* [PATCH v3 1/3] perf evlist: Make uniquifying counter names consistent
2025-05-12 21:50 [PATCH v3 0/3] Fix incorrect counts when count the same uncore event multiple times Chun-Tse Shao
@ 2025-05-12 21:50 ` Chun-Tse Shao
2025-05-12 21:50 ` [PATCH v3 2/3] perf parse-events: Use wildcard processing to set an event to merge into Chun-Tse Shao
` (2 subsequent siblings)
3 siblings, 0 replies; 7+ messages in thread
From: Chun-Tse Shao @ 2025-05-12 21:50 UTC (permalink / raw)
To: linux-kernel
Cc: Ian Rogers, Chun-Tse Shao, peterz, mingo, acme, namhyung,
mark.rutland, alexander.shishkin, jolsa, adrian.hunter, kan.liang,
james.clark, howardchu95, linux, yeoreum.yun, weilin.wang,
linux-perf-users
From: Ian Rogers <irogers@google.com>
perf stat has different uniquification logic to perf record and perf
top. In the case of perf record and perf top all hybrid event
names are uniquified. Perf stat is more disciplined respecting
name config terms, libpfm4 events, etc. Perf stat will uniquify
hybrid events and the non-core PMU cases shouldn't apply to perf
record or perf top. For consistency, remove the uniquification for
perf record and perf top and reuse the perf stat uniquification,
making the code more globally visible for this.
Fix the detection of cross-PMU for disabling uniquify by correctly
setting last_pmu. When setting uniquify on an evsel, make sure the
PMUs between the 2 considered events differ otherwise the uniquify
isn't adding value.
Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Chun-Tse Shao <ctshao@google.com>
---
tools/perf/builtin-record.c | 7 +-
tools/perf/builtin-top.c | 7 +-
tools/perf/util/evlist.c | 66 ++++++++++-----
tools/perf/util/evlist.h | 3 +-
tools/perf/util/evsel.c | 113 +++++++++++++++++++++++++
tools/perf/util/evsel.h | 4 +
tools/perf/util/stat-display.c | 149 +--------------------------------
7 files changed, 176 insertions(+), 173 deletions(-)
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 52cebacc19a7..94d9c97af7b7 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -26,6 +26,7 @@
#include "util/target.h"
#include "util/session.h"
#include "util/tool.h"
+#include "util/stat.h"
#include "util/symbol.h"
#include "util/record.h"
#include "util/cpumap.h"
@@ -2484,7 +2485,11 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)
pr_warning("WARNING: --timestamp-filename option is not available in pipe mode.\n");
}
- evlist__uniquify_name(rec->evlist);
+ /*
+ * Use global stat_config that is zero meaning aggr_mode is AGGR_NONE
+ * and hybrid_merge is false.
+ */
+ evlist__uniquify_evsel_names(rec->evlist, &stat_config);
evlist__config(rec->evlist, opts, &callchain_param);
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index f9f31391bddb..7b6cde87d2af 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -35,6 +35,7 @@
#include "util/mmap.h"
#include "util/session.h"
#include "util/thread.h"
+#include "util/stat.h"
#include "util/symbol.h"
#include "util/synthetic-events.h"
#include "util/top.h"
@@ -1309,7 +1310,11 @@ static int __cmd_top(struct perf_top *top)
}
}
- evlist__uniquify_name(top->evlist);
+ /*
+ * Use global stat_config that is zero meaning aggr_mode is AGGR_NONE
+ * and hybrid_merge is false.
+ */
+ evlist__uniquify_evsel_names(top->evlist, &stat_config);
ret = perf_top__start_counters(top);
if (ret)
return ret;
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 05d1f4e8ee22..dcd1130502df 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -2565,34 +2565,56 @@ void evlist__warn_user_requested_cpus(struct evlist *evlist, const char *cpu_lis
perf_cpu_map__put(user_requested_cpus);
}
-void evlist__uniquify_name(struct evlist *evlist)
+/* Should uniquify be disabled for the evlist? */
+static bool evlist__disable_uniquify(const struct evlist *evlist)
{
- char *new_name, empty_attributes[2] = ":", *attributes;
- struct evsel *pos;
+ struct evsel *counter;
+ struct perf_pmu *last_pmu = NULL;
+ bool first = true;
- if (perf_pmus__num_core_pmus() == 1)
- return;
+ evlist__for_each_entry(evlist, counter) {
+ /* If PMUs vary then uniquify can be useful. */
+ if (!first && counter->pmu != last_pmu)
+ return false;
+ first = false;
+ if (counter->pmu) {
+ /* Allow uniquify for uncore PMUs. */
+ if (!counter->pmu->is_core)
+ return false;
+ /* Keep hybrid event names uniquified for clarity. */
+ if (perf_pmus__num_core_pmus() > 1)
+ return false;
+ }
+ last_pmu = counter->pmu;
+ }
+ return true;
+}
- evlist__for_each_entry(evlist, pos) {
- if (!evsel__is_hybrid(pos))
- continue;
+static bool evlist__set_needs_uniquify(struct evlist *evlist, const struct perf_stat_config *config)
+{
+ struct evsel *counter;
+ bool needs_uniquify = false;
- if (strchr(pos->name, '/'))
- continue;
+ if (evlist__disable_uniquify(evlist)) {
+ evlist__for_each_entry(evlist, counter)
+ counter->uniquified_name = true;
+ return false;
+ }
+
+ evlist__for_each_entry(evlist, counter) {
+ if (evsel__set_needs_uniquify(counter, config))
+ needs_uniquify = true;
+ }
+ return needs_uniquify;
+}
- attributes = strchr(pos->name, ':');
