[linux-2.6-block.git] / tools / perf / bench / futex-wake-parallel.c

// SPDX-License-Identifier: GPL-2.0
/*
 * 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.
 */
#include "bench.h"
#include <linux/compiler.h>
#include "../util/debug.h"

#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 */
/* For the CLR_() macros */
#include <string.h>
#include <pthread.h>

#include <signal.h>
#include "../util/stat.h"
#include <subcmd/parse-options.h>
#include <linux/kernel.h>
#include <linux/time64.h>
#include <errno.h>
#include "futex.h"
#include "cpumap.h"

#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;
static bool done = false, silent = false, fshared = false;
static unsigned int nblocked_threads = 0, nwaking_threads = 0;
static pthread_mutex_t thread_lock;
static pthread_cond_t thread_parent, thread_worker;
static pthread_barrier_t barrier;
static struct stats waketime_stats, wakeup_stats;
static unsigned int threads_starting;
static int futex_flag = 0;

static const struct option options[] = {
	OPT_UINTEGER('t', "threads", &nblocked_threads, "Specify amount of threads"),
	OPT_UINTEGER('w', "nwakers", &nwaking_threads, "Specify amount of waking threads"),
	OPT_BOOLEAN( 's', "silent",  &silent,   "Silent mode: do not display data/details"),
	OPT_BOOLEAN( 'S', "shared",  &fshared,  "Use shared futexes instead of private ones"),
	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;

	pthread_barrier_wait(&barrier);

	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);

	pthread_barrier_init(&barrier, NULL, nwaking_threads + 1);

	/* create and block all threads */
	for (i = 0; i < nwaking_threads; i++) {
		/*
		 * 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");
	}

	pthread_barrier_wait(&barrier);

	for (i = 0; i < nwaking_threads; i++)
		if (pthread_join(td[i].worker, NULL))
			err(EXIT_FAILURE, "pthread_join");

	pthread_barrier_destroy(&barrier);
}

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;
}

static void block_threads(pthread_t *w, pthread_attr_t thread_attr,
			  struct cpu_map *cpu)
{
	cpu_set_t cpuset;
	unsigned int i;

	threads_starting = nblocked_threads;

	/* create and block all threads */
	for (i = 0; i < nblocked_threads; i++) {
		CPU_ZERO(&cpuset);
		CPU_SET(cpu->map[i % cpu->nr], &cpuset);

		if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpuset))
			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);

	for (i = 0; i < nwaking_threads; i++) {
		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,
	       nblocked_threads, waketime_avg / USEC_PER_MSEC,
	       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,
	       nblocked_threads,
	       waketime_avg / USEC_PER_MSEC,
	       rel_stddev_stats(waketime_stddev, waketime_avg));
}


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

	for (i = 0; i < nwaking_threads; i++) {
		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;
}

int bench_futex_wake_parallel(int argc, const char **argv)
{
	int ret = 0;
	unsigned int i, j;
	struct sigaction act;
	pthread_attr_t thread_attr;
	struct thread_data *waking_worker;
	struct cpu_map *cpu;

	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);
	}

	sigfillset(&act.sa_mask);
	act.sa_sigaction = toggle_done;
	sigaction(SIGINT, &act, NULL);

	cpu = cpu_map__new(NULL);
	if (!cpu)
		err(EXIT_FAILURE, "calloc");

	if (!nblocked_threads)
		nblocked_threads = cpu->nr;

	/* some sanity checks */
	if (nwaking_threads > nblocked_threads || !nwaking_threads)
		nwaking_threads = nblocked_threads;

	if (nblocked_threads % nwaking_threads)
		errx(EXIT_FAILURE, "Must be perfectly divisible");
	/*
	 * Each thread will wakeup nwakes tasks in
	 * a single futex_wait call.
	 */
	nwakes = nblocked_threads/nwaking_threads;

	blocked_worker = calloc(nblocked_threads, sizeof(*blocked_worker));
	if (!blocked_worker)
		err(EXIT_FAILURE, "calloc");

	if (!fshared)
		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",
	       getpid(), nblocked_threads, fshared ? "shared":"private",
	       &futex, nwaking_threads, nwakes);

	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++) {
		waking_worker = calloc(nwaking_threads, sizeof(*waking_worker));
		if (!waking_worker)
			err(EXIT_FAILURE, "calloc");

		/* create, launch & block all threads */
		block_threads(blocked_worker, thread_attr, cpu);

