[PATCH v2 0/2] perf stat: Add interval-count and time support

[PATCH v2 0/2] perf stat: Add interval-count and time support

[ufo19890607]

From: yuzhoujian <yuzhoujian@didichuxing.com>

Introduce two new options for perf stat and update perf-stat documentation
accordingly.

The interval-count option can be used to print counts for fixed number of
times, and it should be used specifically with "-I" option.

Show below is the output of the interval-count option for perf stat.

        $ perf stat -I 1000 --interval-count 2 -e cycles -a
        #           time             counts unit events
             1.002827089         93,884,870      cycles
             2.004231506         56,573,446      cycles
        
The time option can be used to print counts after a period of time, and it
should not used with "-I" option.

Show below is the output of the time option for perf stat.

	$ perf stat --time 2000 -e cycles -a
        Performance counter stats for 'system wide':

                157,260,423      cycles

                2.003060766 seconds time elapsed

yuzhoujian (2):
Patch 1: Add support to print counts for fixed times
Patch 2: Add support to print counts after a period of time

Changes since v1: 
- change the name of the new option "times-print" to "interval-count".
- keep the interval-count option interval specifically.

 tools/perf/Documentation/perf-stat.txt | 10 +++++++
 tools/perf/builtin-stat.c              | 53 ++++++++++++++++++++++++++++++++--
 tools/perf/util/stat.h                 |  2 ++
 3 files changed, 62 insertions(+), 3 deletions(-)

-- 
2.14.1

[PATCH v2 1/2] perf stat: Add support to print counts for fixed times

[ufo19890607]

From: yuzhoujian <yuzhoujian@didichuxing.com>

Introduce a new option to print counts for fixed number of times
and update perf-stat documentation accordingly.

Show below is the output of the new option for perf stat.

        $ perf stat -I 1000 --interval-count 2 -e cycles -a
        #           time             counts unit events
             1.002827089         93,884,870      cycles
             2.004231506         56,573,446      cycles

We can just print the counts for several times with this newly introduced
option. The usage of it is a little like vmstat.

        $ vmstat -n 1 2
procs -----------memory---------- ---swap-- -----io---- -system-- ------cpu-----
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st
 0  0      0 78270544 547484 51732076    0    0     0    20    1    1  1  0 99  0  0
 0  0      0 78270512 547484 51732080    0    0     0    16  477 1555  0  0 100  0  0

Changes since v1:
- change the name of the new option "times-print" to "interval-count".
- keep the new option interval specifically.

Signed-off-by: yuzhoujian <yuzhoujian@didichuxing.com>
---
 tools/perf/Documentation/perf-stat.txt |  5 +++++
 tools/perf/builtin-stat.c              | 26 +++++++++++++++++++++++++-
 tools/perf/util/stat.h                 |  1 +
 3 files changed, 31 insertions(+), 1 deletion(-)

diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
index 823fce7674bb..47a21645f60c 100644
--- a/tools/perf/Documentation/perf-stat.txt
+++ b/tools/perf/Documentation/perf-stat.txt
@@ -146,6 +146,11 @@ Print count deltas every N milliseconds (minimum: 10ms)
 The overhead percentage could be high in some cases, for instance with small, sub 100ms intervals.  Use with caution.
 	example: 'perf stat -I 1000 -e cycles -a sleep 5'
 
