perf metricgroup: Avoid scanning unnecessary PMUs for identifier match

[PATCH v1] perf metricgroup: Avoid scanning unnecessary PMUs for identifier match

Posted by Ian Rogers 6 days, 18 hours ago

Only uncore PMUs can have an identifier, so add an optimized
perf_pmus__scan routine for that case to avoid all PMU types being
created.

Signed-off-by: Ian Rogers <irogers@google.com>
---
 tools/perf/util/metricgroup.c |  8 ++------
 tools/perf/util/pmus.c        | 18 +++++++++++++++++-
 tools/perf/util/pmus.h        |  1 +
 3 files changed, 20 insertions(+), 7 deletions(-)

diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
index 7e39d469111b..769b38400832 100644
--- a/tools/perf/util/metricgroup.c
+++ b/tools/perf/util/metricgroup.c
@@ -410,13 +410,9 @@ static int metricgroup__sys_event_iter(const struct pmu_metric *pm,
 	if (!pm->metric_expr || !pm->compat)
 		return 0;
 
-	while ((pmu = perf_pmus__scan(pmu))) {
-
-		if (!pmu->id || !pmu_uncore_identifier_match(pm->compat, pmu->id))
-			continue;
-
+	while ((pmu = perf_pmus__scan_for_uncore_id(pmu, pm->compat)))
 		return d->fn(pm, table, d->data);
-	}
+
 	return 0;
 }
 
diff --git a/tools/perf/util/pmus.c b/tools/perf/util/pmus.c
index 98be2eb8f1f0..9859d7056fb7 100644
--- a/tools/perf/util/pmus.c
+++ b/tools/perf/util/pmus.c
@@ -409,7 +409,7 @@ struct perf_pmu *perf_pmus__scan_matching_wildcard(struct perf_pmu *pmu, const c
 	if (!pmu) {
 		/*
 		 * Core PMUs, other sysfs PMUs and tool PMU can have any name or
-		 * aren't wother optimizing for.
+		 * aren't worth optimizing for.
 		 */
 		unsigned int to_read_pmus =  PERF_TOOL_PMU_TYPE_PE_CORE_MASK |
 			PERF_TOOL_PMU_TYPE_PE_OTHER_MASK |
@@ -486,6 +486,22 @@ static struct perf_pmu *perf_pmus__scan_skip_duplicates(struct perf_pmu *pmu)
 	return NULL;
 }
 
+struct perf_pmu *perf_pmus__scan_for_uncore_id(struct perf_pmu *pmu, const char *compat)
+{
+	if (!pmu) {
+		/* Only uncore PMUs can have identifiers. */
+		unsigned int to_read_pmus = PERF_TOOL_PMU_TYPE_PE_OTHER_MASK;
+
+		pmu_read_sysfs(to_read_pmus);
+		pmu = list_prepare_entry(pmu, &other_pmus, list);
+	}
+	list_for_each_entry_continue(pmu, &other_pmus, list) {
+		if (pmu->id && pmu_uncore_identifier_match(compat, pmu->id))
+			return pmu;
+	}
+	return NULL;
+}
+
 const struct perf_pmu *perf_pmus__pmu_for_pmu_filter(const char *str)
 {
 	struct perf_pmu *pmu = NULL;
diff --git a/tools/perf/util/pmus.h b/tools/perf/util/pmus.h
index 7cb36863711a..0d55edb3f2fc 100644
--- a/tools/perf/util/pmus.h
+++ b/tools/perf/util/pmus.h
@@ -23,6 +23,7 @@ struct perf_pmu *perf_pmus__scan(struct perf_pmu *pmu);
 struct perf_pmu *perf_pmus__scan_core(struct perf_pmu *pmu);
 struct perf_pmu *perf_pmus__scan_for_event(struct perf_pmu *pmu, const char *event);
 struct perf_pmu *perf_pmus__scan_matching_wildcard(struct perf_pmu *pmu, const char *wildcard);
+struct perf_pmu *perf_pmus__scan_for_uncore_id(struct perf_pmu *pmu, const char *compat);
 
 const struct perf_pmu *perf_pmus__pmu_for_pmu_filter(const char *str);
 
-- 
2.53.0.1018.g2bb0e51243-goog

Re: [PATCH v1] perf metricgroup: Avoid scanning unnecessary PMUs for identifier match

