[PATCH] perf bench sched: Add --threaded 0/1 option

[PATCH] perf bench sched: Add --threaded 0/1 option

[Ingo Molnar]

Allow the measurement of thread versus process context switch performance.

The default stays at 'process' based measurement, like lmbench's lat_ctx 
benchmark.

Sample output:

 comet:~/tip/tools/perf> taskset 1 ./perf bench sched pipe --threaded 0
 # Running sched/pipe benchmark...
 # Executed 1000000 pipe operations between two processes

     Total time: 4.138 [sec]

       4.138729 usecs/op
         241620 ops/sec
 comet:~/tip/tools/perf> taskset 1 ./perf bench sched pipe --threaded 1
 # Running sched/pipe benchmark...
 # Executed 1000000 pipe operations between two threads

     Total time: 3.667 [sec]

       3.667667 usecs/op
         272652 ops/sec

Signed-off-by: Ingo Molnar <mingo@kernel.org>

diff --git a/tools/perf/bench/sched-pipe.c b/tools/perf/bench/sched-pipe.c
index 69cfba8..7630dc6 100644
--- a/tools/perf/bench/sched-pipe.c
+++ b/tools/perf/bench/sched-pipe.c
@@ -28,12 +28,25 @@
 #include <sys/time.h>
 #include <sys/types.h>
 
+#include <pthread.h>
+
+struct thread_data {
+	int			nr;
+	int			pipe_read;
+	int			pipe_write;
+	pthread_t		pthread;
+};
+
 #define LOOPS_DEFAULT 1000000
 static int loops = LOOPS_DEFAULT;
 
+static int threaded = 0;
+
 static const struct option options[] = {
 	OPT_INTEGER('l', "loop", &loops,
 		    "Specify number of loops"),
+	OPT_INTEGER('T', "threaded", &threaded,
+		    "Specify threaded/process based task setup"),
 	OPT_END()
 };
 
@@ -42,13 +55,38 @@ static const char * const bench_sched_pipe_usage[] = {
 	NULL
 };
 
+static void *worker_thread(void *__tdata)
+{
+        struct thread_data *td = __tdata;
+	int m = 0, i;
+	int ret;
+
+	for (i = 0; i < loops; i++) {
+		if (!td->nr) {
+			ret = read(td->pipe_read, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+			ret = write(td->pipe_write, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+		} else {
+			ret = write(td->pipe_write, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+			ret = read(td->pipe_read, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+		}
+	}
+
+	return NULL;
+}
+
 int bench_sched_pipe(int argc, const char **argv,
 		     const char *prefix __maybe_unused)
 {
+	struct thread_data threads[2], *td;
 	int pipe_1[2], pipe_2[2];
-	int m = 0, i;
 	struct timeval start, stop, diff;
 	unsigned long long result_usec = 0;
+	int nr_threads = 2;
+	int t;
 
 	/*
 	 * why does "ret" exist?
@@ -58,43 +96,66 @@ int bench_sched_pipe(int argc, const char **argv,
 	int __maybe_unused ret, wait_stat;
 	pid_t pid, retpid __maybe_unused;
 
-	argc = parse_options(argc, argv, options,
-			     bench_sched_pipe_usage, 0);
+	argc = parse_options(argc, argv, options, bench_sched_pipe_usage, 0);
 
 	BUG_ON(pipe(pipe_1));
 	BUG_ON(pipe(pipe_2));
 
-	pid = fork();
-	assert(pid >= 0);
-
 	gettimeofday(&start, NULL);
 
-	if (!pid) {
-		for (i = 0; i < loops; i++) {
-			ret = read(pipe_1[0], &m, sizeof(int));
-			ret = write(pipe_2[1], &m, sizeof(int));
-		}
-	} else {
-		for (i = 0; i < loops; i++) {
-			ret = write(pipe_1[1], &m, sizeof(int));
-			ret = read(pipe_2[0], &m, sizeof(int));
+	for (t = 0; t < nr_threads; t++) {
+		td = threads + t;
+
+		td->nr = t;
+
+		if (t == 0) {
+			td->pipe_read = pipe_1[0];
+			td->pipe_write = pipe_2[1];
+		} else {
+			td->pipe_write = pipe_1[1];
+			td->pipe_read = pipe_2[0];
 		}
 	}
 