		/* 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);

		for (i = 0; i < nblocked_threads; i++) {
			ret = pthread_join(blocked_worker[i], NULL);
			if (ret)
				err(EXIT_FAILURE, "pthread_join");
		}

		do_run_stats(waking_worker);
		if (!silent)
			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;
}
#endif /* HAVE_PTHREAD_BARRIER */
Commit	Line	Data
b2441318	1	// SPDX-License-Identifier: GPL-2.0
d65817b4 DB	2	/*
	3	* Copyright (C) 2015 Davidlohr Bueso.
	4	*
	5	* Block a bunch of threads and let parallel waker threads wakeup an
	6	* equal amount of them. The program output reflects the avg latency
	7	* for each individual thread to service its share of work. Ultimately
	8	* it can be used to measure futex_wake() changes.
	9	*/
8085e5ab JY	10	#include "bench.h"
	11	#include <linux/compiler.h>
	12	#include "../util/debug.h"
d65817b4	13
8085e5ab JY	14	#ifndef HAVE_PTHREAD_BARRIER
	15	int bench_futex_wake_parallel(int argc __maybe_unused, const char **argv __maybe_unused)
	16	{
	17	pr_err("%s: pthread_barrier_t unavailable, disabling this test...\n", __func__);
	18	return 0;
	19	}
	20	#else /* HAVE_PTHREAD_BARRIER */
8a158589	21	/* For the CLR_() macros */
a0f213e1	22	#include <string.h>
8a158589 ACM	23	#include <pthread.h>
8a158589 ACM	24
9c304f6c	25	#include <signal.h>
d65817b4	26	#include "../util/stat.h"
4b6ab94e	27	#include <subcmd/parse-options.h>
9c304f6c	28	#include <linux/kernel.h>
565e6911	29	#include <linux/time64.h>
9c304f6c	30	#include <errno.h>
d65817b4	31	#include "futex.h"
3b2323c2	32	#include "cpumap.h"
d65817b4 DB	33
	34	#include <err.h>
	35	#include <stdlib.h>
	36	#include <sys/time.h>
d65817b4 DB	37
	38	struct thread_data {
	39	pthread_t worker;
	40	unsigned int nwoken;
	41	struct timeval runtime;
	42	};
	43
	44	static unsigned int nwakes = 1;
	45
	46	/* all threads will block on the same futex -- hash bucket chaos ;) */
	47	static u_int32_t futex = 0;
	48
	49	static pthread_t *blocked_worker;
	50	static bool done = false, silent = false, fshared = false;
	51	static unsigned int nblocked_threads = 0, nwaking_threads = 0;
	52	static pthread_mutex_t thread_lock;
	53	static pthread_cond_t thread_parent, thread_worker;
8085e5ab	54	static pthread_barrier_t barrier;
d65817b4	55	static struct stats waketime_stats, wakeup_stats;
3b2323c2	56	static unsigned int threads_starting;
d65817b4 DB	57	static int futex_flag = 0;
	58
	59	static const struct option options[] = {
	60	OPT_UINTEGER('t', "threads", &nblocked_threads, "Specify amount of threads"),
	61	OPT_UINTEGER('w', "nwakers", &nwaking_threads, "Specify amount of waking threads"),
	62	OPT_BOOLEAN( 's', "silent", &silent, "Silent mode: do not display data/details"),
	63	OPT_BOOLEAN( 'S', "shared", &fshared, "Use shared futexes instead of private ones"),
	64	OPT_END()
	65	};
	66
	67	static const char * const bench_futex_wake_parallel_usage[] = {
	68	"perf bench futex wake-parallel <options>",
	69	NULL
	70	};
	71
	72	static void waking_workerfn(void arg)
	73	{
	74	struct thread_data waker = (struct thread_data ) arg;
	75	struct timeval start, end;
	76
8085e5ab JY	77	pthread_barrier_wait(&barrier);
8085e5ab JY	78
d65817b4 DB	79	gettimeofday(&start, NULL);
	80
	81	waker->nwoken = futex_wake(&futex, nwakes, futex_flag);
	82	if (waker->nwoken != nwakes)
	83	warnx("couldn't wakeup all tasks (%d/%d)",
	84	waker->nwoken, nwakes);
	85
	86	gettimeofday(&end, NULL);
	87	timersub(&end, &start, &waker->runtime);
	88
	89	pthread_exit(NULL);
	90	return NULL;
	91	}
	92
	93	static void wakeup_threads(struct thread_data *td, pthread_attr_t thread_attr)
	94	{
	95	unsigned int i;