- if (attributes)
- *attributes = '\0';
- else
- attributes = empty_attributes;
+void evlist__uniquify_evsel_names(struct evlist *evlist, const struct perf_stat_config *config)
+{
+ if (evlist__set_needs_uniquify(evlist, config)) {
+ struct evsel *pos;
- if (asprintf(&new_name, "%s/%s/%s", pos->pmu ? pos->pmu->name : "",
- pos->name, attributes + 1)) {
- free(pos->name);
- pos->name = new_name;
- } else {
- *attributes = ':';
- }
+ evlist__for_each_entry(evlist, pos)
+ evsel__uniquify_counter(pos);
}
}
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index 21f6bff319fd..85859708393e 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -19,6 +19,7 @@
struct pollfd;
struct thread_map;
struct perf_cpu_map;
+struct perf_stat_config;
struct record_opts;
struct strbuf;
struct target;
@@ -434,7 +435,7 @@ struct evsel *evlist__find_evsel(struct evlist *evlist, int idx);
void evlist__format_evsels(struct evlist *evlist, struct strbuf *sb, size_t max_length);
void evlist__check_mem_load_aux(struct evlist *evlist);
void evlist__warn_user_requested_cpus(struct evlist *evlist, const char *cpu_list);
-void evlist__uniquify_name(struct evlist *evlist);
+void evlist__uniquify_evsel_names(struct evlist *evlist, const struct perf_stat_config *config);
bool evlist__has_bpf_output(struct evlist *evlist);
bool evlist__needs_bpf_sb_event(struct evlist *evlist);
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index c06b191e7398..34409828f8ec 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -3954,3 +3954,116 @@ void evsel__remove_from_group(struct evsel *evsel, struct evsel *leader)
leader->core.nr_members--;
}
}
+
+bool evsel__set_needs_uniquify(struct evsel *counter, const struct perf_stat_config *config)
+{
+ struct evsel *evsel;
+
+ if (counter->needs_uniquify) {
+ /* Already set. */
+ return true;
+ }
+
+ if (counter->merged_stat) {
+ /* Counter won't be shown. */
+ return false;
+ }
+
+ if (counter->use_config_name || counter->is_libpfm_event) {
+ /* Original name will be used. */
+ return false;
+ }
+
+ if (!config->hybrid_merge && evsel__is_hybrid(counter)) {
+ /* Unique hybrid counters necessary. */
+ counter->needs_uniquify = true;
+ return true;
+ }
+
+ if (counter->core.attr.type < PERF_TYPE_MAX && counter->core.attr.type != PERF_TYPE_RAW) {
+ /* Legacy event, don't uniquify. */
+ return false;
+ }
+
+ if (counter->pmu && counter->pmu->is_core &&
+ counter->alternate_hw_config != PERF_COUNT_HW_MAX) {
+ /* A sysfs or json event replacing a legacy event, don't uniquify. */
+ return false;
+ }
+
+ if (config->aggr_mode == AGGR_NONE) {
+ /* Always unique with no aggregation. */
+ counter->needs_uniquify = true;
+ return true;
+ }
+
+ /*
+ * Do other non-merged events in the evlist have the same name? If so
+ * uniquify is necessary.
+ */
+ evlist__for_each_entry(counter->evlist, evsel) {
+ if (evsel == counter || evsel->merged_stat || evsel->pmu == counter->pmu)
+ continue;
+
+ if (evsel__name_is(counter, evsel__name(evsel))) {
+ counter->needs_uniquify = true;
+ return true;
+ }
+ }
+ return false;
+}
+
+void evsel__uniquify_counter(struct evsel *counter)
+{
+ const char *name, *pmu_name;
+ char *new_name, *config;
+ int ret;
+
+ /* No uniquification necessary. */
+ if (!counter->needs_uniquify)
+ return;
+
+ /* The evsel was already uniquified. */
+ if (counter->uniquified_name)
+ return;
+
+ /* Avoid checking to uniquify twice. */
+ counter->uniquified_name = true;
+
+ name = evsel__name(counter);
+ pmu_name = counter->pmu->name;
+ /* Already prefixed by the PMU name. */
+ if (!strncmp(name, pmu_name, strlen(pmu_name)))
+ return;
+
+ config = strchr(name, '/');
+ if (config) {
+ int len = config - name;
+
+ if (config[1] == '/') {
+ /* case: event// */
+ ret = asprintf(&new_name, "%s/%.*s/%s", pmu_name, len, name, config + 2);
+ } else {
+ /* case: event/.../ */
+ ret = asprintf(&new_name, "%s/%.*s,%s", pmu_name, len, name, config + 1);
+ }
+ } else {
+ config = strchr(name, ':');
+ if (config) {
+ /* case: event:.. */
+ int len = config - name;
+
+ ret = asprintf(&new_name, "%s/%.*s/%s", pmu_name, len, name, config + 1);
+ } else {
+ /* case: event */
+ ret = asprintf(&new_name, "%s/%s/", pmu_name, name);
+ }
+ }
+ if (ret > 0) {
+ free(counter->name);
+ counter->name = new_name;
+ } else {
+ /* ENOMEM from asprintf. */
+ counter->uniquified_name = false;
+ }
+}
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index 3d47d9318d92..df8d2eea0465 100644
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -16,6 +16,7 @@
struct bpf_object;
struct cgroup;
struct perf_counts;
+struct perf_stat_config;
struct perf_stat_evsel;
union perf_event;
struct bpf_counter_ops;
@@ -548,6 +549,9 @@ void evsel__remove_from_group(struct evsel *evsel, struct evsel *leader);
bool arch_evsel__must_be_in_group(const struct evsel *evsel);
+bool evsel__set_needs_uniquify(struct evsel *counter, const struct perf_stat_config *config);
+void evsel__uniquify_counter(struct evsel *counter);
+
/*
* Macro to swap the bit-field postition and size.