-	gettimeofday(&stop, NULL);
-	timersub(&stop, &start, &diff);
 
-	if (pid) {
+	if (threaded) {
+
+		for (t = 0; t < nr_threads; t++) {
+			td = threads + t;
+
+			ret = pthread_create(&td->pthread, NULL, worker_thread, td);
+			BUG_ON(ret);
+		}
+
+		for (t = 0; t < nr_threads; t++) {
+			td = threads + t;
+
+			ret = pthread_join(td->pthread, NULL);
+			BUG_ON(ret);
+		}
+
+	} else {
+		pid = fork();
+		assert(pid >= 0);
+
+		if (!pid) {
+			worker_thread(threads + 0);
+			exit(0);
+		} else {
+			worker_thread(threads + 1);
+		}
+
 		retpid = waitpid(pid, &wait_stat, 0);
 		assert((retpid == pid) && WIFEXITED(wait_stat));
-	} else {
-		exit(0);
 	}
 
+	gettimeofday(&stop, NULL);
+	timersub(&stop, &start, &diff);
+
 	switch (bench_format) {
 	case BENCH_FORMAT_DEFAULT:
-		printf("# Executed %d pipe operations between two tasks\n\n",
-			loops);
+		printf("# Executed %d pipe operations between two %s\n\n",
+			loops, threaded ? "threads" : "processes");
 
 		result_usec = diff.tv_sec * 1000000;
 		result_usec += diff.tv_usec;

Re: [PATCH] perf bench sched: Add --threaded 0/1 option

[Namhyung Kim]

Hi Ingo,

Just a few nitpicks...

On Tue, 17 Sep 2013 13:42:56 +0200, Ingo Molnar wrote:
> Allow the measurement of thread versus process context switch performance.
>
> The default stays at 'process' based measurement, like lmbench's lat_ctx 
> benchmark.
>
> Sample output:
>
>  comet:~/tip/tools/perf> taskset 1 ./perf bench sched pipe --threaded 0
>  # Running sched/pipe benchmark...
>  # Executed 1000000 pipe operations between two processes
>
>      Total time: 4.138 [sec]
>
>        4.138729 usecs/op
>          241620 ops/sec
>  comet:~/tip/tools/perf> taskset 1 ./perf bench sched pipe --threaded 1

Why not just make it boolean?  IMHO the integer argument can confuse
users as they can think it a number of threads..


>  # Running sched/pipe benchmark...
>  # Executed 1000000 pipe operations between two threads
>
>      Total time: 3.667 [sec]
>
>        3.667667 usecs/op
>          272652 ops/sec
>
> Signed-off-by: Ingo Molnar <mingo@kernel.org>
>
> diff --git a/tools/perf/bench/sched-pipe.c b/tools/perf/bench/sched-pipe.c
> index 69cfba8..7630dc6 100644
> --- a/tools/perf/bench/sched-pipe.c
> +++ b/tools/perf/bench/sched-pipe.c
> @@ -28,12 +28,25 @@
>  #include <sys/time.h>
>  #include <sys/types.h>
>  
> +#include <pthread.h>
> +
> +struct thread_data {
> +	int			nr;
> +	int			pipe_read;
> +	int			pipe_write;
> +	pthread_t		pthread;
> +};
> +
>  #define LOOPS_DEFAULT 1000000
>  static int loops = LOOPS_DEFAULT;
>  
> +static int threaded = 0;
> +
>  static const struct option options[] = {
>  	OPT_INTEGER('l', "loop", &loops,
>  		    "Specify number of loops"),
> +	OPT_INTEGER('T', "threaded", &threaded,
> +		    "Specify threaded/process based task setup"),
>  	OPT_END()
>  };
>  
> @@ -42,13 +55,38 @@ static const char * const bench_sched_pipe_usage[] = {
>  	NULL
>  };
>  
> +static void *worker_thread(void *__tdata)
> +{
> +        struct thread_data *td = __tdata;

Whitespace damaged.  Otherwise looks good to me!

