* [PATCH v5 1/7] perf evsel: Add alternate_hw_config and use in evsel__match
2024-08-28 14:07 [PATCH v5 0/7] Event parsing fixes James Clark
@ 2024-08-28 14:07 ` James Clark
2024-08-28 14:07 ` [PATCH v5 2/7] perf stat: Uniquify event name improvements James Clark
` (6 subsequent siblings)
7 siblings, 0 replies; 14+ messages in thread
From: James Clark @ 2024-08-28 14:07 UTC (permalink / raw)
To: irogers, linux-perf-users, kan.liang, ak
Cc: James Clark, John Garry, Will Deacon, Mike Leach, Leo Yan,
Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo,
Namhyung Kim, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Adrian Hunter, Weilin Wang, Dominique Martinet, Athira Rajeev,
Colin Ian King, Yang Jihong, Ze Gao, Sun Haiyong, Jing Zhang,
Yicong Yang, linux-arm-kernel, linux-kernel
From: Ian Rogers <irogers@google.com>
There are cases where we want to match events like instructions and
cycles with legacy hardware values, in particular in stat-shadow's
hard coded metrics. An evsel's name isn't a good point of reference as
it gets altered, strstr would be too imprecise and re-parsing the
event from its name is silly. Instead, hold the legacy hardware event
name, determined during parsing, in the evsel for this matching case.
Inline evsel__match2 that is only used in builtin-diff.
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: James Clark <james.clark@linaro.org>
---
tools/perf/builtin-diff.c | 6 ++--
tools/perf/util/evsel.c | 21 +++++++++++++
tools/perf/util/evsel.h | 19 ++----------
tools/perf/util/parse-events.c | 57 +++++++++++++++++++++-------------
tools/perf/util/parse-events.h | 8 ++++-
tools/perf/util/parse-events.y | 2 +-
tools/perf/util/pmu.c | 6 +++-
tools/perf/util/pmu.h | 2 +-
8 files changed, 76 insertions(+), 45 deletions(-)
diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c
index 23326dd20333..82fb7773e03e 100644
--- a/tools/perf/builtin-diff.c
+++ b/tools/perf/builtin-diff.c
@@ -469,13 +469,13 @@ static int diff__process_sample_event(const struct perf_tool *tool,
static struct perf_diff pdiff;
-static struct evsel *evsel_match(struct evsel *evsel,
- struct evlist *evlist)
+static struct evsel *evsel_match(struct evsel *evsel, struct evlist *evlist)
{
struct evsel *e;
evlist__for_each_entry(evlist, e) {
- if (evsel__match2(evsel, e))
+ if ((evsel->core.attr.type == e->core.attr.type) &&
+ (evsel->core.attr.config == e->core.attr.config))
return e;
}
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 49cc71511c0c..2928c1f76dad 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -299,6 +299,7 @@ void evsel__init(struct evsel *evsel,
evsel->pmu_name = NULL;
evsel->group_pmu_name = NULL;
evsel->skippable = false;
+ evsel->alternate_hw_config = PERF_COUNT_HW_MAX;
}
struct evsel *evsel__new_idx(struct perf_event_attr *attr, int idx)
@@ -445,6 +446,8 @@ struct evsel *evsel__clone(struct evsel *orig)
if (evsel__copy_config_terms(evsel, orig) < 0)
goto out_err;
+ evsel->alternate_hw_config = orig->alternate_hw_config;
+
return evsel;
out_err:
@@ -1845,6 +1848,24 @@ static int evsel__read_tool(struct evsel *evsel, int cpu_map_idx, int thread)
return 0;
}
+bool __evsel__match(const struct evsel *evsel, u32 type, u64 config)
+{
+
+ u32 e_type = evsel->core.attr.type;
+ u64 e_config = evsel->core.attr.config;
+
+ if (e_type != type) {
+ return type == PERF_TYPE_HARDWARE && evsel->pmu && evsel->pmu->is_core &&
+ evsel->alternate_hw_config == config;
+ }
+
+ if ((type == PERF_TYPE_HARDWARE || type == PERF_TYPE_HW_CACHE) &&
+ perf_pmus__supports_extended_type())
+ e_config &= PERF_HW_EVENT_MASK;
+
+ return e_config == config;
+}
+
int evsel__read_counter(struct evsel *evsel, int cpu_map_idx, int thread)
{
if (evsel__is_tool(evsel))
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index 15acf293e12a..430441fdfbbc 100644
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -102,6 +102,7 @@ struct evsel {
int bpf_fd;
struct bpf_object *bpf_obj;
struct list_head config_terms;
+ u64 alternate_hw_config;
};
/*
@@ -388,26 +389,10 @@ u64 format_field__intval(struct tep_format_field *field, struct perf_sample *sam
struct tep_format_field *evsel__field(struct evsel *evsel, const char *name);
struct tep_format_field *evsel__common_field(struct evsel *evsel, const char *name);
-static inline bool __evsel__match(const struct evsel *evsel, u32 type, u64 config)
-{
- if (evsel->core.attr.type != type)
- return false;
-
- if ((type == PERF_TYPE_HARDWARE || type == PERF_TYPE_HW_CACHE) &&
- perf_pmus__supports_extended_type())
- return (evsel->core.attr.config & PERF_HW_EVENT_MASK) == config;
-
- return evsel->core.attr.config == config;
-}
+bool __evsel__match(const struct evsel *evsel, u32 type, u64 config);
#define evsel__match(evsel, t, c) __evsel__match(evsel, PERF_TYPE_##t, PERF_COUNT_##c)
-static inline bool evsel__match2(struct evsel *e1, struct evsel *e2)
-{
- return (e1->core.attr.type == e2->core.attr.type) &&
- (e1->core.attr.config == e2->core.attr.config);
-}
-
int evsel__read_counter(struct evsel *evsel, int cpu_map_idx, int thread);
int __evsel__read_on_cpu(struct evsel *evsel, int cpu_map_idx, int thread, bool scale);
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index fab01ba54e34..d1355e492df3 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -227,7 +227,7 @@ __add_event(struct list_head *list, int *idx,
bool init_attr,
const char *name, const char *metric_id, struct perf_pmu *pmu,
struct list_head *config_terms, bool auto_merge_stats,
- const char *cpu_list)
+ const char *cpu_list, u64 alternate_hw_config)
{
struct evsel *evsel;
struct perf_cpu_map *cpus = pmu ? perf_cpu_map__get(pmu->cpus) :
@@ -263,6 +263,7 @@ __add_event(struct list_head *list, int *idx,
evsel->auto_merge_stats = auto_merge_stats;
evsel->pmu = pmu;
evsel->pmu_name = pmu ? strdup(pmu->name) : NULL;
+ evsel->alternate_hw_config = alternate_hw_config;
if (name)
evsel->name = strdup(name);
@@ -285,16 +286,19 @@ 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);
+ /*auto_merge_stats=*/false, /*cpu_list=*/NULL,
+ /*alternate_hw_config=*/PERF_COUNT_HW_MAX);
}
static int add_event(struct list_head *list, int *idx,
struct perf_event_attr *attr, const char *name,
- const char *metric_id, struct list_head *config_terms)
+ const char *metric_id, struct list_head *config_terms,
+ u64 alternate_hw_config)
