From 1abec5a096ef4d74bea3782f7844e1f5cf0e90d6 Mon Sep 17 00:00:00 2001 From: Mathieu Desnoyers Date: Sun, 19 Oct 2014 17:31:58 +0200 Subject: [PATCH] workqueue: implement benchmark Signed-off-by: Mathieu Desnoyers --- tests/benchmark/Makefile.am | 6 +- tests/benchmark/test_urcu_workqueue.c | 404 ++++++++++++++++++++++++++ 2 files changed, 409 insertions(+), 1 deletion(-) create mode 100644 tests/benchmark/test_urcu_workqueue.c diff --git a/tests/benchmark/Makefile.am b/tests/benchmark/Makefile.am index 85d454d..1660a29 100644 --- a/tests/benchmark/Makefile.am +++ b/tests/benchmark/Makefile.am @@ -20,7 +20,8 @@ noinst_PROGRAMS = test_urcu test_urcu_dynamic_link test_urcu_timing \ test_urcu_wfq_dynlink test_urcu_wfs_dynlink \ test_urcu_wfcq_dynlink \ test_urcu_lfq_dynlink test_urcu_lfs_dynlink test_urcu_hash \ - test_urcu_lfs_rcu_dynlink + test_urcu_lfs_rcu_dynlink \ + test_urcu_workqueue URCU_COMMON_LIB=$(top_builddir)/liburcu-common.la URCU_LIB=$(top_builddir)/liburcu.la @@ -189,6 +190,9 @@ test_urcu_wfs_dynlink_SOURCES = test_urcu_wfs.c test_urcu_wfs_dynlink_CFLAGS = -DDYNAMIC_LINK_TEST $(AM_CFLAGS) test_urcu_wfs_dynlink_LDADD = $(URCU_COMMON_LIB) +test_urcu_workqueue_SOURCES = test_urcu_workqueue.c +test_urcu_workqueue_LDADD = $(URCU_COMMON_LIB) $(URCU_LIB) + test_urcu_hash_SOURCES = test_urcu_hash.c test_urcu_hash.h \ test_urcu_hash_rw.c test_urcu_hash_unique.c test_urcu_hash_CFLAGS = -DRCU_QSBR $(AM_CFLAGS) diff --git a/tests/benchmark/test_urcu_workqueue.c b/tests/benchmark/test_urcu_workqueue.c new file mode 100644 index 0000000..9ac2350 --- /dev/null +++ b/tests/benchmark/test_urcu_workqueue.c @@ -0,0 +1,404 @@ +/* + * test_urcu_workqueue.c + * + * Userspace RCU library - workqueue test + * + * Copyright February 2010-2014 - Mathieu Desnoyers + * Copyright February 2010 - Paolo Bonzini + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. + */ + +#define _GNU_SOURCE +#include "config.h" +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include "cpuset.h" +#include "thread-id.h" + +/* hardcoded number of CPUs */ +#define NR_CPUS 16384 + +#ifndef DYNAMIC_LINK_TEST +#define _LGPL_SOURCE +#endif +#include +#include +#include + +static volatile int test_go, test_stop_enqueue, test_stop_dequeue; + +static unsigned long work_loops; + +static unsigned long duration; + +static unsigned long dispatch_delay_loops; + +static inline void loop_sleep(unsigned long loops) +{ + while (loops-- != 0) + caa_cpu_relax(); +} + +static int verbose_mode; + +static int test_wait_empty; +static int test_enqueue_stopped; + +#define printf_verbose(fmt, args...) \ + do { \ + if (verbose_mode) \ + fprintf(stderr, fmt, ## args); \ + } while (0) + +static unsigned int cpu_affinities[NR_CPUS]; +static unsigned int next_aff = 0; +static int use_affinity = 0; + +pthread_mutex_t affinity_mutex = PTHREAD_MUTEX_INITIALIZER; + +static void set_affinity(void) +{ +#if HAVE_SCHED_SETAFFINITY + cpu_set_t mask; + int cpu, ret; +#endif /* HAVE_SCHED_SETAFFINITY */ + + if (!use_affinity) + return; + +#if HAVE_SCHED_SETAFFINITY + ret = pthread_mutex_lock(&affinity_mutex); + if (ret) { + perror("Error in pthread mutex lock"); + exit(-1); + } + cpu = cpu_affinities[next_aff++]; + ret = pthread_mutex_unlock(&affinity_mutex); + if (ret) { + perror("Error in pthread mutex unlock"); + exit(-1); + } + + CPU_ZERO(&mask); + CPU_SET(cpu, &mask); +#if