linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] perf sort: only insert overhead && overhead_children when no overhead* field given
@ 2017-06-01  9:03 changbin.du
  2017-06-01 10:21 ` Jiri Olsa
                   ` (2 more replies)
  0 siblings, 3 replies; 10+ messages in thread
From: changbin.du @ 2017-06-01  9:03 UTC (permalink / raw)
  To: peterz, mingo, alexander.shishkin, jolsa; +Cc: linux-kernel, Changbin Du

From: Changbin Du <changbin.du@intel.com>

If we always insert 'overhead' and 'overhead_children' as sort keys,
this make it impossible to sort as overhead (which displayed as Self)
first.Ths will be a problem if the data is collected with call-graph
enabled. Then we never can sort the result as self-overhead on this
data. And sometimes the data is hard to collect.

This patch fix this by only insert overhead && overhead_children
when no overhead* field given.

Signed-off-by: Changbin Du <changbin.du@intel.com>
---
 tools/perf/util/sort.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index 5762ae4..69eea3a 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -2635,6 +2635,9 @@ static char *setup_overhead(char *keys)
 	if (sort__mode == SORT_MODE__DIFF)
 		return keys;
 
+	if (strstr(keys, "overhead"))
+		return keys;
+
 	keys = prefix_if_not_in("overhead", keys);
 
 	if (symbol_conf.cumulate_callchain)
-- 
2.7.4

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* Re: [PATCH] perf sort: only insert overhead && overhead_children when no overhead* field given
  2017-06-01  9:03 [PATCH] perf sort: only insert overhead && overhead_children when no overhead* field given changbin.du
@ 2017-06-01 10:21 ` Jiri Olsa
  2017-06-02  2:52   ` Du, Changbin
  2017-06-02  4:22 ` [PATCH v2] " changbin.du
  2017-07-05  3:35 ` [PATCH v3] " changbin.du
  2 siblings, 1 reply; 10+ messages in thread
From: Jiri Olsa @ 2017-06-01 10:21 UTC (permalink / raw)
  To: changbin.du; +Cc: peterz, mingo, alexander.shishkin, jolsa, linux-kernel

On Thu, Jun 01, 2017 at 05:03:21PM +0800, changbin.du@intel.com wrote:
> From: Changbin Du <changbin.du@intel.com>
> 
> If we always insert 'overhead' and 'overhead_children' as sort keys,
> this make it impossible to sort as overhead (which displayed as Self)
> first.Ths will be a problem if the data is collected with call-graph
> enabled. Then we never can sort the result as self-overhead on this
> data. And sometimes the data is hard to collect.
> 
> This patch fix this by only insert overhead && overhead_children
> when no overhead* field given.
> 
> Signed-off-by: Changbin Du <changbin.du@intel.com>
> ---
>  tools/perf/util/sort.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
> index 5762ae4..69eea3a 100644
> --- a/tools/perf/util/sort.c
> +++ b/tools/perf/util/sort.c
> @@ -2635,6 +2635,9 @@ static char *setup_overhead(char *keys)
>  	if (sort__mode == SORT_MODE__DIFF)
>  		return keys;
>  
> +	if (strstr(keys, "overhead"))
> +		return keys;
> +
>  	keys = prefix_if_not_in("overhead", keys);

hum, you basicaly do what's at begining of prefix_if_not_in function:

static char *prefix_if_not_in(const char *pre, char *str)
{
        char *n;

        if (!str || strstr(str, pre))
                return str;
        ...


could you please provide the example described in changelog?

jirka

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] perf sort: only insert overhead && overhead_children when no overhead* field given
  2017-06-01 10:21 ` Jiri Olsa
@ 2017-06-02  2:52   ` Du, Changbin
  2017-06-02  3:22     ` Du, Changbin
  0 siblings, 1 reply; 10+ messages in thread
From: Du, Changbin @ 2017-06-02  2:52 UTC (permalink / raw)
  To: Jiri Olsa
  Cc: changbin.du, peterz, mingo, alexander.shishkin, jolsa, linux-kernel

[-- Attachment #1: Type: text/plain, Size: 997 bytes --]

On Thu, Jun 01, 2017 at 12:21:39PM +0200, Jiri Olsa wrote:
> On Thu, Jun 01, 2017 at 05:03:21PM +0800, changbin.du@intel.com wrote:
> > diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
> > index 5762ae4..69eea3a 100644
> > --- a/tools/perf/util/sort.c
> > +++ b/tools/perf/util/sort.c
> > @@ -2635,6 +2635,9 @@ static char *setup_overhead(char *keys)
> >  	if (sort__mode == SORT_MODE__DIFF)
> >  		return keys;
> >  
> > +	if (strstr(keys, "overhead"))
> > +		return keys;
> > +
> >  	keys = prefix_if_not_in("overhead", keys);
> 
> hum, you basicaly do what's at begining of prefix_if_not_in function:
> 
> static char *prefix_if_not_in(const char *pre, char *str)
> {
>         char *n;
> 
>         if (!str || strstr(str, pre))
>                 return str;
>         ...
> 
Thanks, will change it.

