[PATCH] perf, tools, stat: Fix perf stat -T

[PATCH] perf, tools, stat: Fix perf stat -T

[Andi Kleen]

From: Andi Kleen <ak@linux.intel.com>

The transaction length metrics in perf stat -T broke recently.
It would not match the metric correctly and always print K/sec.
This was caused by a incorrect update of the cycles_in_tx statistics.
Update the correct variable.

Also the check for zero division was reversed, which resulted
in K/sec being printed for no transactions. Fix this also up.

Cc: jolsa@kernel.org
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
 tools/perf/util/stat-shadow.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c
index 53e8bb7..2a5d8d7 100644
--- a/tools/perf/util/stat-shadow.c
+++ b/tools/perf/util/stat-shadow.c
@@ -85,7 +85,7 @@ void perf_stat__update_shadow_stats(struct perf_evsel *counter, u64 *count,
 	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
 		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
 	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
-		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
+		update_stats(&runtime_cycles_in_tx_stats[ctx][cpu], count[0]);
 	else if (perf_stat_evsel__is(counter, TRANSACTION_START))
 		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
 	else if (perf_stat_evsel__is(counter, ELISION_START))
@@ -398,20 +398,18 @@ void perf_stat__print_shadow_stats(FILE *out, struct perf_evsel *evsel,
 				" #   %5.2f%% aborted cycles         ",
 				100.0 * ((total2-avg) / total));
 	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
-		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
 
-		if (total)
+		if (avg)
 			ratio = total / avg;
 
 		fprintf(out, " # %8.0f cycles / transaction   ", ratio);
 	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
-		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
 
-		if (total)
+		if (avg)
 			ratio = total / avg;
 
 		fprintf(out, " # %8.0f cycles / elision       ", ratio);
-- 
2.4.3

Re: [PATCH] perf, tools, stat: Fix perf stat -T

[Jiri Olsa]

On Mon, Jul 27, 2015 at 04:24:51PM -0700, Andi Kleen wrote:
> From: Andi Kleen <ak@linux.intel.com>
> 
> The transaction length metrics in perf stat -T broke recently.
> It would not match the metric correctly and always print K/sec.
> This was caused by a incorrect update of the cycles_in_tx statistics.
> Update the correct variable.
> 
> Also the check for zero division was reversed, which resulted
> in K/sec being printed for no transactions. Fix this also up.
> 
> Cc: jolsa@kernel.org
> Signed-off-by: Andi Kleen <ak@linux.intel.com>
> ---
>  tools/perf/util/stat-shadow.c | 8 +++-----
>  1 file changed, 3 insertions(+), 5 deletions(-)
> 
> diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c
> index 53e8bb7..2a5d8d7 100644
> --- a/tools/perf/util/stat-shadow.c
> +++ b/tools/perf/util/stat-shadow.c
> @@ -85,7 +85,7 @@ void perf_stat__update_shadow_stats(struct perf_evsel *counter, u64 *count,
>  	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
>  		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
>  	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
> -		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
> +		update_stats(&runtime_cycles_in_tx_stats[ctx][cpu], count[0]);

oops, looks like copy&paste issue.. thanks


>  	else if (perf_stat_evsel__is(counter, TRANSACTION_START))
>  		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
>  	else if (perf_stat_evsel__is(counter, ELISION_START))
> @@ -398,20 +398,18 @@ void perf_stat__print_shadow_stats(FILE *out, struct perf_evsel *evsel,
>  				" #   %5.2f%% aborted cycles         ",
>  				100.0 * ((total2-avg) / total));
>  	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
> -		   avg > 0 &&
>  		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
>  		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
>  
> -		if (total)
> +		if (avg)
>  			ratio = total / avg;
>  
>  		fprintf(out, " # %8.0f cycles / transaction   ", ratio);
>  	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
> -		   avg > 0 &&
>  		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
>  		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
>  
> -		if (total)
> +		if (avg)
>  			ratio = total / avg;
>  
>  		fprintf(out, " # %8.0f cycles / elision       ", ratio);
> -- 

we now print comment line if the avg is 0, but I think it's ok

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

thanks,
jirka

Re: [PATCH] perf, tools, stat: Fix perf stat -T

[Arnaldo Carvalho de Melo]