* Used when,
diff --git a/tools/perf/util/stat-display.c b/tools/perf/util/stat-display.c
index e31d9f64d3ae..c022afb28514 100644
--- a/tools/perf/util/stat-display.c
+++ b/tools/perf/util/stat-display.c
@@ -915,61 +915,6 @@ static void printout(struct perf_stat_config *config, struct outstate *os,
}
}
-static void evsel__uniquify_counter(struct evsel *counter)
-{
- const char *name, *pmu_name;
- char *new_name, *config;
- int ret;
-
- /* No uniquification necessary. */
- if (!counter->needs_uniquify)
- return;
-
- /* The evsel was already uniquified. */
- if (counter->uniquified_name)
- return;
-
- /* Avoid checking to uniquify twice. */
- counter->uniquified_name = true;
-
- name = evsel__name(counter);
- pmu_name = counter->pmu->name;
- /* Already prefixed by the PMU name. */
- if (!strncmp(name, pmu_name, strlen(pmu_name)))
- return;
-
- config = strchr(name, '/');
- if (config) {
- int len = config - name;
-
- if (config[1] == '/') {
- /* case: event// */
- ret = asprintf(&new_name, "%s/%.*s/%s", pmu_name, len, name, config + 2);
- } else {
- /* case: event/.../ */
- ret = asprintf(&new_name, "%s/%.*s,%s", pmu_name, len, name, config + 1);
- }
- } else {
- config = strchr(name, ':');
- if (config) {
- /* case: event:.. */
- int len = config - name;
-
- ret = asprintf(&new_name, "%s/%.*s/%s", pmu_name, len, name, config + 1);
- } else {
- /* case: event */
- ret = asprintf(&new_name, "%s/%s/", pmu_name, name);
- }
- }
- if (ret > 0) {
- free(counter->name);
- counter->name = new_name;
- } else {
- /* ENOMEM from asprintf. */
- counter->uniquified_name = false;
- }
-}
-
/**
* should_skip_zero_count() - Check if the event should print 0 values.
* @config: The perf stat configuration (including aggregation mode).
@@ -1060,8 +1005,6 @@ static void print_counter_aggrdata(struct perf_stat_config *config,
if (counter->merged_stat)
return;
- evsel__uniquify_counter(counter);
-
val = aggr->counts.val;
ena = aggr->counts.ena;
run = aggr->counts.run;
@@ -1636,96 +1579,6 @@ static void print_cgroup_counter(struct perf_stat_config *config, struct evlist
print_metric_end(config, os);
}
-/* Should uniquify be disabled for the evlist? */
-static bool evlist__disable_uniquify(const struct evlist *evlist)
-{
- struct evsel *counter;
- struct perf_pmu *last_pmu = NULL;
- bool first = true;
-
- evlist__for_each_entry(evlist, counter) {
- /* If PMUs vary then uniquify can be useful. */
- if (!first && counter->pmu != last_pmu)
- return false;
- first = false;
- if (counter->pmu) {
- /* Allow uniquify for uncore PMUs. */
- if (!counter->pmu->is_core)
- return false;
- /* Keep hybrid event names uniquified for clarity. */
- if (perf_pmus__num_core_pmus() > 1)
- return false;
- }
- }
- return true;
-}
-
-static void evsel__set_needs_uniquify(struct evsel *counter, const struct perf_stat_config *config)
-{
- struct evsel *evsel;
-
- if (counter->merged_stat) {
- /* Counter won't be shown. */
- return;
- }
-
- if (counter->use_config_name || counter->is_libpfm_event) {
- /* Original name will be used. */
- return;
- }
-
- if (!config->hybrid_merge && evsel__is_hybrid(counter)) {
- /* Unique hybrid counters necessary. */
- counter->needs_uniquify = true;
- return;
- }
-
- if (counter->core.attr.type < PERF_TYPE_MAX && counter->core.attr.type != PERF_TYPE_RAW) {
- /* Legacy event, don't uniquify. */
- return;
- }
-
- if (counter->pmu && counter->pmu->is_core &&
- counter->alternate_hw_config != PERF_COUNT_HW_MAX) {
- /* A sysfs or json event replacing a legacy event, don't uniquify. */
- return;
- }
-
- if (config->aggr_mode == AGGR_NONE) {
- /* Always unique with no aggregation. */
- counter->needs_uniquify = true;
- return;
- }
-
- /*
- * Do other non-merged events in the evlist have the same name? If so
- * uniquify is necessary.
