* [PATCH 0/1] perf: Add --pmu-filter option for filtering PMUs
@ 2026-02-25 3:41 Qinxin Xia
2026-02-25 3:41 ` [PATCH 1/1] " Qinxin Xia
0 siblings, 1 reply; 4+ messages in thread
From: Qinxin Xia @ 2026-02-25 3:41 UTC (permalink / raw)
To: irogers, peterz, mingo, acme, namhyung, wangyushan12, hejunhao3,
jonathan.cameron, xiaqinxin
Cc: linux-perf-users, linux-kernel, mark.rutland, alexander.shishkin,
jolsa, adrian.hunter, "james.clark
This patch adds a new --pmu-filter option to perf-stat command to allow
filtering events on specific PMUs. This is useful when there are
multiple PMUs with same type (e.g. hisi_sicl2_cpa0 and hisi_sicl0_cpa0).
perf event:
[root@localhost tmp]# perf stat -e cpa_cycles
Performance counter stats for 'system wide':
3,005,509,659 hisi_sicl0_cpa0/cpa_cycles/
3,005,502,319 hisi_sicl2_cpa0/cpa_cycles/
3.005460970 seconds time elapsed
[root@localhost tmp]# perf stat --pmu-filter hisi_sicl0_cpa0 -e cpa_cycles
Performance counter stats for 'system wide':
3,003,899,571 cpa_cycles
3.003864470 seconds time elapsed
perf metric:
[root@localhost tmp]# perf stat -M cpa_p0_avg_bw
Performance counter stats for 'system wide':
19,417,779,115 hisi_sicl0_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
0 hisi_sicl0_cpa0/cpa_p0_wr_dat/
0 hisi_sicl0_cpa0/cpa_p0_rd_dat_64b/
0 hisi_sicl0_cpa0/cpa_p0_rd_dat_32b/
19,417,751,103 hisi_sicl10_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
0 hisi_sicl10_cpa0/cpa_p0_wr_dat/
0 hisi_sicl10_cpa0/cpa_p0_rd_dat_64b/
0 hisi_sicl10_cpa0/cpa_p0_rd_dat_32b/
19,417,730,679 hisi_sicl2_cpa0/cpa_cycles/ # 0.31 cpa_p0_avg_bw
75,635,749 hisi_sicl2_cpa0/cpa_p0_wr_dat/
18,520,640 hisi_sicl2_cpa0/cpa_p0_rd_dat_64b/
0 hisi_sicl2_cpa0/cpa_p0_rd_dat_32b/
19,417,674,227 hisi_sicl8_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
0 hisi_sicl8_cpa0/cpa_p0_wr_dat/
0 hisi_sicl8_cpa0/cpa_p0_rd_dat_64b/
0 hisi_sicl8_cpa0/cpa_p0_rd_dat_32b/
19.417734480 seconds time elapsed
[root@localhost tmp]# perf stat --pmu-filter hisi_sicl2_cpa0 -M cpa_p0_avg_bw
Performance counter stats for 'system wide':
6,234,093,559 cpa_cycles # 0.60 cpa_p0_avg_bw
50,548,465 cpa_p0_wr_dat
Changes since RFC:
- Address the comments from Ian, replace '--uncorepmu' with '--pmu-filter'.
- Link: https://lore.kernel.org/linux-perf-users/20260120095058.2778834-1-xiaqinxin@huawei.com/
Qinxin Xia (1):
perf: Add --pmu-filter option for filtering PMUs
tools/perf/Documentation/perf-stat.txt | 4 ++++
tools/perf/builtin-stat.c | 26 ++++++++++++++++++++++++++
tools/perf/util/metricgroup.c | 18 +++++++++++++-----
3 files changed, 43 insertions(+), 5 deletions(-)
--
2.33.0
^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH 1/1] perf: Add --pmu-filter option for filtering PMUs
2026-02-25 3:41 [PATCH 0/1] perf: Add --pmu-filter option for filtering PMUs Qinxin Xia
@ 2026-02-25 3:41 ` Qinxin Xia
2026-03-04 1:19 ` Namhyung Kim
0 siblings, 1 reply; 4+ messages in thread
From: Qinxin Xia @ 2026-02-25 3:41 UTC (permalink / raw)
To: irogers, peterz, mingo, acme, namhyung, wangyushan12, hejunhao3,
jonathan.cameron, xiaqinxin
Cc: linux-perf-users, linux-kernel, mark.rutland, alexander.shishkin,
jolsa, adrian.hunter, "james.clark
This patch adds a new --pmu-filter option to perf-stat command to allow
filtering events on specific PMUs. This is useful when there are
multiple PMUs with same type (e.g. hisi_sicl2_cpa0 and hisi_sicl0_cpa0).