Posted by James Clark 3 days, 2 hours ago


On 26/03/2026 10:44 pm, Ian Rogers wrote:
> Only uncore PMUs can have an identifier, so add an optimized
> perf_pmus__scan routine for that case to avoid all PMU types being
> created.
> 
> Signed-off-by: Ian Rogers <irogers@google.com>

Reviewed-by: James Clark <james.clark@linaro.org>

> ---
>   tools/perf/util/metricgroup.c |  8 ++------
>   tools/perf/util/pmus.c        | 18 +++++++++++++++++-
>   tools/perf/util/pmus.h        |  1 +
>   3 files changed, 20 insertions(+), 7 deletions(-)
> 
> diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
> index 7e39d469111b..769b38400832 100644
> --- a/tools/perf/util/metricgroup.c
> +++ b/tools/perf/util/metricgroup.c
> @@ -410,13 +410,9 @@ static int metricgroup__sys_event_iter(const struct pmu_metric *pm,
>   	if (!pm->metric_expr || !pm->compat)
>   		return 0;
>   
> -	while ((pmu = perf_pmus__scan(pmu))) {
> -
> -		if (!pmu->id || !pmu_uncore_identifier_match(pm->compat, pmu->id))
> -			continue;
> -
> +	while ((pmu = perf_pmus__scan_for_uncore_id(pmu, pm->compat)))
>   		return d->fn(pm, table, d->data);
> -	}
> +
>   	return 0;
>   }
>   
> diff --git a/tools/perf/util/pmus.c b/tools/perf/util/pmus.c
> index 98be2eb8f1f0..9859d7056fb7 100644
> --- a/tools/perf/util/pmus.c
> +++ b/tools/perf/util/pmus.c
> @@ -409,7 +409,7 @@ struct perf_pmu *perf_pmus__scan_matching_wildcard(struct perf_pmu *pmu, const c
>   	if (!pmu) {
>   		/*
>   		 * Core PMUs, other sysfs PMUs and tool PMU can have any name or
> -		 * aren't wother optimizing for.
> +		 * aren't worth optimizing for.
>   		 */
>   		unsigned int to_read_pmus =  PERF_TOOL_PMU_TYPE_PE_CORE_MASK |
>   			PERF_TOOL_PMU_TYPE_PE_OTHER_MASK |
> @@ -486,6 +486,22 @@ static struct perf_pmu *perf_pmus__scan_skip_duplicates(struct perf_pmu *pmu)
>   	return NULL;
>   }
>   
> +struct perf_pmu *perf_pmus__scan_for_uncore_id(struct perf_pmu *pmu, const char *compat)
> +{
> +	if (!pmu) {
> +		/* Only uncore PMUs can have identifiers. */
> +		unsigned int to_read_pmus = PERF_TOOL_PMU_TYPE_PE_OTHER_MASK;
> +
> +		pmu_read_sysfs(to_read_pmus);
> +		pmu = list_prepare_entry(pmu, &other_pmus, list);
> +	}
> +	list_for_each_entry_continue(pmu, &other_pmus, list) {
> +		if (pmu->id && pmu_uncore_identifier_match(compat, pmu->id))
> +			return pmu;
> +	}
> +	return NULL;
> +}
> +
>   const struct perf_pmu *perf_pmus__pmu_for_pmu_filter(const char *str)
>   {
>   	struct perf_pmu *pmu = NULL;
> diff --git a/tools/perf/util/pmus.h b/tools/perf/util/pmus.h
> index 7cb36863711a..0d55edb3f2fc 100644
> --- a/tools/perf/util/pmus.h
> +++ b/tools/perf/util/pmus.h
> @@ -23,6 +23,7 @@ struct perf_pmu *perf_pmus__scan(struct perf_pmu *pmu);
>   struct perf_pmu *perf_pmus__scan_core(struct perf_pmu *pmu);
>   struct perf_pmu *perf_pmus__scan_for_event(struct perf_pmu *pmu, const char *event);
>   struct perf_pmu *perf_pmus__scan_matching_wildcard(struct perf_pmu *pmu, const char *wildcard);
> +struct perf_pmu *perf_pmus__scan_for_uncore_id(struct perf_pmu *pmu, const char *compat);
>   
>   const struct perf_pmu *perf_pmus__pmu_for_pmu_filter(const char *str);
>

