[PATCH 1/2] perf tools: Add --quiet option to perf stat

[PATCH 1/2] perf tools: Add --quiet option to perf stat

[Andi Kleen]

Add a new --quiet option to perf stat. This is useful with perf stat
record to write the data only to the perf.data file, which can lower
measurement overhead because the data doesn't need to be formatted.

On my 4C desktop:

% time ./perf stat record  -e $(python -c 'print ",".join(["cycles"]*1000)')  -a -I 1000 sleep 5
...
real    0m5.377s
user    0m0.238s
sys     0m0.452s
% time ./perf stat record --quiet -e $(python -c 'print ",".join(["cycles"]*1000)')  -a -I 1000 sleep 5

real    0m5.452s
user    0m0.183s
sys     0m0.423s

In this example it cuts the user time by 20%. On systems with more cores
the savings are higher.

Signed-off-by: Andi Kleen <andi@firstfloor.org>
---
 tools/perf/Documentation/perf-stat.txt | 4 ++++
 tools/perf/builtin-stat.c              | 6 +++++-
 tools/perf/util/stat.h                 | 1 +
 3 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
index 9f9f29025e49..b138dd192423 100644
--- a/tools/perf/Documentation/perf-stat.txt
+++ b/tools/perf/Documentation/perf-stat.txt
@@ -316,6 +316,10 @@ small group that need not have multiplexing is lowered. This option
 forbids the event merging logic from sharing events between groups and
 may be used to increase accuracy in this case.
 
+--quiet::
+Don't print output. This is useful with perf stat record below to only
+write data to the perf.data file.
+
 STAT RECORD
 -----------
 Stores stat data into perf data file.
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index b01af171d94f..743fe47e7a88 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -972,6 +972,8 @@ static void print_counters(struct timespec *ts, int argc, const char **argv)
 	/* Do not print anything if we record to the pipe. */
 	if (STAT_RECORD && perf_stat.data.is_pipe)
 		return;
+	if (stat_config.quiet)
+		return;
 
 	perf_evlist__print_counters(evsel_list, &stat_config, &target,
 				    ts, argc, argv);
@@ -1171,6 +1173,8 @@ static struct option stat_options[] = {
 		    "threads of same physical core"),
 	OPT_BOOLEAN(0, "summary", &stat_config.summary,
 		       "print summary for interval mode"),
+	OPT_BOOLEAN(0, "quiet", &stat_config.quiet,
+			"don't print output (useful with record)"),
 #ifdef HAVE_LIBPFM
 	OPT_CALLBACK(0, "pfm-events", &evsel_list, "event",
 		"libpfm4 event selector. use 'perf list' to list available events",
@@ -2132,7 +2136,7 @@ int cmd_stat(int argc, const char **argv)
 		goto out;
 	}
 
-	if (!output) {
+	if (!output && !stat_config.quiet) {
 		struct timespec tm;
 		mode = append_file ? "a" : "w";
 
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index 487010c624be..05adf8165025 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -122,6 +122,7 @@ struct perf_stat_config {
 	bool			 metric_no_group;
 	bool			 metric_no_merge;
 	bool			 stop_read_counter;
+	bool			 quiet;
 	FILE			*output;
 	unsigned int		 interval;
 	unsigned int		 timeout;
-- 
2.28.0

[PATCH 2/2] perf tools: Support -x for perf stat report

[Andi Kleen]

Add support for the -x, option to enable CSV output with perf stat
report. Useful to parse the information with other programs.

% perf stat record --quiet -a -I 1000 sleep 5
% perf stat report -x,
     1.000838672,4003.55,msec,cpu-clock,4003548736,100.00,,
     1.000838672,11243,,context-switches,4003631885,100.00,0.003,M/sec
     1.000838672,1682,,cpu-migrations,4003672150,100.00,0.420,K/sec
     1.000838672,13244,,page-faults,4003697471,100.00,0.003,M/sec
     1.000838672,2953214077,,cycles,4003715495,100.00,0.738,GHz
     1.000838672,4380820799,,instructions,4003738270,100.00,1.48,insn per cycle
     1.000838672,809865653,,branches,4003760616,100.00,202.287,M/sec
     1.000838672,12439843,,branch-misses,4003780785,100.00,1.54,of all branches
...

Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
 tools/perf/builtin-stat.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 743fe47e7a88..31e7bd877f1d 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -1988,6 +1988,8 @@ static int __cmd_report(int argc, const char **argv)
 		     "aggregate counts per numa node", AGGR_NODE),
 	OPT_SET_UINT('A', "no-aggr", &perf_stat.aggr_mode,
 		     "disable CPU count aggregation", AGGR_NONE),
+	OPT_STRING('x', "field-separator", &stat_config.csv_sep, "separator",
+		   "print counts with custom separator"),
 	OPT_END()
 	};
 	struct stat st;
@@ -2002,6 +2004,11 @@ static int __cmd_report(int argc, const char **argv)
 			input_name = "perf.data";
 	}
 
+	if (stat_config.csv_sep) {
+		stat_config.csv_output = true;
+		stat_config.big_num = false;
+	}
+
 	perf_stat.data.path = input_name;
 	perf_stat.data.mode = PERF_DATA_MODE_READ;
 
-- 
2.28.0

Re: [PATCH 2/2] perf tools: Support -x for perf stat report

[Jiri Olsa]

On Mon, Oct 26, 2020 at 05:27:37PM -0700, Andi Kleen wrote:
> Add support for the -x, option to enable CSV output with perf stat
> report. Useful to parse the information with other programs.
> 
> % perf stat record --quiet -a -I 1000 sleep 5
> % perf stat report -x,
>      1.000838672,4003.55,msec,cpu-clock,4003548736,100.00,,
>      1.000838672,11243,,context-switches,4003631885,100.00,0.003,M/sec
>      1.000838672,1682,,cpu-migrations,4003672150,100.00,0.420,K/sec
>      1.000838672,13244,,page-faults,4003697471,100.00,0.003,M/sec
>      1.000838672,2953214077,,cycles,4003715495,100.00,0.738,GHz
>      1.000838672,4380820799,,instructions,4003738270,100.00,1.48,insn per cycle
>      1.000838672,809865653,,branches,4003760616,100.00,202.287,M/sec
>      1.000838672,12439843,,branch-misses,4003780785,100.00,1.54,of all branches
> ...
> 
> Signed-off-by: Andi Kleen <ak@linux.intel.com>
> ---
>  tools/perf/builtin-stat.c | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
> index 743fe47e7a88..31e7bd877f1d 100644
> --- a/tools/perf/builtin-stat.c
> +++ b/tools/perf/builtin-stat.c
> @@ -1988,6 +1988,8 @@ static int __cmd_report(int argc, const char **argv)
>  		     "aggregate counts per numa node", AGGR_NODE),
>  	OPT_SET_UINT('A', "no-aggr", &perf_stat.aggr_mode,
>  		     "disable CPU count aggregation", AGGR_NONE),
> +	OPT_STRING('x', "field-separator", &stat_config.csv_sep, "separator",
> +		   "print counts with custom separator"),
>  	OPT_END()
>  	};
>  	struct stat st;
> @@ -2002,6 +2004,11 @@ static int __cmd_report(int argc, const char **argv)
>  			input_name = "perf.data";
>  	}
>  
> +	if (stat_config.csv_sep) {
> +		stat_config.csv_output = true;
> +		stat_config.big_num = false;
> +	}

stat_config.csv_sep is by default ' ' so this will force
the standard output to be with -x

jirka

> +
>  	perf_stat.data.path = input_name;
>  	perf_stat.data.mode = PERF_DATA_MODE_READ;
>  
> -- 
> 2.28.0
> 