+--interval-count times::
+Print count deltas for fixed number of times.
+This option should be used together with "-I" option.
+	example: 'perf stat -I 1000 --interval-count 2 -e cycles -a'
+
 --metric-only::
 Only print computed metrics. Print them in a single line.
 Don't show any raw values. Not supported with --per-thread.
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 98bf9d32f222..406f546ad74c 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -168,6 +168,7 @@ static struct timespec		ref_time;
 static struct cpu_map		*aggr_map;
 static aggr_get_id_t		aggr_get_id;
 static bool			append_file;
+static bool			interval_count;
 static const char		*output_name;
 static int			output_fd;
 static int			print_free_counters_hint;
@@ -571,6 +572,7 @@ static struct perf_evsel *perf_evsel__reset_weak_group(struct perf_evsel *evsel)
 static int __run_perf_stat(int argc, const char **argv)
 {
 	int interval = stat_config.interval;
+	int times = stat_config.times;
 	char msg[BUFSIZ];
 	unsigned long long t0, t1;
 	struct perf_evsel *counter;
@@ -700,6 +702,10 @@ static int __run_perf_stat(int argc, const char **argv)
 			while (!waitpid(child_pid, &status, WNOHANG)) {
 				nanosleep(&ts, NULL);
 				process_interval();
+				if (interval_count == true) {
+					if (--times == 0)
+						break;
+				}
 			}
 		}
 		waitpid(child_pid, &status, 0);
@@ -716,8 +722,13 @@ static int __run_perf_stat(int argc, const char **argv)
 		enable_counters();
 		while (!done) {
 			nanosleep(&ts, NULL);
-			if (interval)
+			if (interval) {
 				process_interval();
+				if (interval_count == true) {
+					if (--times == 0)
+						break;
+				}
+			}
 		}
 	}
 
@@ -1891,6 +1902,8 @@ static const struct option stat_options[] = {
 			"command to run after to the measured command"),
 	OPT_UINTEGER('I', "interval-print", &stat_config.interval,
 		    "print counts at regular interval in ms (>= 10)"),
+	OPT_INTEGER(0, "interval-count", &stat_config.times,
+		    "print counts for fixed number of times"),
 	OPT_SET_UINT(0, "per-socket", &stat_config.aggr_mode,
 		     "aggregate counts per processor socket", AGGR_SOCKET),
 	OPT_SET_UINT(0, "per-core", &stat_config.aggr_mode,
@@ -2689,6 +2702,7 @@ int cmd_stat(int argc, const char **argv)
 	const char *mode;
 	FILE *output = stderr;
 	unsigned int interval;
+	int times;
 	const char * const stat_subcommands[] = { "record", "report" };
 
 	setlocale(LC_ALL, "");
@@ -2719,6 +2733,7 @@ int cmd_stat(int argc, const char **argv)
 		return __cmd_report(argc, argv);
 
 	interval = stat_config.interval;
+	times = stat_config.times;
 
 	/*
 	 * For record command the -o is already taken care of.
@@ -2870,6 +2885,15 @@ int cmd_stat(int argc, const char **argv)
 				   "The overhead percentage could be high in some cases. "
 				   "Please proceed with caution.\n");
 	}
+	if (times && interval)
+		interval_count = true;
+	else if (times && !interval) {
+		pr_err("interval-count option should be used together with "
+				"interval-print.\n");
+		parse_options_usage(stat_usage, stat_options, "interval-count", 0);
+		parse_options_usage(stat_usage, stat_options, "I", 1);
+		goto out;
+	}
 
 	if (perf_evlist__alloc_stats(evsel_list, interval))
 		goto out;
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index dbc6f7134f61..b2f8a347d358 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -90,6 +90,7 @@ struct perf_stat_config {
 	bool		scale;
 	FILE		*output;
 	unsigned int	interval;
+	int				times;
 	struct runtime_stat *stats;
 	int		stats_num;
 };
-- 
2.14.1

[PATCH v2 2/2] perf stat: Add support to print counts after a period of time

[ufo19890607]

From: yuzhoujian <yuzhoujian@didichuxing.com>

Introduce a new option to print counts after N milliseconds
and update perf-stat documentation accordingly.

Show below is the output of the new option for perf stat.

        $ perf stat --time 2000 -e cycles -a
        Performance counter stats for 'system wide':

                157,260,423      cycles

                2.003060766 seconds time elapsed

We can print the count deltas after N milliseconds with this new
introduced option. This option is not supported with "-I" option.