{
return __add_event(list, idx, attr, /*init_attr*/true, name, metric_id,
/*pmu=*/NULL, config_terms,
- /*auto_merge_stats=*/false, /*cpu_list=*/NULL) ? 0 : -ENOMEM;
+ /*auto_merge_stats=*/false, /*cpu_list=*/NULL,
+ alternate_hw_config) ? 0 : -ENOMEM;
}
static int add_event_tool(struct list_head *list, int *idx,
@@ -314,7 +318,9 @@ static int add_event_tool(struct list_head *list, int *idx,
evsel = __add_event(list, idx, &attr, /*init_attr=*/true, /*name=*/NULL,
/*metric_id=*/NULL, /*pmu=*/NULL,
/*config_terms=*/NULL, /*auto_merge_stats=*/false,
- cpu_list);
+ cpu_list,
+ /*alternate_hw_config=*/PERF_COUNT_HW_MAX);
+
if (!evsel)
return -ENOMEM;
evsel->tool_event = tool_event;
@@ -448,7 +454,7 @@ 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);
+ bool auto_merge_stats, u64 alternate_hw_config);
int parse_events_add_cache(struct list_head *list, int *idx, const char *name,
struct parse_events_state *parse_state,
@@ -474,7 +480,8 @@ 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));
+ perf_pmu__auto_merge_stats(pmu),
+ /*alternate_hw_config=*/PERF_COUNT_HW_MAX);
if (ret)
return ret;
continue;
@@ -505,7 +512,8 @@ 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,
- /*cpu_list=*/NULL) == NULL)
+ /*cpu_list=*/NULL,
+ /*alternate_hw_config=*/PERF_COUNT_HW_MAX) == NULL)
return -ENOMEM;
free_config_terms(&config_terms);
@@ -750,7 +758,7 @@ int parse_events_add_breakpoint(struct parse_events_state *parse_state,
name = get_config_name(head_config);
return add_event(list, &parse_state->idx, &attr, name, /*mertic_id=*/NULL,
- &config_terms);
+ &config_terms, /*alternate_hw_config=*/PERF_COUNT_HW_MAX);
}
static int check_type_val(struct parse_events_term *term,
@@ -1050,6 +1058,7 @@ static int config_term_pmu(struct perf_event_attr *attr,
if (perf_pmu__have_event(pmu, term->config)) {
term->type_term = PARSE_EVENTS__TERM_TYPE_USER;
term->no_value = true;
+ term->alternate_hw_config = true;
} else {
attr->type = PERF_TYPE_HARDWARE;
attr->config = term->val.num;
@@ -1362,8 +1371,9 @@ static int __parse_events_add_numeric(struct parse_events_state *parse_state,
name = get_config_name(head_config);
metric_id = get_config_metric_id(head_config);
ret = __add_event(list, &parse_state->idx, &attr, /*init_attr*/true, name,
- metric_id, pmu, &config_terms, /*auto_merge_stats=*/false,
- /*cpu_list=*/NULL) ? 0 : -ENOMEM;
+ metric_id, pmu, &config_terms, /*auto_merge_stats=*/false,
+ /*cpu_list=*/NULL, /*alternate_hw_config=*/PERF_COUNT_HW_MAX
+ ) == NULL ? -ENOMEM : 0;
free_config_terms(&config_terms);
return ret;
}
@@ -1421,7 +1431,7 @@ 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)
+ bool auto_merge_stats, u64 alternate_hw_config)
{
struct perf_event_attr attr;
struct perf_pmu_info info;
@@ -1458,7 +1468,7 @@ static int parse_events_add_pmu(struct parse_events_state *parse_state,
/*init_attr=*/true, /*name=*/NULL,
/*metric_id=*/NULL, pmu,
/*config_terms=*/NULL, auto_merge_stats,
- /*cpu_list=*/NULL);
+ /*cpu_list=*/NULL, alternate_hw_config);
return evsel ? 0 : -ENOMEM;
}
@@ -1479,7 +1489,8 @@ static int parse_events_add_pmu(struct parse_events_state *parse_state,
/* Look for event names in the terms and rewrite into format based terms. */
if (!parse_state->fake_pmu && perf_pmu__check_alias(pmu, &parsed_terms,
- &info, &alias_rewrote_terms, err)) {
+ &info, &alias_rewrote_terms,
+ &alternate_hw_config, err)) {
parse_events_terms__exit(&parsed_terms);
return -EINVAL;
}
@@ -1525,7 +1536,8 @@ 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, /*cpu_list=*/NULL);
+ &config_terms, auto_merge_stats, /*cpu_list=*/NULL,
+ alternate_hw_config);
if (!evsel) {
parse_events_terms__exit(&parsed_terms);
return -ENOMEM;
@@ -1551,7 +1563,7 @@ static int parse_events_add_pmu(struct parse_events_state *parse_state,
}
int parse_events_multi_pmu_add(struct parse_events_state *parse_state,
- const char *event_name,
+ const char *event_name, u64 hw_config,
const struct parse_events_terms *const_parsed_terms,
struct list_head **listp, void *loc_)
{
@@ -1604,7 +1616,7 @@ int parse_events_multi_pmu_add(struct parse_events_state *parse_state,
auto_merge_stats = perf_pmu__auto_merge_stats(pmu);
if (!parse_events_add_pmu(parse_state, list, pmu,
- &parsed_terms, auto_merge_stats)) {
+ &parsed_terms, auto_merge_stats, hw_config)) {
struct strbuf sb;
strbuf_init(&sb, /*hint=*/ 0);
@@ -1617,7 +1629,7 @@ int parse_events_multi_pmu_add(struct parse_events_state *parse_state,
if (parse_state->fake_pmu) {
if (!parse_events_add_pmu(parse_state, list, parse_state->fake_pmu, &parsed_terms,
- /*auto_merge_stats=*/true)) {
+ /*auto_merge_stats=*/true, hw_config)) {
struct strbuf sb;
strbuf_init(&sb, /*hint=*/ 0);
@@ -1658,7 +1670,8 @@ 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 = parse_state->fake_pmu ?: perf_pmus__find(event_or_pmu);
if (pmu && !parse_events_add_pmu(parse_state, *listp, pmu, const_parsed_terms,
- /*auto_merge_stats=*/false))
+ /*auto_merge_stats=*/false,
+ /*alternate_hw_config=*/PERF_COUNT_HW_MAX))
return 0;
pmu = NULL;
@@ -1670,7 +1683,8 @@ int parse_events_multi_pmu_add_or_add_pmu(struct parse_events_state *parse_state
if (!parse_events_add_pmu(parse_state, *listp, pmu,
const_parsed_terms,
- auto_merge_stats)) {
+ auto_merge_stats,
+ /*alternate_hw_config=*/PERF_COUNT_HW_MAX)) {
ok++;
parse_state->wild_card_pmus = true;
}
@@ -1681,7 +1695,8 @@ int parse_events_multi_pmu_add_or_add_pmu(struct parse_events_state *parse_state
/* Failure to add, assume event_or_pmu is an event name. */
zfree(listp);
- if (!parse_events_multi_pmu_add(parse_state, event_or_pmu, const_parsed_terms, listp, loc))
+ if (!parse_events_multi_pmu_add(parse_state, event_or_pmu, PERF_COUNT_HW_MAX,
+ const_parsed_terms, listp, loc))
return 0;
if (asprintf(&help, "Unable to find PMU or event on a PMU of '%s'", event_or_pmu) < 0)
diff --git a/tools/perf/util/parse-events.h b/tools/perf/util/parse-events.h
index b735cd9e0acf..703cb6d9f970 100644
--- a/tools/perf/util/parse-events.h
+++ b/tools/perf/util/parse-events.h
@@ -127,6 +127,12 @@ struct parse_events_term {
* value is assumed to be 1. An event name also has no value.
*/
bool no_value;
+ /**
+ * @alternate_hw_config: config is the event name but num is an
+ * alternate PERF_TYPE_HARDWARE config value which is often nice for the
+ * sake of quick matching.