perf event:
[root@localhost tmp]# perf stat -e cpa_cycles
Performance counter stats for 'system wide':
3,005,509,659 hisi_sicl0_cpa0/cpa_cycles/
3,005,502,319 hisi_sicl2_cpa0/cpa_cycles/
3.005460970 seconds time elapsed
[root@localhost tmp]# perf stat --pmu-filter hisi_sicl0_cpa0 -e cpa_cycles
Performance counter stats for 'system wide':
3,003,899,571 cpa_cycles
3.003864470 seconds time elapsed
perf metric:
[root@localhost tmp]# perf stat -M cpa_p0_avg_bw
Performance counter stats for 'system wide':
19,417,779,115 hisi_sicl0_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
0 hisi_sicl0_cpa0/cpa_p0_wr_dat/
0 hisi_sicl0_cpa0/cpa_p0_rd_dat_64b/
0 hisi_sicl0_cpa0/cpa_p0_rd_dat_32b/
19,417,751,103 hisi_sicl10_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
0 hisi_sicl10_cpa0/cpa_p0_wr_dat/
0 hisi_sicl10_cpa0/cpa_p0_rd_dat_64b/
0 hisi_sicl10_cpa0/cpa_p0_rd_dat_32b/
19,417,730,679 hisi_sicl2_cpa0/cpa_cycles/ # 0.31 cpa_p0_avg_bw
75,635,749 hisi_sicl2_cpa0/cpa_p0_wr_dat/
18,520,640 hisi_sicl2_cpa0/cpa_p0_rd_dat_64b/
0 hisi_sicl2_cpa0/cpa_p0_rd_dat_32b/
19,417,674,227 hisi_sicl8_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
0 hisi_sicl8_cpa0/cpa_p0_wr_dat/
0 hisi_sicl8_cpa0/cpa_p0_rd_dat_64b/
0 hisi_sicl8_cpa0/cpa_p0_rd_dat_32b/
19.417734480 seconds time elapsed
[root@localhost tmp]# perf stat --pmu-filter hisi_sicl2_cpa0 -M cpa_p0_avg_bw
Performance counter stats for 'system wide':
6,234,093,559 cpa_cycles # 0.60 cpa_p0_avg_bw
50,548,465 cpa_p0_wr_dat
7,552,182 cpa_p0_rd_dat_64b
0 cpa_p0_rd_dat_32b
6.234139320 seconds time elapsed
Signed-off-by: Qinxin Xia <xiaqinxin@huawei.com>
---
tools/perf/Documentation/perf-stat.txt | 4 ++++
tools/perf/builtin-stat.c | 26 ++++++++++++++++++++++++++
tools/perf/util/metricgroup.c | 18 +++++++++++++-----
3 files changed, 43 insertions(+), 5 deletions(-)
diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
index 7cccc3a847d1..b72a29c9223c 100644
--- a/tools/perf/Documentation/perf-stat.txt
+++ b/tools/perf/Documentation/perf-stat.txt
@@ -578,6 +578,10 @@ $ perf config stat.no-csv-summary=true
Only enable events on applying cpu with this type for hybrid platform
(e.g. core or atom)"
+--pmu-filter::
+Only enable events on applying pmu with specified for multiple
+pmus with same type (e.g. hisi_sicl2_cpa0 or hisi_sicl0_cpa0)
+
EXAMPLES
--------
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 73c2ba7e3076..67fb91816c0b 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -1214,6 +1214,28 @@ static int parse_cputype(const struct option *opt,
return 0;
}
+static int parse_pmu_filter(const struct option *opt,
+ const char *str,
+ int unset __maybe_unused)
+{
+ const struct perf_pmu *pmu;
+ struct evlist *evlist = *(struct evlist **)opt->value;
+
+ if (!list_empty(&evlist->core.entries)) {
+ fprintf(stderr, "Must define pmu-filter before events/metrics\n");
+ return -1;
+ }
+
+ pmu = perf_pmus__pmu_for_pmu_filter(str);
+ if (!pmu) {
+ fprintf(stderr, "--pmu-filter %s is not supported!\n", str);
+ return -1;
+ }
+ parse_events_option_args.pmu_filter = pmu->name;
+
+ return 0;
+}
+
static int parse_cache_level(const struct option *opt,
const char *str,
int unset __maybe_unused)
@@ -2561,6 +2583,10 @@ int cmd_stat(int argc, const char **argv)
"Only enable events on applying cpu with this type "
