aboutsummaryrefslogtreecommitdiffstats
path: root/tools/perf/bench
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2014-03-31 14:13:25 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2014-03-31 14:13:25 -0400
commit8c292f11744297dfb3a69f4a0bccbe4a6417b50d (patch)
treef1a89560de25a69b697d459a9b5cf2e738038d9f /tools/perf/bench
parentd31605dc8a63f1df28443ddb3560b1079417af92 (diff)
parent538592ff0b008237ae88f5ce5fb1247127dc3ce5 (diff)
Merge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull perf changes from Ingo Molnar: "Main changes: Kernel side changes: - Add SNB/IVB/HSW client uncore memory controller support (Stephane Eranian) - Fix various x86/P4 PMU driver bugs (Don Zickus) Tooling, user visible changes: - Add several futex 'perf bench' microbenchmarks (Davidlohr Bueso) - Speed up thread map generation (Don Zickus) - Introduce 'perf kvm --list-cmds' command line option for use by scripts (Ramkumar Ramachandra) - Print the evsel name in the annotate stdio output, prep to fix support outputting annotation for multiple events, not just for the first one (Arnaldo Carvalho de Melo) - Allow setting preferred callchain method in .perfconfig (Jiri Olsa) - Show in what binaries/modules 'perf probe's are set (Masami Hiramatsu) - Support distro-style debuginfo for uprobe in 'perf probe' (Masami Hiramatsu) Tooling, internal changes and fixes: - Use tid in mmap/mmap2 events to find maps (Don Zickus) - Record the reason for filtering an address_location (Namhyung Kim) - Apply all filters to an addr_location (Namhyung Kim) - Merge al->filtered with hist_entry->filtered in report/hists (Namhyung Kim) - Fix memory leak when synthesizing thread records (Namhyung Kim) - Use ui__has_annotation() in 'report' (Namhyung Kim) - hists browser refactorings to reuse code accross UIs (Namhyung Kim) - Add support for the new DWARF unwinder library in elfutils (Jiri Olsa) - Fix build race in the generation of bison files (Jiri Olsa) - Further streamline the feature detection display, trimming it a bit to show just the libraries detected, using VF=1 gets a more verbose output, showing the less interesting feature checks as well (Jiri Olsa). - Check compatible symtab type before loading dso (Namhyung Kim) - Check return value of filename__read_debuglink() (Stephane Eranian) - Move some hashing and fs related code from tools/perf/util/ to tools/lib/ so that it can be used by more tools/ living utilities (Borislav Petkov) - Prepare DWARF unwinding code for using an elfutils alternative unwinding library (Jiri Olsa) - Fix DWARF unwind max_stack processing (Jiri Olsa) - Add dwarf unwind 'perf test' entry (Jiri Olsa) - 'perf probe' improvements including memory leak fixes, sharing the intlist class with other tools, uprobes/kprobes code sharing and use of ref_reloc_sym (Masami Hiramatsu) - Shorten sample symbol resolving by adding cpumode to struct addr_location (Arnaldo Carvalho de Melo) - Fix synthesizing mmaps for threads (Don Zickus) - Fix invalid output on event group stdio report (Namhyung Kim) - Fixup header alignment in 'perf sched latency' output (Ramkumar Ramachandra) - Fix off-by-one error in 'perf timechart record' argv handling (Ramkumar Ramachandra) Tooling, cleanups: - Remove unused thread__find_map function (Jiri Olsa) - Remove unused simple_strtoul() function (Ramkumar Ramachandra) Tooling, documentation updates: - Update function names in debug messages (Ramkumar Ramachandra) - Update some code references in design.txt (Ramkumar Ramachandra) - Clarify load-latency information in the 'perf mem' docs (Andi Kleen) - Clarify x86 register naming in 'perf probe' docs (Andi Kleen)" * 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (96 commits) perf tools: Remove unused simple_strtoul() function perf tools: Update some code references in design.txt perf evsel: Update function names in debug messages perf tools: Remove thread__find_map function perf annotate: Print the evsel name in the stdio output perf report: Use ui__has_annotation() perf tools: Fix memory leak when synthesizing thread records perf tools: Use tid in mmap/mmap2 events to find maps perf report: Merge al->filtered with hist_entry->filtered perf symbols: Apply all filters to an addr_location perf symbols: Record the reason for filtering an address_location perf sched: Fixup header alignment in 'latency' output perf timechart: Fix off-by-one error in 'record' argv handling perf machine: Factor machine__find_thread to take tid argument perf tools: Speed up thread map generation perf kvm: introduce --list-cmds for use by scripts perf ui hists: Pass evsel to hpp->header/width functions explicitly perf symbols: Introduce thread__find_cpumode_addr_location perf session: Change header.misc dump from decimal to hex perf ui/tui: Reuse generic __hpp__fmt() code ...