+ */
+ bool alternate_hw_config;
};
struct parse_events_error {
@@ -238,7 +244,7 @@ struct evsel *parse_events__add_event(int idx, struct perf_event_attr *attr,
struct perf_pmu *pmu);
int parse_events_multi_pmu_add(struct parse_events_state *parse_state,
- const char *event_name,
+ const char *event_name, u64 hw_config,
const struct parse_events_terms *const_parsed_terms,
struct list_head **listp, void *loc);
diff --git a/tools/perf/util/parse-events.y b/tools/perf/util/parse-events.y
index b3c51f06cbdc..dcf47fabdfdd 100644
--- a/tools/perf/util/parse-events.y
+++ b/tools/perf/util/parse-events.y
@@ -292,7 +292,7 @@ PE_NAME sep_dc
struct list_head *list;
int err;
- err = parse_events_multi_pmu_add(_parse_state, $1, NULL, &list, &@1);
+ err = parse_events_multi_pmu_add(_parse_state, $1, PERF_COUNT_HW_MAX, NULL, &list, &@1);
if (err < 0) {
struct parse_events_state *parse_state = _parse_state;
struct parse_events_error *error = parse_state->error;
diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
index 0b38c51bd6eb..6f3ca69b71f1 100644
--- a/tools/perf/util/pmu.c
+++ b/tools/perf/util/pmu.c
@@ -1602,7 +1602,7 @@ static int check_info_data(struct perf_pmu *pmu,
*/
int perf_pmu__check_alias(struct perf_pmu *pmu, struct parse_events_terms *head_terms,
struct perf_pmu_info *info, bool *rewrote_terms,
- struct parse_events_error *err)
+ u64 *alternate_hw_config, struct parse_events_error *err)
{
struct parse_events_term *term, *h;
struct perf_pmu_alias *alias;
@@ -1630,6 +1630,7 @@ int perf_pmu__check_alias(struct perf_pmu *pmu, struct parse_events_terms *head_
NULL);
return ret;
}
+
*rewrote_terms = true;
ret = check_info_data(pmu, alias, info, err, term->err_term);
if (ret)
@@ -1638,6 +1639,9 @@ int perf_pmu__check_alias(struct perf_pmu *pmu, struct parse_events_terms *head_
if (alias->per_pkg)
info->per_pkg = true;
+ if (term->alternate_hw_config)
+ *alternate_hw_config = term->val.num;
+
list_del_init(&term->list);
parse_events_term__delete(term);
}
diff --git a/tools/perf/util/pmu.h b/tools/perf/util/pmu.h
index b2d3fd291f02..b5f65b705a1e 100644
--- a/tools/perf/util/pmu.h
+++ b/tools/perf/util/pmu.h
@@ -213,7 +213,7 @@ __u64 perf_pmu__format_bits(struct perf_pmu *pmu, const char *name);
int perf_pmu__format_type(struct perf_pmu *pmu, const char *name);
int perf_pmu__check_alias(struct perf_pmu *pmu, struct parse_events_terms *head_terms,
struct perf_pmu_info *info, bool *rewrote_terms,
- struct parse_events_error *err);
+ u64 *alternate_hw_config, struct parse_events_error *err);
int perf_pmu__find_event(struct perf_pmu *pmu, const char *event, void *state, pmu_event_callback cb);
void perf_pmu_format__set_value(void *format, int config, unsigned long *bits);
--
2.34.1
^ permalink raw reply related [flat|nested] 14+ messages in thread* [PATCH v5 2/7] perf stat: Uniquify event name improvements
2024-08-28 14:07 [PATCH v5 0/7] Event parsing fixes James Clark
2024-08-28 14:07 ` [PATCH v5 1/7] perf evsel: Add alternate_hw_config and use in evsel__match James Clark
@ 2024-08-28 14:07 ` James Clark
2024-08-28 14:07 ` [PATCH v5 3/7] perf stat: Remove evlist__add_default_attrs use strings James Clark
` (5 subsequent siblings)
7 siblings, 0 replies; 14+ messages in thread
From: James Clark @ 2024-08-28 14:07 UTC (permalink / raw)
To: irogers, linux-perf-users, kan.liang, ak
Cc: James Clark, John Garry, Will Deacon, Mike Leach, Leo Yan,
Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo,
Namhyung Kim, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Adrian Hunter, Weilin Wang, Dominique Martinet, Athira Rajeev,
Yang Jihong, Colin Ian King, Ze Gao, Sun Haiyong, Jing Zhang,
Yicong Yang, linux-arm-kernel, linux-kernel
From: Ian Rogers <irogers@google.com>
Without aggregation on Intel:
```
$ perf stat -e instructions,cycles ...
```
Will use "cycles" for the name of the legacy cycles event but as
"instructions" has a sysfs name it will and a "[cpu]" PMU suffix. This
often breaks things as the space between the event and the PMU name
look like an extra column. The existing uniquify logic was also
uniquifying in cases when all events are core and not with uncore
events, it was not correctly handling modifiers, etc.
Change the logic so that an initial pass that can disable
uniquification is run. For individual counters, disable uniquification
in more cases such as for consistency with legacy events or for
libpfm4 events. Don't use the "[pmu]" style suffix in uniquification,
always use "pmu/.../". Change how modifiers/terms are handled in the
uniquification so that they look like parse-able events.
This fixes "102: perf stat metrics (shadow stat) test:" that has been
failing due to "instructions [cpu]" breaking its column/awk logic when
values aren't aggregated. This started happening when instructions
could match a sysfs rather than a legacy event, so the fixes tag
reflects this.
Fixes: 617824a7f0f7 ("perf parse-events: Prefer sysfs/JSON hardware events over legacy")
Signed-off-by: Ian Rogers <irogers@google.com>
[ Fix Intel TPEBS counting mode test ]
Signed-off-by: James Clark <james.clark@linaro.org>
---
.../perf/tests/shell/test_stat_intel_tpebs.sh | 11 +-
tools/perf/util/stat-display.c | 101 ++++++++++++++----
2 files changed, 85 insertions(+), 27 deletions(-)
diff --git a/tools/perf/tests/shell/test_stat_intel_tpebs.sh b/tools/perf/tests/shell/test_stat_intel_tpebs.sh
index c60b29add980..9a11f42d153c 100755
--- a/tools/perf/tests/shell/test_stat_intel_tpebs.sh
+++ b/tools/perf/tests/shell/test_stat_intel_tpebs.sh
@@ -8,12 +8,15 @@ grep -q GenuineIntel /proc/cpuinfo || { echo Skipping non-Intel; exit 2; }
# Use this event for testing because it should exist in all platforms
event=cache-misses:R
+# Hybrid platforms output like "cpu_atom/cache-misses/R", rather than as above
+alt_name=/cache-misses/R
+
# Without this cmd option, default value or zero is returned
-echo "Testing without --record-tpebs"
-result=$(perf stat -e "$event" true 2>&1)
-[[ "$result" =~ $event ]] || exit 1
+#echo "Testing without --record-tpebs"
+#result=$(perf stat -e "$event" true 2>&1)