SCHED_SETAFFINITY_ARGS == 2 + sched_setaffinity(0, &mask); +#else + sched_setaffinity(0, sizeof(mask), &mask); +#endif +#endif /* HAVE_SCHED_SETAFFINITY */ +} + +/* + * returns 0 if test should end. + */ +static int test_duration_dequeue(void) +{ + return !test_stop_dequeue; +} + +static int test_duration_enqueue(void) +{ + return !test_stop_enqueue; +} + +static DEFINE_URCU_TLS(unsigned long long, nr_dequeues); +static DEFINE_URCU_TLS(unsigned long long, nr_enqueues); + +static unsigned int nr_dispatchers; +static unsigned int nr_workers; + +static struct urcu_workqueue workqueue; + +struct test_work { + struct urcu_work w; +}; + +static void *thr_dispatcher(void *_count) +{ + unsigned long long *count = _count; + bool was_nonempty; + + printf_verbose("thread_begin %s, tid %lu\n", + "dispatcher", urcu_get_thread_id()); + + set_affinity(); + + while (!test_go) + { + } + cmm_smp_mb(); + + for (;;) { + struct test_work *work = malloc(sizeof(*work)); + if (!work) + goto fail; + printf_verbose("queue work %p\n", work); + urcu_queue_work(&workqueue, &work->w); + URCU_TLS(nr_enqueues)++; + + if (caa_unlikely(dispatch_delay_loops)) + loop_sleep(dispatch_delay_loops); +fail: + if (caa_unlikely(!test_duration_enqueue())) + break; + } + + uatomic_inc(&test_enqueue_stopped); + count[0] = URCU_TLS(nr_enqueues); + printf_verbose("dispatcher thread_end, tid %lu, " + "enqueues %llu\n", + urcu_get_thread_id(), + URCU_TLS(nr_enqueues)); + return ((void*)1); +} + +static void *thr_worker(void *_count) +{ + unsigned long long *count = _count; + unsigned int counter = 0; + struct urcu_worker worker; + int blocking = 1; + + printf_verbose("thread_begin %s, tid %lu\n", + "worker", urcu_get_thread_id()); + + set_affinity(); + + rcu_register_thread(); + urcu_worker_init(&worker, URCU_WORKER_STEAL); + //urcu_worker_init(&worker, 0); + urcu_worker_register(&workqueue, &worker); + + while (!test_go) + { + } + cmm_smp_mb(); + + for (;;) { + int batch_work_count = 0; + + urcu_accept_work(&workqueue, &worker, blocking); + for (;;) { + struct urcu_work *work; + struct test_work *t; + + work = urcu_dequeue_work(&worker); + if (!work) + break; + t = caa_container_of(work, struct test_work, w); + printf_verbose("dequeue work %p\n", t); + batch_work_count++; + URCU_TLS(nr_dequeues)++; + if (caa_unlikely(work_loops)) + loop_sleep(work_loops); + free(t); + } + if (!test_duration_dequeue()) + blocking = 0; + if (caa_unlikely(!test_duration_dequeue() + && !batch_work_count)) + break; + } +end: + urcu_worker_unregister(&workqueue, &worker); + rcu_unregister_thread(); + + printf_verbose("worker thread_end, tid %lu, " + "dequeues %llu\n", + urcu_get_thread_id(), + URCU_TLS(nr_dequeues)); + count[0] = URCU_TLS(nr_dequeues); + return ((void*)2); +} + +static void show_usage(int argc, char **argv) +{ + printf("Usage : %s nr_workers nr_dispatchers duration (s) \n", + argv[0]); + printf("OPTIONS:\n"); + printf(" [-d delay] (dispatcher period (in loops))\n"); + printf(" [-c duration] (worker period (in loops))\n"); + printf(" [-v] (verbose output)\n"); + printf(" [-a cpu#] [-a cpu#]... (affinity)\n"); + printf(" [-w] Wait for worker to empty stack\n"); + printf("\n"); +} + +int main(int argc, char **argv) +{ + int err; + pthread_t *tid_dispatcher, *tid_worker; + void *tret; + unsigned long long *count_dispatcher, *count_worker; + unsigned