futex-wake-parallel.c 8.6 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2 3 4 5 6 7 8 9
/*
 * Copyright (C) 2015 Davidlohr Bueso.
 *
 * Block a bunch of threads and let parallel waker threads wakeup an
 * equal amount of them. The program output reflects the avg latency
 * for each individual thread to service its share of work. Ultimately
 * it can be used to measure futex_wake() changes.
 */
J
James Yang 已提交
10 11 12
#include "bench.h"
#include <linux/compiler.h>
#include "../util/debug.h"
13

J
James Yang 已提交
14 15 16 17 18 19 20
#ifndef HAVE_PTHREAD_BARRIER
int bench_futex_wake_parallel(int argc __maybe_unused, const char **argv __maybe_unused)
{
	pr_err("%s: pthread_barrier_t unavailable, disabling this test...\n", __func__);
	return 0;
}
#else /* HAVE_PTHREAD_BARRIER */
21
/* For the CLR_() macros */
22
#include <string.h>
23 24
#include <pthread.h>

25
#include <signal.h>
26
#include "../util/stat.h"
27
#include <subcmd/parse-options.h>
28
#include <linux/kernel.h>
29
#include <linux/time64.h>
30
#include <errno.h>
31
#include "futex.h"
32
#include <perf/cpumap.h>
33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49

#include <err.h>
#include <stdlib.h>
#include <sys/time.h>

struct thread_data {
	pthread_t worker;
	unsigned int nwoken;
	struct timeval runtime;
};

static unsigned int nwakes = 1;

/* all threads will block on the same futex -- hash bucket chaos ;) */
static u_int32_t futex = 0;

static pthread_t *blocked_worker;
50
static bool done = false;
51 52
static pthread_mutex_t thread_lock;
static pthread_cond_t thread_parent, thread_worker;
J
James Yang 已提交
53
static pthread_barrier_t barrier;
54
static struct stats waketime_stats, wakeup_stats;
D
Davidlohr Bueso 已提交
55
static unsigned int threads_starting;
56 57
static int futex_flag = 0;

58 59
static struct bench_futex_parameters params;

60
static const struct option options[] = {
61 62 63 64
	OPT_UINTEGER('t', "threads", &params.nthreads, "Specify amount of threads"),
	OPT_UINTEGER('w', "nwakers", &params.nwakes, "Specify amount of waking threads"),
	OPT_BOOLEAN( 's', "silent",  &params.silent, "Silent mode: do not display data/details"),
	OPT_BOOLEAN( 'S', "shared",  &params.fshared, "Use shared futexes instead of private ones"),
65 66 67 68 69 70 71 72 73 74 75 76 77
	OPT_END()
};

static const char * const bench_futex_wake_parallel_usage[] = {
	"perf bench futex wake-parallel <options>",
	NULL
};

static void *waking_workerfn(void *arg)
{
	struct thread_data *waker = (struct thread_data *) arg;
	struct timeval start, end;

J
James Yang 已提交
78 79
	pthread_barrier_wait(&barrier);

80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
	gettimeofday(&start, NULL);

	waker->nwoken = futex_wake(&futex, nwakes, futex_flag);
	if (waker->nwoken != nwakes)
		warnx("couldn't wakeup all tasks (%d/%d)",
		      waker->nwoken, nwakes);

	gettimeofday(&end, NULL);
	timersub(&end, &start, &waker->runtime);

	pthread_exit(NULL);
	return NULL;
}

static void wakeup_threads(struct thread_data *td, pthread_attr_t thread_attr)
{
	unsigned int i;

	pthread_attr_setdetachstate(&thread_attr, PTHREAD_CREATE_JOINABLE);

100
	pthread_barrier_init(&barrier, NULL, params.nwakes + 1);
J
James Yang 已提交
101

102
	/* create and block all threads */
103
	for (i = 0; i < params.nwakes; i++) {
104 105 106 107 108 109 110 111 112 113
		/*
		 * Thread creation order will impact per-thread latency
		 * as it will affect the order to acquire the hb spinlock.
		 * For now let the scheduler decide.
		 */
		if (pthread_create(&td[i].worker, &thread_attr,
				   waking_workerfn, (void *)&td[i]))
			err(EXIT_FAILURE, "pthread_create");
	}