+#[[ "$result" =~ $event || "$result" =~ $alt_name ]] || exit 1
# In platforms that do not support TPEBS, it should execute without error.
echo "Testing with --record-tpebs"
result=$(perf stat -e "$event" --record-tpebs -a sleep 0.01 2>&1)
-[[ "$result" =~ "perf record" && "$result" =~ $event ]] || exit 1
+[[ "$result" =~ "perf record" && "$result" =~ $event || "$result" =~ $alt_name ]] || exit 1
diff --git a/tools/perf/util/stat-display.c b/tools/perf/util/stat-display.c
index ea96e4ebad8c..cbff43ff8d0f 100644
--- a/tools/perf/util/stat-display.c
+++ b/tools/perf/util/stat-display.c
@@ -871,38 +871,66 @@ static void printout(struct perf_stat_config *config, struct outstate *os,
static void uniquify_event_name(struct evsel *counter)
{
- char *new_name;
- char *config;
- int ret = 0;
+ const char *name, *pmu_name;
+ char *new_name, *config;
+ int ret;
- if (counter->uniquified_name || counter->use_config_name ||
- !counter->pmu_name || !strncmp(evsel__name(counter), counter->pmu_name,
- strlen(counter->pmu_name)))
+ /* The evsel was already uniquified. */
+ if (counter->uniquified_name)
return;
- config = strchr(counter->name, '/');
+ /* Avoid checking to uniquify twice. */
+ counter->uniquified_name = true;
+
+ /* The evsel has a "name=" config term or is from libpfm. */
+ if (counter->use_config_name || counter->is_libpfm_event)
+ return;
+
+ /* Legacy no PMU event, don't uniquify. */
+ if (!counter->pmu ||
+ (counter->pmu->type < PERF_TYPE_MAX && counter->pmu->type != PERF_TYPE_RAW))
+ return;
+
+ /* A sysfs or json event replacing a legacy event, don't uniquify. */
+ if (counter->pmu->is_core && counter->alternate_hw_config != PERF_COUNT_HW_MAX)
+ return;
+
+ 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) {
- if (asprintf(&new_name,
- "%s%s", counter->pmu_name, config) > 0) {
- free(counter->name);
- counter->name = new_name;
- }
- } else {
- if (evsel__is_hybrid(counter)) {
- ret = asprintf(&new_name, "%s/%s/",
- counter->pmu_name, counter->name);
+ int len = config - name;
+
+ if (config[1] == '/') {
+ /* case: event// */
+ ret = asprintf(&new_name, "%s/%.*s/%s", pmu_name, len, name, config + 2);
} else {
- ret = asprintf(&new_name, "%s [%s]",
- counter->name, counter->pmu_name);
+ /* 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;
- if (ret) {
- free(counter->name);
- counter->name = new_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);
}
}
-
- counter->uniquified_name = true;
+ if (ret > 0) {
+ free(counter->name);
+ counter->name = new_name;
+ } else {
+ /* ENOMEM from asprintf. */
+ counter->uniquified_name = false;
+ }
}
static bool hybrid_uniquify(struct evsel *evsel, struct perf_stat_config *config)
@@ -1559,6 +1587,31 @@ static void print_cgroup_counter(struct perf_stat_config *config, struct evlist
print_metric_end(config, os);
}
+static void disable_uniquify(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;
+ first = false;
+ if (counter->pmu) {
+ /* Allow uniquify for uncore PMUs. */
+ if (!counter->pmu->is_core)
+ return;
+ /* Keep hybrid event names uniquified for clarity. */
+ if (perf_pmus__num_core_pmus() > 1)
+ return;
+ }
+ }
+ evlist__for_each_entry_continue(evlist, counter) {
+ counter->uniquified_name = true;
+ }
+}
+
void evlist__print_counters(struct evlist *evlist, struct perf_stat_config *config,
struct target *_target, struct timespec *ts,
int argc, const char **argv)
@@ -1572,6 +1625,8 @@ void evlist__print_counters(struct evlist *evlist, struct perf_stat_config *conf
.first = true,
};
+ disable_uniquify(evlist);
+
if (config->iostat_run)
evlist->selected = evlist__first(evlist);
--
2.34.1
^ permalink raw reply related [flat|nested] 14+ messages in thread* [PATCH v5 3/7] perf stat: Remove evlist__add_default_attrs use strings
2024-08-28 14:07 [PATCH v5 0/7] Event parsing fixes James Clark
2024-08-28 14:07 ` [PATCH v5 1/7] perf evsel: Add alternate_hw_config and use in evsel__match James Clark
2024-08-28 14:07 ` [PATCH v5 2/7] perf stat: Uniquify event name improvements James Clark
@ 2024-08-28 14:07 ` James Clark
2024-08-28 14:07 ` [PATCH v5 4/7] perf evsel x86: Make evsel__has_perf_metrics work for legacy events James Clark
` (4 subsequent siblings)
7 siblings, 0 replies; 14+ messages in thread
From: James Clark @ 2024-08-28 14:07 UTC (permalink / raw)
To: irogers, linux-perf-users, kan.liang, ak
Cc: James Clark, John Garry, Will Deacon, Mike Leach, Leo Yan,
Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo,
Namhyung Kim, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Adrian Hunter, Weilin Wang, Dominique Martinet, Athira Rajeev,
Yang Jihong, Colin Ian King, Ze Gao, Yanteng Si, Sun Haiyong,
Jing Zhang, Yicong Yang, linux-arm-kernel, linux-kernel
From: Ian Rogers <irogers@google.com>
add_default_atttributes would add evsels by having pre-created
perf_event_attr, however, this needed fixing for hybrid as the
extended PMU type was necessary for each core PMU. The logic for this
was in an arch specific x86 function and wasn't present for ARM,
meaning that default events weren't being opened on all PMUs on
ARM. Change the creation of the default events to use parse_events and
strings as that will open the events on all PMUs.
Rather than try to detect events on PMUs before parsing, parse the
event but skip its output in stat-display.
The previous order of hardware events was: cycles,
stalled-cycles-frontend, stalled-cycles-backend, instructions. As
instructions is a more fundamental concept the order is changed to:
instructions, cycles, stalled-cycles-frontend, stalled-cycles-backend.
Closes: https://lore.kernel.org/lkml/CAP-5=fVABSBZnsmtRn1uF-k-G1GWM-L5SgiinhPTfHbQsKXb_g@mail.gmail.com/
Signed-off-by: Ian Rogers <irogers@google.com>
[Don't display unsupported default events]
Signed-off-by: James Clark <james.clark@linaro.org>
---
tools/perf/arch/x86/util/evlist.c | 74 +-------
tools/perf/builtin-stat.c | 291 ++++++++++++------------------
tools/perf/util/evlist.c | 43 -----
tools/perf/util/evlist.h | 12 --
tools/perf/util/stat-display.c | 8 +
tools/perf/util/stat-shadow.c | 4 +-
6 files changed, 125 insertions(+), 307 deletions(-)
diff --git a/tools/perf/arch/x86/util/evlist.c b/tools/perf/arch/x86/util/evlist.c
index cebdd483149e..3a4bf13b1759 100644
--- a/tools/perf/arch/x86/util/evlist.c
+++ b/tools/perf/arch/x86/util/evlist.c
@@ -1,78 +1,10 @@
// SPDX-License-Identifier: GPL-2.0
-#include <stdio.h>
-#include "util/pmu.h"
-#include "util/pmus.h"
-#include "util/evlist.h"
-#include "util/parse-events.h"
-#include "util/event.h"
+#include <string.h>
+#include "../../../util/evlist.h"
+#include "../../../util/evsel.h"
#include "topdown.h"
#include "evsel.h"
-static int ___evlist__add_default_attrs(struct evlist *evlist,
- struct perf_event_attr *attrs,
- size_t nr_attrs)
-{
- LIST_HEAD(head);
- size_t i = 0;
-
- for (i = 0; i < nr_attrs; i++)
- event_attr_init(attrs + i);
-
- if (perf_pmus__num_core_pmus() == 1)
- return evlist__add_attrs(evlist, attrs, nr_attrs);
-
- for (i = 0; i < nr_attrs; i++) {
- struct perf_pmu *pmu = NULL;
-
- if (attrs[i].type == PERF_TYPE_SOFTWARE) {
- struct evsel *evsel = evsel__new(attrs + i);
-
- if (evsel == NULL)
- goto out_delete_partial_list;
- list_add_tail(&evsel->core.node, &head);
- continue;
- }
-
- while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
- struct perf_cpu_map *cpus;
- struct evsel *evsel;
-
- evsel = evsel__new(attrs + i);
- if (evsel == NULL)
- goto out_delete_partial_list;
- evsel->core.attr.config |= (__u64)pmu->type << PERF_PMU_TYPE_SHIFT;
- cpus = perf_cpu_map__get(pmu->cpus);
- evsel->core.cpus = cpus;
- evsel->core.own_cpus = perf_cpu_map__get(cpus);
- evsel->pmu_name = strdup(pmu->name);
- list_add_tail(&evsel->core.node, &head);
- }
- }
-
- evlist__splice_list_tail(evlist, &head);
-
- return 0;
-
-out_delete_partial_list:
- {
- struct evsel *evsel, *n;
-
- __evlist__for_each_entry_safe(&head, n, evsel)
- evsel__delete(evsel);
- }
- return -1;
-}
-
-int arch_evlist__add_default_attrs(struct evlist *evlist,
- struct perf_event_attr *attrs,