- */
- evlist__for_each_entry(counter->evlist, evsel) {
- if (evsel == counter || evsel->merged_stat)
- continue;
-
- if (evsel__name_is(counter, evsel__name(evsel))) {
- counter->needs_uniquify = true;
- return;
- }
- }
-}
-
-static void evlist__set_needs_uniquify(struct evlist *evlist, const struct perf_stat_config *config)
-{
- struct evsel *counter;
-
- if (evlist__disable_uniquify(evlist)) {
- evlist__for_each_entry(evlist, counter)
- counter->uniquified_name = true;
- return;
- }
-
- evlist__for_each_entry(evlist, counter)
- evsel__set_needs_uniquify(counter, config);
-}
-
void evlist__print_counters(struct evlist *evlist, struct perf_stat_config *config,
struct target *_target, struct timespec *ts,
int argc, const char **argv)
@@ -1737,7 +1590,7 @@ void evlist__print_counters(struct evlist *evlist, struct perf_stat_config *conf
.first = true,
};
- evlist__set_needs_uniquify(evlist, config);
+ evlist__uniquify_evsel_names(evlist, config);
if (config->iostat_run)
evlist->selected = evlist__first(evlist);
--
2.49.0.1045.g170613ef41-goog
^ permalink raw reply related [flat|nested] 7+ messages in thread* [PATCH v3 2/3] perf parse-events: Use wildcard processing to set an event to merge into
2025-05-12 21:50 [PATCH v3 0/3] Fix incorrect counts when count the same uncore event multiple times Chun-Tse Shao
2025-05-12 21:50 ` [PATCH v3 1/3] perf evlist: Make uniquifying counter names consistent Chun-Tse Shao
@ 2025-05-12 21:50 ` Chun-Tse Shao
2025-05-12 21:50 ` [PATCH v3 3/3] perf test: Add stat uniquifying test Chun-Tse Shao
2025-05-13 20:02 ` [PATCH v3 0/3] Fix incorrect counts when count the same uncore event multiple times Arnaldo Carvalho de Melo
3 siblings, 0 replies; 7+ messages in thread
From: Chun-Tse Shao @ 2025-05-12 21:50 UTC (permalink / raw)
To: linux-kernel
Cc: Ian Rogers, Chun-Tse Shao, peterz, mingo, acme, namhyung,
mark.rutland, alexander.shishkin, jolsa, adrian.hunter, kan.liang,
howardchu95, yeoreum.yun, linux, james.clark, weilin.wang,
linux-perf-users
From: Ian Rogers <irogers@google.com>
The merge stat code fails for uncore events if they are repeated
twice, for example `perf stat -e clockticks,clockticks -I 1000` as the
counts of the second set of uncore events will be merged into the
first counter. Reimplement the logic to have a first_wildcard_match so
that merged later events correctly merge into the first wildcard event
that they will be aggregated into.
Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Chun-Tse Shao <ctshao@google.com>
---
tools/perf/util/evsel.c | 18 ++++---
tools/perf/util/evsel.h | 7 ++-
tools/perf/util/parse-events.c | 86 ++++++++++++++++++++++++----------
tools/perf/util/stat-display.c | 11 ++++-
tools/perf/util/stat.c | 40 ++--------------
5 files changed, 89 insertions(+), 73 deletions(-)
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 34409828f8ec..d55482f094bf 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -552,11 +552,11 @@ struct evsel *evsel__clone(struct evsel *dest, struct evsel *orig)
evsel->exclude_GH = orig->exclude_GH;
evsel->sample_read = orig->sample_read;
- evsel->auto_merge_stats = orig->auto_merge_stats;
evsel->collect_stat = orig->collect_stat;
evsel->weak_group = orig->weak_group;
evsel->use_config_name = orig->use_config_name;
evsel->pmu = orig->pmu;
+ evsel->first_wildcard_match = orig->first_wildcard_match;
if (evsel__copy_config_terms(evsel, orig) < 0)
goto out_err;
@@ -3964,11 +3964,6 @@ bool evsel__set_needs_uniquify(struct evsel *counter, const struct perf_stat_con
return true;
}
- if (counter->merged_stat) {
- /* Counter won't be shown. */
- return false;
- }
-
if (counter->use_config_name || counter->is_libpfm_event) {
/* Original name will be used. */
return false;
@@ -3997,12 +3992,21 @@ bool evsel__set_needs_uniquify(struct evsel *counter, const struct perf_stat_con
return true;
}
+ if (counter->first_wildcard_match != NULL) {
+ /*
+ * If stats are merged then only the first_wildcard_match is
+ * displayed, there is no need to uniquify this evsel as the
+ * name won't be shown.
+ */
+ return false;
+ }
+
/*
* Do other non-merged events in the evlist have the same name? If so
* uniquify is necessary.
*/
evlist__for_each_entry(counter->evlist, evsel) {
- if (evsel == counter || evsel->merged_stat || evsel->pmu == counter->pmu)
+ if (evsel == counter || evsel->first_wildcard_match || evsel->pmu == counter->pmu)
continue;
if (evsel__name_is(counter, evsel__name(evsel))) {
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index df8d2eea0465..6dbc9690e0c9 100644
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -70,6 +70,11 @@ struct evsel {
const char *unit;
struct cgroup *cgrp;
const char *metric_id;
+ /*
+ * This point to the first evsel with the same name, intended to store the
+ * aggregated counts in aggregation mode.
+ */
+ struct evsel *first_wildcard_match;
/* parse modifier helper */
int exclude_GH;
int sample_read;
@@ -78,7 +83,6 @@ struct evsel {
bool percore;
bool precise_max;
bool is_libpfm_event;
- bool auto_merge_stats;
bool collect_stat;
bool weak_group;
bool bpf_counter;
@@ -115,7 +119,6 @@ struct evsel {
bool ignore_missing_thread;
bool forced_leader;
bool cmdline_group_boundary;
- bool merged_stat;
bool reset_group;
bool errored;
bool needs_auxtrace_mmap;
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 177d045577c4..2380de56a207 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -250,13 +250,34 @@ __add_event(struct list_head *list, int *idx,
struct perf_event_attr *attr,
bool init_attr,
const char *name, const char *metric_id, struct perf_pmu *pmu,
- struct list_head *config_terms, bool auto_merge_stats,
+ struct list_head *config_terms, struct evsel *first_wildcard_match,
struct perf_cpu_map *cpu_list, u64 alternate_hw_config)
{
struct evsel *evsel;
bool is_pmu_core;
struct perf_cpu_map *cpus;
+ /*
+ * Ensure the first_wildcard_match's PMU matches that of the new event
+ * being added. Otherwise try to match with another event further down
+ * the evlist.