Thanks,
Namhyung


> +	int m = 0, i;
> +	int ret;
> +
> +	for (i = 0; i < loops; i++) {
> +		if (!td->nr) {
> +			ret = read(td->pipe_read, &m, sizeof(int));
> +			BUG_ON(ret != sizeof(int));
> +			ret = write(td->pipe_write, &m, sizeof(int));
> +			BUG_ON(ret != sizeof(int));
> +		} else {
> +			ret = write(td->pipe_write, &m, sizeof(int));
> +			BUG_ON(ret != sizeof(int));
> +			ret = read(td->pipe_read, &m, sizeof(int));
> +			BUG_ON(ret != sizeof(int));
> +		}
> +	}
> +
> +	return NULL;
> +}

[PATCH] perf bench sched: Add --threaded option

[Ingo Molnar]

* Namhyung Kim <namhyung@kernel.org> wrote:

> Hi Ingo,
> 
> Just a few nitpicks...
> 
> On Tue, 17 Sep 2013 13:42:56 +0200, Ingo Molnar wrote:
> > Allow the measurement of thread versus process context switch performance.
> >
> > The default stays at 'process' based measurement, like lmbench's lat_ctx 
> > benchmark.
> >
> > Sample output:
> >
> >  comet:~/tip/tools/perf> taskset 1 ./perf bench sched pipe --threaded 0
> >  # Running sched/pipe benchmark...
> >  # Executed 1000000 pipe operations between two processes
> >
> >      Total time: 4.138 [sec]
> >
> >        4.138729 usecs/op
> >          241620 ops/sec
> >  comet:~/tip/tools/perf> taskset 1 ./perf bench sched pipe --threaded 1
> 
> Why not just make it boolean?  IMHO the integer argument can confuse 
> users as they can think it a number of threads..

Agreed, fixed.

Btw., we should allow "--flag 0/1" type of input for bool options as well, 
it's easier for scripting. Any other integer should be rejected, and 
non-integer should be interpreted as the next option.

In addition, it would be nice to allow the following two aliases as well:

	--non-flag
	--dont-flag

So that the following can be written:

	--threaded
	--non-threaded

or, if the option name is a verb:

	--use-threads
	--dont-use-threads

Instead of forcing the following incorrect grammar on users:

	--no-threaded
	--no-use-threads

> > +static void *worker_thread(void *__tdata)
> > +{
> > +        struct thread_data *td = __tdata;
> 
> Whitespace damaged.  Otherwise looks good to me!

Fixed this too - new version attached.

Thanks,

	Ingo

--------------------------------->
Subject: perf bench sched: Add --threaded option
From: Ingo Molnar <mingo@kernel.org>
Date: Tue, 17 Sep 2013 13:42:56 +0200

Allow the measurement of thread versus process context switch
performance.

The default stays at 'process' based measurement, like lmbench's
lat_ctx benchmark.

Sample output:

 comet:~/tip/tools/perf> taskset 1 ./perf bench sched pipe
 # Running sched/pipe benchmark...
 # Executed 1000000 pipe operations between two processes

     Total time: 4.138 [sec]

       4.138729 usecs/op
         241620 ops/sec
 comet:~/tip/tools/perf> taskset 1 ./perf bench sched pipe --threaded
 # Running sched/pipe benchmark...
 # Executed 1000000 pipe operations between two threads

     Total time: 3.667 [sec]

       3.667667 usecs/op
         272652 ops/sec

Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Link: http://lkml.kernel.org/r/20130917114256.GA31159@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 tools/perf/bench/sched-pipe.c |  115 +++++++++++++++++++++++++++++++-----------
 1 file changed, 86 insertions(+), 29 deletions(-)