"for hybrid platform (e.g. core or atom)",
parse_cputype),
+ OPT_CALLBACK(0, "pmu-filter", &evsel_list, "pmu",
+ "Only enable events on applying pmu with specified "
+ "for multiple pmus with same type(e.g. hisi_sicl2_cpa0 or hisi_sicl0_cpa0)",
+ parse_pmu_filter),
#ifdef HAVE_LIBPFM
OPT_CALLBACK(0, "pfm-events", &evsel_list, "event",
"libpfm4 event selector. use 'perf list' to list available events",
diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
index 46bf4dfeebc8..89b83ca38483 100644
--- a/tools/perf/util/metricgroup.c
+++ b/tools/perf/util/metricgroup.c
@@ -387,8 +387,13 @@ static bool match_pm_metric_or_groups(const struct pmu_metric *pm, const char *p
const char *metric_or_groups)
{
const char *pm_pmu = pm->pmu ?: "cpu";
+ struct perf_pmu *perf_pmu = NULL;
- if (strcmp(pmu, "all") && strcmp(pm_pmu, pmu))
+ if (pm->pmu)
+ perf_pmu = perf_pmus__find(pm->pmu);
+
+ if (strcmp(pmu, "all") && strcmp(pm_pmu, pmu) &&
+ (perf_pmu && !perf_pmu__name_wildcard_match(perf_pmu, pmu)))
return false;
return match_metric_or_groups(pm->metric_group, metric_or_groups) ||
@@ -1259,7 +1264,8 @@ static int build_combined_expr_ctx(const struct list_head *metric_list,
static int parse_ids(bool metric_no_merge, bool fake_pmu,
struct expr_parse_ctx *ids, const char *modifier,
bool group_events, const bool tool_events[TOOL_PMU__EVENT_MAX],
- struct evlist **out_evlist)
+ struct evlist **out_evlist,
+ const char *filter_pmu)
{
struct parse_events_error parse_error;
struct evlist *parsed_evlist;
@@ -1313,7 +1319,7 @@ static int parse_ids(bool metric_no_merge, bool fake_pmu,
}
pr_debug("Parsing metric events '%s'\n", events.buf);
parse_events_error__init(&parse_error);
- ret = __parse_events(parsed_evlist, events.buf, /*pmu_filter=*/NULL,
+ ret = __parse_events(parsed_evlist, events.buf, filter_pmu,
&parse_error, fake_pmu, /*warn_if_reordered=*/false,
/*fake_tp=*/false);
if (ret) {
@@ -1416,7 +1422,8 @@ static int parse_groups(struct evlist *perf_evlist,
/*modifier=*/NULL,
/*group_events=*/false,
tool_events,
- &combined_evlist);
+ &combined_evlist,
+ (pmu && strcmp(pmu, "all") == 0) ? NULL : pmu);
}
if (combined)
expr__ctx_free(combined);
@@ -1471,7 +1478,8 @@ static int parse_groups(struct evlist *perf_evlist,
}
if (!metric_evlist) {
ret = parse_ids(metric_no_merge, fake_pmu, m->pctx, m->modifier,
- m->group_events, tool_events, &m->evlist);
+ m->group_events, tool_events, &m->evlist,
+ (pmu && strcmp(pmu, "all") == 0) ? NULL : pmu);
if (ret)
goto out;
--
2.33.0
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH 1/1] perf: Add --pmu-filter option for filtering PMUs
2026-02-25 3:41 ` [PATCH 1/1] " Qinxin Xia
@ 2026-03-04 1:19 ` Namhyung Kim
2026-03-05 6:14 ` Qinxin Xia
0 siblings, 1 reply; 4+ messages in thread
From: Namhyung Kim @ 2026-03-04 1:19 UTC (permalink / raw)
To: Qinxin Xia
Cc: irogers, peterz, mingo, acme, wangyushan12, hejunhao3,
jonathan.cameron, linux-perf-users, linux-kernel, mark.rutland,
alexander.shishkin, jolsa, adrian.hunter
On Wed, Feb 25, 2026 at 11:41:25AM +0800, Qinxin Xia wrote:
> This patch adds a new --pmu-filter option to perf-stat command to allow
> filtering events on specific PMUs. This is useful when there are
> multiple PMUs with same type (e.g. hisi_sicl2_cpa0 and hisi_sicl0_cpa0).