+ */
+ if (first_wildcard_match) {
+ struct evsel *pos = list_prev_entry(first_wildcard_match, core.node);
+
+ first_wildcard_match = NULL;
+ list_for_each_entry_continue(pos, list, core.node) {
+ if (perf_pmu__name_no_suffix_match(pos->pmu, pmu->name)) {
+ first_wildcard_match = pos;
+ break;
+ }
+ if (pos->pmu->is_core && (!pmu || pmu->is_core)) {
+ first_wildcard_match = pos;
+ break;
+ }
+ }
+ }
+
if (pmu) {
is_pmu_core = pmu->is_core;
cpus = perf_cpu_map__get(perf_cpu_map__is_empty(cpu_list) ? pmu->cpus : cpu_list);
@@ -293,9 +314,9 @@ __add_event(struct list_head *list, int *idx,
evsel->core.own_cpus = perf_cpu_map__get(cpus);
evsel->core.requires_cpu = pmu ? pmu->is_uncore : false;
evsel->core.is_pmu_core = is_pmu_core;
- evsel->auto_merge_stats = auto_merge_stats;
evsel->pmu = pmu;
evsel->alternate_hw_config = alternate_hw_config;
+ evsel->first_wildcard_match = first_wildcard_match;
if (name)
evsel->name = strdup(name);
@@ -318,7 +339,7 @@ struct evsel *parse_events__add_event(int idx, struct perf_event_attr *attr,
{
return __add_event(/*list=*/NULL, &idx, attr, /*init_attr=*/false, name,
metric_id, pmu, /*config_terms=*/NULL,
- /*auto_merge_stats=*/false, /*cpu_list=*/NULL,
+ /*first_wildcard_match=*/NULL, /*cpu_list=*/NULL,
/*alternate_hw_config=*/PERF_COUNT_HW_MAX);
}
@@ -329,7 +350,7 @@ static int add_event(struct list_head *list, int *idx,
{
return __add_event(list, idx, attr, /*init_attr*/true, name, metric_id,
/*pmu=*/NULL, config_terms,
- /*auto_merge_stats=*/false, /*cpu_list=*/NULL,
+ /*first_wildcard_match=*/NULL, /*cpu_list=*/NULL,
alternate_hw_config) ? 0 : -ENOMEM;
}
@@ -454,7 +475,7 @@ bool parse_events__filter_pmu(const struct parse_events_state *parse_state,
static int parse_events_add_pmu(struct parse_events_state *parse_state,
struct list_head *list, struct perf_pmu *pmu,
const struct parse_events_terms *const_parsed_terms,
- bool auto_merge_stats, u64 alternate_hw_config);
+ struct evsel *first_wildcard_match, u64 alternate_hw_config);
int parse_events_add_cache(struct list_head *list, int *idx, const char *name,
struct parse_events_state *parse_state,
@@ -466,6 +487,7 @@ int parse_events_add_cache(struct list_head *list, int *idx, const char *name,
const char *metric_id = get_config_metric_id(parsed_terms);
struct perf_cpu_map *cpus = get_config_cpu(parsed_terms);
int ret = 0;
+ struct evsel *first_wildcard_match = NULL;
while ((pmu = perf_pmus__scan(pmu)) != NULL) {
LIST_HEAD(config_terms);
@@ -481,10 +503,13 @@ int parse_events_add_cache(struct list_head *list, int *idx, const char *name,
*/
ret = parse_events_add_pmu(parse_state, list, pmu,
parsed_terms,
- perf_pmu__auto_merge_stats(pmu),
+ first_wildcard_match,
/*alternate_hw_config=*/PERF_COUNT_HW_MAX);
if (ret)
goto out_err;
+ if (first_wildcard_match == NULL)
+ first_wildcard_match =
+ container_of(list->prev, struct evsel, core.node);
continue;
}
@@ -515,10 +540,12 @@ int parse_events_add_cache(struct list_head *list, int *idx, const char *name,
}
if (__add_event(list, idx, &attr, /*init_attr*/true, config_name ?: name,
- metric_id, pmu, &config_terms, /*auto_merge_stats=*/false,
+ metric_id, pmu, &config_terms, first_wildcard_match,
cpus, /*alternate_hw_config=*/PERF_COUNT_HW_MAX) == NULL)
ret = -ENOMEM;
+ if (first_wildcard_match == NULL)
+ first_wildcard_match = container_of(list->prev, struct evsel, core.node);
free_config_terms(&config_terms);
if (ret)
goto out_err;
@@ -1387,7 +1414,8 @@ int parse_events_add_tracepoint(struct parse_events_state *parse_state,
static int __parse_events_add_numeric(struct parse_events_state *parse_state,
struct list_head *list,
struct perf_pmu *pmu, u32 type, u32 extended_type,
- u64 config, const struct parse_events_terms *head_config)
+ u64 config, const struct parse_events_terms *head_config,
+ struct evsel *first_wildcard_match)
{
struct perf_event_attr attr;
LIST_HEAD(config_terms);
@@ -1416,7 +1444,7 @@ static int __parse_events_add_numeric(struct parse_events_state *parse_state,
metric_id = get_config_metric_id(head_config);
cpus = get_config_cpu(head_config);
ret = __add_event(list, &parse_state->idx, &attr, /*init_attr*/true, name,
- metric_id, pmu, &config_terms, /*auto_merge_stats=*/false,
+ metric_id, pmu, &config_terms, first_wildcard_match,
cpus, /*alternate_hw_config=*/PERF_COUNT_HW_MAX) ? 0 : -ENOMEM;
perf_cpu_map__put(cpus);
free_config_terms(&config_terms);
@@ -1434,6 +1462,7 @@ int parse_events_add_numeric(struct parse_events_state *parse_state,