Index: tip/tools/perf/bench/sched-pipe.c
===================================================================
--- tip.orig/tools/perf/bench/sched-pipe.c
+++ tip/tools/perf/bench/sched-pipe.c
@@ -7,9 +7,7 @@
  * Based on pipe-test-1m.c by Ingo Molnar <mingo@redhat.com>
  *  http://people.redhat.com/mingo/cfs-scheduler/tools/pipe-test-1m.c
  * Ported to perf by Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
- *
  */
-
 #include "../perf.h"
 #include "../util/util.h"
 #include "../util/parse-options.h"
@@ -28,12 +26,24 @@
 #include <sys/time.h>
 #include <sys/types.h>
 
+#include <pthread.h>
+
+struct thread_data {
+	int			nr;
+	int			pipe_read;
+	int			pipe_write;
+	pthread_t		pthread;
+};
+
 #define LOOPS_DEFAULT 1000000
-static int loops = LOOPS_DEFAULT;
+static	int			loops = LOOPS_DEFAULT;
+
+/* Use processes by default: */
+static bool			threaded;
 
 static const struct option options[] = {
-	OPT_INTEGER('l', "loop", &loops,
-		    "Specify number of loops"),
+	OPT_INTEGER('l', "loop",	&loops,		"Specify number of loops"),
+	OPT_BOOLEAN('T', "threaded",	&threaded,	"Specify threads/process based task setup"),
 	OPT_END()
 };
 
@@ -42,13 +52,37 @@ static const char * const bench_sched_pi
 	NULL
 };
 
-int bench_sched_pipe(int argc, const char **argv,
-		     const char *prefix __maybe_unused)
+static void *worker_thread(void *__tdata)
 {
-	int pipe_1[2], pipe_2[2];
+	struct thread_data *td = __tdata;
 	int m = 0, i;
+	int ret;
+
+	for (i = 0; i < loops; i++) {
+		if (!td->nr) {
+			ret = read(td->pipe_read, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+			ret = write(td->pipe_write, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+		} else {
+			ret = write(td->pipe_write, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+			ret = read(td->pipe_read, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+		}
+	}
+
+	return NULL;
+}
+
+int bench_sched_pipe(int argc, const char **argv, const char *prefix __maybe_unused)
+{
+	struct thread_data threads[2], *td;
+	int pipe_1[2], pipe_2[2];
 	struct timeval start, stop, diff;
 	unsigned long long result_usec = 0;
+	int nr_threads = 2;
+	int t;
 
 	/*
 	 * why does "ret" exist?
@@ -58,43 +92,66 @@ int bench_sched_pipe(int argc, const cha
 	int __maybe_unused ret, wait_stat;
 	pid_t pid, retpid __maybe_unused;
 
-	argc = parse_options(argc, argv, options,
-			     bench_sched_pipe_usage, 0);
+	argc = parse_options(argc, argv, options, bench_sched_pipe_usage, 0);
 
 	BUG_ON(pipe(pipe_1));
 	BUG_ON(pipe(pipe_2));
 
-	pid = fork();
-	assert(pid >= 0);
-
 	gettimeofday(&start, NULL);
 
-	if (!pid) {
-		for (i = 0; i < loops; i++) {
-			ret = read(pipe_1[0], &m, sizeof(int));
-			ret = write(pipe_2[1], &m, sizeof(int));
-		}
-	} else {
-		for (i = 0; i < loops; i++) {
-			ret = write(pipe_1[1], &m, sizeof(int));
-			ret = read(pipe_2[0], &m, sizeof(int));
+	for (t = 0; t < nr_threads; t++) {
+		td = threads + t;
+
+		td->nr = t;
+
+		if (t == 0) {
+			td->pipe_read = pipe_1[0];
+			td->pipe_write = pipe_2[1];
+		} else {
+			td->pipe_write = pipe_1[1];
+			td->pipe_read = pipe_2[0];
 		}
 	}
 
-	gettimeofday(&stop, NULL);
-	timersub(&stop, &start, &diff);
 