> 
> could you please provide the example described in changelog?
> 
Will add example cmdline there, Thanks.

> jirka

-- 
Thanks,
Changbin Du

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] perf sort: only insert overhead && overhead_children when no overhead* field given
  2017-06-02  2:52   ` Du, Changbin
@ 2017-06-02  3:22     ` Du, Changbin
  0 siblings, 0 replies; 10+ messages in thread
From: Du, Changbin @ 2017-06-02  3:22 UTC (permalink / raw)
  To: Du, Changbin
  Cc: Jiri Olsa, peterz, mingo, alexander.shishkin, jolsa, linux-kernel

[-- Attachment #1: Type: text/plain, Size: 1592 bytes --]

On Fri, Jun 02, 2017 at 10:52:24AM +0800, Du, Changbin wrote:
> On Thu, Jun 01, 2017 at 12:21:39PM +0200, Jiri Olsa wrote:
> > On Thu, Jun 01, 2017 at 05:03:21PM +0800, changbin.du@intel.com wrote:
> > > diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
> > > index 5762ae4..69eea3a 100644
> > > --- a/tools/perf/util/sort.c
> > > +++ b/tools/perf/util/sort.c
> > > @@ -2635,6 +2635,9 @@ static char *setup_overhead(char *keys)
> > >  	if (sort__mode == SORT_MODE__DIFF)
> > >  		return keys;
> > >  
> > > +	if (strstr(keys, "overhead"))
> > > +		return keys;
> > > +
> > >  	keys = prefix_if_not_in("overhead", keys);
> > 
> > hum, you basicaly do what's at begining of prefix_if_not_in function:
> > 
> > static char *prefix_if_not_in(const char *pre, char *str)
> > {
> >         char *n;
> > 
> >         if (!str || strstr(str, pre))
> >                 return str;
> >         ...
> > 
> Thanks, will change it.
>
Misunderstood your words. This is not equal to prefix_if_not_in.
# perf record -ag
# perf report -s overhead,sym

Samples: 7K of event 'cycles', Event count (approx.): 865138253
  Children      Self  Symbol
  +   26.41%     0.00%  [k] verify_cpu
  +   26.37%     0.04%  [k] cpu_startup_entry
  +   25.93%     0.27%  [k] do_idle
  +   19.88%     0.00%  [k] start_secondary
  ....

'Children' should not be here.

> > 
> > could you please provide the example described in changelog?
> > 
> Will add example cmdline there, Thanks.
> 
> > jirka
> 
> -- 
> Thanks,
> Changbin Du



-- 
Thanks,
Changbin Du

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

^ permalink raw reply	[flat|nested] 10+ messages in thread

* [PATCH v2] perf sort: only insert overhead && overhead_children when no overhead* field given
  2017-06-01  9:03 [PATCH] perf sort: only insert overhead && overhead_children when no overhead* field given changbin.du
  2017-06-01 10:21 ` Jiri Olsa
@ 2017-06-02  4:22 ` changbin.du
  2017-06-26  8:06   ` Du, Changbin
  2017-07-05  3:35 ` [PATCH v3] " changbin.du
  2 siblings, 1 reply; 10+ messages in thread
From: changbin.du @ 2017-06-02  4:22 UTC (permalink / raw)
  To: peterz, mingo, alexander.shishkin, jolsa; +Cc: linux-kernel, Changbin Du

From: Changbin Du <changbin.du@intel.com>

If we always insert 'overhead' and 'overhead_children' as sort keys,
this make it impossible to sort as overhead (which displayed as Self)
first.Ths will be a problem if the data is collected with call-graph
enabled. Then we never can sort the result as self-overhead on this
data. And sometimes the data is hard to collect.