/* Wildcards on numeric values are only supported by core PMUs. */
if (wildcard && perf_pmus__supports_extended_type()) {
+ struct evsel *first_wildcard_match = NULL;
while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
int ret;
@@ -1443,15 +1472,20 @@ int parse_events_add_numeric(struct parse_events_state *parse_state,
ret = __parse_events_add_numeric(parse_state, list, pmu,
type, pmu->type,
- config, head_config);
+ config, head_config,
+ first_wildcard_match);
if (ret)
return ret;
+ if (first_wildcard_match == NULL)
+ first_wildcard_match =
+ container_of(list->prev, struct evsel, core.node);
}
if (found_supported)
return 0;
}
return __parse_events_add_numeric(parse_state, list, perf_pmus__find_by_type(type),
- type, /*extended_type=*/0, config, head_config);
+ type, /*extended_type=*/0, config, head_config,
+ /*first_wildcard_match=*/NULL);
}
static bool config_term_percore(struct list_head *config_terms)
@@ -1469,7 +1503,7 @@ static bool config_term_percore(struct list_head *config_terms)
static int parse_events_add_pmu(struct parse_events_state *parse_state,
struct list_head *list, struct perf_pmu *pmu,
const struct parse_events_terms *const_parsed_terms,
- bool auto_merge_stats, u64 alternate_hw_config)
+ struct evsel *first_wildcard_match, u64 alternate_hw_config)
{
struct perf_event_attr attr;
struct perf_pmu_info info;
@@ -1506,7 +1540,7 @@ static int parse_events_add_pmu(struct parse_events_state *parse_state,
evsel = __add_event(list, &parse_state->idx, &attr,
/*init_attr=*/true, /*name=*/NULL,
/*metric_id=*/NULL, pmu,
- /*config_terms=*/NULL, auto_merge_stats,
+ /*config_terms=*/NULL, first_wildcard_match,
/*cpu_list=*/NULL, alternate_hw_config);
return evsel ? 0 : -ENOMEM;
}
@@ -1577,7 +1611,7 @@ static int parse_events_add_pmu(struct parse_events_state *parse_state,
evsel = __add_event(list, &parse_state->idx, &attr, /*init_attr=*/true,
get_config_name(&parsed_terms),
get_config_metric_id(&parsed_terms), pmu,
- &config_terms, auto_merge_stats, term_cpu, alternate_hw_config);
+ &config_terms, first_wildcard_match, term_cpu, alternate_hw_config);
perf_cpu_map__put(term_cpu);
if (!evsel) {
parse_events_terms__exit(&parsed_terms);
@@ -1614,6 +1648,7 @@ int parse_events_multi_pmu_add(struct parse_events_state *parse_state,
int ok = 0;
const char *config;
struct parse_events_terms parsed_terms;
+ struct evsel *first_wildcard_match = NULL;
*listp = NULL;
@@ -1646,17 +1681,14 @@ int parse_events_multi_pmu_add(struct parse_events_state *parse_state,
INIT_LIST_HEAD(list);
while ((pmu = perf_pmus__scan(pmu)) != NULL) {
- bool auto_merge_stats;
-
if (parse_events__filter_pmu(parse_state, pmu))
continue;
if (!perf_pmu__have_event(pmu, event_name))
continue;
- auto_merge_stats = perf_pmu__auto_merge_stats(pmu);
if (!parse_events_add_pmu(parse_state, list, pmu,
- &parsed_terms, auto_merge_stats, hw_config)) {
+ &parsed_terms, first_wildcard_match, hw_config)) {
struct strbuf sb;
strbuf_init(&sb, /*hint=*/ 0);
@@ -1665,11 +1697,13 @@ int parse_events_multi_pmu_add(struct parse_events_state *parse_state,
strbuf_release(&sb);
ok++;
}
+ if (first_wildcard_match == NULL)
+ first_wildcard_match = container_of(list->prev, struct evsel, core.node);
}
if (parse_state->fake_pmu) {
if (!parse_events_add_pmu(parse_state, list, perf_pmus__fake_pmu(), &parsed_terms,
- /*auto_merge_stats=*/true, hw_config)) {
+ first_wildcard_match, hw_config)) {
struct strbuf sb;
strbuf_init(&sb, /*hint=*/ 0);
@@ -1700,6 +1734,7 @@ int parse_events_multi_pmu_add_or_add_pmu(struct parse_events_state *parse_state
struct perf_pmu *pmu;
int ok = 0;
char *help;
+ struct evsel *first_wildcard_match = NULL;
*listp = malloc(sizeof(**listp));
if (!*listp)
@@ -1710,14 +1745,14 @@ int parse_events_multi_pmu_add_or_add_pmu(struct parse_events_state *parse_state
/* Attempt to add to list assuming event_or_pmu is a PMU name. */
pmu = perf_pmus__find(event_or_pmu);
if (pmu && !parse_events_add_pmu(parse_state, *listp, pmu, const_parsed_terms,
- /*auto_merge_stats=*/false,
+ first_wildcard_match,
/*alternate_hw_config=*/PERF_COUNT_HW_MAX))
return 0;
if (parse_state->fake_pmu) {
if (!parse_events_add_pmu(parse_state, *listp, perf_pmus__fake_pmu(),
const_parsed_terms,
- /*auto_merge_stats=*/false,
+ first_wildcard_match,
/*alternate_hw_config=*/PERF_COUNT_HW_MAX))
return 0;
}
@@ -1727,15 +1762,16 @@ int parse_events_multi_pmu_add_or_add_pmu(struct parse_events_state *parse_state
while ((pmu = perf_pmus__scan(pmu)) != NULL) {
if (!parse_events__filter_pmu(parse_state, pmu) &&