-	if (pid) {
+	if (threaded) {
+
+		for (t = 0; t < nr_threads; t++) {
+			td = threads + t;
+
+			ret = pthread_create(&td->pthread, NULL, worker_thread, td);
+			BUG_ON(ret);
+		}
+
+		for (t = 0; t < nr_threads; t++) {
+			td = threads + t;
+
+			ret = pthread_join(td->pthread, NULL);
+			BUG_ON(ret);
+		}
+
+	} else {
+		pid = fork();
+		assert(pid >= 0);
+
+		if (!pid) {
+			worker_thread(threads + 0);
+			exit(0);
+		} else {
+			worker_thread(threads + 1);
+		}
+
 		retpid = waitpid(pid, &wait_stat, 0);
 		assert((retpid == pid) && WIFEXITED(wait_stat));
-	} else {
-		exit(0);
 	}
 
+	gettimeofday(&stop, NULL);
+	timersub(&stop, &start, &diff);
+
 	switch (bench_format) {
 	case BENCH_FORMAT_DEFAULT:
-		printf("# Executed %d pipe operations between two tasks\n\n",
-			loops);
+		printf("# Executed %d pipe operations between two %s\n\n",
+			loops, threaded ? "threads" : "processes");
 
 		result_usec = diff.tv_sec * 1000000;
 		result_usec += diff.tv_usec;

[tip:perf/core] perf bench sched: Add --threaded option

[tip-bot for Ingo Molnar]

Commit-ID:  a9faa0cab619fad380c2669825aa84212943d528
Gitweb:     http://git.kernel.org/tip/a9faa0cab619fad380c2669825aa84212943d528
Author:     Ingo Molnar <mingo@kernel.org>
AuthorDate: Mon, 23 Sep 2013 10:17:35 +0200
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Fri, 11 Oct 2013 12:17:25 -0300

perf bench sched: Add --threaded option

Allow the measurement of thread versus process context switch
performance.

The default stays at 'process' based measurement, like lmbench's lat_ctx
benchmark.

Sample output:

 comet:~/tip/tools/perf> taskset 1 ./perf bench sched pipe
 # Running sched/pipe benchmark...
 # Executed 1000000 pipe operations between two processes

     Total time: 4.138 [sec]

       4.138729 usecs/op
         241620 ops/sec
 comet:~/tip/tools/perf> taskset 1 ./perf bench sched pipe --threaded
 # Running sched/pipe benchmark...
 # Executed 1000000 pipe operations between two threads

     Total time: 3.667 [sec]

       3.667667 usecs/op
         272652 ops/sec

Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Link: http://lkml.kernel.org/r/20130917114256.GA31159@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/bench/sched-pipe.c | 115 +++++++++++++++++++++++++++++++-----------
 1 file changed, 86 insertions(+), 29 deletions(-)

diff --git a/tools/perf/bench/sched-pipe.c b/tools/perf/bench/sched-pipe.c
index 69cfba8..07a8d76 100644
--- a/tools/perf/bench/sched-pipe.c
+++ b/tools/perf/bench/sched-pipe.c
@@ -7,9 +7,7 @@
  * Based on pipe-test-1m.c by Ingo Molnar <mingo@redhat.com>
  *  http://people.redhat.com/mingo/cfs-scheduler/tools/pipe-test-1m.c
  * Ported to perf by Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
- *
  */
-
 #include "../perf.h"
 #include "../util/util.h"
 #include "../util/parse-options.h"
@@ -28,12 +26,24 @@
 #include <sys/time.h>
 #include <sys/types.h>
 
+#include <pthread.h>
+
+struct thread_data {
+	int			nr;
+	int			pipe_read;
+	int			pipe_write;
+	pthread_t		pthread;
+};
+
 #define LOOPS_DEFAULT 1000000
-static int loops = LOOPS_DEFAULT;
+static	int			loops = LOOPS_DEFAULT;
+
+/* Use processes by default: */
+static bool			threaded;
 
 static const struct option options[] = {
-	OPT_INTEGER('l', "loop", &loops,
-		    "Specify number of loops"),
+	OPT_INTEGER('l', "loop",	&loops,		"Specify number of loops"),
+	OPT_BOOLEAN('T', "threaded",	&threaded,	"Specify threads/process based task setup"),
 	OPT_END()
 };
 