Diffstat (limited to 'tools/perf/bench')
-rw-r--r--tools/perf/bench/bench.h3
-rw-r--r--tools/perf/bench/futex-hash.c212
-rw-r--r--tools/perf/bench/futex-requeue.c211
-rw-r--r--tools/perf/bench/futex-wake.c201
-rw-r--r--tools/perf/bench/futex.h71
5 files changed, 698 insertions, 0 deletions
diff --git a/tools/perf/bench/bench.h b/tools/perf/bench/bench.h
index 0fdc85269c4d..eba46709b279 100644
--- a/tools/perf/bench/bench.h
+++ b/tools/perf/bench/bench.h
@@ -31,6 +31,9 @@ extern int bench_sched_pipe(int argc, const char **argv, const char *prefix);
31extern int bench_mem_memcpy(int argc, const char **argv, 31extern int bench_mem_memcpy(int argc, const char **argv,
32 const char *prefix __maybe_unused); 32 const char *prefix __maybe_unused);
33extern int bench_mem_memset(int argc, const char **argv, const char *prefix); 33extern int bench_mem_memset(int argc, const char **argv, const char *prefix);
34extern int bench_futex_hash(int argc, const char **argv, const char *prefix);
35extern int bench_futex_wake(int argc, const char **argv, const char *prefix);
36extern int bench_futex_requeue(int argc, const char **argv, const char *prefix);
34 37
35#define BENCH_FORMAT_DEFAULT_STR "default" 38#define BENCH_FORMAT_DEFAULT_STR "default"
36#define BENCH_FORMAT_DEFAULT 0 39#define BENCH_FORMAT_DEFAULT 0
diff --git a/tools/perf/bench/futex-hash.c b/tools/perf/bench/futex-hash.c
new file mode 100644
index 000000000000..a84206e9c4aa
--- /dev/null
+++ b/tools/perf/bench/futex-hash.c
@@ -0,0 +1,212 @@
1/*
2 * Copyright (C) 2013 Davidlohr Bueso <davidlohr@hp.com>
3 *
4 * futex-hash: Stress the hell out of the Linux kernel futex uaddr hashing.
5 *
6 * This program is particularly useful for measuring the kernel's futex hash
7 * table/function implementation. In order for it to make sense, use with as
8 * many threads and futexes as possible.