perf_pmu__wildcard_match(pmu, event_or_pmu)) {
- bool auto_merge_stats = perf_pmu__auto_merge_stats(pmu);
-
if (!parse_events_add_pmu(parse_state, *listp, pmu,
const_parsed_terms,
- auto_merge_stats,
+ first_wildcard_match,
/*alternate_hw_config=*/PERF_COUNT_HW_MAX)) {
ok++;
parse_state->wild_card_pmus = true;
}
+ if (first_wildcard_match == NULL)
+ first_wildcard_match =
+ container_of((*listp)->prev, struct evsel, core.node);
}
}
if (ok)
@@ -2267,7 +2303,7 @@ int __parse_events(struct evlist *evlist, const char *str, const char *pmu_filte
if (verbose > 0) {
struct strbuf sb = STRBUF_INIT;
- evlist__uniquify_name(evlist);
+ evlist__uniquify_evsel_names(evlist, &stat_config);
evlist__format_evsels(evlist, &sb, 2048);
pr_debug("evlist after sorting/fixing: '%s'\n", sb.buf);
strbuf_release(&sb);
diff --git a/tools/perf/util/stat-display.c b/tools/perf/util/stat-display.c
index c022afb28514..729ad5cd52cb 100644
--- a/tools/perf/util/stat-display.c
+++ b/tools/perf/util/stat-display.c
@@ -1002,8 +1002,15 @@ static void print_counter_aggrdata(struct perf_stat_config *config,
os->evsel = counter;
/* Skip already merged uncore/hybrid events */
- if (counter->merged_stat)
- return;
+ if (config->aggr_mode != AGGR_NONE) {
+ if (evsel__is_hybrid(counter)) {
+ if (config->hybrid_merge && counter->first_wildcard_match != NULL)
+ return;
+ } else {
+ if (counter->first_wildcard_match != NULL)
+ return;
+ }
+ }
val = aggr->counts.val;
ena = aggr->counts.ena;
diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c
index 1f7abd8754c7..355a7d5c8ab8 100644
--- a/tools/perf/util/stat.c
+++ b/tools/perf/util/stat.c
@@ -535,35 +535,6 @@ static int evsel__merge_aggr_counters(struct evsel *evsel, struct evsel *alias)
return 0;
}
-/*
- * Events should have the same name, scale, unit, cgroup but on different core
- * PMUs or on different but matching uncore PMUs.
- */
-static bool evsel__is_alias(struct evsel *evsel_a, struct evsel *evsel_b)
-{
- if (strcmp(evsel__name(evsel_a), evsel__name(evsel_b)))
- return false;
-
- if (evsel_a->scale != evsel_b->scale)
- return false;
-
- if (evsel_a->cgrp != evsel_b->cgrp)
- return false;
-
- if (strcmp(evsel_a->unit, evsel_b->unit))
- return false;
-
- if (evsel__is_clock(evsel_a) != evsel__is_clock(evsel_b))
- return false;
-
- if (evsel_a->pmu == evsel_b->pmu || evsel_a->pmu == NULL || evsel_b->pmu == NULL)
- return false;
-
- if (evsel_a->pmu->is_core)
- return evsel_b->pmu->is_core;
-
- return perf_pmu__name_no_suffix_match(evsel_a->pmu, evsel_b->pmu->name);
-}
static void evsel__merge_aliases(struct evsel *evsel)
{
@@ -572,10 +543,9 @@ static void evsel__merge_aliases(struct evsel *evsel)
alias = list_prepare_entry(evsel, &(evlist->core.entries), core.node);
list_for_each_entry_continue(alias, &evlist->core.entries, core.node) {
- /* Merge the same events on different PMUs. */
- if (evsel__is_alias(evsel, alias)) {
+ if (alias->first_wildcard_match == evsel) {
+ /* Merge the same events on different PMUs. */
evsel__merge_aggr_counters(evsel, alias);
- alias->merged_stat = true;
}
}
}
@@ -588,11 +558,7 @@ static bool evsel__should_merge_hybrid(const struct evsel *evsel,
static void evsel__merge_stats(struct evsel *evsel, struct perf_stat_config *config)
{
- /* this evsel is already merged */
- if (evsel->merged_stat)
- return;
-
- if (evsel->auto_merge_stats || evsel__should_merge_hybrid(evsel, config))
+ if (!evsel->pmu || !evsel->pmu->is_core || evsel__should_merge_hybrid(evsel, config))
evsel__merge_aliases(evsel);
}
--
2.49.0.1045.g170613ef41-goog
^ permalink raw reply related [flat|nested] 7+ messages in thread* Re: [PATCH v3 0/3] Fix incorrect counts when count the same uncore event multiple times
2025-05-12 21:50 [PATCH v3 0/3] Fix incorrect counts when count the same uncore event multiple times Chun-Tse Shao
` (2 preceding siblings ...)
2025-05-12 21:50 ` [PATCH v3 3/3] perf test: Add stat uniquifying test Chun-Tse Shao
@ 2025-05-13 20:02 ` Arnaldo Carvalho de Melo
2025-05-13 20:04 ` Arnaldo Carvalho de Melo
3 siblings, 1 reply; 7+ messages in thread
From: Arnaldo Carvalho de Melo @ 2025-05-13 20:02 UTC (permalink / raw)
To: Chun-Tse Shao
Cc: linux-kernel, peterz, mingo, namhyung, mark.rutland,
alexander.shishkin, jolsa, irogers, adrian.hunter, kan.liang,
james.clark, howardchu95, weilin.wang, yeoreum.yun, linux,
linux-perf-users
On Mon, May 12, 2025 at 02:50:29PM -0700, Chun-Tse Shao wrote:
> Let's take a look an example, the machine is SKX with 6 IMC devices.