>
> perf event:
> [root@localhost tmp]# perf stat -e cpa_cycles
> Performance counter stats for 'system wide':
>
> 3,005,509,659 hisi_sicl0_cpa0/cpa_cycles/
> 3,005,502,319 hisi_sicl2_cpa0/cpa_cycles/
>
> 3.005460970 seconds time elapsed
>
> [root@localhost tmp]# perf stat --pmu-filter hisi_sicl0_cpa0 -e cpa_cycles
> Performance counter stats for 'system wide':
>
> 3,003,899,571 cpa_cycles
>
> 3.003864470 seconds time elapsed
In this case, you can just use -e /hisi_sicl0_cpa0/cpa_cycles/.
>
> perf metric:
> [root@localhost tmp]# perf stat -M cpa_p0_avg_bw
> Performance counter stats for 'system wide':
>
> 19,417,779,115 hisi_sicl0_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
> 0 hisi_sicl0_cpa0/cpa_p0_wr_dat/
> 0 hisi_sicl0_cpa0/cpa_p0_rd_dat_64b/
> 0 hisi_sicl0_cpa0/cpa_p0_rd_dat_32b/
> 19,417,751,103 hisi_sicl10_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
> 0 hisi_sicl10_cpa0/cpa_p0_wr_dat/
> 0 hisi_sicl10_cpa0/cpa_p0_rd_dat_64b/
> 0 hisi_sicl10_cpa0/cpa_p0_rd_dat_32b/
> 19,417,730,679 hisi_sicl2_cpa0/cpa_cycles/ # 0.31 cpa_p0_avg_bw
> 75,635,749 hisi_sicl2_cpa0/cpa_p0_wr_dat/
> 18,520,640 hisi_sicl2_cpa0/cpa_p0_rd_dat_64b/
> 0 hisi_sicl2_cpa0/cpa_p0_rd_dat_32b/
> 19,417,674,227 hisi_sicl8_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
> 0 hisi_sicl8_cpa0/cpa_p0_wr_dat/
> 0 hisi_sicl8_cpa0/cpa_p0_rd_dat_64b/
> 0 hisi_sicl8_cpa0/cpa_p0_rd_dat_32b/
>
> 19.417734480 seconds time elapsed
>
> [root@localhost tmp]# perf stat --pmu-filter hisi_sicl2_cpa0 -M cpa_p0_avg_bw
> Performance counter stats for 'system wide':
>
> 6,234,093,559 cpa_cycles # 0.60 cpa_p0_avg_bw
> 50,548,465 cpa_p0_wr_dat
> 7,552,182 cpa_p0_rd_dat_64b
> 0 cpa_p0_rd_dat_32b
>
> 6.234139320 seconds time elapsed
But right, metrics seem to need this. Does this support wildcard match
for multiple PMUs?