- size_t nr_attrs)
-{
- if (!nr_attrs)
- return 0;
-
- return ___evlist__add_default_attrs(evlist, attrs, nr_attrs);
-}
-
int arch_evlist__cmp(const struct evsel *lhs, const struct evsel *rhs)
{
if (topdown_sys_has_perf_metrics() &&
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index cf985cdb9a6e..6302822853b0 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -1814,130 +1814,25 @@ static int perf_stat_init_aggr_mode_file(struct perf_stat *st)
}
/*
- * Add default attributes, if there were no attributes specified or
+ * Add default events, if there were no attributes specified or
* if -d/--detailed, -d -d or -d -d -d is used:
*/
-static int add_default_attributes(void)
+static int add_default_events(void)
{
- struct perf_event_attr default_attrs0[] = {
-
- { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_TASK_CLOCK },
- { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CONTEXT_SWITCHES },
- { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_CPU_MIGRATIONS },
- { .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_PAGE_FAULTS },
-
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES },
-};
- struct perf_event_attr frontend_attrs[] = {
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_FRONTEND },
-};
- struct perf_event_attr backend_attrs[] = {
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_BACKEND },
-};
- struct perf_event_attr default_attrs1[] = {
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS },
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
- { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES },
-
-};
-
-/*
- * Detailed stats (-d), covering the L1 and last level data caches:
- */
- struct perf_event_attr detailed_attrs[] = {
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_L1D << 0 |
- (PERF_COUNT_HW_CACHE_OP_READ << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_L1D << 0 |
- (PERF_COUNT_HW_CACHE_OP_READ << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_LL << 0 |
- (PERF_COUNT_HW_CACHE_OP_READ << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_LL << 0 |
- (PERF_COUNT_HW_CACHE_OP_READ << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
-};
-
-/*
- * Very detailed stats (-d -d), covering the instruction cache and the TLB caches:
- */
- struct perf_event_attr very_detailed_attrs[] = {
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_L1I << 0 |
- (PERF_COUNT_HW_CACHE_OP_READ << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_L1I << 0 |
- (PERF_COUNT_HW_CACHE_OP_READ << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_DTLB << 0 |
- (PERF_COUNT_HW_CACHE_OP_READ << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_DTLB << 0 |
- (PERF_COUNT_HW_CACHE_OP_READ << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_ITLB << 0 |
- (PERF_COUNT_HW_CACHE_OP_READ << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_ITLB << 0 |
- (PERF_COUNT_HW_CACHE_OP_READ << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
-
-};
+ const char *pmu = parse_events_option_args.pmu_filter ?: "all";
+ struct parse_events_error err;
+ struct evlist *evlist = evlist__new();
+ struct evsel *evsel;
+ int ret = 0;
-/*
- * Very, very detailed stats (-d -d -d), adding prefetch events:
- */
- struct perf_event_attr very_very_detailed_attrs[] = {
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_L1D << 0 |
- (PERF_COUNT_HW_CACHE_OP_PREFETCH << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_ACCESS << 16) },
-
- { .type = PERF_TYPE_HW_CACHE,
- .config =
- PERF_COUNT_HW_CACHE_L1D << 0 |
- (PERF_COUNT_HW_CACHE_OP_PREFETCH << 8) |
- (PERF_COUNT_HW_CACHE_RESULT_MISS << 16) },
-};
+ if (!evlist)
+ return -ENOMEM;
- struct perf_event_attr default_null_attrs[] = {};
- const char *pmu = parse_events_option_args.pmu_filter ?: "all";
+ parse_events_error__init(&err);
/* Set attrs if no event is selected and !null_run: */
if (stat_config.null_run)
- return 0;
+ goto out;
if (transaction_run) {
/* Handle -T as -M transaction. Once platform specific metrics
@@ -1947,9 +1842,10 @@ static int add_default_attributes(void)
*/
if (!metricgroup__has_metric(pmu, "transaction")) {
pr_err("Missing transaction metrics\n");
- return -1;
+ ret = -1;
+ goto out;
}
- return metricgroup__parse_groups(evsel_list, pmu, "transaction",
+ ret = metricgroup__parse_groups(evlist, pmu, "transaction",
stat_config.metric_no_group,
stat_config.metric_no_merge,
stat_config.metric_no_threshold,
@@ -1957,6 +1853,7 @@ static int add_default_attributes(void)
stat_config.system_wide,
stat_config.hardware_aware_grouping,
&stat_config.metric_events);
+ goto out;
}
if (smi_cost) {
@@ -1964,26 +1861,29 @@ static int add_default_attributes(void)
if (sysfs__read_int(FREEZE_ON_SMI_PATH, &smi) < 0) {
pr_err("freeze_on_smi is not supported.\n");
- return -1;
+ ret = -1;
+ goto out;
}
if (!smi) {
if (sysfs__write_int(FREEZE_ON_SMI_PATH, 1) < 0) {
- fprintf(stderr, "Failed to set freeze_on_smi.\n");
- return -1;
+ pr_err("Failed to set freeze_on_smi.\n");
+ ret = -1;
+ goto out;
}
smi_reset = true;
}
if (!metricgroup__has_metric(pmu, "smi")) {
pr_err("Missing smi metrics\n");
- return -1;
+ ret = -1;
+ goto out;
}
if (!force_metric_only)
stat_config.metric_only = true;
- return metricgroup__parse_groups(evsel_list, pmu, "smi",
+ ret = metricgroup__parse_groups(evlist, pmu, "smi",
stat_config.metric_no_group,
stat_config.metric_no_merge,
stat_config.metric_no_threshold,
@@ -1991,6 +1891,7 @@ static int add_default_attributes(void)
stat_config.system_wide,
stat_config.hardware_aware_grouping,
&stat_config.metric_events);
+ goto out;
}
if (topdown_run) {
@@ -2003,21 +1904,23 @@ static int add_default_attributes(void)
if (!max_level) {
pr_err("Topdown requested but the topdown metric groups aren't present.\n"
"(See perf list the metric groups have names like TopdownL1)\n");
- return -1;
+ ret = -1;
+ goto out;
}
if (stat_config.topdown_level > max_level) {
pr_err("Invalid top-down metrics level. The max level is %u.\n", max_level);
- return -1;
- } else if (!stat_config.topdown_level)
+ ret = -1;
+ goto out;
+ } else if (!stat_config.topdown_level) {
stat_config.topdown_level = 1;
-
+ }
if (!stat_config.interval && !stat_config.metric_only) {
fprintf(stat_config.output,
"Topdown accuracy may decrease when measuring long periods.\n"
"Please print the result regularly, e.g. -I1000\n");
}
str[8] = stat_config.topdown_level + '0';
- if (metricgroup__parse_groups(evsel_list,
+ if (metricgroup__parse_groups(evlist,
pmu, str,
/*metric_no_group=*/false,
/*metric_no_merge=*/false,
@@ -2025,41 +1928,49 @@ static int add_default_attributes(void)
stat_config.user_requested_cpu_list,
stat_config.system_wide,
stat_config.hardware_aware_grouping,
- &stat_config.metric_events) < 0)
- return -1;
+ &stat_config.metric_events) < 0) {
+ ret = -1;
+ goto out;
+ }
}
if (!stat_config.topdown_level)
stat_config.topdown_level = 1;
- if (!evsel_list->core.nr_entries) {
+ if (!evlist->core.nr_entries && !evsel_list->core.nr_entries) {
/* No events so add defaults. */
if (target__has_cpu(&target))
- default_attrs0[0].config = PERF_COUNT_SW_CPU_CLOCK;
+ ret = parse_events(evlist, "cpu-clock", &err);
+ else
+ ret = parse_events(evlist, "task-clock", &err);
+ if (ret)
+ goto out;
+
+ ret = parse_events(evlist,
+ "context-switches,"
+ "cpu-migrations,"
+ "page-faults,"
+ "instructions,"
+ "cycles,"
+ "stalled-cycles-frontend,"
+ "stalled-cycles-backend,"
+ "branches,"
+ "branch-misses",
+ &err);
+ if (ret)
+ goto out;
- if (evlist__add_default_attrs(evsel_list, default_attrs0) < 0)
- return -1;
- if (perf_pmus__have_event("cpu", "stalled-cycles-frontend")) {
- if (evlist__add_default_attrs(evsel_list, frontend_attrs) < 0)
- return -1;
- }
- if (perf_pmus__have_event("cpu", "stalled-cycles-backend")) {
- if (evlist__add_default_attrs(evsel_list, backend_attrs) < 0)
- return -1;
- }
- if (evlist__add_default_attrs(evsel_list, default_attrs1) < 0)
- return -1;
/*
* Add TopdownL1 metrics if they exist. To minimize
* multiplexing, don't request threshold computation.
