perf bench: add benchmark for evlist open/close operations
PATCH 1/1 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)
--- tools/perf/bench/Build | 1 + tools/perf/bench/bench.h | 1 + tools/perf/bench/evlist-open-close.c | 275 +++++++++++++++++++++++++++ tools/perf/builtin-bench.c | 1 + 4 files changed, 278 insertions(+) create mode 100644 tools/perf/bench/evlist-open-close.c
tools/perf/bench/Build +1 -0
@@ -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
tools/perf/bench/bench.h +1 -0
@@ -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
tools/perf/bench/evlist-open-close.c new file +275 -0
275 lines not on this page · raw message
tools/perf/builtin-bench.c +1 -0
@@ -88,6 +88,7 @@ static struct bench internals_benchmarks[] = { { "synthesize", "Benchmark perf event synthesis", bench_synthesize }, { "kallsyms-parse", "Benchmark kallsyms parsing", bench_kallsyms_parse }, { "inject-build-id", "Benchmark build-id injection", bench_inject_build_id }, + { "evlist-open-close", "Benchmark evlist open and close", bench_evlist_open_close }, { NULL, NULL, NULL } };
Signature · 1 line
2.31.1
Arnaldo Carvalho de Melo
maintainerEm Mon, Aug 09, 2021 at 10:11:02PM +0200, Riccardo Mancini escreveu:
Quoted 158 lines
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)Signed-off-by Riccardo Mancini <[email protected]>--- tools/perf/bench/Build | 1 + tools/perf/bench/bench.h | 1 + tools/perf/bench/evlist-open-close.c | 275 +++++++++++++++++++++++++++ tools/perf/builtin-bench.c | 1 + 4 files changed, 278 insertions(+) create mode 100644 tools/perf/bench/evlist-open-close.ctools/perf/bench/Build +1
-0@@ -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.otools/perf/bench/bench.h +1
-0@@ -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 0tools/perf/bench/evlist-open-close.c new file +84
-0@@ -0,0 +1,275 @@ +// 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 evlist *evlist;
tools/perf/bench/evlist-open-close.c line 94
Quoted 10 lines
tools/perf/bench/evlist-open-close.c new file +10
-0+struct parse_events_error err; +int ret; + + evlist =evlist__new (); +if (!evlist) { +pr_err ("Not enough memory to create evlist\n" ); +return NULL ; + } + +bzero (&err,sizeof (err));
man bzero
The bzero() function is deprecated (marked as LEGACY in POSIX.1-2001); use memset(3) in new programs. POSIX.1-2008 removes the specification of bzero(). The bzero() function first appeared in 4.3BSD.
I'm replacing it with a memset().
Quoted 196 lines
tools/perf/bench/evlist-open-close.c new file +181
-0+ 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) +{ +int err =-1 ; +char sbuf[STRERR_BUFSIZE ]; + + 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) +{ +struct evlist *evlist; +struct timeval start, end, diff; +double time_average, time_stddev; +int i; +int err; +u64 runtime_us; +struct stats time_stats; + +init_stats (&time_stats); + +// used to print statistics only + evlist =bench__create_evlist (evstr); + +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); +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) +{ +int err, i, final_size, str_size; +struct strbuf buf; +char sbuf[STRERR_BUFSIZE ]; + + 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) +{ +int err =0 ; +char *evstr, errbuf[BUFSIZ ]; + + 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,BUFSIZ ); +pr_err ("%s\n" , errbuf); + goto out; + } + + err =target__parse_uid (&opts.target); +if (err) { +target__strerror (&opts.target, err, errbuf,BUFSIZ ); +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; +}tools/perf/builtin-bench.c +1
