diff options
author | Riccardo Mancini <rickyman7@gmail.com> | 2021-08-09 22:11:02 +0200 |
---|---|---|
committer | Arnaldo Carvalho de Melo <acme@redhat.com> | 2021-08-10 11:32:37 -0300 |
commit | 4241eabf59d5b7e9b5b567526f6e319f81dff894 (patch) | |
tree | fcf5d2ee9a848a4450505aaa4059d270f1abc31d /tools/perf/bench | |
parent | f2c24ebadd90b4165ce2c49c3c6a100312266f54 (diff) | |
download | linux-stable-4241eabf59d5b7e9b5b567526f6e319f81dff894.tar.gz linux-stable-4241eabf59d5b7e9b5b567526f6e319f81dff894.tar.bz2 linux-stable-4241eabf59d5b7e9b5b567526f6e319f81dff894.zip |
perf bench: Add benchmark for evlist open/close operations
This new benchmark finds the total time that is taken to open, mmap,
enable, disable, munmap, close an evlist (time taken for new,
create_maps, config, delete is not counted in).
The evlist can be configured as in perf-record using the
-a,-C,-e,-u,--per-thread,-t,-p options.
The events can be duplicated in the evlist to quickly test performance
with many events using the -n options.
Furthermore, also the number of iterations used to calculate the
statistics is customizable.
Examples:
- Open one dummy event system-wide:
$ sudo ./perf bench internals evlist-open-close
Number of cpus: 4
Number of threads: 1
Number of events: 1 (4 fds)
Number of iterations: 100
Average open-close took: 613.870 usec (+- 32.852 usec)
- Open the group '{cs,cycles}' on CPU 0
$ sudo ./perf bench internals evlist-open-close -e '{cs,cycles}' -C 0
Number of cpus: 1
Number of threads: 1
Number of events: 2 (2 fds)
Number of iterations: 100
Average open-close took: 8503.220 usec (+- 252.652 usec)
- Open 10 'cycles' events for user 0, calculate average over 100 runs
$ sudo ./perf bench internals evlist-open-close -e cycles -n 10 -u 0 -i 100
Number of cpus: 4
Number of threads: 328
Number of events: 10 (13120 fds)
Number of iterations: 100
Average open-close took: 180043.140 usec (+- 2295.889 usec)
Committer notes:
Replaced a deprecated bzero() call with designated initialized zeroing.
Added some missing evlist allocation checks, one noted by Riccardo on
the mailing list.
Minor cosmetic changes (sent in private).
Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/20210809201101.277594-1-rickyman7@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Diffstat (limited to 'tools/perf/bench')
-rw-r--r-- | tools/perf/bench/Build | 1 | ||||
-rw-r--r-- | tools/perf/bench/bench.h | 1 | ||||
-rw-r--r-- | tools/perf/bench/evlist-open-close.c | 257 |
3 files changed, 259 insertions, 0 deletions
diff --git a/tools/perf/bench/Build b/tools/perf/bench/Build index e43f46931b41..61d45fcb4057 100644 --- a/tools/perf/bench/Build +++ b/tools/perf/bench/Build @@ -13,6 +13,7 @@ perf-y += synthesize.o perf-y += kallsyms-parse.o perf-y += find-bit-bench.o perf-y += inject-buildid.o +perf-y += evlist-open-close.o perf-$(CONFIG_X86_64) += mem-memcpy-x86-64-asm.o perf-$(CONFIG_X86_64) += mem-memset-x86-64-asm.o diff --git a/tools/perf/bench/bench.h b/tools/perf/bench/bench.h index eac36afab2b3..b3480bc33fe8 100644 --- a/tools/perf/bench/bench.h +++ b/tools/perf/bench/bench.h @@ -48,6 +48,7 @@ int bench_epoll_ctl(int argc, const char **argv); int bench_synthesize(int argc, const char **argv); int bench_kallsyms_parse(int argc, const char **argv); int bench_inject_build_id(int argc, const char **argv); +int bench_evlist_open_close(int argc, const char **argv); #define BENCH_FORMAT_DEFAULT_STR "default" #define