*/
if (metricgroup__has_metric(pmu, "Default")) {
struct evlist *metric_evlist = evlist__new();
- struct evsel *metric_evsel;
-
- if (!metric_evlist)
- return -1;
+ if (!metric_evlist) {
+ ret = -ENOMEM;
+ goto out;
+ }
if (metricgroup__parse_groups(metric_evlist, pmu, "Default",
/*metric_no_group=*/false,
/*metric_no_merge=*/false,
@@ -2067,43 +1978,65 @@ static int add_default_attributes(void)
stat_config.user_requested_cpu_list,
stat_config.system_wide,
stat_config.hardware_aware_grouping,
- &stat_config.metric_events) < 0)
- return -1;
-
- evlist__for_each_entry(metric_evlist, metric_evsel) {
- metric_evsel->skippable = true;
- metric_evsel->default_metricgroup = true;
+ &stat_config.metric_events) < 0) {
+ ret = -1;
+ goto out;
}
- evlist__splice_list_tail(evsel_list, &metric_evlist->core.entries);
+
+ evlist__for_each_entry(metric_evlist, evsel)
+ evsel->default_metricgroup = true;
+
+ evlist__splice_list_tail(evlist, &metric_evlist->core.entries);
evlist__delete(metric_evlist);
}
-
- /* Platform specific attrs */
- if (evlist__add_default_attrs(evsel_list, default_null_attrs) < 0)
- return -1;
}
/* Detailed events get appended to the event list: */
- if (detailed_run < 1)
- return 0;
-
- /* Append detailed run extra attributes: */
- if (evlist__add_default_attrs(evsel_list, detailed_attrs) < 0)
- return -1;
-
- if (detailed_run < 2)
- return 0;
-
- /* Append very detailed run extra attributes: */
- if (evlist__add_default_attrs(evsel_list, very_detailed_attrs) < 0)
- return -1;
-
- if (detailed_run < 3)
- return 0;
-
- /* Append very, very detailed run extra attributes: */
- return evlist__add_default_attrs(evsel_list, very_very_detailed_attrs);
+ if (!ret && detailed_run >= 1) {
+ /*
+ * Detailed stats (-d), covering the L1 and last level data
+ * caches:
+ */
+ ret = parse_events(evlist,
+ "L1-dcache-loads,"
+ "L1-dcache-load-misses,"
+ "LLC-loads,"
+ "LLC-load-misses",
+ &err);
+ }
+ if (!ret && detailed_run >= 2) {
+ /*
+ * Very detailed stats (-d -d), covering the instruction cache
+ * and the TLB caches:
+ */
+ ret = parse_events(evlist,
+ "L1-icache-loads,"
+ "L1-icache-load-misses,"
+ "dTLB-loads,"
+ "dTLB-load-misses,"
+ "iTLB-loads,"
+ "iTLB-load-misses",
+ &err);
+ }
+ if (!ret && detailed_run >= 3) {
+ /*
+ * Very, very detailed stats (-d -d -d), adding prefetch events:
+ */
+ ret = parse_events(evlist,
+ "L1-dcache-prefetches,"
+ "L1-dcache-prefetch-misses",
+ &err);
+ }
+out:
+ if (!ret) {
+ evlist__for_each_entry(evlist, evsel)
+ evsel->skippable = true;
+ }
+ parse_events_error__exit(&err);
+ evlist__splice_list_tail(evsel_list, &evlist->core.entries);
+ evlist__delete(evlist);
+ return ret;
}
static const char * const stat_record_usage[] = {
@@ -2760,7 +2693,7 @@ int cmd_stat(int argc, const char **argv)
}
}
- if (add_default_attributes())
+ if (add_default_events())
goto out;
if (stat_config.cgroup_list) {
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index c0379fa1ccfe..a941f46c958d 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -319,49 +319,6 @@ struct evsel *evlist__add_sched_switch(struct evlist *evlist, bool system_wide)
}
#endif
-int evlist__add_attrs(struct evlist *evlist, struct perf_event_attr *attrs, size_t nr_attrs)
-{
- struct evsel *evsel, *n;
- LIST_HEAD(head);
- size_t i;
-
- for (i = 0; i < nr_attrs; i++) {
- evsel = evsel__new_idx(attrs + i, evlist->core.nr_entries + i);
- if (evsel == NULL)
- goto out_delete_partial_list;
- list_add_tail(&evsel->core.node, &head);
- }
-
- evlist__splice_list_tail(evlist, &head);
-
- return 0;
-
-out_delete_partial_list:
- __evlist__for_each_entry_safe(&head, n, evsel)
- evsel__delete(evsel);
- return -1;
-}
-
-int __evlist__add_default_attrs(struct evlist *evlist, struct perf_event_attr *attrs, size_t nr_attrs)
-{
- size_t i;
-
- for (i = 0; i < nr_attrs; i++)
- event_attr_init(attrs + i);
-
- return evlist__add_attrs(evlist, attrs, nr_attrs);
-}
-
-__weak int arch_evlist__add_default_attrs(struct evlist *evlist,
- struct perf_event_attr *attrs,
- size_t nr_attrs)
-{
- if (!nr_attrs)
- return 0;
-
- return __evlist__add_default_attrs(evlist, attrs, nr_attrs);
-}
-
struct evsel *evlist__find_tracepoint_by_id(struct evlist *evlist, int id)
{
struct evsel *evsel;
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index b46f1a320783..ca80c970bcd6 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -102,18 +102,6 @@ void evlist__delete(struct evlist *evlist);
void evlist__add(struct evlist *evlist, struct evsel *entry);
void evlist__remove(struct evlist *evlist, struct evsel *evsel);
-int evlist__add_attrs(struct evlist *evlist, struct perf_event_attr *attrs, size_t nr_attrs);
-
-int __evlist__add_default_attrs(struct evlist *evlist,
- struct perf_event_attr *attrs, size_t nr_attrs);
-
-int arch_evlist__add_default_attrs(struct evlist *evlist,
- struct perf_event_attr *attrs,
- size_t nr_attrs);
-
-#define evlist__add_default_attrs(evlist, array) \
- arch_evlist__add_default_attrs(evlist, array, ARRAY_SIZE(array))
-
int arch_evlist__cmp(const struct evsel *lhs, const struct evsel *rhs);
int evlist__add_dummy(struct evlist *evlist);
diff --git a/tools/perf/util/stat-display.c b/tools/perf/util/stat-display.c
index cbff43ff8d0f..5402998881c4 100644
--- a/tools/perf/util/stat-display.c
+++ b/tools/perf/util/stat-display.c
@@ -7,6 +7,7 @@
#include <perf/cpumap.h>
#include "color.h"
#include "counts.h"
+#include "debug.h"
#include "evlist.h"
#include "evsel.h"
#include "stat.h"
@@ -967,6 +968,13 @@ static bool should_skip_zero_counter(struct perf_stat_config *config,
struct perf_cpu cpu;
int idx;
+ /*
+ * Skip unsupported default events when not verbose. (default events
+ * are all marked 'skippable').