-0@@ -88,6 +88,7 @@ static struct bench internals_benchmarks[] = { { "synthesize", "Benchmark perf event synthesis", bench_synthesize }, { "kallsyms-parse", "Benchmark kallsyms parsing", bench_kallsyms_parse }, { "inject-build-id", "Benchmark build-id injection", bench_inject_build_id }, + { "evlist-open-close", "Benchmark evlist open and close", bench_evlist_open_close }, { NULL, NULL, NULL } };Signature · 1 line
2.31.1
Signature · 1 line
- Arnaldo
Arnaldo Carvalho de Melo
maintainerEm Mon, Aug 09, 2021 at 05:23:31PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 13 lines
Em Mon, Aug 09, 2021 at 10:11:02PM +0200, Riccardo Mancini escreveu:
Quoted 11 lines
tools/perf/bench/evlist-open-close.c new file +10
-0+static struct evlist *bench__create_evlist (char *evstr) +{ +struct evlist *evlist; +struct parse_events_error err; +int ret; + evlist =evlist__new (); +if (!evlist) { +pr_err ("Not enough memory to create evlist\n" ); +return NULL ; + }
Quoted 7 lines
tools/perf/bench/evlist-open-close.c line 94
Quoted 1 line
tools/perf/bench/evlist-open-close.c new file +1
-0+bzero (&err,sizeof (err));man bzero
The
bzero()function is deprecated (marked as LEGACY in POSIX.1-2001); use memset(3) in new programs. POSIX.1-2008 removes the specification ofbzero(). Thebzero()function first appeared in 4.3BSD.I'm replacing it with a
memset().
This one is also equivalent:
tools/perf/tests/pmu-events.c: struct parse_events_error error = { .idx = 0, };
https://gcc.gnu.org/onlinedocs/gcc/Designated-Inits.html
That text is a bit roundabout, as it says that the members that are not explicitely initialized will be initialized as variables with static storage duration, i.e. zeroed.
- Arnaldo
Hi Arnaldo,
Quoted 21 lines · Arnaldo Carvalho de Melo
Em Mon, Aug 09, 2021 at 05:23:31PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 20 lines
Em Mon, Aug 09, 2021 at 10:11:02PM +0200, Riccardo Mancini escreveu:
Review
Quoted 13 lines
(quoted diff) +11
-0+static struct evlist *bench__create_evlist(char *evstr) +{ + struct evlist *evlist; + struct parse_events_error err; + int ret; + evlist = evlist__new(); + if (!evlist) { + pr_err("Not enough memory to create evlist\n"); + return NULL; + } + bzero(&err, sizeof(err));man bzero
The
bzero()function is deprecated (marked as LEGACY in POSIX.1-2001); use memset(3) in new programs. POSIX.1-2008 removes the specification of bzero(). Thebzero()function first appeared in 4.3BSD.
Oops, I didn't know, but I saw it is being used in some parts in perf, maybe we should get rid of them:
$ rg -c bzero builtin-lock.c:1 arch/powerpc/util/kvm-stat.c:1 builtin-stat.c:1 builtin-trace.c:2 bench/evlist-open-close.c:1 bench/numa.c:5 tests/parse-events.c:1 tests/backward-ring-buffer.c:1 tests/bpf.c:2 util/metricgroup.c:1 util/parse-events.c:1
Quoted 5 lines
Quoted 2 lines
I'm replacing it with a
memset().This one is also equivalent:
Quoted 7 lines
tools/perf/tests/pmu-events.c: struct
parse_events_errorerror = { .idx = 0, };https://gcc.gnu.org/onlinedocs/gcc/Designated-Inits.html
That text is a bit roundabout, as it says that the members that are not explicitely initialized will be initialized as variables with static storage duration, i.e. zeroed.
Would it be the same doing the shorter {0}. It would be a general solution for these init-to-zero cases.