9 */
10
11#include "../perf.h"
12#include "../util/util.h"
13#include "../util/stat.h"
14#include "../util/parse-options.h"
15#include "../util/header.h"
16#include "bench.h"
17#include "futex.h"
18
19#include <err.h>
20#include <stdlib.h>
21#include <sys/time.h>
22#include <pthread.h>
23
24static unsigned int nthreads = 0;
25static unsigned int nsecs = 10;
26/* amount of futexes per thread */
27static unsigned int nfutexes = 1024;
28static bool fshared = false, done = false, silent = false;
29
30struct timeval start, end, runtime;
31static pthread_mutex_t thread_lock;
32static unsigned int threads_starting;
33static struct stats throughput_stats;
34static pthread_cond_t thread_parent, thread_worker;
35
36struct worker {
37 int tid;
38 u_int32_t *futex;
39 pthread_t thread;
40 unsigned long ops;
41};
42
43static const struct option options[] = {
44 OPT_UINTEGER('t', "threads", &nthreads, "Specify amount of threads"),
45 OPT_UINTEGER('r', "runtime", &nsecs, "Specify runtime (in seconds)"),
46 OPT_UINTEGER('f', "futexes", &nfutexes, "Specify amount of futexes per threads"),
47 OPT_BOOLEAN( 's', "silent", &silent, "Silent mode: do not display data/details"),
48 OPT_BOOLEAN( 'S', "shared", &fshared, "Use shared futexes instead of private ones"),
49 OPT_END()
50};
51
52static const char * const bench_futex_hash_usage[] = {
53 "perf bench futex hash <options>",
54 NULL
55};
56
57static void *workerfn(void *arg)
58{
59 int ret;
60 unsigned int i;
61 struct worker *w = (struct worker *) arg;
62
63 pthread_mutex_lock(&thread_lock);
64 threads_starting--;
65 if (!threads_starting)
66 pthread_cond_signal(&thread_parent);
67 pthread_cond_wait(&thread_worker, &thread_lock);
68 pthread_mutex_unlock(&thread_lock);
69
70 do {
71 for (i = 0; i < nfutexes; i++, w->ops++) {
72 /*
73 * We want the futex calls to fail in order to stress
74 * the hashing of uaddr and not measure other steps,
75 * such as internal waitqueue handling, thus enlarging
76 * the critical region protected by hb->lock.
77 */
78 ret = futex_wait(&w->futex[i], 1234, NULL,
79 fshared ? 0 : FUTEX_PRIVATE_FLAG);
80 if (!silent &&
81 (!ret || errno != EAGAIN || errno != EWOULDBLOCK))
82 warn("Non-expected futex return call");
83 }
84 } while (!done);
85
86 return NULL;
87}
88
89static void toggle_done(int sig __maybe_unused,
90 siginfo_t *info __maybe_unused,
91 void *uc __maybe_unused)
92{
93 /* inform all threads that we're done for the day */
94 done = true;
95 gettimeofday(&end, NULL);
96 timersub(&end, &start, &runtime);
97}
98
99static void print_summary(void)
100{
101 unsigned long avg = avg_stats(&throughput_stats);
102 double stddev = stddev_stats(&throughput_stats);
103
104 printf("%sAveraged %ld operations/sec (+- %.2f%%), total secs = %d\n",
105 !silent ? "\n" : "", avg, rel_stddev_stats(stddev, avg),
106 (int) runtime.tv_sec);
107}
108
109int bench_futex_hash(int argc, const char **argv,
110 const char *prefix __maybe_unused)
111{
112 int ret = 0;
113 cpu_set_t cpu;
114 struct sigaction act;
115 unsigned int i, ncpus;
116 pthread_attr_t thread_attr;
117 struct worker *worker = NULL;
118
119 argc = parse_options(argc, argv, options, bench_futex_hash_usage, 0);
120 if (argc) {
121 usage_with_options(bench_futex_hash_usage, options);
122 exit(EXIT_FAILURE);
123 }
124
125 ncpus = sysconf(_SC_NPROCESSORS_ONLN);
126
127 sigfillset(&act.sa_mask);
128 act.sa_sigaction = toggle_done;
129 sigaction(SIGINT, &act, NULL);
130
131 if (!nthreads) /* default to the number of CPUs */