Changes since v1:
- none.

Signed-off-by: yuzhoujian <yuzhoujian@didichuxing.com>
---
 tools/perf/Documentation/perf-stat.txt |  7 ++++++-
 tools/perf/builtin-stat.c              | 29 ++++++++++++++++++++++++++---
 tools/perf/util/stat.h                 |  1 +
 3 files changed, 33 insertions(+), 4 deletions(-)

diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
index 47a21645f60c..8270a0708ea2 100644
--- a/tools/perf/Documentation/perf-stat.txt
+++ b/tools/perf/Documentation/perf-stat.txt
@@ -149,7 +149,12 @@ The overhead percentage could be high in some cases, for instance with small, su
 --interval-count times::
 Print count deltas for fixed number of times.
 This option should be used together with "-I" option.
-	example: 'perf stat -I 1000 --interval-count 2 -e cycles -a'
+	example: 'perf stat -I 1000 --times-print 2 -e cycles -a'
+
+--time msecs::
+Print count deltas after N milliseconds (minimum: 10 ms).
+This option is not supported with "-I" option.
+	example: 'perf stat --time 2000 -e cycles -a'
 
 --metric-only::
 Only print computed metrics. Print them in a single line.
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 406f546ad74c..427f06dc35cc 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -573,6 +573,7 @@ static int __run_perf_stat(int argc, const char **argv)
 {
 	int interval = stat_config.interval;
 	int times = stat_config.times;
+	int time = stat_config.time;
 	char msg[BUFSIZ];
 	unsigned long long t0, t1;
 	struct perf_evsel *counter;
@@ -586,6 +587,9 @@ static int __run_perf_stat(int argc, const char **argv)
 	if (interval) {
 		ts.tv_sec  = interval / USEC_PER_MSEC;
 		ts.tv_nsec = (interval % USEC_PER_MSEC) * NSEC_PER_MSEC;
+	} else if (time) {
+		ts.tv_sec  = time / USEC_PER_MSEC;
+		ts.tv_nsec = (time % USEC_PER_MSEC) * NSEC_PER_MSEC;
 	} else {
 		ts.tv_sec  = 1;
 		ts.tv_nsec = 0;
@@ -698,9 +702,11 @@ static int __run_perf_stat(int argc, const char **argv)
 		perf_evlist__start_workload(evsel_list);
 		enable_counters();
 
-		if (interval) {
+		if (interval || time) {
 			while (!waitpid(child_pid, &status, WNOHANG)) {
 				nanosleep(&ts, NULL);
+				if (time)
+					break;
 				process_interval();
 				if (interval_count == true) {
 					if (--times == 0)
@@ -722,7 +728,9 @@ static int __run_perf_stat(int argc, const char **argv)
 		enable_counters();
 		while (!done) {
 			nanosleep(&ts, NULL);
-			if (interval) {
+			if (interval || time) {
+				if (time)
+					break;
 				process_interval();
 				if (interval_count == true) {
 					if (--times == 0)
@@ -1904,6 +1912,8 @@ static const struct option stat_options[] = {
 		    "print counts at regular interval in ms (>= 10)"),
 	OPT_INTEGER(0, "interval-count", &stat_config.times,
 		    "print counts for fixed number of times"),
+	OPT_UINTEGER(0, "time", &stat_config.time,
+		    "print counts after a period of time in ms (>= 10)"),
 	OPT_SET_UINT(0, "per-socket", &stat_config.aggr_mode,
 		     "aggregate counts per processor socket", AGGR_SOCKET),
 	OPT_SET_UINT(0, "per-core", &stat_config.aggr_mode,
@@ -2701,7 +2711,7 @@ int cmd_stat(int argc, const char **argv)
 	int status = -EINVAL, run_idx;
 	const char *mode;
 	FILE *output = stderr;
-	unsigned int interval;
+	unsigned int interval, time;
 	int times;
 	const char * const stat_subcommands[] = { "record", "report" };
 