> perf record -ag
> perf report -s overhead,sym

Samples: 7K of event 'cycles', Event count (approx.): 865138253
  Children      Self  Symbol
  +   26.41%     0.00%  [k] verify_cpu
  +   26.37%     0.04%  [k] cpu_startup_entry
  +   25.93%     0.27%  [k] do_idle
  +   19.88%     0.00%  [k] start_secondary
  ....

I intend to sort as 'Self', but actually it sort as 'Children'.

This patch fix this by only insert overhead && overhead_children
when no overhead* field given.

Signed-off-by: Changbin Du <changbin.du@intel.com>
---
v2: Add the example in commit message.

---
 tools/perf/util/sort.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index 5762ae4..69eea3a 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -2635,6 +2635,9 @@ static char *setup_overhead(char *keys)
 	if (sort__mode == SORT_MODE__DIFF)
 		return keys;
 
+	if (strstr(keys, "overhead"))
+		return keys;
+
 	keys = prefix_if_not_in("overhead", keys);
 
 	if (symbol_conf.cumulate_callchain)
-- 
2.7.4

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* Re: [PATCH v2] perf sort: only insert overhead && overhead_children when no overhead* field given
  2017-06-02  4:22 ` [PATCH v2] " changbin.du
@ 2017-06-26  8:06   ` Du, Changbin
  2017-06-29 15:22     ` Jiri Olsa
  0 siblings, 1 reply; 10+ messages in thread
From: Du, Changbin @ 2017-06-26  8:06 UTC (permalink / raw)
  To: jolsa; +Cc: peterz, mingo, alexander.shishkin, jolsa, linux-kernel

[-- Attachment #1: Type: text/plain, Size: 1758 bytes --]


Hi, Jiri,
what is the status of this one? I didn't get a response of v2. thanks.

On Fri, Jun 02, 2017 at 12:22:00PM +0800, changbin.du@intel.com wrote:
> From: Changbin Du <changbin.du@intel.com>
> 
> If we always insert 'overhead' and 'overhead_children' as sort keys,
> this make it impossible to sort as overhead (which displayed as Self)
> first.Ths will be a problem if the data is collected with call-graph
> enabled. Then we never can sort the result as self-overhead on this
> data. And sometimes the data is hard to collect.
> 
> > perf record -ag
> > perf report -s overhead,sym
> 
> Samples: 7K of event 'cycles', Event count (approx.): 865138253
>   Children      Self  Symbol
>   +   26.41%     0.00%  [k] verify_cpu
>   +   26.37%     0.04%  [k] cpu_startup_entry
>   +   25.93%     0.27%  [k] do_idle
>   +   19.88%     0.00%  [k] start_secondary
>   ....
> 
> I intend to sort as 'Self', but actually it sort as 'Children'.
> 
> This patch fix this by only insert overhead && overhead_children
> when no overhead* field given.
> 
> Signed-off-by: Changbin Du <changbin.du@intel.com>
> ---
> v2: Add the example in commit message.
> 
> ---
>  tools/perf/util/sort.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
> index 5762ae4..69eea3a 100644
> --- a/tools/perf/util/sort.c
> +++ b/tools/perf/util/sort.c
> @@ -2635,6 +2635,9 @@ static char *setup_overhead(char *keys)
>  	if (sort__mode == SORT_MODE__DIFF)
>  		return keys;
>  
> +	if (strstr(keys, "overhead"))
> +		return keys;
> +
>  	keys = prefix_if_not_in("overhead", keys);
>  
>  	if (symbol_conf.cumulate_callchain)
> -- 
> 2.7.4
> 

-- 
Thanks,
Changbin Du

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH v2] perf sort: only insert overhead && overhead_children when no overhead* field given
  2017-06-26  8:06   ` Du, Changbin
@ 2017-06-29 15:22     ` Jiri Olsa
  2017-07-05  3:02       ` Du, Changbin
  0 siblings, 1 reply; 10+ messages in thread
From: Jiri Olsa @ 2017-06-29 15:22 UTC (permalink / raw)
  To: Du, Changbin; +Cc: jolsa, peterz, mingo, alexander.shishkin, linux-kernel