	96
	97	pthread_attr_setdetachstate(&thread_attr, PTHREAD_CREATE_JOINABLE);
	98
8085e5ab JY	99	pthread_barrier_init(&barrier, NULL, nwaking_threads + 1);
8085e5ab JY	100
d65817b4 DB	101	/* create and block all threads */
	102	for (i = 0; i < nwaking_threads; i++) {
	103	/*
	104	* Thread creation order will impact per-thread latency
	105	* as it will affect the order to acquire the hb spinlock.
	106	* For now let the scheduler decide.
	107	*/
	108	if (pthread_create(&td[i].worker, &thread_attr,
	109	waking_workerfn, (void *)&td[i]))
	110	err(EXIT_FAILURE, "pthread_create");
	111	}
	112
8085e5ab JY	113	pthread_barrier_wait(&barrier);
8085e5ab JY	114
d65817b4 DB	115	for (i = 0; i < nwaking_threads; i++)
	116	if (pthread_join(td[i].worker, NULL))
	117	err(EXIT_FAILURE, "pthread_join");
8085e5ab JY	118
8085e5ab JY	119	pthread_barrier_destroy(&barrier);
d65817b4 DB	120	}
	121
	122	static void blocked_workerfn(void arg __maybe_unused)
	123	{
	124	pthread_mutex_lock(&thread_lock);
	125	threads_starting--;
	126	if (!threads_starting)
	127	pthread_cond_signal(&thread_parent);
	128	pthread_cond_wait(&thread_worker, &thread_lock);
	129	pthread_mutex_unlock(&thread_lock);
	130
	131	while (1) { /* handle spurious wakeups */
	132	if (futex_wait(&futex, 0, NULL, futex_flag) != EINTR)
	133	break;
	134	}
	135
	136	pthread_exit(NULL);
	137	return NULL;
	138	}
	139
3b2323c2 DB	140	static void block_threads(pthread_t *w, pthread_attr_t thread_attr,
3b2323c2 DB	141	struct cpu_map *cpu)
d65817b4	142	{
3b2323c2	143	cpu_set_t cpuset;
d65817b4 DB	144	unsigned int i;
	145
	146	threads_starting = nblocked_threads;
	147
	148	/* create and block all threads */
	149	for (i = 0; i < nblocked_threads; i++) {
3b2323c2 DB	150	CPU_ZERO(&cpuset);
3b2323c2 DB	151	CPU_SET(cpu->map[i % cpu->nr], &cpuset);
d65817b4	152
3b2323c2	153	if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpuset))
d65817b4 DB	154	err(EXIT_FAILURE, "pthread_attr_setaffinity_np");
	155
	156	if (pthread_create(&w[i], &thread_attr, blocked_workerfn, NULL))
	157	err(EXIT_FAILURE, "pthread_create");
	158	}
	159	}
	160
	161	static void print_run(struct thread_data *waking_worker, unsigned int run_num)
	162	{
	163	unsigned int i, wakeup_avg;
	164	double waketime_avg, waketime_stddev;
	165	struct stats __waketime_stats, __wakeup_stats;
	166
	167	init_stats(&__wakeup_stats);
	168	init_stats(&__waketime_stats);
	169
	170	for (i = 0; i < nwaking_threads; i++) {
	171	update_stats(&__waketime_stats, waking_worker[i].runtime.tv_usec);
	172	update_stats(&__wakeup_stats, waking_worker[i].nwoken);
	173	}
	174
	175	waketime_avg = avg_stats(&__waketime_stats);
	176	waketime_stddev = stddev_stats(&__waketime_stats);
	177	wakeup_avg = avg_stats(&__wakeup_stats);
	178
	179	printf("[Run %d]: Avg per-thread latency (waking %d/%d threads) "
	180	"in %.4f ms (+-%.2f%%)\n", run_num + 1, wakeup_avg,
565e6911	181	nblocked_threads, waketime_avg / USEC_PER_MSEC,
d65817b4 DB	182	rel_stddev_stats(waketime_stddev, waketime_avg));
	183	}
	184
	185	static void print_summary(void)
	186	{
	187	unsigned int wakeup_avg;
	188	double waketime_avg, waketime_stddev;
	189
	190	waketime_avg = avg_stats(&waketime_stats);
	191	waketime_stddev = stddev_stats(&waketime_stats);
	192	wakeup_avg = avg_stats(&wakeup_stats);
	193
	194	printf("Avg per-thread latency (waking %d/%d threads) in %.4f ms (+-%.2f%%)\n",
	195	wakeup_avg,
	196	nblocked_threads,
565e6911	197	waketime_avg / USEC_PER_MSEC,
d65817b4 DB	198	rel_stddev_stats(waketime_stddev, waketime_avg));
	199	}
	200
	201
	202	static void do_run_stats(struct thread_data *waking_worker)
	203	{