@@ -2734,6 +2744,7 @@ int cmd_stat(int argc, const char **argv)
 
 	interval = stat_config.interval;
 	times = stat_config.times;
+	time = stat_config.time;
 
 	/*
 	 * For record command the -o is already taken care of.
@@ -2885,6 +2896,7 @@ int cmd_stat(int argc, const char **argv)
 				   "The overhead percentage could be high in some cases. "
 				   "Please proceed with caution.\n");
 	}
+
 	if (times && interval)
 		interval_count = true;
 	else if (times && !interval) {
@@ -2895,6 +2907,17 @@ int cmd_stat(int argc, const char **argv)
 		goto out;
 	}
 
+	if (time && time < 10) {
+		pr_err("time must be >= 10ms.\n");
+		parse_options_usage(stat_usage, stat_options, "time", 0);
+		goto out;
+	} else if (time && interval) {
+		pr_err("time option is not supported with interval-print.\n");
+		parse_options_usage(stat_usage, stat_options, "time", 0);
+		parse_options_usage(stat_usage, stat_options, "I", 1);
+		goto out;
+	}
+
 	if (perf_evlist__alloc_stats(evsel_list, interval))
 		goto out;
 
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index b2f8a347d358..6c621951fda8 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -90,6 +90,7 @@ struct perf_stat_config {
 	bool		scale;
 	FILE		*output;
 	unsigned int	interval;
+	unsigned int	time;
 	int				times;
 	struct runtime_stat *stats;
 	int		stats_num;
-- 
2.14.1

Re: [PATCH v2 2/2] perf stat: Add support to print counts after a period of time

[Jiri Olsa]

On Thu, Jan 25, 2018 at 10:10:04AM +0100, ufo19890607 wrote:

SNIP

>  --metric-only::
>  Only print computed metrics. Print them in a single line.
> diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
> index 406f546ad74c..427f06dc35cc 100644
> --- a/tools/perf/builtin-stat.c
> +++ b/tools/perf/builtin-stat.c
> @@ -573,6 +573,7 @@ static int __run_perf_stat(int argc, const char **argv)
>  {
>  	int interval = stat_config.interval;
>  	int times = stat_config.times;
> +	int time = stat_config.time;
>  	char msg[BUFSIZ];
>  	unsigned long long t0, t1;
>  	struct perf_evsel *counter;
> @@ -586,6 +587,9 @@ static int __run_perf_stat(int argc, const char **argv)
>  	if (interval) {
>  		ts.tv_sec  = interval / USEC_PER_MSEC;
>  		ts.tv_nsec = (interval % USEC_PER_MSEC) * NSEC_PER_MSEC;
> +	} else if (time) {
> +		ts.tv_sec  = time / USEC_PER_MSEC;
> +		ts.tv_nsec = (time % USEC_PER_MSEC) * NSEC_PER_MSEC;

I like the idea.. it won't work with -I option, but let's
keep it like that until someone needs it ;-)

>  	} else {
>  		ts.tv_sec  = 1;
>  		ts.tv_nsec = 0;
> @@ -698,9 +702,11 @@ static int __run_perf_stat(int argc, const char **argv)
>  		perf_evlist__start_workload(evsel_list);
>  		enable_counters();
>  
> -		if (interval) {
> +		if (interval || time) {
>  			while (!waitpid(child_pid, &status, WNOHANG)) {
>  				nanosleep(&ts, NULL);
> +				if (time)
> +					break;
>  				process_interval();
>  				if (interval_count == true) {
>  					if (--times == 0)
> @@ -722,7 +728,9 @@ static int __run_perf_stat(int argc, const char **argv)
>  		enable_counters();
>  		while (!done) {
>  			nanosleep(&ts, NULL);
> -			if (interval) {
> +			if (interval || time) {
> +				if (time)
> +					break;

you can put the time check with break directly after nanosleep
to keep some consistency with the workload case