Re: [PATCH v1] perf metricgroup: Avoid scanning unnecessary PMUs for identifier match

Posted by Namhyung Kim 2 days, 10 hours ago

On Mon, Mar 30, 2026 at 03:44:21PM +0100, James Clark wrote:
> 
> 
> On 26/03/2026 10:44 pm, Ian Rogers wrote:
> > Only uncore PMUs can have an identifier, so add an optimized
> > perf_pmus__scan routine for that case to avoid all PMU types being
> > created.
> > 
> > Signed-off-by: Ian Rogers <irogers@google.com>
> 
> Reviewed-by: James Clark <james.clark@linaro.org>
> 
> > ---
> >   tools/perf/util/metricgroup.c |  8 ++------
> >   tools/perf/util/pmus.c        | 18 +++++++++++++++++-
> >   tools/perf/util/pmus.h        |  1 +
> >   3 files changed, 20 insertions(+), 7 deletions(-)
> > 
> > diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
> > index 7e39d469111b..769b38400832 100644
> > --- a/tools/perf/util/metricgroup.c
> > +++ b/tools/perf/util/metricgroup.c
> > @@ -410,13 +410,9 @@ static int metricgroup__sys_event_iter(const struct pmu_metric *pm,
> >   	if (!pm->metric_expr || !pm->compat)
> >   		return 0;
> > -	while ((pmu = perf_pmus__scan(pmu))) {
> > -
> > -		if (!pmu->id || !pmu_uncore_identifier_match(pm->compat, pmu->id))
> > -			continue;
> > -
> > +	while ((pmu = perf_pmus__scan_for_uncore_id(pmu, pm->compat)))
> >   		return d->fn(pm, table, d->data);

Sashiko review: it's natural to convert it to 'if'.

Thanks,
Namhyung


> > -	}
> > +
> >   	return 0;
> >   }
> > diff --git a/tools/perf/util/pmus.c b/tools/perf/util/pmus.c
> > index 98be2eb8f1f0..9859d7056fb7 100644
> > --- a/tools/perf/util/pmus.c
> > +++ b/tools/perf/util/pmus.c
> > @@ -409,7 +409,7 @@ struct perf_pmu *perf_pmus__scan_matching_wildcard(struct perf_pmu *pmu, const c
> >   	if (!pmu) {
> >   		/*
> >   		 * Core PMUs, other sysfs PMUs and tool PMU can have any name or
> > -		 * aren't wother optimizing for.
> > +		 * aren't worth optimizing for.
> >   		 */
> >   		unsigned int to_read_pmus =  PERF_TOOL_PMU_TYPE_PE_CORE_MASK |
> >   			PERF_TOOL_PMU_TYPE_PE_OTHER_MASK |
> > @@ -486,6 +486,22 @@ static struct perf_pmu *perf_pmus__scan_skip_duplicates(struct perf_pmu *pmu)
> >   	return NULL;
> >   }
> > +struct perf_pmu *perf_pmus__scan_for_uncore_id(struct perf_pmu *pmu, const char *compat)
> > +{
> > +	if (!pmu) {
> > +		/* Only uncore PMUs can have identifiers. */
> > +		unsigned int to_read_pmus = PERF_TOOL_PMU_TYPE_PE_OTHER_MASK;
> > +
> > +		pmu_read_sysfs(to_read_pmus);
> > +		pmu = list_prepare_entry(pmu, &other_pmus, list);
> > +	}
> > +	list_for_each_entry_continue(pmu, &other_pmus, list) {
> > +		if (pmu->id && pmu_uncore_identifier_match(compat, pmu->id))
> > +			return pmu;
> > +	}
> > +	return NULL;
> > +}
> > +
> >   const struct perf_pmu *perf_pmus__pmu_for_pmu_filter(const char *str)
> >   {
> >   	struct perf_pmu *pmu = NULL;
> > diff --git a/tools/perf/util/pmus.h b/tools/perf/util/pmus.h
> > index 7cb36863711a..0d55edb3f2fc 100644
> > --- a/tools/perf/util/pmus.h
> > +++ b/tools/perf/util/pmus.h
> > @@ -23,6 +23,7 @@ struct perf_pmu *perf_pmus__scan(struct perf_pmu *pmu);
> >   struct perf_pmu *perf_pmus__scan_core(struct perf_pmu *pmu);
> >   struct perf_pmu *perf_pmus__scan_for_event(struct perf_pmu *pmu, const char *event);
> >   struct perf_pmu *perf_pmus__scan_matching_wildcard(struct perf_pmu *pmu, const char *wildcard);
> > +struct perf_pmu *perf_pmus__scan_for_uncore_id(struct perf_pmu *pmu, const char *compat);
> >   const struct perf_pmu *perf_pmus__pmu_for_pmu_filter(const char *str);
>

