tools/perf/util/stat-display.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Aggregation is different for core events compared to uncore. For
uncore we don't want the counts for events not in the aggregation
id. Switch the !counter->pmu->is_uncore test to a
counter->pmu->is_core as the hwmon PMU doesn't set the uncore
boolean. There are 2 booleans as some PMUs forget to set either
cpumask or cpus that are used to determine whether a PMU is uncore or
core. Also check config->aggr_get_id is present as the non-core
'software' PMU doesn't have it set.
Before:
```
$ perf stat --per-core -e temp_cpu -a true
Performance counter stats for 'system wide':
S0-D0-C0 1 63.00 'C temp_cpu
S0-D0-C1 0 <not counted> 'C temp_cpu
S0-D0-C2 0 <not counted> 'C temp_cpu
S0-D0-C3 0 <not counted> 'C temp_cpu
S0-D0-C4 0 <not counted> 'C temp_cpu
S0-D0-C5 0 <not counted> 'C temp_cpu
S0-D0-C6 0 <not counted> 'C temp_cpu
S0-D0-C7 0 <not counted> 'C temp_cpu
0.001375790 seconds time elapsed
```
After:
```
$ perf stat --per-core -e temp_cpu -a true
Performance counter stats for 'system wide':
S0-D0-C0 1 38.00 'C temp_cpu
0.001260575 seconds time elapsed
```
Signed-off-by: Ian Rogers <irogers@google.com>
---
tools/perf/util/stat-display.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/perf/util/stat-display.c b/tools/perf/util/stat-display.c
index 91386429115d..ed3270ab2557 100644
--- a/tools/perf/util/stat-display.c
+++ b/tools/perf/util/stat-display.c
@@ -981,7 +981,7 @@ static bool should_skip_zero_counter(struct perf_stat_config *config,
* Skip value 0 when it's an uncore event and the given aggr id
* does not belong to the PMU cpumask.
*/
- if (!counter->pmu || !counter->pmu->is_uncore)
+ if (!counter->pmu || counter->pmu->is_core || !config->aggr_get_id)
return false;
perf_cpu_map__for_each_cpu(cpu, idx, counter->pmu->cpus) {
--
2.49.0.472.ge94155a9ec-goog
On 28/03/2025 4:53 am, Ian Rogers wrote:
> Aggregation is different for core events compared to uncore. For
> uncore we don't want the counts for events not in the aggregation
> id. Switch the !counter->pmu->is_uncore test to a
> counter->pmu->is_core as the hwmon PMU doesn't set the uncore
> boolean. There are 2 booleans as some PMUs forget to set either
> cpumask or cpus that are used to determine whether a PMU is uncore or
> core. Also check config->aggr_get_id is present as the non-core
> 'software' PMU doesn't have it set.
>
> Before:
> ```
> $ perf stat --per-core -e temp_cpu -a true
>
> Performance counter stats for 'system wide':
>
> S0-D0-C0 1 63.00 'C temp_cpu
> S0-D0-C1 0 <not counted> 'C temp_cpu
> S0-D0-C2 0 <not counted> 'C temp_cpu
> S0-D0-C3 0 <not counted> 'C temp_cpu
> S0-D0-C4 0 <not counted> 'C temp_cpu
> S0-D0-C5 0 <not counted> 'C temp_cpu
> S0-D0-C6 0 <not counted> 'C temp_cpu
> S0-D0-C7 0 <not counted> 'C temp_cpu
>
> 0.001375790 seconds time elapsed
> ```
>
> After:
> ```
> $ perf stat --per-core -e temp_cpu -a true
>
> Performance counter stats for 'system wide':
>
> S0-D0-C0 1 38.00 'C temp_cpu
>
> 0.001260575 seconds time elapsed
> ```
>
> Signed-off-by: Ian Rogers <irogers@google.com>
> ---
> tools/perf/util/stat-display.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/perf/util/stat-display.c b/tools/perf/util/stat-display.c
> index 91386429115d..ed3270ab2557 100644
> --- a/tools/perf/util/stat-display.c
> +++ b/tools/perf/util/stat-display.c
> @@ -981,7 +981,7 @@ static bool should_skip_zero_counter(struct perf_stat_config *config,
> * Skip value 0 when it's an uncore event and the given aggr id
> * does not belong to the PMU cpumask.
> */
> - if (!counter->pmu || !counter->pmu->is_uncore)
> + if (!counter->pmu || counter->pmu->is_core || !config->aggr_get_id)
> return false;
>
> perf_cpu_map__for_each_cpu(cpu, idx, counter->pmu->cpus) {
Reviewed-by: James Clark <james.clark@linaro.org>
© 2016 - 2025 Red Hat, Inc.