On Mon, Jun 26, 2017 at 04:06:26PM +0800, Du, Changbin wrote:
> 
> Hi, Jiri,
> what is the status of this one? I didn't get a response of v2. thanks.
> 
> On Fri, Jun 02, 2017 at 12:22:00PM +0800, changbin.du@intel.com wrote:
> > From: Changbin Du <changbin.du@intel.com>
> > 
> > If we always insert 'overhead' and 'overhead_children' as sort keys,
> > this make it impossible to sort as overhead (which displayed as Self)
> > first.Ths will be a problem if the data is collected with call-graph
> > enabled. Then we never can sort the result as self-overhead on this
> > data. And sometimes the data is hard to collect.
> > 
> > > perf record -ag
> > > perf report -s overhead,sym
> > 
> > Samples: 7K of event 'cycles', Event count (approx.): 865138253
> >   Children      Self  Symbol
> >   +   26.41%     0.00%  [k] verify_cpu
> >   +   26.37%     0.04%  [k] cpu_startup_entry
> >   +   25.93%     0.27%  [k] do_idle
> >   +   19.88%     0.00%  [k] start_secondary
> >   ....
> > 
> > I intend to sort as 'Self', but actually it sort as 'Children'.
> > 
> > This patch fix this by only insert overhead && overhead_children
> > when no overhead* field given.
> > 
> > Signed-off-by: Changbin Du <changbin.du@intel.com>
> > ---
> > v2: Add the example in commit message.
> > 
> > ---
> >  tools/perf/util/sort.c | 3 +++
> >  1 file changed, 3 insertions(+)
> > 
> > diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
> > index 5762ae4..69eea3a 100644
> > --- a/tools/perf/util/sort.c
> > +++ b/tools/perf/util/sort.c
> > @@ -2635,6 +2635,9 @@ static char *setup_overhead(char *keys)
> >  	if (sort__mode == SORT_MODE__DIFF)
> >  		return keys;
> >  
> > +	if (strstr(keys, "overhead"))
> > +		return keys;
> > +

ah ok, I think it's right.. basically you're forcing
precedence of the -s option over the --children option

you could have your example working by running:
  $ perf report -s overhead,sym --no-children

please state something like:

/*
 * User already stated overhead within -s option,
 * do not mangle with that.
 */

thanks,
jirka

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH v2] perf sort: only insert overhead && overhead_children when no overhead* field given
  2017-06-29 15:22     ` Jiri Olsa
@ 2017-07-05  3:02       ` Du, Changbin
  0 siblings, 0 replies; 10+ messages in thread
From: Du, Changbin @ 2017-07-05  3:02 UTC (permalink / raw)
  To: Jiri Olsa
  Cc: Du, Changbin, jolsa, peterz, mingo, alexander.shishkin, linux-kernel

[-- Attachment #1: Type: text/plain, Size: 843 bytes --]

On Thu, Jun 29, 2017 at 05:22:52PM +0200, Jiri Olsa wrote:
> > > --- a/tools/perf/util/sort.c
> > > +++ b/tools/perf/util/sort.c
> > > @@ -2635,6 +2635,9 @@ static char *setup_overhead(char *keys)
> > >  	if (sort__mode == SORT_MODE__DIFF)
> > >  		return keys;
> > >  
> > > +	if (strstr(keys, "overhead"))
> > > +		return keys;
> > > +
> 
> ah ok, I think it's right.. basically you're forcing
> precedence of the -s option over the --children option
> 
> you could have your example working by running:
>   $ perf report -s overhead,sym --no-children
> 
Thanks for the knowledge.

> please state something like:
> 
> /*
>  * User already stated overhead within -s option,
>  * do not mangle with that.
>  */
> 
I'll add this as comment around the change for v3.

> thanks,
> jirka

-- 
Thanks,
Changbin Du

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

^ permalink raw reply	[flat|nested] 10+ messages in thread

* [PATCH v3] perf sort: only insert overhead && overhead_children when no overhead* field given
  2017-06-01  9:03 [PATCH] perf sort: only insert overhead && overhead_children when no overhead* field given changbin.du
  2017-06-01 10:21 ` Jiri Olsa
  2017-06-02  4:22 ` [PATCH v2] " changbin.du
@ 2017-07-05  3:35 ` changbin.du
  2017-07-06 16:04   ` Jiri Olsa
  2 siblings, 1 reply; 10+ messages in thread
From: changbin.du @ 2017-07-05  3:35 UTC (permalink / raw)
  To: jolsa; +Cc: peterz, mingo, alexander.shishkin, linux-kernel, changbin.du

From: Changbin Du <changbin.du@intel.com>

If we always insert 'overhead' and 'overhead_children' as sort keys,
this make it impossible to sort as overhead (which displayed as Self)
first.Ths will be a problem if the data is collected with call-graph
enabled. Then we never can sort the result as self-overhead on this
data. And sometimes the data is hard to collect.