+ */
+ if (verbose == 0 && counter->skippable && !counter->supported)
+ return true;
+
/*
* Skip value 0 when enabling --per-thread globally,
* otherwise it will have too many 0 output.
diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c
index 6bb975e46de3..5548f751e56e 100644
--- a/tools/perf/util/stat-shadow.c
+++ b/tools/perf/util/stat-shadow.c
@@ -76,7 +76,7 @@ void perf_stat__reset_shadow_stats(void)
memset(&ru_stats, 0, sizeof(ru_stats));
}
-static enum stat_type evsel__stat_type(const struct evsel *evsel)
+static enum stat_type evsel__stat_type(struct evsel *evsel)
{
/* Fake perf_hw_cache_op_id values for use with evsel__match. */
u64 PERF_COUNT_hw_cache_l1d_miss = PERF_COUNT_HW_CACHE_L1D |
@@ -152,7 +152,7 @@ static const char *get_ratio_color(const double ratios[3], double val)
static double find_stat(const struct evsel *evsel, int aggr_idx, enum stat_type type)
{
- const struct evsel *cur;
+ struct evsel *cur;
int evsel_ctx = evsel_context(evsel);
evlist__for_each_entry(evsel->evlist, cur) {
--
2.34.1
^ permalink raw reply related [flat|nested] 14+ messages in thread* [PATCH v5 4/7] perf evsel x86: Make evsel__has_perf_metrics work for legacy events
2024-08-28 14:07 [PATCH v5 0/7] Event parsing fixes James Clark
` (2 preceding siblings ...)
2024-08-28 14:07 ` [PATCH v5 3/7] perf stat: Remove evlist__add_default_attrs use strings James Clark
@ 2024-08-28 14:07 ` James Clark
2024-08-28 14:07 ` [PATCH v5 5/7] perf evsel: Remove pmu_name James Clark
` (3 subsequent siblings)
7 siblings, 0 replies; 14+ messages in thread
From: James Clark @ 2024-08-28 14:07 UTC (permalink / raw)
To: irogers, linux-perf-users, kan.liang, ak
Cc: James Clark, John Garry, Will Deacon, Mike Leach, Leo Yan,
Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo,
Namhyung Kim, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Adrian Hunter, Weilin Wang, Athira Rajeev, Dominique Martinet,
Colin Ian King, Yang Jihong, Ze Gao, Jing Zhang, Sun Haiyong,
Yicong Yang, linux-arm-kernel, linux-kernel
From: Ian Rogers <irogers@google.com>
Use PMU interface to better detect core PMU for legacy events. Look
for slots event on core PMU if it is appropriate for the event.
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: James Clark <james.clark@linaro.org>
---
tools/perf/arch/x86/util/evsel.c | 31 ++++++++++++++++++++++++++-----
1 file changed, 26 insertions(+), 5 deletions(-)
diff --git a/tools/perf/arch/x86/util/evsel.c b/tools/perf/arch/x86/util/evsel.c
index 090d0f371891..1eaae8819c5e 100644
--- a/tools/perf/arch/x86/util/evsel.c
+++ b/tools/perf/arch/x86/util/evsel.c
@@ -21,7 +21,8 @@ void arch_evsel__set_sample_weight(struct evsel *evsel)
/* Check whether the evsel's PMU supports the perf metrics */
bool evsel__sys_has_perf_metrics(const struct evsel *evsel)
{
- const char *pmu_name = evsel->pmu_name ? evsel->pmu_name : "cpu";
+ struct perf_pmu *pmu;
+ u32 type = evsel->core.attr.type;
/*
* The PERF_TYPE_RAW type is the core PMU type, e.g., "cpu" PMU
@@ -31,11 +32,31 @@ bool evsel__sys_has_perf_metrics(const struct evsel *evsel)
* Checking both the PERF_TYPE_RAW type and the slots event
* should be good enough to detect the perf metrics feature.
*/
- if ((evsel->core.attr.type == PERF_TYPE_RAW) &&
- perf_pmus__have_event(pmu_name, "slots"))
- return true;
+again:
+ switch (type) {
+ case PERF_TYPE_HARDWARE:
+ case PERF_TYPE_HW_CACHE:
+ type = evsel->core.attr.config >> PERF_PMU_TYPE_SHIFT;
+ if (type)
+ goto again;
+ break;
+ case PERF_TYPE_RAW:
+ break;
+ default:
+ return false;
+ }
+
+ pmu = evsel->pmu;
+ if (pmu == &perf_pmu__fake)
+ pmu = NULL;
- return false;
+ if (!pmu) {
+ while ((pmu = perf_pmus__scan_core(pmu)) != NULL) {
+ if (pmu->type == PERF_TYPE_RAW)
+ break;
+ }
+ }
+ return pmu && perf_pmu__have_event(pmu, "slots");
}
bool arch_evsel__must_be_in_group(const struct evsel *evsel)
--
2.34.1
^ permalink raw reply related [flat|nested] 14+ messages in thread* [PATCH v5 5/7] perf evsel: Remove pmu_name
2024-08-28 14:07 [PATCH v5 0/7] Event parsing fixes James Clark
` (3 preceding siblings ...)
2024-08-28 14:07 ` [PATCH v5 4/7] perf evsel x86: Make evsel__has_perf_metrics work for legacy events James Clark
@ 2024-08-28 14:07 ` James Clark
2024-08-28 14:07 ` [PATCH v5 6/7] perf test: Make stat test work on DT devices James Clark
` (2 subsequent siblings)
7 siblings, 0 replies; 14+ messages in thread
From: James Clark @ 2024-08-28 14:07 UTC (permalink / raw)
To: irogers, linux-perf-users, kan.liang, ak
Cc: James Clark, John Garry, Will Deacon, Mike Leach, Leo Yan,
Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo,
Namhyung Kim, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Adrian Hunter, Weilin Wang, Dominique Martinet, Athira Rajeev,
Yang Jihong, Colin Ian King, Ze Gao, Sun Haiyong, Jing Zhang,
Yicong Yang, linux-arm-kernel, linux-kernel
From: Ian Rogers <irogers@google.com>
"evsel->pmu_name" is only ever assigned a strdup of "pmu->name", a
strdup of "evsel->pmu_name" or NULL. As such, prefer to use
"pmu->name" directly and even to directly compare PMUs than PMU
names. For safety, add some additional NULL tests.