Re: [PATCH v1] perf metricgroup: Avoid scanning unnecessary PMUs for identifier match

Posted by Ian Rogers 2 days, 3 hours ago

On Mon, Mar 30, 2026 at 11:30 PM Namhyung Kim <namhyung@kernel.org> wrote:
>
> On Mon, Mar 30, 2026 at 03:44:21PM +0100, James Clark wrote:
> >
> >
> > On 26/03/2026 10:44 pm, Ian Rogers wrote:
> > > Only uncore PMUs can have an identifier, so add an optimized
> > > perf_pmus__scan routine for that case to avoid all PMU types being
> > > created.
> > >
> > > Signed-off-by: Ian Rogers <irogers@google.com>
> >
> > Reviewed-by: James Clark <james.clark@linaro.org>
> >
> > > ---
> > >   tools/perf/util/metricgroup.c |  8 ++------
> > >   tools/perf/util/pmus.c        | 18 +++++++++++++++++-
> > >   tools/perf/util/pmus.h        |  1 +
> > >   3 files changed, 20 insertions(+), 7 deletions(-)
> > >
> > > diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
> > > index 7e39d469111b..769b38400832 100644
> > > --- a/tools/perf/util/metricgroup.c
> > > +++ b/tools/perf/util/metricgroup.c
> > > @@ -410,13 +410,9 @@ static int metricgroup__sys_event_iter(const struct pmu_metric *pm,
> > >     if (!pm->metric_expr || !pm->compat)
> > >             return 0;
> > > -   while ((pmu = perf_pmus__scan(pmu))) {
> > > -
> > > -           if (!pmu->id || !pmu_uncore_identifier_match(pm->compat, pmu->id))
> > > -                   continue;
> > > -
> > > +   while ((pmu = perf_pmus__scan_for_uncore_id(pmu, pm->compat)))
> > >             return d->fn(pm, table, d->data);
>
> Sashiko review: it's natural to convert it to 'if'.

Yeah, I saw and disagreed. The pattern with the  "while ((pmu =
perf_pmus__scan...(pmu)))" functions is for them to be a while loop.
With an "if" it would read:
```
pmu = perf_pmus__scan...(pmu);
if (pmu)
   return ...
```
and we lose the consistency of having while loops in the code.