Unrelated to this small issue, I noticed I forgot to check the return of bench__create_evlist. Would you like me to send a v2 fixing both issues or are you able to apply this other small change yourself?
tools/perf/bench/evlist-open-close.c +4 -0
@@ -168,7 +168,11 @@ static int bench_evlist_open_close__run(char *evstr)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) {
Thanks,
Riccardo
Quoted 2 lines
- Arnaldo
Arnaldo Carvalho de Melo
maintainerEm Tue, Aug 10, 2021 at 12:31:55PM +0200, Riccardo Mancini escreveu:
Quoted 27 lines
Hi Arnaldo,
Quoted 21 lines · Arnaldo Carvalho de Melo
Em Mon, Aug 09, 2021 at 05:23:31PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 20 lines
Em Mon, Aug 09, 2021 at 10:11:02PM +0200, Riccardo Mancini escreveu:
Review
Quoted 13 lines
(quoted diff) +11
-0+static struct evlist *bench__create_evlist(char *evstr) +{ + struct evlist *evlist; + struct parse_events_error err; + int ret; + evlist = evlist__new(); + if (!evlist) { + pr_err("Not enough memory to create evlist\n"); + return NULL; + } + bzero(&err, sizeof(err));man bzero
The
bzero()function is deprecated (marked as LEGACY in POSIX.1-2001); use memset(3) in new programs. POSIX.1-2008 removes the specification of bzero(). Thebzero()function first appeared in 4.3BSD.Oops, I didn't know, but I saw it is being used in some parts in perf, maybe we should get rid of them:
Quoted 12 lines
$ rg -c bzero builtin-lock.c:1 arch/powerpc/util/kvm-stat.c:1 builtin-stat.c:1 builtin-trace.c:2 bench/evlist-open-close.c:1 bench/numa.c:5 tests/parse-events.c:1 tests/backward-ring-buffer.c:1 tests/bpf.c:2 util/metricgroup.c:1 util/parse-events.c:1
Yeah, patches are welcome, but at least lets not add new ones :-)
Quoted 6 lines
Quoted 6 lines
Quoted 1 line
I'm replacing it with a
memset().This one is also equivalent:
tools/perf/tests/pmu-events.c: struct
parse_events_errorerror = { .idx = 0, };
Quoted 8 lines
Quoted 5 lines
https://gcc.gnu.org/onlinedocs/gcc/Designated-Inits.html
That text is a bit roundabout, as it says that the members that are not explicitely initialized will be initialized as variables with static storage duration, i.e. zeroed.
Would it be the same doing the shorter {0}. It would be a general solution for these init-to-zero cases.
I'd have to do some extra research to remember why that is also not optimal, IIRC the '= { .a = 0, };' is the optimal one.
Quoted 3 lines
Unrelated to this small issue, I noticed I forgot to check the return of
bench__create_evlist. Would you like me to send a v2 fixing both issues or are you able to apply this other small change yourself?
Nah, as this is the HEAD right now in my local branch, I'll apply it myself, thanks!
- Arnaldo
Quoted 24 lines
tools/perf/bench/evlist-open-close.c +4
-0@@ -168,7 +168,11 @@ static int bench_evlist_open_close__run(char *evstr)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) {Thanks,
RiccardoQuoted 2 lines
- Arnaldo
Signature · 1 line
- Arnaldo
Arnaldo Carvalho de Melo
maintainerEm Tue, Aug 10, 2021 at 11:03:22AM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 7 lines
Em Tue, Aug 10, 2021 at 12:31:55PM +0200, Riccardo Mancini escreveu:
Quoted 3 lines
Unrelated to this small issue, I noticed I forgot to check the return of
bench__create_evlist. Would you like me to send a v2 fixing both issues or are you able to apply this other small change yourself?Nah, as this is the HEAD right now in my local branch, I'll apply it myself, thanks!
Also you forgot another check for this same function (bench__create_evlist()), before that loop. I'm fixing that as well.
Quoted 17 lines
- Arnaldo
Quoted 15 lines
tools/perf/bench/evlist-open-close.c +4
-0@@ -168,7 +168,11 @@ static int bench_evlist_open_close__run(char *evstr)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);