132 nthreads = ncpus;
133
134 worker = calloc(nthreads, sizeof(*worker));
135 if (!worker)
136 goto errmem;
137
138 printf("Run summary [PID %d]: %d threads, each operating on %d [%s] futexes for %d secs.\n\n",
139 getpid(), nthreads, nfutexes, fshared ? "shared":"private", nsecs);
140
141 init_stats(&throughput_stats);
142 pthread_mutex_init(&thread_lock, NULL);
143 pthread_cond_init(&thread_parent, NULL);
144 pthread_cond_init(&thread_worker, NULL);
145
146 threads_starting = nthreads;
147 pthread_attr_init(&thread_attr);
148 gettimeofday(&start, NULL);
149 for (i = 0; i < nthreads; i++) {
150 worker[i].tid = i;
151 worker[i].futex = calloc(nfutexes, sizeof(*worker[i].futex));
152 if (!worker[i].futex)
153 goto errmem;
154
155 CPU_ZERO(&cpu);
156 CPU_SET(i % ncpus, &cpu);
157
158 ret = pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpu);
159 if (ret)
160 err(EXIT_FAILURE, "pthread_attr_setaffinity_np");
161
162 ret = pthread_create(&worker[i].thread, &thread_attr, workerfn,
163 (void *)(struct worker *) &worker[i]);
164 if (ret)
165 err(EXIT_FAILURE, "pthread_create");
166
167 }
168 pthread_attr_destroy(&thread_attr);
169
170 pthread_mutex_lock(&thread_lock);
171 while (threads_starting)
172 pthread_cond_wait(&thread_parent, &thread_lock);
173 pthread_cond_broadcast(&thread_worker);
174 pthread_mutex_unlock(&thread_lock);
175
176 sleep(nsecs);
177 toggle_done(0, NULL, NULL);
178
179 for (i = 0; i < nthreads; i++) {
180 ret = pthread_join(worker[i].thread, NULL);
181 if (ret)
182 err(EXIT_FAILURE, "pthread_join");
183 }
184
185 /* cleanup & report results */
186 pthread_cond_destroy(&thread_parent);
187 pthread_cond_destroy(&thread_worker);
188 pthread_mutex_destroy(&thread_lock);
189
190 for (i = 0; i < nthreads; i++) {
191 unsigned long t = worker[i].ops/runtime.tv_sec;
192 update_stats(&throughput_stats, t);
193 if (!silent) {
194 if (nfutexes == 1)
195 printf("[thread %2d] futex: %p [ %ld ops/sec ]\n",
196 worker[i].tid, &worker[i].futex[0], t);
197 else
198 printf("[thread %2d] futexes: %p ... %p [ %ld ops/sec ]\n",
199 worker[i].tid, &worker[i].futex[0],
200 &worker[i].futex[nfutexes-1], t);
201 }
202
203 free(worker[i].futex);
204 }
205
206 print_summary();
207
208 free(worker);
209 return ret;
210errmem:
211 err(EXIT_FAILURE, "calloc");
212}
diff --git a/tools/perf/bench/futex-requeue.c b/tools/perf/bench/futex-requeue.c
new file mode 100644
index 000000000000..a16255876f1d
--- /dev/null
+++ b/tools/perf/bench/futex-requeue.c
@@ -0,0 +1,211 @@
1/*
2 * Copyright (C) 2013 Davidlohr Bueso <davidlohr@hp.com>
3 *
4 * futex-requeue: Block a bunch of threads on futex1 and requeue them
5 * on futex2, N at a time.
6 *
7 * This program is particularly useful to measure the latency of nthread
8 * requeues without waking up any tasks -- thus mimicking a regular futex_wait.
9 */
10
11#include "../perf.h"
12#include "../util/util.h"
13#include "../util/stat.h"
14#include "../util/parse-options.h"
15#include "../util/header.h"
16#include "bench.h"
17#include "futex.h"
18
19#include <err.h>
20#include <stdlib.h>
21#include <sys/time.h>
22#include <pthread.h>
23
24static u_int32_t futex1 = 0, futex2 = 0;
25
26/*
27 * How many tasks to requeue at a time.
28 * Default to 1 in order to make the kernel work more.
29 */
30static unsigned int nrequeue = 1;
31
32/*
33 * There can be significant variance from run to run,
34 * the more repeats, the more exact the overall avg and
35 * the better idea of the futex latency.