Thanks,
Ian

> Thanks,
> Namhyung
>
>
> > > -   }
> > > +
> > >     return 0;
> > >   }
> > > diff --git a/tools/perf/util/pmus.c b/tools/perf/util/pmus.c
> > > index 98be2eb8f1f0..9859d7056fb7 100644
> > > --- a/tools/perf/util/pmus.c
> > > +++ b/tools/perf/util/pmus.c
> > > @@ -409,7 +409,7 @@ struct perf_pmu *perf_pmus__scan_matching_wildcard(struct perf_pmu *pmu, const c
> > >     if (!pmu) {
> > >             /*
> > >              * Core PMUs, other sysfs PMUs and tool PMU can have any name or
> > > -            * aren't wother optimizing for.
> > > +            * aren't worth optimizing for.
> > >              */
> > >             unsigned int to_read_pmus =  PERF_TOOL_PMU_TYPE_PE_CORE_MASK |
> > >                     PERF_TOOL_PMU_TYPE_PE_OTHER_MASK |
> > > @@ -486,6 +486,22 @@ static struct perf_pmu *perf_pmus__scan_skip_duplicates(struct perf_pmu *pmu)
> > >     return NULL;
> > >   }
> > > +struct perf_pmu *perf_pmus__scan_for_uncore_id(struct perf_pmu *pmu, const char *compat)
> > > +{
> > > +   if (!pmu) {
> > > +           /* Only uncore PMUs can have identifiers. */
> > > +           unsigned int to_read_pmus = PERF_TOOL_PMU_TYPE_PE_OTHER_MASK;
> > > +
> > > +           pmu_read_sysfs(to_read_pmus);
> > > +           pmu = list_prepare_entry(pmu, &other_pmus, list);
> > > +   }
> > > +   list_for_each_entry_continue(pmu, &other_pmus, list) {
> > > +           if (pmu->id && pmu_uncore_identifier_match(compat, pmu->id))
> > > +                   return pmu;
> > > +   }
> > > +   return NULL;
> > > +}
> > > +
> > >   const struct perf_pmu *perf_pmus__pmu_for_pmu_filter(const char *str)
> > >   {
> > >     struct perf_pmu *pmu = NULL;
> > > diff --git a/tools/perf/util/pmus.h b/tools/perf/util/pmus.h
> > > index 7cb36863711a..0d55edb3f2fc 100644
> > > --- a/tools/perf/util/pmus.h
> > > +++ b/tools/perf/util/pmus.h
> > > @@ -23,6 +23,7 @@ struct perf_pmu *perf_pmus__scan(struct perf_pmu *pmu);
> > >   struct perf_pmu *perf_pmus__scan_core(struct perf_pmu *pmu);
> > >   struct perf_pmu *perf_pmus__scan_for_event(struct perf_pmu *pmu, const char *event);
> > >   struct perf_pmu *perf_pmus__scan_matching_wildcard(struct perf_pmu *pmu, const char *wildcard);
> > > +struct perf_pmu *perf_pmus__scan_for_uncore_id(struct perf_pmu *pmu, const char *compat);
> > >   const struct perf_pmu *perf_pmus__pmu_for_pmu_filter(const char *str);
> >

Re: [PATCH v1] perf metricgroup: Avoid scanning unnecessary PMUs for identifier match

Posted by Namhyung Kim 1 day, 13 hours ago

On Tue, Mar 31, 2026 at 07:05:58AM -0700, Ian Rogers wrote:
> On Mon, Mar 30, 2026 at 11:30 PM Namhyung Kim <namhyung@kernel.org> wrote:
> >
> > On Mon, Mar 30, 2026 at 03:44:21PM +0100, James Clark wrote:
> > >
> > >
> > > On 26/03/2026 10:44 pm, Ian Rogers wrote:
> > > > Only uncore PMUs can have an identifier, so add an optimized
> > > > perf_pmus__scan routine for that case to avoid all PMU types being
> > > > created.
> > > >
> > > > Signed-off-by: Ian Rogers <irogers@google.com>
> > >
> > > Reviewed-by: James Clark <james.clark@linaro.org>
> > >
> > > > ---
> > > >   tools/perf/util/metricgroup.c |  8 ++------
> > > >   tools/perf/util/pmus.c        | 18 +++++++++++++++++-
> > > >   tools/perf/util/pmus.h        |  1 +
> > > >   3 files changed, 20 insertions(+), 7 deletions(-)
> > > >
> > > > diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
> > > > index 7e39d469111b..769b38400832 100644
> > > > --- a/tools/perf/util/metricgroup.c
> > > > +++ b/tools/perf/util/metricgroup.c
> > > > @@ -410,13 +410,9 @@ static int metricgroup__sys_event_iter(const struct pmu_metric *pm,
> > > >     if (!pm->metric_expr || !pm->compat)
> > > >             return 0;
> > > > -   while ((pmu = perf_pmus__scan(pmu))) {
> > > > -
> > > > -           if (!pmu->id || !pmu_uncore_identifier_match(pm->compat, pmu->id))
> > > > -                   continue;
> > > > -
> > > > +   while ((pmu = perf_pmus__scan_for_uncore_id(pmu, pm->compat)))
> > > >             return d->fn(pm, table, d->data);
> >
> > Sashiko review: it's natural to convert it to 'if'.
> 
> Yeah, I saw and disagreed. The pattern with the  "while ((pmu =
> perf_pmus__scan...(pmu)))" functions is for them to be a while loop.
> With an "if" it would read:
> ```
> pmu = perf_pmus__scan...(pmu);
> if (pmu)
>    return ...
> ```
> and we lose the consistency of having while loops in the code.

I understand it's a pattern to scan PMUs but having an unused
variable assignment in a while loop with a return statement seems
unnatural and maybe someone would send a patch for it later.

At least we can add a comment saying it's intended?

Thanks,
Namhyung