@@ -42,13 +52,37 @@ static const char * const bench_sched_pipe_usage[] = {
 	NULL
 };
 
-int bench_sched_pipe(int argc, const char **argv,
-		     const char *prefix __maybe_unused)
+static void *worker_thread(void *__tdata)
 {
-	int pipe_1[2], pipe_2[2];
+	struct thread_data *td = __tdata;
 	int m = 0, i;
+	int ret;
+
+	for (i = 0; i < loops; i++) {
+		if (!td->nr) {
+			ret = read(td->pipe_read, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+			ret = write(td->pipe_write, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+		} else {
+			ret = write(td->pipe_write, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+			ret = read(td->pipe_read, &m, sizeof(int));
+			BUG_ON(ret != sizeof(int));
+		}
+	}
+
+	return NULL;
+}
+
+int bench_sched_pipe(int argc, const char **argv, const char *prefix __maybe_unused)
+{
+	struct thread_data threads[2], *td;
+	int pipe_1[2], pipe_2[2];
 	struct timeval start, stop, diff;
 	unsigned long long result_usec = 0;
+	int nr_threads = 2;
+	int t;
 
 	/*
 	 * why does "ret" exist?
@@ -58,43 +92,66 @@ int bench_sched_pipe(int argc, const char **argv,
 	int __maybe_unused ret, wait_stat;
 	pid_t pid, retpid __maybe_unused;
 
-	argc = parse_options(argc, argv, options,
-			     bench_sched_pipe_usage, 0);
+	argc = parse_options(argc, argv, options, bench_sched_pipe_usage, 0);
 
 	BUG_ON(pipe(pipe_1));
 	BUG_ON(pipe(pipe_2));
 
-	pid = fork();
-	assert(pid >= 0);
-
 	gettimeofday(&start, NULL);
 
-	if (!pid) {
-		for (i = 0; i < loops; i++) {
-			ret = read(pipe_1[0], &m, sizeof(int));
-			ret = write(pipe_2[1], &m, sizeof(int));
-		}
-	} else {
-		for (i = 0; i < loops; i++) {
-			ret = write(pipe_1[1], &m, sizeof(int));
-			ret = read(pipe_2[0], &m, sizeof(int));
+	for (t = 0; t < nr_threads; t++) {
+		td = threads + t;
+
+		td->nr = t;
+
+		if (t == 0) {
+			td->pipe_read = pipe_1[0];
+			td->pipe_write = pipe_2[1];
+		} else {
+			td->pipe_write = pipe_1[1];
+			td->pipe_read = pipe_2[0];
 		}
 	}
 
-	gettimeofday(&stop, NULL);
-	timersub(&stop, &start, &diff);
 
-	if (pid) {
+	if (threaded) {
+
+		for (t = 0; t < nr_threads; t++) {
+			td = threads + t;
+
+			ret = pthread_create(&td->pthread, NULL, worker_thread, td);
+			BUG_ON(ret);
+		}
+
+		for (t = 0; t < nr_threads; t++) {
+			td = threads + t;
+
+			ret = pthread_join(td->pthread, NULL);
+			BUG_ON(ret);
+		}
+
+	} else {
+		pid = fork();
+		assert(pid >= 0);
+
+		if (!pid) {
+			worker_thread(threads + 0);
+			exit(0);
+		} else {
+			worker_thread(threads + 1);
+		}
+
 		retpid = waitpid(pid, &wait_stat, 0);
 		assert((retpid == pid) && WIFEXITED(wait_stat));
-	} else {
-		exit(0);
 	}
 
+	gettimeofday(&stop, NULL);
+	timersub(&stop, &start, &diff);
+
 	switch (bench_format) {
 	case BENCH_FORMAT_DEFAULT:
-		printf("# Executed %d pipe operations between two tasks\n\n",
-			loops);
+		printf("# Executed %d pipe operations between two %s\n\n",
+			loops, threaded ? "threads" : "processes");
 
 		result_usec = diff.tv_sec * 1000000;
 		result_usec += diff.tv_usec;