Re: [PATCH 1/2] perf tools: Add --quiet option to perf stat

[Arnaldo Carvalho de Melo]

Em Mon, Oct 26, 2020 at 05:27:36PM -0700, Andi Kleen escreveu:
> Add a new --quiet option to perf stat. This is useful with perf stat
> record to write the data only to the perf.data file, which can lower
> measurement overhead because the data doesn't need to be formatted.
> 
> On my 4C desktop:
> 
> % time ./perf stat record  -e $(python -c 'print ",".join(["cycles"]*1000)')  -a -I 1000 sleep 5
> ...
> real    0m5.377s
> user    0m0.238s
> sys     0m0.452s
> % time ./perf stat record --quiet -e $(python -c 'print ",".join(["cycles"]*1000)')  -a -I 1000 sleep 5
> 
> real    0m5.452s
> user    0m0.183s
> sys     0m0.423s
> 
> In this example it cuts the user time by 20%. On systems with more cores
> the savings are higher.

Applied 1/2,

Thanks,

- Arnaldo
 
> Signed-off-by: Andi Kleen <andi@firstfloor.org>
> ---
>  tools/perf/Documentation/perf-stat.txt | 4 ++++
>  tools/perf/builtin-stat.c              | 6 +++++-
>  tools/perf/util/stat.h                 | 1 +
>  3 files changed, 10 insertions(+), 1 deletion(-)
> 
> diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
> index 9f9f29025e49..b138dd192423 100644
> --- a/tools/perf/Documentation/perf-stat.txt
> +++ b/tools/perf/Documentation/perf-stat.txt
> @@ -316,6 +316,10 @@ small group that need not have multiplexing is lowered. This option
>  forbids the event merging logic from sharing events between groups and
>  may be used to increase accuracy in this case.
>  
> +--quiet::
> +Don't print output. This is useful with perf stat record below to only
> +write data to the perf.data file.
> +
>  STAT RECORD
>  -----------
>  Stores stat data into perf data file.
> diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
> index b01af171d94f..743fe47e7a88 100644
> --- a/tools/perf/builtin-stat.c
> +++ b/tools/perf/builtin-stat.c
> @@ -972,6 +972,8 @@ static void print_counters(struct timespec *ts, int argc, const char **argv)
>  	/* Do not print anything if we record to the pipe. */
>  	if (STAT_RECORD && perf_stat.data.is_pipe)
>  		return;
> +	if (stat_config.quiet)
> +		return;
>  
>  	perf_evlist__print_counters(evsel_list, &stat_config, &target,
>  				    ts, argc, argv);
> @@ -1171,6 +1173,8 @@ static struct option stat_options[] = {
>  		    "threads of same physical core"),
>  	OPT_BOOLEAN(0, "summary", &stat_config.summary,
>  		       "print summary for interval mode"),
> +	OPT_BOOLEAN(0, "quiet", &stat_config.quiet,
> +			"don't print output (useful with record)"),
>  #ifdef HAVE_LIBPFM
>  	OPT_CALLBACK(0, "pfm-events", &evsel_list, "event",
>  		"libpfm4 event selector. use 'perf list' to list available events",
> @@ -2132,7 +2136,7 @@ int cmd_stat(int argc, const char **argv)
>  		goto out;
>  	}
>  
> -	if (!output) {
> +	if (!output && !stat_config.quiet) {
>  		struct timespec tm;
>  		mode = append_file ? "a" : "w";
>  
> diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
> index 487010c624be..05adf8165025 100644
> --- a/tools/perf/util/stat.h
> +++ b/tools/perf/util/stat.h
> @@ -122,6 +122,7 @@ struct perf_stat_config {
>  	bool			 metric_no_group;
>  	bool			 metric_no_merge;
>  	bool			 stop_read_counter;
> +	bool			 quiet;
>  	FILE			*output;
>  	unsigned int		 interval;
>  	unsigned int		 timeout;
> -- 
> 2.28.0
> 

-- 

- Arnaldo