BENCH_FORMAT_DEFAULT 0 diff --git a/tools/perf/bench/evlist-open-close.c b/tools/perf/bench/evlist-open-close.c new file mode 100644 index 000000000000..674cb14cbaa9 --- /dev/null +++ b/tools/perf/bench/evlist-open-close.c @@ -0,0 +1,257 @@ +// SPDX-License-Identifier: GPL-2.0 +#include <stdio.h> +#include <stdlib.h> +#include <limits.h> +#include "bench.h" +#include "../util/debug.h" +#include "../util/stat.h" +#include "../util/evlist.h" +#include "../util/evsel.h" +#include "../util/strbuf.h" +#include "../util/record.h" +#include "../util/parse-events.h" +#include "internal/threadmap.h" +#include "internal/cpumap.h" +#include <linux/perf_event.h> +#include <linux/kernel.h> +#include <linux/time64.h> +#include <linux/string.h> +#include <subcmd/parse-options.h> + +#define MMAP_FLUSH_DEFAULT 1 + +static int iterations = 100; +static int nr_events = 1; +static const char *event_string = "dummy"; + +static struct record_opts opts = { + .sample_time = true, + .mmap_pages = UINT_MAX, + .user_freq = UINT_MAX, + .user_interval = ULLONG_MAX, + .freq = 4000, + .target = { + .uses_mmap = true, + .default_per_cpu = true, + }, + .mmap_flush = MMAP_FLUSH_DEFAULT, + .nr_threads_synthesize = 1, + .ctl_fd = -1, + .ctl_fd_ack = -1, +}; + +static const struct option options[] = { + OPT_STRING('e', "event", &event_string, "event", "event selector. use 'perf list' to list available events"), + OPT_INTEGER('n', "nr-events", &nr_events, + "number of dummy events to create (default 1). If used with -e, it clones those events n times (1 = no change)"), + OPT_INTEGER('i', "iterations", &iterations, "Number of iterations used to compute average (default=100)"), + OPT_BOOLEAN('a', "all-cpus", &opts.target.system_wide, "system-wide collection from all CPUs"), + OPT_STRING('C', "cpu", &opts.target.cpu_list, "cpu", "list of cpus where to open events"), + OPT_STRING('p', "pid", &opts.target.pid, "pid", "record events on existing process id"), + OPT_STRING('t', "tid", &opts.target.tid, "tid", "record events on existing thread id"), + OPT_STRING('u', "uid", &opts.target.uid_str, "user", "user to profile"), + OPT_BOOLEAN(0, "per-thread", &opts.target.per_thread, "use per-thread mmaps"), + OPT_END() +}; + +static const char *const bench_usage[] = { + "perf bench internals evlist-open-close <options>", + NULL +}; + +static int evlist__count_evsel_fds(struct evlist *evlist) +{ + struct evsel *evsel; + int cnt = 0; + + evlist__for_each_entry(evlist, evsel) + cnt += evsel->core.threads->nr * evsel->core.cpus->nr; + + return cnt; +} + +static struct evlist *bench__create_evlist(char *evstr) +{ + struct parse_events_error err = { .idx = 0, }; + struct evlist *evlist = evlist__new(); + int ret; + + if (!evlist) { + pr_err("Not enough memory to create evlist\n"); + return NULL; + } + + ret = parse_events(evlist, evstr, &err); + if (ret) { + parse_events_print_error(&err, evstr); + pr_err("Run 'perf list' for a list of valid events\n"); + ret = 1; + goto out_delete_evlist; + } + + ret = evlist__create_maps(evlist, &opts.target); + if (ret < 0) { + pr_err("Not enough memory to create thread/cpu maps\n"); + goto out_delete_evlist; + } + + evlist__config(evlist, &opts, NULL); + + return evlist; + +out_delete_evlist: + evlist__delete(evlist); + return NULL; +} + +static int bench__do_evlist_open_close(struct evlist *evlist) +{ + char sbuf[STRERR_BUFSIZE]; + int err = evlist__open(evlist); + + if (err < 0) { + pr_err("evlist__open: %s\n", str_error_r(errno, sbuf, sizeof(sbuf))); + return err; + } + + err = evlist__mmap(evlist, opts.mmap_pages); + if (err < 0) { + pr_err("evlist__mmap: %s\n", str_error_r(errno, sbuf, sizeof(sbuf))); + return err; + } + + evlist__enable(evlist); + evlist__disable(evlist); + evlist__munmap(evlist); + evlist__close(evlist); + + return 0; +} + +static int bench_evlist_open_close__run(char *evstr) +{ + // used to print statistics only + struct evlist *evlist = bench__create_evlist(evstr); + double time_average, time_stddev; + struct timeval start, end, diff; + struct stats time_stats; + u64 runtime_us; + int i, err; + + if (!evlist) + return -ENOMEM; + + init_stats(&time_stats); + + printf(" Number of cpus:\t%d\n", evlist->core.cpus->nr); + printf(" Number of threads:\t%d\n", evlist->core.threads->nr); + printf(" Number of events:\t%d (%d fds)\n", + evlist->core.nr_entries, evlist__count_evsel_fds(evlist)); + printf(" Number of iterations:\t%d\n", iterations); + + evlist__delete(evlist); + + for (i = 0; i < iterations; i++) { + pr_debug("Started iteration %d\n", i); + evlist = bench__create_evlist(evstr); + if (!evlist) + return -ENOMEM; + + gettimeofday(&start, NULL); + err = bench__do_evlist_open_close(evlist); + if (err) { + evlist__delete(evlist); + return err; + } + + gettimeofday(&end, NULL); + timersub(&end, &start, &diff); + runtime_us = diff.tv_sec * USEC_PER_SEC + diff.tv_usec; + update_stats(&time_stats, runtime_us); + + evlist__delete(evlist); + pr_debug("Iteration %d took:\t%ldus\n", i, runtime_us); + } + + time_average = avg_stats(&time_stats); + time_stddev = stddev_stats(&time_stats); + printf(" Average open-close took: %.3f usec (+- %.3f usec)\n", time_average, time_stddev); + + return 0; +} + +static char *bench__repeat_event_string(const char *evstr, int n) +{ + char sbuf[STRERR_BUFSIZE]; + struct strbuf buf; + int i, str_size = strlen(evstr), + final_size = str_size * n + n, + err = strbuf_init(&buf, final_size); + + if (err) { + pr_err("strbuf_init: %s\n", str_error_r(err, sbuf, sizeof(sbuf))); + goto out_error; + } + + for (i = 0; i < n; i++) { + err = strbuf_add(&buf, evstr, str_size); + if (err) { + pr_err("strbuf_add: %s\n", str_error_r(err, sbuf, sizeof(sbuf))); + goto out_error; + } + + err = strbuf_addch(&buf, i == n-1 ? '\0' : ','); + if (err) { + pr_err("strbuf_addch: %s\n", str_error_r(err, sbuf, sizeof(sbuf))); + goto out_error; + } + } + + return strbuf_detach(&buf, NULL); + +out_error: + strbuf_release(&buf); + return NULL; +} + + +int bench_evlist_open_close(int argc, const char **argv) +{ + char *evstr, errbuf[BUFSIZ]; + int err; + + argc = parse_options(argc, argv, options, bench_usage, 0); + if (argc) { + usage_with_options(bench_usage, options); + exit(EXIT_FAILURE); + } + + err = target__validate(&opts.target); + if (err) { + target__strerror(&opts.target, err, errbuf, sizeof(errbuf)); + pr_err("%s\n", errbuf); + goto out; + } + + err = target__parse_uid(&opts.target); + if (err) { + target__strerror(&opts.target, err, errbuf, sizeof(errbuf)); + pr_err("%s", errbuf); + goto out; + } + + /* Enable ignoring missing threads when -u/-p option is defined. */ + opts.ignore_missing_thread = opts.target.uid != UINT_MAX || opts.target.pid; + + evstr = bench__repeat_event_string(event_string, nr_events); + if (!evstr) { + err = -ENOMEM; + goto out; + } + + err = bench_evlist_open_close__run(evstr); + + free(evstr); +out: + return err; +} |