Thanks,
Namhyung
>
> Signed-off-by: Qinxin Xia <xiaqinxin@huawei.com>
> ---
> tools/perf/Documentation/perf-stat.txt | 4 ++++
> tools/perf/builtin-stat.c | 26 ++++++++++++++++++++++++++
> tools/perf/util/metricgroup.c | 18 +++++++++++++-----
> 3 files changed, 43 insertions(+), 5 deletions(-)
>
> diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
> index 7cccc3a847d1..b72a29c9223c 100644
> --- a/tools/perf/Documentation/perf-stat.txt
> +++ b/tools/perf/Documentation/perf-stat.txt
> @@ -578,6 +578,10 @@ $ perf config stat.no-csv-summary=true
> Only enable events on applying cpu with this type for hybrid platform
> (e.g. core or atom)"
>
> +--pmu-filter::
> +Only enable events on applying pmu with specified for multiple
> +pmus with same type (e.g. hisi_sicl2_cpa0 or hisi_sicl0_cpa0)
> +
> EXAMPLES
> --------
>
> diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
> index 73c2ba7e3076..67fb91816c0b 100644
> --- a/tools/perf/builtin-stat.c
> +++ b/tools/perf/builtin-stat.c
> @@ -1214,6 +1214,28 @@ static int parse_cputype(const struct option *opt,
> return 0;
> }
>
> +static int parse_pmu_filter(const struct option *opt,
> + const char *str,
> + int unset __maybe_unused)
> +{
> + const struct perf_pmu *pmu;
> + struct evlist *evlist = *(struct evlist **)opt->value;
> +
> + if (!list_empty(&evlist->core.entries)) {
> + fprintf(stderr, "Must define pmu-filter before events/metrics\n");
> + return -1;
> + }
> +
> + pmu = perf_pmus__pmu_for_pmu_filter(str);
> + if (!pmu) {
> + fprintf(stderr, "--pmu-filter %s is not supported!\n", str);
> + return -1;
> + }
> + parse_events_option_args.pmu_filter = pmu->name;
> +
> + return 0;
> +}
> +
> static int parse_cache_level(const struct option *opt,
> const char *str,
> int unset __maybe_unused)
> @@ -2561,6 +2583,10 @@ int cmd_stat(int argc, const char **argv)
> "Only enable events on applying cpu with this type "
> "for hybrid platform (e.g. core or atom)",
> parse_cputype),
> + OPT_CALLBACK(0, "pmu-filter", &evsel_list, "pmu",
> + "Only enable events on applying pmu with specified "
> + "for multiple pmus with same type(e.g. hisi_sicl2_cpa0 or hisi_sicl0_cpa0)",
> + parse_pmu_filter),
> #ifdef HAVE_LIBPFM
> OPT_CALLBACK(0, "pfm-events", &evsel_list, "event",
> "libpfm4 event selector. use 'perf list' to list available events",
> diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
> index 46bf4dfeebc8..89b83ca38483 100644
> --- a/tools/perf/util/metricgroup.c
> +++ b/tools/perf/util/metricgroup.c
> @@ -387,8 +387,13 @@ static bool match_pm_metric_or_groups(const struct pmu_metric *pm, const char *p
> const char *metric_or_groups)
> {
> const char *pm_pmu = pm->pmu ?: "cpu";
> + struct perf_pmu *perf_pmu = NULL;
>
> - if (strcmp(pmu, "all") && strcmp(pm_pmu, pmu))
> + if (pm->pmu)
> + perf_pmu = perf_pmus__find(pm->pmu);
> +
> + if (strcmp(pmu, "all") && strcmp(pm_pmu, pmu) &&
> + (perf_pmu && !perf_pmu__name_wildcard_match(perf_pmu, pmu)))
> return false;
>