J
James Yang 已提交
114 115
	pthread_barrier_wait(&barrier);

116
	for (i = 0; i < params.nwakes; i++)
117 118
		if (pthread_join(td[i].worker, NULL))
			err(EXIT_FAILURE, "pthread_join");
J
James Yang 已提交
119 120

	pthread_barrier_destroy(&barrier);
121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
}

static void *blocked_workerfn(void *arg __maybe_unused)
{
	pthread_mutex_lock(&thread_lock);
	threads_starting--;
	if (!threads_starting)
		pthread_cond_signal(&thread_parent);
	pthread_cond_wait(&thread_worker, &thread_lock);
	pthread_mutex_unlock(&thread_lock);

	while (1) { /* handle spurious wakeups */
		if (futex_wait(&futex, 0, NULL, futex_flag) != EINTR)
			break;
	}

	pthread_exit(NULL);
	return NULL;
}

D
Davidlohr Bueso 已提交
141
static void block_threads(pthread_t *w, pthread_attr_t thread_attr,
142
			  struct perf_cpu_map *cpu)
143
{
D
Davidlohr Bueso 已提交
144
	cpu_set_t cpuset;
145 146
	unsigned int i;

147
	threads_starting = params.nthreads;
148 149

	/* create and block all threads */
150
	for (i = 0; i < params.nthreads; i++) {
D
Davidlohr Bueso 已提交
151 152
		CPU_ZERO(&cpuset);
		CPU_SET(cpu->map[i % cpu->nr], &cpuset);
153

D
Davidlohr Bueso 已提交
154
		if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpuset))
155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170
			err(EXIT_FAILURE, "pthread_attr_setaffinity_np");

		if (pthread_create(&w[i], &thread_attr, blocked_workerfn, NULL))
			err(EXIT_FAILURE, "pthread_create");
	}
}

static void print_run(struct thread_data *waking_worker, unsigned int run_num)
{
	unsigned int i, wakeup_avg;
	double waketime_avg, waketime_stddev;
	struct stats __waketime_stats, __wakeup_stats;

	init_stats(&__wakeup_stats);
	init_stats(&__waketime_stats);

171
	for (i = 0; i < params.nwakes; i++) {
172 173 174 175 176 177 178 179 180 181
		update_stats(&__waketime_stats, waking_worker[i].runtime.tv_usec);
		update_stats(&__wakeup_stats, waking_worker[i].nwoken);
	}

	waketime_avg = avg_stats(&__waketime_stats);
	waketime_stddev = stddev_stats(&__waketime_stats);
	wakeup_avg = avg_stats(&__wakeup_stats);

	printf("[Run %d]: Avg per-thread latency (waking %d/%d threads) "
	       "in %.4f ms (+-%.2f%%)\n", run_num + 1, wakeup_avg,
182
	       params.nthreads, waketime_avg / USEC_PER_MSEC,
183 184 185 186 187 188 189 190 191 192 193 194 195 196
	       rel_stddev_stats(waketime_stddev, waketime_avg));
}

static void print_summary(void)
{
	unsigned int wakeup_avg;
	double waketime_avg, waketime_stddev;

	waketime_avg = avg_stats(&waketime_stats);
	waketime_stddev = stddev_stats(&waketime_stats);
	wakeup_avg = avg_stats(&wakeup_stats);

	printf("Avg per-thread latency (waking %d/%d threads) in %.4f ms (+-%.2f%%)\n",
	       wakeup_avg,
197
	       params.nthreads,
198
	       waketime_avg / USEC_PER_MSEC,
199 200 201 202 203 204 205 206
	       rel_stddev_stats(waketime_stddev, waketime_avg));
}


static void do_run_stats(struct thread_data *waking_worker)
{
	unsigned int i;

207
	for (i = 0; i < params.nwakes; i++) {
208 209 210 211 212 213 214 215 216 217 218 219 220
		update_stats(&waketime_stats, waking_worker[i].runtime.tv_usec);
		update_stats(&wakeup_stats, waking_worker[i].nwoken);
	}

}

static void toggle_done(int sig __maybe_unused,
			siginfo_t *info __maybe_unused,
			void *uc __maybe_unused)
{
	done = true;
}