long long tot_enqueues = 0, tot_dequeues = 0; + unsigned long long end_dequeues = 0; + int i, a, retval = 0; + + if (argc < 4) { + show_usage(argc, argv); + return -1; + } + + err = sscanf(argv[1], "%u", &nr_workers); + if (err != 1) { + show_usage(argc, argv); + return -1; + } + + err = sscanf(argv[2], "%u", &nr_dispatchers); + if (err != 1) { + show_usage(argc, argv); + return -1; + } + + err = sscanf(argv[3], "%lu", &duration); + if (err != 1) { + show_usage(argc, argv); + return -1; + } + + for (i = 4; i < argc; i++) { + if (argv[i][0] != '-') + continue; + switch (argv[i][1]) { + case 'a': + if (argc < i + 2) { + show_usage(argc, argv); + return -1; + } + a = atoi(argv[++i]); + cpu_affinities[next_aff++] = a; + use_affinity = 1; + printf_verbose("Adding CPU %d affinity\n", a); + break; + case 'c': + if (argc < i + 2) { + show_usage(argc, argv); + return -1; + } + work_loops = atol(argv[++i]); + break; + case 'd': + if (argc < i + 2) { + show_usage(argc, argv); + return -1; + } + dispatch_delay_loops = atol(argv[++i]); + break; + case 'v': + verbose_mode = 1; + break; + case 'w': + test_wait_empty = 1; + break; + } + } + + printf_verbose("running test for %lu seconds, %u dispatchers, " + "%u workers.\n", + duration, nr_dispatchers, nr_workers); + if (test_wait_empty) + printf_verbose("Wait for workers to empty workqueue.\n"); + printf_verbose("Work duration: %lu loops.\n", work_loops); + printf_verbose("Dispatcher arrival delay: %lu loops.\n", dispatch_delay_loops); + printf_verbose("thread %-6s, tid %lu\n", + "main", urcu_get_thread_id()); + + tid_dispatcher = calloc(nr_dispatchers, sizeof(*tid_dispatcher)); + tid_worker = calloc(nr_workers, sizeof(*tid_worker)); + count_dispatcher = calloc(nr_dispatchers, sizeof(*count_dispatcher)); + count_worker = calloc(nr_workers, sizeof(*count_worker)); + urcu_workqueue_init(&workqueue); + + next_aff = 0; + + for (i = 0; i < nr_dispatchers; i++) { + err = pthread_create(&tid_dispatcher[i], NULL, thr_dispatcher, + &count_dispatcher[i]); + if (err != 0) + exit(1); + } + for (i = 0; i < nr_workers; i++) { + err = pthread_create(&tid_worker[i], NULL, thr_worker, + &count_worker[i]); + if (err != 0) + exit(1); + } + + cmm_smp_mb(); + + test_go = 1; + + for (i = 0; i < duration; i++) { + sleep(1); + if (verbose_mode) + (void) write(1, ".", 1); + } + + test_stop_enqueue = 1; + while (nr_dispatchers != uatomic_read(&test_enqueue_stopped)) { + sleep(1); + } + + if (test_wait_empty) { + while (!cds_wfcq_empty(&workqueue.head, &workqueue.tail)) { + sleep(1); + } + } + test_stop_dequeue = 1; + + /* Send finish to all workers */ + urcu_workqueue_wakeup_all(&workqueue); + + for (i = 0; i < nr_dispatchers; i++) { + err = pthread_join(tid_dispatcher[i], &tret); + if (err != 0) + exit(1); + tot_enqueues += count_dispatcher[i]; + } + for (i = 0; i < nr_workers; i++) { + err = pthread_join(tid_worker[i], &tret); + if (err != 0) + exit(1); + tot_dequeues += count_worker[i]; + } + + printf("SUMMARY %-25s testdur %4lu nr_dispatchers %3u dispatch_delay_loops %6lu " + "work_loops %lu nr_workers %3u " + "nr_enqueues %12llu nr_dequeues %12llu\n", + argv[0], duration, nr_dispatchers, dispatch_delay_loops, work_loops, + nr_workers, tot_enqueues, tot_dequeues); + free(count_dispatcher); + free(count_worker); + free(tid_dispatcher); + free(tid_worker); + return retval; +} -- 2.34.1