> return match_metric_or_groups(pm->metric_group, metric_or_groups) ||
> @@ -1259,7 +1264,8 @@ static int build_combined_expr_ctx(const struct list_head *metric_list,
> static int parse_ids(bool metric_no_merge, bool fake_pmu,
> struct expr_parse_ctx *ids, const char *modifier,
> bool group_events, const bool tool_events[TOOL_PMU__EVENT_MAX],
> - struct evlist **out_evlist)
> + struct evlist **out_evlist,
> + const char *filter_pmu)
> {
> struct parse_events_error parse_error;
> struct evlist *parsed_evlist;
> @@ -1313,7 +1319,7 @@ static int parse_ids(bool metric_no_merge, bool fake_pmu,
> }
> pr_debug("Parsing metric events '%s'\n", events.buf);
> parse_events_error__init(&parse_error);
> - ret = __parse_events(parsed_evlist, events.buf, /*pmu_filter=*/NULL,
> + ret = __parse_events(parsed_evlist, events.buf, filter_pmu,
> &parse_error, fake_pmu, /*warn_if_reordered=*/false,
> /*fake_tp=*/false);
> if (ret) {
> @@ -1416,7 +1422,8 @@ static int parse_groups(struct evlist *perf_evlist,
> /*modifier=*/NULL,
> /*group_events=*/false,
> tool_events,
> - &combined_evlist);
> + &combined_evlist,
> + (pmu && strcmp(pmu, "all") == 0) ? NULL : pmu);
> }
> if (combined)
> expr__ctx_free(combined);
> @@ -1471,7 +1478,8 @@ static int parse_groups(struct evlist *perf_evlist,
> }
> if (!metric_evlist) {
> ret = parse_ids(metric_no_merge, fake_pmu, m->pctx, m->modifier,
> - m->group_events, tool_events, &m->evlist);
> + m->group_events, tool_events, &m->evlist,
> + (pmu && strcmp(pmu, "all") == 0) ? NULL : pmu);
> if (ret)
> goto out;
>
> --
> 2.33.0
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH 1/1] perf: Add --pmu-filter option for filtering PMUs
2026-03-04 1:19 ` Namhyung Kim
@ 2026-03-05 6:14 ` Qinxin Xia
0 siblings, 0 replies; 4+ messages in thread
From: Qinxin Xia @ 2026-03-05 6:14 UTC (permalink / raw)
To: Namhyung Kim
Cc: irogers, peterz, mingo, acme, wangyushan12, hejunhao3,
jonathan.cameron, linux-perf-users, linux-kernel, mark.rutland,
alexander.shishkin, jolsa, adrian.hunter
On 2026/3/4 09:19:48, Namhyung Kim <namhyung@kernel.org> wrote:
> On Wed, Feb 25, 2026 at 11:41:25AM +0800, Qinxin Xia wrote:
>> This patch adds a new --pmu-filter option to perf-stat command to allow
>> filtering events on specific PMUs. This is useful when there are
>> multiple PMUs with same type (e.g. hisi_sicl2_cpa0 and hisi_sicl0_cpa0).
>>
>> perf event:
>> [root@localhost tmp]# perf stat -e cpa_cycles
>> Performance counter stats for 'system wide':
>>
>> 3,005,509,659 hisi_sicl0_cpa0/cpa_cycles/
>> 3,005,502,319 hisi_sicl2_cpa0/cpa_cycles/
>>
>> 3.005460970 seconds time elapsed
>>
>> [root@localhost tmp]# perf stat --pmu-filter hisi_sicl0_cpa0 -e cpa_cycles
>> Performance counter stats for 'system wide':
>>
>> 3,003,899,571 cpa_cycles
>>
>> 3.003864470 seconds time elapsed
>
> In this case, you can just use -e /hisi_sicl0_cpa0/cpa_cycles/.
>
Oh, I overlooked that, thanks for pointing it out. In the next version,
I'll delete this description in commit.