Em Tue, Jul 28, 2015 at 04:10:06PM +0200, Jiri Olsa escreveu:
> On Mon, Jul 27, 2015 at 04:24:51PM -0700, Andi Kleen wrote:
> > From: Andi Kleen <ak@linux.intel.com>
> > 
> > The transaction length metrics in perf stat -T broke recently.
> > It would not match the metric correctly and always print K/sec.
> > This was caused by a incorrect update of the cycles_in_tx statistics.
> > Update the correct variable.
> > 
> > Also the check for zero division was reversed, which resulted
> > in K/sec being printed for no transactions. Fix this also up.
> > 
> > Cc: jolsa@kernel.org
> > Signed-off-by: Andi Kleen <ak@linux.intel.com>
> > ---
> >  tools/perf/util/stat-shadow.c | 8 +++-----
> >  1 file changed, 3 insertions(+), 5 deletions(-)
> > 
> > diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c
> > index 53e8bb7..2a5d8d7 100644
> > --- a/tools/perf/util/stat-shadow.c
> > +++ b/tools/perf/util/stat-shadow.c
> > @@ -85,7 +85,7 @@ void perf_stat__update_shadow_stats(struct perf_evsel *counter, u64 *count,
> >  	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
> >  		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
> >  	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
> > -		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
> > +		update_stats(&runtime_cycles_in_tx_stats[ctx][cpu], count[0]);
> 
> oops, looks like copy&paste issue.. thanks
> 
> 
> we now print comment line if the avg is 0, but I think it's ok
> 
> Acked-by: Jiri Olsa <jolsa@kernel.org>

This affects just perf/core, right?

Re: [PATCH] perf, tools, stat: Fix perf stat -T

[Jiri Olsa]

On Tue, Jul 28, 2015 at 11:21:32AM -0300, Arnaldo Carvalho de Melo wrote:
> Em Tue, Jul 28, 2015 at 04:10:06PM +0200, Jiri Olsa escreveu:
> > On Mon, Jul 27, 2015 at 04:24:51PM -0700, Andi Kleen wrote:
> > > From: Andi Kleen <ak@linux.intel.com>
> > > 
> > > The transaction length metrics in perf stat -T broke recently.
> > > It would not match the metric correctly and always print K/sec.
> > > This was caused by a incorrect update of the cycles_in_tx statistics.
> > > Update the correct variable.
> > > 
> > > Also the check for zero division was reversed, which resulted
> > > in K/sec being printed for no transactions. Fix this also up.
> > > 
> > > Cc: jolsa@kernel.org
> > > Signed-off-by: Andi Kleen <ak@linux.intel.com>
> > > ---
> > >  tools/perf/util/stat-shadow.c | 8 +++-----
> > >  1 file changed, 3 insertions(+), 5 deletions(-)
> > > 
> > > diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c
> > > index 53e8bb7..2a5d8d7 100644
> > > --- a/tools/perf/util/stat-shadow.c
> > > +++ b/tools/perf/util/stat-shadow.c
> > > @@ -85,7 +85,7 @@ void perf_stat__update_shadow_stats(struct perf_evsel *counter, u64 *count,
> > >  	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
> > >  		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
> > >  	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
> > > -		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
> > > +		update_stats(&runtime_cycles_in_tx_stats[ctx][cpu], count[0]);
> > 
> > oops, looks like copy&paste issue.. thanks
> > 
> > 
> > we now print comment line if the avg is 0, but I think it's ok
> > 
> > Acked-by: Jiri Olsa <jolsa@kernel.org>
> 
> This affects just perf/core, right?

I guess, dont think it's an urgent one.. Andi, please speak up ;-)

jirka

Re: [PATCH] perf, tools, stat: Fix perf stat -T

[Arnaldo Carvalho de Melo]

Em Tue, Jul 28, 2015 at 04:30:18PM +0200, Jiri Olsa escreveu:
> On Tue, Jul 28, 2015 at 11:21:32AM -0300, Arnaldo Carvalho de Melo wrote:
> > Em Tue, Jul 28, 2015 at 04:10:06PM +0200, Jiri Olsa escreveu:
> > > On Mon, Jul 27, 2015 at 04:24:51PM -0700, Andi Kleen wrote:
> > > > +++ b/tools/perf/util/stat-shadow.c
> > > > +		update_stats(&runtime_cycles_in_tx_stats[ctx][cpu], count[0]);

> > > oops, looks like copy&paste issue.. thanks

> > > we now print comment line if the avg is 0, but I think it's ok

> > > Acked-by: Jiri Olsa <jolsa@kernel.org>
 
> > This affects just perf/core, right?
 
> I guess, dont think it's an urgent one.. Andi, please speak up ;-)

I'm not talking about it being urgent or not, I'm talking about it being
a bug present only in perf/core or if this is something that is in
perf/urgent, i.e. affects what is in Linus's upstream tree, etc.