221
int bench_futex_wake_parallel(int argc, const char **argv)
222 223 224 225 226 227
{
	int ret = 0;
	unsigned int i, j;
	struct sigaction act;
	pthread_attr_t thread_attr;
	struct thread_data *waking_worker;
228
	struct perf_cpu_map *cpu;
229 230 231 232 233 234 235 236

	argc = parse_options(argc, argv, options,
			     bench_futex_wake_parallel_usage, 0);
	if (argc) {
		usage_with_options(bench_futex_wake_parallel_usage, options);
		exit(EXIT_FAILURE);
	}

237
	memset(&act, 0, sizeof(act));
238 239 240 241
	sigfillset(&act.sa_mask);
	act.sa_sigaction = toggle_done;
	sigaction(SIGINT, &act, NULL);

242
	cpu = perf_cpu_map__new(NULL);
D
Davidlohr Bueso 已提交
243 244 245
	if (!cpu)
		err(EXIT_FAILURE, "calloc");

246 247
	if (!params.nthreads)
		params.nthreads = cpu->nr;
248 249

	/* some sanity checks */
250 251 252
	if (params.nwakes > params.nthreads ||
	    !params.nwakes)
		params.nwakes = params.nthreads;
253

254
	if (params.nthreads % params.nwakes)
255 256 257 258 259
		errx(EXIT_FAILURE, "Must be perfectly divisible");
	/*
	 * Each thread will wakeup nwakes tasks in
	 * a single futex_wait call.
	 */
260
	nwakes = params.nthreads/params.nwakes;
261

262
	blocked_worker = calloc(params.nthreads, sizeof(*blocked_worker));
263 264 265
	if (!blocked_worker)
		err(EXIT_FAILURE, "calloc");

266
	if (!params.fshared)
267 268 269 270
		futex_flag = FUTEX_PRIVATE_FLAG;

	printf("Run summary [PID %d]: blocking on %d threads (at [%s] "
	       "futex %p), %d threads waking up %d at a time.\n\n",
271 272
	       getpid(), params.nthreads, params.fshared ? "shared":"private",
	       &futex, params.nwakes, nwakes);
273 274 275 276 277 278 279 280 281 282

	init_stats(&wakeup_stats);
	init_stats(&waketime_stats);

	pthread_attr_init(&thread_attr);
	pthread_mutex_init(&thread_lock, NULL);
	pthread_cond_init(&thread_parent, NULL);
	pthread_cond_init(&thread_worker, NULL);

	for (j = 0; j < bench_repeat && !done; j++) {
283
		waking_worker = calloc(params.nwakes, sizeof(*waking_worker));
284 285 286 287
		if (!waking_worker)
			err(EXIT_FAILURE, "calloc");

		/* create, launch & block all threads */
D
Davidlohr Bueso 已提交
288
		block_threads(blocked_worker, thread_attr, cpu);
289 290 291 292 293 294 295 296 297 298 299 300 301

		/* make sure all threads are already blocked */
		pthread_mutex_lock(&thread_lock);
		while (threads_starting)
			pthread_cond_wait(&thread_parent, &thread_lock);
		pthread_cond_broadcast(&thread_worker);
		pthread_mutex_unlock(&thread_lock);

		usleep(100000);

		/* Ok, all threads are patiently blocked, start waking folks up */
		wakeup_threads(waking_worker, thread_attr);

302
		for (i = 0; i < params.nthreads; i++) {
303 304 305 306 307 308
			ret = pthread_join(blocked_worker[i], NULL);
			if (ret)
				err(EXIT_FAILURE, "pthread_join");
		}

		do_run_stats(waking_worker);
309
		if (!params.silent)
310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
			print_run(waking_worker, j);

		free(waking_worker);
	}

	/* cleanup & report results */
	pthread_cond_destroy(&thread_parent);
	pthread_cond_destroy(&thread_worker);
	pthread_mutex_destroy(&thread_lock);
	pthread_attr_destroy(&thread_attr);

	print_summary();

	free(blocked_worker);
	return ret;
}
J
James Yang 已提交
326
#endif /* HAVE_PTHREAD_BARRIER */