>>
>> perf metric:
>> [root@localhost tmp]# perf stat -M cpa_p0_avg_bw
>> Performance counter stats for 'system wide':
>>
>> 19,417,779,115 hisi_sicl0_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
>> 0 hisi_sicl0_cpa0/cpa_p0_wr_dat/
>> 0 hisi_sicl0_cpa0/cpa_p0_rd_dat_64b/
>> 0 hisi_sicl0_cpa0/cpa_p0_rd_dat_32b/
>> 19,417,751,103 hisi_sicl10_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
>> 0 hisi_sicl10_cpa0/cpa_p0_wr_dat/
>> 0 hisi_sicl10_cpa0/cpa_p0_rd_dat_64b/
>> 0 hisi_sicl10_cpa0/cpa_p0_rd_dat_32b/
>> 19,417,730,679 hisi_sicl2_cpa0/cpa_cycles/ # 0.31 cpa_p0_avg_bw
>> 75,635,749 hisi_sicl2_cpa0/cpa_p0_wr_dat/
>> 18,520,640 hisi_sicl2_cpa0/cpa_p0_rd_dat_64b/
>> 0 hisi_sicl2_cpa0/cpa_p0_rd_dat_32b/
>> 19,417,674,227 hisi_sicl8_cpa0/cpa_cycles/ # 0.00 cpa_p0_avg_bw
>> 0 hisi_sicl8_cpa0/cpa_p0_wr_dat/
>> 0 hisi_sicl8_cpa0/cpa_p0_rd_dat_64b/
>> 0 hisi_sicl8_cpa0/cpa_p0_rd_dat_32b/
>>
>> 19.417734480 seconds time elapsed
>>
>> [root@localhost tmp]# perf stat --pmu-filter hisi_sicl2_cpa0 -M cpa_p0_avg_bw
>> Performance counter stats for 'system wide':
>>
>> 6,234,093,559 cpa_cycles # 0.60 cpa_p0_avg_bw
>> 50,548,465 cpa_p0_wr_dat
>> 7,552,182 cpa_p0_rd_dat_64b
>> 0 cpa_p0_rd_dat_32b
>>
>> 6.234139320 seconds time elapsed
>
> But right, metrics seem to need this. Does this support wildcard match
> for multiple PMUs?
>
> Thanks,
> Namhyung
>
This patch does not support wildcard match. I will try to support it in
the next version.
Thank you for your review.
>>
>> Signed-off-by: Qinxin Xia <xiaqinxin@huawei.com>
>> ---
>> tools/perf/Documentation/perf-stat.txt | 4 ++++
>> tools/perf/builtin-stat.c | 26 ++++++++++++++++++++++++++
>> tools/perf/util/metricgroup.c | 18 +++++++++++++-----
>> 3 files changed, 43 insertions(+), 5 deletions(-)
>>
>> diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
>> index 7cccc3a847d1..b72a29c9223c 100644
>> --- a/tools/perf/Documentation/perf-stat.txt
>> +++ b/tools/perf/Documentation/perf-stat.txt
>> @@ -578,6 +578,10 @@ $ perf config stat.no-csv-summary=true
>> Only enable events on applying cpu with this type for hybrid platform
>> (e.g. core or atom)"
>>
>> +--pmu-filter::
>> +Only enable events on applying pmu with specified for multiple
>> +pmus with same type (e.g. hisi_sicl2_cpa0 or hisi_sicl0_cpa0)
>> +
>> EXAMPLES
>> --------
>>
>> diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
>> index 73c2ba7e3076..67fb91816c0b 100644
>> --- a/tools/perf/builtin-stat.c
>> +++ b/tools/perf/builtin-stat.c
>> @@ -1214,6 +1214,28 @@ static int parse_cputype(const struct option *opt,
>> return 0;
>> }
>>
>> +static int parse_pmu_filter(const struct option *opt,
>> + const char *str,
>> + int unset __maybe_unused)
>> +{
>> + const struct perf_pmu *pmu;
>> + struct evlist *evlist = *(struct evlist **)opt->value;
>> +
>> + if (!list_empty(&evlist->core.entries)) {
>> + fprintf(stderr, "Must define pmu-filter before events/metrics\n");
>> + return -1;
>> + }
>> +
>> + pmu = perf_pmus__pmu_for_pmu_filter(str);
>> + if (!pmu) {
>> + fprintf(stderr, "--pmu-filter %s is not supported!\n", str);
>> + return -1;
>> + }
>> + parse_events_option_args.pmu_filter = pmu->name;
>> +
>> + return 0;
>> +}
>> +
>> static int parse_cache_level(const struct option *opt,