36 */
37static unsigned int repeat = 10;
38
39static pthread_t *worker;
40static bool done = 0, silent = 0;
41static pthread_mutex_t thread_lock;
42static pthread_cond_t thread_parent, thread_worker;
43static struct stats requeuetime_stats, requeued_stats;
44static unsigned int ncpus, threads_starting, nthreads = 0;
45
46static const struct option options[] = {
47 OPT_UINTEGER('t', "threads", &nthreads, "Specify amount of threads"),
48 OPT_UINTEGER('q', "nrequeue", &nrequeue, "Specify amount of threads to requeue at once"),
49 OPT_UINTEGER('r', "repeat", &repeat, "Specify amount of times to repeat the run"),
50 OPT_BOOLEAN( 's', "silent", &silent, "Silent mode: do not display data/details"),
51 OPT_END()
52};
53
54static const char * const bench_futex_requeue_usage[] = {
55 "perf bench futex requeue <options>",
56 NULL
57};
58
59static void print_summary(void)
60{
61 double requeuetime_avg = avg_stats(&requeuetime_stats);
62 double requeuetime_stddev = stddev_stats(&requeuetime_stats);
63 unsigned int requeued_avg = avg_stats(&requeued_stats);
64
65 printf("Requeued %d of %d threads in %.4f ms (+-%.2f%%)\n",
66 requeued_avg,
67 nthreads,
68 requeuetime_avg/1e3,
69 rel_stddev_stats(requeuetime_stddev, requeuetime_avg));
70}
71
72static void *workerfn(void *arg __maybe_unused)
73{
74 pthread_mutex_lock(&thread_lock);
75 threads_starting--;
76 if (!threads_starting)
77 pthread_cond_signal(&thread_parent);
78 pthread_cond_wait(&thread_worker, &thread_lock);
79 pthread_mutex_unlock(&thread_lock);
80
81 futex_wait(&futex1, 0, NULL, FUTEX_PRIVATE_FLAG);
82 return NULL;
83}
84
85static void block_threads(pthread_t *w,
86 pthread_attr_t thread_attr)
87{
88 cpu_set_t cpu;
89 unsigned int i;
90
91 threads_starting = nthreads;
92
93 /* create and block all threads */
94 for (i = 0; i < nthreads; i++) {
95 CPU_ZERO(&cpu);
96 CPU_SET(i % ncpus, &cpu);
97
98 if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpu))
99 err(EXIT_FAILURE, "pthread_attr_setaffinity_np");
100
101 if (pthread_create(&w[i], &thread_attr, workerfn, NULL))
102 err(EXIT_FAILURE, "pthread_create");
103 }
104}
105
106static void toggle_done(int sig __maybe_unused,
107 siginfo_t *info __maybe_unused,
108 void *uc __maybe_unused)
109{
110 done = true;
111}
112
113int bench_futex_requeue(int argc, const char **argv,
114 const char *prefix __maybe_unused)
115{
116 int ret = 0;
117 unsigned int i, j;
118 struct sigaction act;
119 pthread_attr_t thread_attr;
120
121 argc = parse_options(argc, argv, options, bench_futex_requeue_usage, 0);
122 if (argc)
123 goto err;
124
125 ncpus = sysconf(_SC_NPROCESSORS_ONLN);
126
127 sigfillset(&act.sa_mask);
128 act.sa_sigaction = toggle_done;
129 sigaction(SIGINT, &act, NULL);
130
131 if (!nthreads)
132 nthreads = ncpus;
133
134 worker = calloc(nthreads, sizeof(*worker));
135 if (!worker)
136 err(EXIT_FAILURE, "calloc");
137
138 printf("Run summary [PID %d]: Requeuing %d threads (from %p to %p), "
139 "%d at a time.\n\n",
140 getpid(), nthreads, &futex1, &futex2, nrequeue);
141
142 init_stats(&requeued_stats);
143 init_stats(&requeuetime_stats);
144 pthread_attr_init(&thread_attr);
145 pthread_mutex_init(&thread_lock, NULL);
146 pthread_cond_init(&thread_parent, NULL);
147 pthread_cond_init(&thread_worker, NULL);
148
149 for (j = 0; j < repeat && !done; j++) {
150 unsigned int nrequeued = 0;
151 struct timeval start, end, runtime;
152
153 /* create, launch & block all threads */
154 block_threads(worker, thread_attr);
155
156 /* make sure all threads are already blocked */
157 pthread_mutex_lock(&thread_lock);
158 while (threads_starting)
159 pthread_cond_wait(&thread_parent, &thread_lock);
160 pthread_cond_broadcast(&thread_worker);
161 pthread_mutex_unlock(&thread_lock);
162
163 usleep(100000);
164
165 /* Ok, all threads are patiently blocked, start requeueing */
166 gettimeofday(&start, NULL);
167 for (nrequeued = 0; nrequeued < nthreads; nrequeued += nrequeue)
168 /*
169 * Do not wakeup any tasks blocked on futex1, allowing
170 * us to really measure futex_wait functionality.