- Arnaldo

Re: [PATCH] perf, tools, stat: Fix perf stat -T

[Jiri Olsa]

On Tue, Jul 28, 2015 at 11:32:57AM -0300, Arnaldo Carvalho de Melo wrote:
> Em Tue, Jul 28, 2015 at 04:30:18PM +0200, Jiri Olsa escreveu:
> > On Tue, Jul 28, 2015 at 11:21:32AM -0300, Arnaldo Carvalho de Melo wrote:
> > > Em Tue, Jul 28, 2015 at 04:10:06PM +0200, Jiri Olsa escreveu:
> > > > On Mon, Jul 27, 2015 at 04:24:51PM -0700, Andi Kleen wrote:
> > > > > +++ b/tools/perf/util/stat-shadow.c
> > > > > +		update_stats(&runtime_cycles_in_tx_stats[ctx][cpu], count[0]);
> 
> > > > oops, looks like copy&paste issue.. thanks
> 
> > > > we now print comment line if the avg is 0, but I think it's ok
> 
> > > > Acked-by: Jiri Olsa <jolsa@kernel.org>
>  
> > > This affects just perf/core, right?
>  
> > I guess, dont think it's an urgent one.. Andi, please speak up ;-)
> 
> I'm not talking about it being urgent or not, I'm talking about it being
> a bug present only in perf/core or if this is something that is in
> perf/urgent, i.e. affects what is in Linus's upstream tree, etc.

ah ok, it affects perf/urgent as well

jirka

Re: [PATCH] perf, tools, stat: Fix perf stat -T

[Andi Kleen]

> > I'm not talking about it being urgent or not, I'm talking about it being
> > a bug present only in perf/core or if this is something that is in
> > perf/urgent, i.e. affects what is in Linus's upstream tree, etc.
> 
> ah ok, it affects perf/urgent as well

I think there are some more typos in the cache statistics. 
Compare perf stat -d -d -d -d output with an old perf.

-Andi

-- 
ak@linux.intel.com -- Speaking for myself only

[tip:perf/core] perf stat: Fix transaction lenght metrics

[tip-bot for Andi Kleen]

Commit-ID:  5497628576a3c5f3dbab224fa5a5d027f43d8b50
Gitweb:     http://git.kernel.org/tip/5497628576a3c5f3dbab224fa5a5d027f43d8b50
Author:     Andi Kleen <ak@linux.intel.com>
AuthorDate: Mon, 27 Jul 2015 16:24:51 -0700
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Tue, 28 Jul 2015 12:05:04 -0300

perf stat: Fix transaction lenght metrics

The transaction length metrics in perf stat -T broke recently.

It would not match the metric correctly and always print K/sec.

This was caused by a incorrect update of the cycles_in_tx statistics.

Update the correct variable.

Also the check for zero division was reversed, which resulted in K/sec
being printed for no transactions. Fix this also up.

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Link: http://lkml.kernel.org/r/1438039491-22091-1-git-send-email-andi@firstfloor.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/util/stat-shadow.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c
index 53e8bb7..2a5d8d7 100644
--- a/tools/perf/util/stat-shadow.c
+++ b/tools/perf/util/stat-shadow.c
@@ -85,7 +85,7 @@ void perf_stat__update_shadow_stats(struct perf_evsel *counter, u64 *count,
 	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
 		update_stats(&runtime_cycles_stats[ctx][cpu], count[0]);
 	else if (perf_stat_evsel__is(counter, CYCLES_IN_TX))
-		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
+		update_stats(&runtime_cycles_in_tx_stats[ctx][cpu], count[0]);
 	else if (perf_stat_evsel__is(counter, TRANSACTION_START))
 		update_stats(&runtime_transaction_stats[ctx][cpu], count[0]);
 	else if (perf_stat_evsel__is(counter, ELISION_START))
@@ -398,20 +398,18 @@ void perf_stat__print_shadow_stats(FILE *out, struct perf_evsel *evsel,
 				" #   %5.2f%% aborted cycles         ",
 				100.0 * ((total2-avg) / total));
 	} else if (perf_stat_evsel__is(evsel, TRANSACTION_START) &&
-		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
 
-		if (total)
+		if (avg)
 			ratio = total / avg;
 
 		fprintf(out, " # %8.0f cycles / transaction   ", ratio);
 	} else if (perf_stat_evsel__is(evsel, ELISION_START) &&
-		   avg > 0 &&
 		   runtime_cycles_in_tx_stats[ctx][cpu].n != 0) {
 		total = avg_stats(&runtime_cycles_in_tx_stats[ctx][cpu]);
 
-		if (total)
+		if (avg)
 			ratio = total / avg;
 
 		fprintf(out, " # %8.0f cycles / elision       ", ratio);