=%d\n", __func__,\ 106 rte_lcore_id(), dev, port) 107 108 static inline __attribute__((always_inline)) int 109 perf_process_last_stage(struct rte_mempool *const pool, 110 struct rte_event *const ev, struct worker_data *const w, 111 void *bufs[], int const buf_sz, uint8_t count) 112 { 113 bufs[count++] = ev->event_ptr; 114 w->processed_pkts++; 115 rte_smp_wmb(); 116 117 if (unlikely(count == buf_sz)) { 118 count = 0; 119 rte_mempool_put_bulk(pool, bufs, buf_sz); 120 } 121 return count; 122 } 123 124 static inline __attribute__((always_inline)) uint8_t 125 perf_process_last_stage_latency(struct rte_mempool *const pool, 126 struct rte_event *const ev, struct worker_data *const w, 127 void *bufs[], int const buf_sz, uint8_t count) 128 { 129 uint64_t latency; 130 struct perf_elt *const m = ev->event_ptr; 131 132 bufs[count++] = ev->event_ptr; 133 w->processed_pkts++; 134 135 if (unlikely(count == buf_sz)) { 136 count = 0; 137 latency = rte_get_timer_cycles() - m->timestamp; 138 rte_mempool_put_bulk(pool, bufs, buf_sz); 139 } else { 140 latency = rte_get_timer_cycles() - m->timestamp; 141 } 142 143 w->latency += latency; 144 rte_smp_wmb(); 145 return count; 146 } 147 148 149 static inline int 150 perf_nb_event_ports(struct evt_options *opt) 151 { 152 return evt_nr_active_lcores(opt->wlcores) + 153 evt_nr_active_lcores(opt->plcores); 154 } 155 156 int perf_test_result(struct evt_test *test, struct evt_options *opt); 157 int perf_opt_check(struct evt_options *opt, uint64_t nb_queues); 158 int perf_test_setup(struct evt_test *test, struct evt_options *opt); 159 int perf_mempool_setup(struct evt_test *test, struct evt_options *opt); 160 int perf_event_dev_port_setup(struct evt_test *test, struct evt_options *opt, 161 uint8_t stride, uint8_t nb_queues); 162 int perf_event_dev_service_setup(uint8_t dev_id); 163 int perf_launch_lcores(struct evt_test *test, struct evt_options *opt, 164 int (*worker)(void *)); 165 void perf_opt_dump(struct evt_options *opt, uint8_t nb_queues); 166 void perf_test_destroy(struct evt_test *test, struct evt_options *opt); 167 void perf_eventdev_destroy(struct evt_test *test, struct evt_options *opt); 168 void perf_mempool_destroy(struct evt_test *test, struct evt_options *opt); 169 170 #endif /* _TEST_PERF_COMMON_ */ 171