>> const char *str,
>> int unset __maybe_unused)
>> @@ -2561,6 +2583,10 @@ int cmd_stat(int argc, const char **argv)
>> "Only enable events on applying cpu with this type "
>> "for hybrid platform (e.g. core or atom)",
>> parse_cputype),
>> + OPT_CALLBACK(0, "pmu-filter", &evsel_list, "pmu",
>> + "Only enable events on applying pmu with specified "
>> + "for multiple pmus with same type(e.g. hisi_sicl2_cpa0 or hisi_sicl0_cpa0)",
>> + parse_pmu_filter),
>> #ifdef HAVE_LIBPFM
>> OPT_CALLBACK(0, "pfm-events", &evsel_list, "event",
>> "libpfm4 event selector. use 'perf list' to list available events",
>> diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
>> index 46bf4dfeebc8..89b83ca38483 100644
>> --- a/tools/perf/util/metricgroup.c
>> +++ b/tools/perf/util/metricgroup.c
>> @@ -387,8 +387,13 @@ static bool match_pm_metric_or_groups(const struct pmu_metric *pm, const char *p
>> const char *metric_or_groups)
>> {
>> const char *pm_pmu = pm->pmu ?: "cpu";
>> + struct perf_pmu *perf_pmu = NULL;
>>
>> - if (strcmp(pmu, "all") && strcmp(pm_pmu, pmu))
>> + if (pm->pmu)
>> + perf_pmu = perf_pmus__find(pm->pmu);
>> +
>> + if (strcmp(pmu, "all") && strcmp(pm_pmu, pmu) &&
>> + (perf_pmu && !perf_pmu__name_wildcard_match(perf_pmu, pmu)))
>> return false;
>>
>> return match_metric_or_groups(pm->metric_group, metric_or_groups) ||
>> @@ -1259,7 +1264,8 @@ static int build_combined_expr_ctx(const struct list_head *metric_list,
>> static int parse_ids(bool metric_no_merge, bool fake_pmu,
>> struct expr_parse_ctx *ids, const char *modifier,
>> bool group_events, const bool tool_events[TOOL_PMU__EVENT_MAX],
>> - struct evlist **out_evlist)
>> + struct evlist **out_evlist,
>> + const char *filter_pmu)
>> {
>> struct parse_events_error parse_error;
>> struct evlist *parsed_evlist;
>> @@ -1313,7 +1319,7 @@ static int parse_ids(bool metric_no_merge, bool fake_pmu,
>> }
>> pr_debug("Parsing metric events '%s'\n", events.buf);
>> parse_events_error__init(&parse_error);
>> - ret = __parse_events(parsed_evlist, events.buf, /*pmu_filter=*/NULL,
>> + ret = __parse_events(parsed_evlist, events.buf, filter_pmu,
>> &parse_error, fake_pmu, /*warn_if_reordered=*/false,
>> /*fake_tp=*/false);
>> if (ret) {
>> @@ -1416,7 +1422,8 @@ static int parse_groups(struct evlist *perf_evlist,
>> /*modifier=*/NULL,
>> /*group_events=*/false,
>> tool_events,
>> - &combined_evlist);
>> + &combined_evlist,
>> + (pmu && strcmp(pmu, "all") == 0) ? NULL : pmu);
>> }
>> if (combined)
>> expr__ctx_free(combined);
>> @@ -1471,7 +1478,8 @@ static int parse_groups(struct evlist *perf_evlist,
>> }
>> if (!metric_evlist) {
>> ret = parse_ids(metric_no_merge, fake_pmu, m->pctx, m->modifier,
>> - m->group_events, tool_events, &m->evlist);
>> + m->group_events, tool_events, &m->evlist,
>> + (pmu && strcmp(pmu, "all") == 0) ? NULL : pmu);
>> if (ret)
>> goto out;
>>
>> --
>> 2.33.0
>>
>
--
Thanks,
Qinxin
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2026-03-05 6:14 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-02-25 3:41 [PATCH 0/1] perf: Add --pmu-filter option for filtering PMUs Qinxin Xia
2026-02-25 3:41 ` [PATCH 1/1] " Qinxin Xia
2026-03-04 1:19 ` Namhyung Kim
2026-03-05 6:14 ` Qinxin Xia
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox