2009-09-17 18:08:04

by Ingo Molnar

Subject: [tip:perfcounters/core] perf sched: Determine the number of CPUs automatically

Commit-ID: 5349256ab12dcbc766495cd1cf3651f7203f8564
Gitweb: http://git.kernel.org/tip/5349256ab12dcbc766495cd1cf3651f7203f8564
Author: Ingo Molnar <[email protected]>
AuthorDate: Thu, 17 Sep 2009 18:24:55 +0200
Committer: Ingo Molnar <[email protected]>
CommitDate: Thu, 17 Sep 2009 18:24:55 +0200

perf sched: Determine the number of CPUs automatically

For 'perf sched map' output, determine max_cpu automatically,
instead of the static default of 15.

Cc: Peter Zijlstra <[email protected]>
Cc: Mike Galbraith <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Frederic Weisbecker <[email protected]>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <[email protected]>


---
tools/perf/builtin-sched.c | 5 ++++-
1 files changed, 4 insertions(+), 1 deletions(-)

diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c
index f67e351..947fd71 100644
--- a/tools/perf/builtin-sched.c
+++ b/tools/perf/builtin-sched.c
@@ -1347,7 +1347,7 @@ process_sched_wakeup_event(struct raw_event_sample *raw,
* Track the current task - that way we can know whether there's any
* weird events, such as a task being switched away that is not current.
*/
-static int max_cpu = 15;
+static int max_cpu;

static u32 curr_pid[MAX_CPUS] = { [0 ... MAX_CPUS-1] = -1 };

@@ -1818,6 +1818,9 @@ static struct trace_sched_handler map_ops = {

static void __cmd_map(void)
{
+ max_cpu = system("exit `grep ^processor /proc/cpuinfo | wc -l`");
+ max_cpu = WEXITSTATUS(max_cpu)-1;
+
setup_pager();
read_events();
print_bad_events();

2009-09-17 19:34:29

by Arjan van de Ven

Subject: Re: [tip:perfcounters/core] perf sched: Determine the number of CPUs automatically

On Thu, 17 Sep 2009 18:07:09 GMT
>
> perf sched: Determine the number of CPUs automatically
>
> For 'perf sched map' output, determine max_cpu automatically,
> instead of the static default of 15.
>
> static u32 curr_pid[MAX_CPUS] = { [0 ... MAX_CPUS-1] = -1 };
>
> @@ -1818,6 +1818,9 @@ static struct trace_sched_handler map_ops = {
>
> static void __cmd_map(void)
> {
> + max_cpu = system("exit `grep ^processor /proc/cpuinfo | wc
> -l`");

YUCK.

what is wrong with
max_cpu = sysconf(_SC_NPROCESSORS_CONF);
??

--
Arjan van de Ven Intel Open Source Technology Centre
For development, discussion and tips for power savings,
visit http://www.lesswatts.org

2009-09-17 19:45:46

by Ingo Molnar

Subject: Re: [tip:perfcounters/core] perf sched: Determine the number of CPUs automatically

* Arjan van de Ven <[email protected]> wrote:

> On Thu, 17 Sep 2009 18:07:09 GMT
> >
> > perf sched: Determine the number of CPUs automatically
> >
> > For 'perf sched map' output, determine max_cpu automatically,
> > instead of the static default of 15.
> >
> > static u32 curr_pid[MAX_CPUS] = { [0 ... MAX_CPUS-1] = -1 };
> >
> > @@ -1818,6 +1818,9 @@ static struct trace_sched_handler map_ops = {
> >
> > static void __cmd_map(void)
> > {
> > + max_cpu = system("exit `grep ^processor /proc/cpuinfo | wc
> > -l`");
>
> YUCK.
>
> what is wrong with
> max_cpu = sysconf(_SC_NPROCESSORS_CONF);
> ??

That's nicer - mind sending a patch?

Ingo

2009-09-17 20:09:33

by Ingo Molnar

Subject: Re: [tip:perfcounters/core] perf sched: Determine the number of CPUs automatically

* Ingo Molnar <[email protected]> wrote:

>
> * Arjan van de Ven <[email protected]> wrote:
>
> > On Thu, 17 Sep 2009 18:07:09 GMT
> > >
> > > perf sched: Determine the number of CPUs automatically
> > >
> > > For 'perf sched map' output, determine max_cpu automatically,
> > > instead of the static default of 15.
> > >
> > > static u32 curr_pid[MAX_CPUS] = { [0 ... MAX_CPUS-1] = -1 };
> > >
> > > @@ -1818,6 +1818,9 @@ static struct trace_sched_handler map_ops = {
> > >
> > > static void __cmd_map(void)
> > > {
> > > + max_cpu = system("exit `grep ^processor /proc/cpuinfo | wc
> > > -l`");
> >
> > YUCK.
> >
> > what is wrong with
> > max_cpu = sysconf(_SC_NPROCESSORS_CONF);
> > ??
>
> That's nicer - mind sending a patch?

fixed this up myself.

Ingo