171 */
172 futex_cmp_requeue(&futex1, 0, &futex2, 0, nrequeue,
173 FUTEX_PRIVATE_FLAG);
174 gettimeofday(&end, NULL);
175 timersub(&end, &start, &runtime);
176
177 update_stats(&requeued_stats, nrequeued);
178 update_stats(&requeuetime_stats, runtime.tv_usec);
179
180 if (!silent) {
181 printf("[Run %d]: Requeued %d of %d threads in %.4f ms\n",
182 j + 1, nrequeued, nthreads, runtime.tv_usec/1e3);
183 }
184
185 /* everybody should be blocked on futex2, wake'em up */
186 nrequeued = futex_wake(&futex2, nthreads, FUTEX_PRIVATE_FLAG);
187 if (nthreads != nrequeued)
188 warnx("couldn't wakeup all tasks (%d/%d)", nrequeued, nthreads);
189
190 for (i = 0; i < nthreads; i++) {
191 ret = pthread_join(worker[i], NULL);
192 if (ret)
193 err(EXIT_FAILURE, "pthread_join");
194 }
195
196 }
197
198 /* cleanup & report results */
199 pthread_cond_destroy(&thread_parent);
200 pthread_cond_destroy(&thread_worker);
201 pthread_mutex_destroy(&thread_lock);
202 pthread_attr_destroy(&thread_attr);
203
204 print_summary();
205
206 free(worker);
207 return ret;
208err:
209 usage_with_options(bench_futex_requeue_usage, options);
210 exit(EXIT_FAILURE);
211}
diff --git a/tools/perf/bench/futex-wake.c b/tools/perf/bench/futex-wake.c
new file mode 100644
index 000000000000..d096169b161e
--- /dev/null
+++ b/tools/perf/bench/futex-wake.c
@@ -0,0 +1,201 @@
1/*
2 * Copyright (C) 2013 Davidlohr Bueso <davidlohr@hp.com>
3 *
4 * futex-wake: Block a bunch of threads on a futex and wake'em up, N at a time.
5 *
6 * This program is particularly useful to measure the latency of nthread wakeups
7 * in non-error situations: all waiters are queued and all wake calls wakeup
8 * one or more tasks, and thus the waitqueue is never empty.
9 */
10
11#include "../perf.h"
12#include "../util/util.h"
13#include "../util/stat.h"
14#include "../util/parse-options.h"
15#include "../util/header.h"
16#include "bench.h"
17#include "futex.h"
18
19#include <err.h>
20#include <stdlib.h>
21#include <sys/time.h>
22#include <pthread.h>
23
24/* all threads will block on the same futex */
25static u_int32_t futex1 = 0;
26
27/*
28 * How many wakeups to do at a time.
29 * Default to 1 in order to make the kernel work more.
30 */
31static unsigned int nwakes = 1;
32
33/*
34 * There can be significant variance from run to run,
35 * the more repeats, the more exact the overall avg and
36 * the better idea of the futex latency.