>  				process_interval();
>  				if (interval_count == true) {
>  					if (--times == 0)
> @@ -1904,6 +1912,8 @@ static const struct option stat_options[] = {
>  		    "print counts at regular interval in ms (>= 10)"),
>  	OPT_INTEGER(0, "interval-count", &stat_config.times,
>  		    "print counts for fixed number of times"),
> +	OPT_UINTEGER(0, "time", &stat_config.time,
> +		    "print counts after a period of time in ms (>= 10)"),
>  	OPT_SET_UINT(0, "per-socket", &stat_config.aggr_mode,
>  		     "aggregate counts per processor socket", AGGR_SOCKET),
>  	OPT_SET_UINT(0, "per-core", &stat_config.aggr_mode,
> @@ -2701,7 +2711,7 @@ int cmd_stat(int argc, const char **argv)
>  	int status = -EINVAL, run_idx;
>  	const char *mode;
>  	FILE *output = stderr;
> -	unsigned int interval;
> +	unsigned int interval, time;
>  	int times;
>  	const char * const stat_subcommands[] = { "record", "report" };
>  
> @@ -2734,6 +2744,7 @@ int cmd_stat(int argc, const char **argv)
>  
>  	interval = stat_config.interval;
>  	times = stat_config.times;
> +	time = stat_config.time;
>  
>  	/*
>  	 * For record command the -o is already taken care of.
> @@ -2885,6 +2896,7 @@ int cmd_stat(int argc, const char **argv)
>  				   "The overhead percentage could be high in some cases. "
>  				   "Please proceed with caution.\n");
>  	}
> +
>  	if (times && interval)
>  		interval_count = true;
>  	else if (times && !interval) {
> @@ -2895,6 +2907,17 @@ int cmd_stat(int argc, const char **argv)
>  		goto out;
>  	}
>  
> +	if (time && time < 10) {
> +		pr_err("time must be >= 10ms.\n");

what is this limitation for? please mentioned
that in comment and changelog

thanks,
jirka

Re: [PATCH v2 2/2] perf stat: Add support to print counts after a period of time

[Jiri Olsa]

On Fri, Jan 26, 2018 at 09:59:02AM +0800, 禹舟键 wrote:
> 2018-01-25 20:27 GMT+08:00 Jiri Olsa <jolsa@redhat.com>:
> 
> > On Thu, Jan 25, 2018 at 10:10:04AM +0100, ufo19890607 wrote:
> >
> > SNIP
> >
> > >  --metric-only::
> > >  Only print computed metrics. Print them in a single line.
> > > diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
> > > index 406f546ad74c..427f06dc35cc 100644
> > > --- a/tools/perf/builtin-stat.c
> > > +++ b/tools/perf/builtin-stat.c
> > > @@ -573,6 +573,7 @@ static int __run_perf_stat(int argc, const char
> > **argv)
> > >  {
> > >       int interval = stat_config.interval;
> > >       int times = stat_config.times;
> > > +     int time = stat_config.time;
> > >       char msg[BUFSIZ];
> > >       unsigned long long t0, t1;
> > >       struct perf_evsel *counter;
> > > @@ -586,6 +587,9 @@ static int __run_perf_stat(int argc, const char
> > **argv)
> > >       if (interval) {
> > >               ts.tv_sec  = interval / USEC_PER_MSEC;
> > >               ts.tv_nsec = (interval % USEC_PER_MSEC) * NSEC_PER_MSEC;
> > > +     } else if (time) {
> > > +             ts.tv_sec  = time / USEC_PER_MSEC;
> > > +             ts.tv_nsec = (time % USEC_PER_MSEC) * NSEC_PER_MSEC;
> >
> > I like the idea.. it won't work with -I option, but let's
> > keep it like that until someone needs it ;-)
> >
> >
> Hi Jirka
> 
> What do you mean by this sentence？？？

that I'm ok with the patch with small fixes I mentioned

jirka