>
> perf stat -e clockticks,clockticks -I 1000
> # time counts unit events
> 1.001127430 6,901,503,174 uncore_imc_0/clockticks/
> 1.001127430 3,940,896,301 uncore_imc_0/clockticks/
> 2.002649722 988,376,876 uncore_imc_0/clockticks/
> 2.002649722 988,376,141 uncore_imc_0/clockticks/
> 3.004071319 1,000,292,675 uncore_imc_0/clockticks/
> 3.004071319 1,000,294,160 uncore_imc_0/clockticks/
>
> 1) The events name should not be uniquified.
> 2) The initial count for the first `clockticks` is doubled.
> 3) Subsequent count only report for the first IMC device.
>
> The first patch fixes 1) and 3), and the second patch fixes 2).
So, after having just the first patch applied I'm getting:
CC /tmp/build/perf-tools-next/util/bpf-filter-flex.o
util/parse-events.c: In function ‘__parse_events’:
util/parse-events.c:2270:25: error: implicit declaration of function ‘evlist__uniquify_name’; did you mean ‘evlist__uniquify_evsel_names’? [-Wimplicit-function-declaration]
2270 | evlist__uniquify_name(evlist);
| ^~~~~~~~~~~~~~~~~~~~~
| evlist__uniquify_evsel_names
make[4]: *** [/home/acme/git/perf-tools-next/tools/build/Makefile.build:85: /tmp/build/perf-tools-next/util/parse-events.o] Error 1
make[4]: *** Waiting for unfinished jobs....
make[3]: *** [/home/acme/git/perf-tools-next/tools/build/Makefile.build:142: util] Error 2
make[2]: *** [Makefile.perf:798: /tmp/build/perf-tools-next/perf-util-in.o] Error 2
make[2]: *** Waiting for unfinished jobs....
CC /tmp/build/perf-tools-next/pmu-events/pmu-events.o
LD /tmp/build/perf-tools-next/pmu-events/pmu-events-in.o
make[1]: *** [Makefile.perf:290: sub-make] Error 2
make: *** [Makefile:119: install-bin] Error 2
make: Leaving directory '/home/acme/git/perf-tools-next/tools/perf'
⬢ [acme@toolbx perf-tools-next]$
⬢ [acme@toolbx perf-tools-next]$
⬢ [acme@toolbx perf-tools-next]$
⬢ [acme@toolbx perf-tools-next]$
⬢ [acme@toolbx perf-tools-next]$
⬢ [acme@toolbx perf-tools-next]$
⬢ [acme@toolbx perf-tools-next]$
⬢ [acme@toolbx perf-tools-next]$ git log --oneline -3
6ffcaec3ac0d055a (HEAD) perf evlist: Make uniquifying counter names consistent
4102ff8b1fdaa588 perf metricgroup: Binary search when resolving referred to metrics
754baf426e099fbf perf pmu: Change aliases from list to hashmap
⬢ [acme@toolbx perf-tools-next]$
When test building the second patch, it builds, so I'm now looking if
you used things from the future or if the second patch removes the
problem.
- Arnaldo
> After these fix:
>
> perf stat -e clockticks,clockticks -I 1000
> # time counts unit events
> 1.001127586 4,126,938,857 clockticks
> 1.001127586 4,121,564,277 clockticks
> 2.001686014 3,953,806,350 clockticks
> 2.001686014 3,953,809,541 clockticks
> 3.003121403 4,137,750,252 clockticks
> 3.003121403 4,137,749,048 clockticks
>
> I also tested `-A`, `--per-socket`, `--per-die` and `--per-core`, all
> looks good.
>
> Ian tested `hybrid-merge` and `hwmon`, all looks good as well.
>
> Chun-Tse Shao (1):
> perf test: Add stat uniquifying test
>
> Ian Rogers (2):
> perf evlist: Make uniquifying counter names consistent
> perf parse-events: Use wildcard processing to set an event to merge
> into
>
> tools/perf/builtin-record.c | 7 +-
> tools/perf/builtin-top.c | 7 +-
> .../tests/shell/stat+event_uniquifying.sh | 69 ++++++++
> tools/perf/util/evlist.c | 66 +++++---
> tools/perf/util/evlist.h | 3 +-
> tools/perf/util/evsel.c | 119 ++++++++++++-
> tools/perf/util/evsel.h | 11 +-
> tools/perf/util/parse-events.c | 90 +++++++---
> tools/perf/util/stat-display.c | 160 ++----------------
> tools/perf/util/stat.c | 40 +----
> 10 files changed, 332 insertions(+), 240 deletions(-)
> create mode 100755 tools/perf/tests/shell/stat+event_uniquifying.sh
>
> --
> v3: Rebase with tmp.perf-tools-next. Since most of the conflicts are from
> lore.kernel.org/20250403194337.40202-5-irogers@google.com, tested v3
> patches with:
>
> perf stat -A -C 0,4-5,8 -e "instructions/cpu=0/,l1d-misses/cpu=4,cpu=5/,inst_retired.any/cpu=8/,cycles" -a sleep 0.1
>
> Performance counter stats for 'system wide':
>
> CPU0 682,860 instructions/cpu=0/ # 0.27 insn per cycle
> CPU4 53,774 l1d-misses
> CPU5 18,725 l1d-misses
> CPU8 608,698 inst_retired.any/cpu=8/
> CPU0 2,574,325 cycles
> CPU4 4,267,115 cycles
> CPU5 1,741,536 cycles
> CPU8 1,969,547 cycles
>
> 0.102746958 seconds time elapsed
>
> v2: lore.kernel.org/20250327225651.642965-1-ctshao@google.com
> - Fixes for `hwmon` and `--hybrid-merge`.
> - Add a test for event uniquifying.
>
> v1: lore.kernel.org/20250326234758.480431-1-ctshao@google.com
>
> 2.49.0.1045.g170613ef41-goog
^ permalink raw reply [flat|nested] 7+ messages in thread