37 */
38static unsigned int repeat = 10;
39
40pthread_t *worker;
41static bool done = 0, silent = 0;
42static pthread_mutex_t thread_lock;
43static pthread_cond_t thread_parent, thread_worker;
44static struct stats waketime_stats, wakeup_stats;
45static unsigned int ncpus, threads_starting, nthreads = 0;
46
47static const struct option options[] = {
48 OPT_UINTEGER('t', "threads", &nthreads, "Specify amount of threads"),
49 OPT_UINTEGER('w', "nwakes", &nwakes, "Specify amount of threads to wake at once"),
50 OPT_UINTEGER('r', "repeat", &repeat, "Specify amount of times to repeat the run"),
51 OPT_BOOLEAN( 's', "silent", &silent, "Silent mode: do not display data/details"),
52 OPT_END()
53};
54
55static const char * const bench_futex_wake_usage[] = {
56 "perf bench futex wake <options>",
57 NULL
58};
59
60static void *workerfn(void *arg __maybe_unused)
61{
62 pthread_mutex_lock(&thread_lock);
63 threads_starting--;
64 if (!threads_starting)
65 pthread_cond_signal(&thread_parent);
66 pthread_cond_wait(&thread_worker, &thread_lock);
67 pthread_mutex_unlock(&thread_lock);
68
69 futex_wait(&futex1, 0, NULL, FUTEX_PRIVATE_FLAG);
70 return NULL;
71}
72
73static void print_summary(void)
74{
75 double waketime_avg = avg_stats(&waketime_stats);
76 double waketime_stddev = stddev_stats(&waketime_stats);
77 unsigned int wakeup_avg = avg_stats(&wakeup_stats);
78
79 printf("Wokeup %d of %d threads in %.4f ms (+-%.2f%%)\n",
80 wakeup_avg,
81 nthreads,
82 waketime_avg/1e3,
83 rel_stddev_stats(waketime_stddev, waketime_avg));
84}
85
86static void block_threads(pthread_t *w,
87 pthread_attr_t thread_attr)
88{
89 cpu_set_t cpu;
90 unsigned int i;
91
92 threads_starting = nthreads;
93
94 /* create and block all threads */
95 for (i = 0; i < nthreads; i++) {
96 CPU_ZERO(&cpu);
97 CPU_SET(i % ncpus, &cpu);
98
99 if (pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t), &cpu))
100 err(EXIT_FAILURE, "pthread_attr_setaffinity_np");
101
102 if (pthread_create(&w[i], &thread_attr, workerfn, NULL))
103 err(EXIT_FAILURE, "pthread_create");
104 }
105}
106
107static void toggle_done(int sig __maybe_unused,
108 siginfo_t *info __maybe_unused,
109 void *uc __maybe_unused)
110{
111 done = true;
112}
113
114int bench_futex_wake(int argc, const char **argv,
115 const char *prefix __maybe_unused)
116{
117 int ret = 0;
118 unsigned int i, j;
119 struct sigaction act;
120 pthread_attr_t thread_attr;
121
122 argc = parse_options(argc, argv, options, bench_futex_wake_usage, 0);
123 if (argc) {
124 usage_with_options(bench_futex_wake_usage, options);
125 exit(EXIT_FAILURE);
126 }
127
128 ncpus = sysconf(_SC_NPROCESSORS_ONLN);
129
130 sigfillset(&act.sa_mask);
131 act.sa_sigaction = toggle_done;
132 sigaction(SIGINT, &act, NULL);
133
134 if (!nthreads)
135 nthreads = ncpus;
136
137 worker = calloc(nthreads, sizeof(*worker));
138 if (!worker)
139 err(EXIT_FAILURE, "calloc");
140
141 printf("Run summary [PID %d]: blocking on %d threads (at futex %p), "
142 "waking up %d at a time.\n\n",
143 getpid(), nthreads, &futex1, nwakes);
144
145 init_stats(&wakeup_stats);
146 init_stats(&waketime_stats);
147 pthread_attr_init(&thread_attr);
148 pthread_mutex_init(&thread_lock, NULL);
149 pthread_cond_init(&thread_parent, NULL);
150 pthread_cond_init(&thread_worker, NULL);
151
152 for (j = 0; j < repeat && !done; j++) {
153 unsigned int nwoken = 0;
154 struct timeval start, end, runtime;
155
156 /* create, launch & block all threads */
157 block_threads(worker, thread_attr);
158
159 /* make sure all threads are already blocked */
160 pthread_mutex_lock(&thread_lock);
161 while (threads_starting)
162 pthread_cond_wait(&thread_parent, &thread_lock);
163 pthread_cond_broadcast(&thread_worker);
164 pthread_mutex_unlock(&thread_lock);
165
166 usleep(100000);
167
168 /* Ok, all threads are patiently blocked, start waking folks up */
169 gettimeofday(&start, NULL);
170 while (nwoken != nthreads)
171 nwoken += futex_wake(&futex1, nwakes, FUTEX_PRIVATE_FLAG);
172 gettimeofday(&end, NULL);
173 timersub(&end, &start, &runtime);
174
175 update_stats(&wakeup_stats, nwoken);
176 update_stats(&waketime_stats, runtime.tv_usec);
177
178 if (!silent) {
179 printf("[Run %d]: Wokeup %d of %d threads in %.4f ms\n",
180 j + 1, nwoken, nthreads, runtime.tv_usec/1e3);
181 }
182
183 for (i = 0; i < nthreads; i++) {
184 ret = pthread_join(worker[i], NULL);
185 if (ret)
186 err(EXIT_FAILURE, "pthread_join");
187 }
188
189 }
190
191 /* cleanup & report results */
192 pthread_cond_destroy(&thread_parent);
193 pthread_cond_destroy(&thread_worker);
194 pthread_mutex_destroy(&thread_lock);
195 pthread_attr_destroy(&thread_attr);
196
197 print_summary();
198
199 free(worker);
200 return ret;
201}
diff --git a/tools/perf/bench/futex.h b/tools/perf/bench/futex.h
new file mode 100644
index 000000000000..71f2844cf97f
--- /dev/null
+++ b/tools/perf/bench/futex.h
@@ -0,0 +1,71 @@
1/*
2 * Glibc independent futex library for testing kernel functionality.
3 * Shamelessly stolen from Darren Hart <dvhltc@us.ibm.com>
4 * http://git.kernel.org/cgit/linux/kernel/git/dvhart/futextest.git/
5 */
6
7#ifndef _FUTEX_H
8#define _FUTEX_H
9
10#include <unistd.h>
11#include <sys/syscall.h>
12#include <sys/types.h>
13#include <linux/futex.h>
14
15/**
16 * futex() - SYS_futex syscall wrapper
17 * @uaddr: address of first futex
18 * @op: futex op code
19 * @val: typically expected value of uaddr, but varies by op
20 * @timeout: typically an absolute struct timespec (except where noted
21 * otherwise). Overloaded by some ops
22 * @uaddr2: address of second futex for some ops\
23 * @val3: varies by op
24 * @opflags: flags to be bitwise OR'd with op, such as FUTEX_PRIVATE_FLAG
25 *
26 * futex() is used by all the following futex op wrappers. It can also be
27 * used for misuse and abuse testing. Generally, the specific op wrappers
28 * should be used instead. It is a macro instead of an static inline function as
29 * some of the types over overloaded (timeout is used for nr_requeue for
30 * example).
31 *
32 * These argument descriptions are the defaults for all
33 * like-named arguments in the following wrappers except where noted below.
34 */
35#define futex(uaddr, op, val, timeout, uaddr2, val3, opflags) \
36 syscall(SYS_futex, uaddr, op | opflags, val, timeout, uaddr2, val3)
37
38/**
39 * futex_wait() - block on uaddr with optional timeout
40 * @timeout: relative timeout
41 */
42static inline int
43futex_wait(u_int32_t *uaddr, u_int32_t val, struct timespec *timeout, int opflags)
44{
45 return futex(uaddr, FUTEX_WAIT, val, timeout, NULL, 0, opflags);
46}
47
48/**
49 * futex_wake() - wake one or more tasks blocked on uaddr
50 * @nr_wake: wake up to this many tasks
51 */
52static inline int
53futex_wake(u_int32_t *uaddr, int nr_wake, int opflags)
54{
55 return futex(uaddr, FUTEX_WAKE, nr_wake, NULL, NULL, 0, opflags);
56}
57
58/**
59* futex_cmp_requeue() - requeue tasks from uaddr to uaddr2
60* @nr_wake: wake up to this many tasks
61* @nr_requeue: requeue up to this many tasks
62*/
63static inline int
64futex_cmp_requeue(u_int32_t *uaddr, u_int32_t val, u_int32_t *uaddr2, int nr_wake,
65 int nr_requeue, int opflags)
66{
67 return futex(uaddr, FUTEX_CMP_REQUEUE, nr_wake, nr_requeue, uaddr2,
68 val, opflags);
69}
70
71#endif /* _FUTEX_H */
generated by cgit v1.2.2 (git 2.25.0) at 2025-08-28 00:04:44 -0400