	204	unsigned int i;
	205
	206	for (i = 0; i < nwaking_threads; i++) {
	207	update_stats(&waketime_stats, waking_worker[i].runtime.tv_usec);
	208	update_stats(&wakeup_stats, waking_worker[i].nwoken);
	209	}
	210
	211	}
	212
	213	static void toggle_done(int sig __maybe_unused,
	214	siginfo_t *info __maybe_unused,
	215	void *uc __maybe_unused)
	216	{
	217	done = true;
	218	}
	219
b0ad8ea6	220	int bench_futex_wake_parallel(int argc, const char **argv)
d65817b4 DB	221	{
	222	int ret = 0;
	223	unsigned int i, j;
	224	struct sigaction act;
	225	pthread_attr_t thread_attr;
	226	struct thread_data *waking_worker;
3b2323c2	227	struct cpu_map *cpu;
d65817b4 DB	228
	229	argc = parse_options(argc, argv, options,
	230	bench_futex_wake_parallel_usage, 0);
	231	if (argc) {
	232	usage_with_options(bench_futex_wake_parallel_usage, options);
	233	exit(EXIT_FAILURE);
	234	}
	235
	236	sigfillset(&act.sa_mask);
	237	act.sa_sigaction = toggle_done;
	238	sigaction(SIGINT, &act, NULL);
	239
3b2323c2 DB	240	cpu = cpu_map__new(NULL);
	241	if (!cpu)
	242	err(EXIT_FAILURE, "calloc");
	243
d65817b4	244	if (!nblocked_threads)
3b2323c2	245	nblocked_threads = cpu->nr;
d65817b4 DB	246
	247	/* some sanity checks */
	248	if (nwaking_threads > nblocked_threads \|\| !nwaking_threads)
	249	nwaking_threads = nblocked_threads;
	250
	251	if (nblocked_threads % nwaking_threads)
	252	errx(EXIT_FAILURE, "Must be perfectly divisible");
	253	/*
	254	* Each thread will wakeup nwakes tasks in
	255	* a single futex_wait call.
	256	*/
	257	nwakes = nblocked_threads/nwaking_threads;
	258
	259	blocked_worker = calloc(nblocked_threads, sizeof(*blocked_worker));
	260	if (!blocked_worker)
	261	err(EXIT_FAILURE, "calloc");
	262
	263	if (!fshared)
	264	futex_flag = FUTEX_PRIVATE_FLAG;
	265
	266	printf("Run summary [PID %d]: blocking on %d threads (at [%s] "
	267	"futex %p), %d threads waking up %d at a time.\n\n",
	268	getpid(), nblocked_threads, fshared ? "shared":"private",
	269	&futex, nwaking_threads, nwakes);
	270
	271	init_stats(&wakeup_stats);
	272	init_stats(&waketime_stats);
	273
	274	pthread_attr_init(&thread_attr);
	275	pthread_mutex_init(&thread_lock, NULL);
	276	pthread_cond_init(&thread_parent, NULL);
	277	pthread_cond_init(&thread_worker, NULL);
	278
	279	for (j = 0; j < bench_repeat && !done; j++) {
	280	waking_worker = calloc(nwaking_threads, sizeof(*waking_worker));
	281	if (!waking_worker)
	282	err(EXIT_FAILURE, "calloc");
	283
	284	/* create, launch & block all threads */
3b2323c2	285	block_threads(blocked_worker, thread_attr, cpu);
d65817b4 DB	286
	287	/* make sure all threads are already blocked */
	288	pthread_mutex_lock(&thread_lock);
	289	while (threads_starting)
	290	pthread_cond_wait(&thread_parent, &thread_lock);
	291	pthread_cond_broadcast(&thread_worker);
	292	pthread_mutex_unlock(&thread_lock);
	293
	294	usleep(100000);
	295
	296	/* Ok, all threads are patiently blocked, start waking folks up */
	297	wakeup_threads(waking_worker, thread_attr);
	298
	299	for (i = 0; i < nblocked_threads; i++) {
	300	ret = pthread_join(blocked_worker[i], NULL);
	301	if (ret)
	302	err(EXIT_FAILURE, "pthread_join");
	303	}
	304
	305	do_run_stats(waking_worker);
	306	if (!silent)
	307	print_run(waking_worker, j);
	308
	309	free(waking_worker);
	310	}
	311
	312	/* cleanup & report results */
	313	pthread_cond_destroy(&thread_parent);
	314	pthread_cond_destroy(&thread_worker);
	315	pthread_mutex_destroy(&thread_lock);
	316	pthread_attr_destroy(&thread_attr);
	317
	318	print_summary();
	319
	320	free(blocked_worker);
	321	return ret;
	322	}
8085e5ab	323	#endif /* HAVE_PTHREAD_BARRIER */