> perf record -ag
> perf report -s overhead,sym

Samples: 7K of event 'cycles', Event count (approx.): 865138253
  Children      Self  Symbol
  +   26.41%     0.00%  [k] verify_cpu
  +   26.37%     0.04%  [k] cpu_startup_entry
  +   25.93%     0.27%  [k] do_idle
  +   19.88%     0.00%  [k] start_secondary
  ....

I intend to sort as 'Self', but actually it sort as 'Children'.

This patch fix this by only insert overhead && overhead_children
when no overhead* field given.

Signed-off-by: Changbin Du <changbin.du@intel.com>
---
v3: Add comment in code from Jiri
v2: Add the example in commit message.

---
 tools/perf/util/sort.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index 8b327c9..9e087fb 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -2635,6 +2635,13 @@ static char *setup_overhead(char *keys)
 	if (sort__mode == SORT_MODE__DIFF)
 		return keys;
 
+	/**
+	 * User already stated overhead within -s option,
+	 * do not mangle with that.
+	 */
+	if (strstr(keys, "overhead"))
+		return keys;
+
 	keys = prefix_if_not_in("overhead", keys);
 
 	if (symbol_conf.cumulate_callchain)
-- 
2.7.4

^ permalink raw reply related	[flat|nested] 10+ messages in thread

* Re: [PATCH v3] perf sort: only insert overhead && overhead_children when no overhead* field given
  2017-07-05  3:35 ` [PATCH v3] " changbin.du
@ 2017-07-06 16:04   ` Jiri Olsa
  0 siblings, 0 replies; 10+ messages in thread
From: Jiri Olsa @ 2017-07-06 16:04 UTC (permalink / raw)
  To: changbin.du; +Cc: jolsa, peterz, mingo, alexander.shishkin, linux-kernel

On Wed, Jul 05, 2017 at 11:35:52AM +0800, changbin.du@intel.com wrote:
> From: Changbin Du <changbin.du@intel.com>
> 
> If we always insert 'overhead' and 'overhead_children' as sort keys,
> this make it impossible to sort as overhead (which displayed as Self)
> first.Ths will be a problem if the data is collected with call-graph
> enabled. Then we never can sort the result as self-overhead on this
> data. And sometimes the data is hard to collect.
> 
> > perf record -ag
> > perf report -s overhead,sym
> 
> Samples: 7K of event 'cycles', Event count (approx.): 865138253
>   Children      Self  Symbol
>   +   26.41%     0.00%  [k] verify_cpu
>   +   26.37%     0.04%  [k] cpu_startup_entry
>   +   25.93%     0.27%  [k] do_idle
>   +   19.88%     0.00%  [k] start_secondary
>   ....
> 
> I intend to sort as 'Self', but actually it sort as 'Children'.
> 
> This patch fix this by only insert overhead && overhead_children
> when no overhead* field given.
> 
> Signed-off-by: Changbin Du <changbin.du@intel.com>

Acked-by: Jiri Olsa <jolsa@kernel.org>

thanks,
jirka

> ---
> v3: Add comment in code from Jiri
> v2: Add the example in commit message.
> 
> ---
>  tools/perf/util/sort.c | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
> index 8b327c9..9e087fb 100644
> --- a/tools/perf/util/sort.c
> +++ b/tools/perf/util/sort.c
> @@ -2635,6 +2635,13 @@ static char *setup_overhead(char *keys)
>  	if (sort__mode == SORT_MODE__DIFF)
>  		return keys;
>  
> +	/**
> +	 * User already stated overhead within -s option,
> +	 * do not mangle with that.
> +	 */
> +	if (strstr(keys, "overhead"))
> +		return keys;
> +
>  	keys = prefix_if_not_in("overhead", keys);
>  
>  	if (symbol_conf.cumulate_callchain)
> -- 
> 2.7.4
> 

^ permalink raw reply	[flat|nested] 10+ messages in thread

end of thread, other threads:[~2017-07-06 16:04 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-06-01  9:03 [PATCH] perf sort: only insert overhead && overhead_children when no overhead* field given changbin.du
2017-06-01 10:21 ` Jiri Olsa
2017-06-02  2:52   ` Du, Changbin
2017-06-02  3:22     ` Du, Changbin
2017-06-02  4:22 ` [PATCH v2] " changbin.du
2017-06-26  8:06   ` Du, Changbin
2017-06-29 15:22     ` Jiri Olsa
2017-07-05  3:02       ` Du, Changbin
2017-07-05  3:35 ` [PATCH v3] " changbin.du
2017-07-06 16:04   ` Jiri Olsa

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).