Signed-off-by: Ian Rogers <irogers@google.com>
[ Fix arm-spe.c usage of pmu_name ]
Signed-off-by: James Clark <james.clark@linaro.org>
---
tools/perf/arch/arm64/util/arm-spe.c | 4 ++--
tools/perf/arch/x86/util/evsel.c | 4 ++--
tools/perf/tests/parse-events.c | 2 +-
tools/perf/util/evlist.c | 3 ++-
tools/perf/util/evsel.c | 7 -------
tools/perf/util/evsel.h | 3 +--
tools/perf/util/metricgroup.c | 4 ++--
tools/perf/util/parse-events.c | 1 -
tools/perf/util/stat-shadow.c | 10 +++++-----
tools/perf/util/stat.c | 2 +-
10 files changed, 16 insertions(+), 24 deletions(-)
diff --git a/tools/perf/arch/arm64/util/arm-spe.c b/tools/perf/arch/arm64/util/arm-spe.c
index d59f6ca499f2..81dcfeccdaf8 100644
--- a/tools/perf/arch/arm64/util/arm-spe.c
+++ b/tools/perf/arch/arm64/util/arm-spe.c
@@ -188,9 +188,9 @@ static int arm_spe_recording_options(struct auxtrace_record *itr,
evlist__for_each_entry(evlist, evsel) {
if (evsel__is_aux_event(evsel)) {
- if (!strstarts(evsel->pmu_name, ARM_SPE_PMU_NAME)) {
+ if (!strstarts(evsel->pmu->name, ARM_SPE_PMU_NAME)) {
pr_err("Found unexpected auxtrace event: %s\n",
- evsel->pmu_name);
+ evsel->pmu->name);
return -EINVAL;
}
opts->full_auxtrace = true;
diff --git a/tools/perf/arch/x86/util/evsel.c b/tools/perf/arch/x86/util/evsel.c
index 1eaae8819c5e..fe8a817d2061 100644
--- a/tools/perf/arch/x86/util/evsel.c
+++ b/tools/perf/arch/x86/util/evsel.c
@@ -84,7 +84,7 @@ int arch_evsel__hw_name(struct evsel *evsel, char *bf, size_t size)
return scnprintf(bf, size, "%s", event_name);
return scnprintf(bf, size, "%s/%s/",
- evsel->pmu_name ? evsel->pmu_name : "cpu",
+ evsel->pmu ? evsel->pmu->name : "cpu",
event_name);
}
@@ -129,7 +129,7 @@ int arch_evsel__open_strerror(struct evsel *evsel, char *msg, size_t size)
return 0;
if (!evsel->core.attr.precise_ip &&
- !(evsel->pmu_name && !strncmp(evsel->pmu_name, "ibs", 3)))
+ !(evsel->pmu && !strncmp(evsel->pmu->name, "ibs", 3)))
return 0;
/* More verbose IBS errors. */
diff --git a/tools/perf/tests/parse-events.c b/tools/perf/tests/parse-events.c
index edc2adcf1bae..c38fa4e16670 100644
--- a/tools/perf/tests/parse-events.c
+++ b/tools/perf/tests/parse-events.c
@@ -730,7 +730,7 @@ static int test__checkevent_pmu_events(struct evlist *evlist)
TEST_ASSERT_VAL("wrong number of entries", 1 == evlist->core.nr_entries);
TEST_ASSERT_VAL("wrong type", PERF_TYPE_RAW == evsel->core.attr.type ||
- strcmp(evsel->pmu_name, "cpu"));
+ strcmp(evsel->pmu->name, "cpu"));
TEST_ASSERT_VAL("wrong exclude_user",
!evsel->core.attr.exclude_user);
TEST_ASSERT_VAL("wrong exclude_kernel",
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index a941f46c958d..fff7fd4ae785 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -2576,7 +2576,8 @@ void evlist__uniquify_name(struct evlist *evlist)
else
attributes = empty_attributes;
- if (asprintf(&new_name, "%s/%s/%s", pos->pmu_name, pos->name, attributes + 1)) {
+ if (asprintf(&new_name, "%s/%s/%s", pos->pmu ? "" : pos->pmu->name,
+ pos->name, attributes + 1)) {
free(pos->name);
pos->name = new_name;
} else {
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 2928c1f76dad..082159fb2091 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -296,7 +296,6 @@ void evsel__init(struct evsel *evsel,
evsel->metric_events = NULL;
evsel->per_pkg_mask = NULL;
evsel->collect_stat = false;
- evsel->pmu_name = NULL;
evsel->group_pmu_name = NULL;
evsel->skippable = false;
evsel->alternate_hw_config = PERF_COUNT_HW_MAX;
@@ -394,11 +393,6 @@ struct evsel *evsel__clone(struct evsel *orig)
if (evsel->group_name == NULL)
goto out_err;
}
- if (orig->pmu_name) {
- evsel->pmu_name = strdup(orig->pmu_name);
- if (evsel->pmu_name == NULL)
- goto out_err;
- }
if (orig->group_pmu_name) {
evsel->group_pmu_name = strdup(orig->group_pmu_name);
if (evsel->group_pmu_name == NULL)
@@ -1497,7 +1491,6 @@ void evsel__exit(struct evsel *evsel)
zfree(&evsel->group_name);
zfree(&evsel->name);
zfree(&evsel->filter);
- zfree(&evsel->pmu_name);
zfree(&evsel->group_pmu_name);
zfree(&evsel->unit);
zfree(&evsel->metric_id);
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index 430441fdfbbc..936aafb046b6 100644
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -72,7 +72,6 @@ struct evsel {
struct {
char *name;
char *group_name;
- const char *pmu_name;
const char *group_pmu_name;
#ifdef HAVE_LIBTRACEEVENT
struct tep_event *tp_format;
@@ -184,7 +183,7 @@ struct evsel {
unsigned long open_flags;
int precise_ip_original;
- /* for missing_features */
+ /* The PMU the event is from. Used for missing_features, PMU name, etc. */
struct perf_pmu *pmu;
/* For tool events */
diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
index 69f6a46402c3..bbbb2e7eb8db 100644
--- a/tools/perf/util/metricgroup.c
+++ b/tools/perf/util/metricgroup.c
@@ -297,8 +297,8 @@ static int setup_metric_events(const char *pmu, struct hashmap *ids,
struct expr_id_data *val_ptr;
/* Don't match events for the wrong hybrid PMU. */
- if (!all_pmus && ev->pmu_name && evsel__is_hybrid(ev) &&
- strcmp(ev->pmu_name, pmu))
+ if (!all_pmus && ev->pmu && evsel__is_hybrid(ev) &&
+ strcmp(ev->pmu->name, pmu))
continue;
/*
* Check for duplicate events with the same name. For
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index d1355e492df3..fa37567e0323 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -262,7 +262,6 @@ __add_event(struct list_head *list, int *idx,
evsel->core.is_pmu_core = pmu ? pmu->is_core : false;
evsel->auto_merge_stats = auto_merge_stats;
evsel->pmu = pmu;
- evsel->pmu_name = pmu ? strdup(pmu->name) : NULL;
evsel->alternate_hw_config = alternate_hw_config;
if (name)
diff --git a/tools/perf/util/stat-shadow.c b/tools/perf/util/stat-shadow.c
index 5548f751e56e..72275806e6e2 100644
--- a/tools/perf/util/stat-shadow.c
+++ b/tools/perf/util/stat-shadow.c
@@ -573,7 +573,7 @@ static void perf_stat__print_metricgroup_header(struct perf_stat_config *config,
{
bool need_full_name = perf_pmus__num_core_pmus() > 1;
static const char *last_name;
- static const char *last_pmu;
+ static const struct perf_pmu *last_pmu;
char full_name[64];
/*
@@ -584,21 +584,21 @@ static void perf_stat__print_metricgroup_header(struct perf_stat_config *config,
* different metric events.
*/
if (last_name && !strcmp(last_name, name)) {
- if (!need_full_name || !strcmp(last_pmu, evsel->pmu_name)) {
+ if (!need_full_name || last_pmu != evsel->pmu) {
out->print_metricgroup_header(config, ctxp, NULL);
return;
}
}
- if (need_full_name)
- scnprintf(full_name, sizeof(full_name), "%s (%s)", name, evsel->pmu_name);
+ if (need_full_name && evsel->pmu)
+ scnprintf(full_name, sizeof(full_name), "%s (%s)", name, evsel->pmu->name);
else
scnprintf(full_name, sizeof(full_name), "%s", name);
out->print_metricgroup_header(config, ctxp, full_name);
last_name = name;
- last_pmu = evsel->pmu_name;
+ last_pmu = evsel->pmu;
}
/**
diff --git a/tools/perf/util/stat.c b/tools/perf/util/stat.c
index 0bd5467389e4..7c2ccdcc3fdb 100644
--- a/tools/perf/util/stat.c
+++ b/tools/perf/util/stat.c
@@ -553,7 +553,7 @@ static bool evsel__is_alias(struct evsel *evsel_a, struct evsel *evsel_b)
if (evsel__is_clock(evsel_a) != evsel__is_clock(evsel_b))
return false;
- return !!strcmp(evsel_a->pmu_name, evsel_b->pmu_name);
+ return evsel_a->pmu != evsel_b->pmu;
}
static void evsel__merge_aliases(struct evsel *evsel)
--
2.34.1
^ permalink raw reply related [flat|nested] 14+ messages in thread* [PATCH v5 6/7] perf test: Make stat test work on DT devices
2024-08-28 14:07 [PATCH v5 0/7] Event parsing fixes James Clark
` (4 preceding siblings ...)
2024-08-28 14:07 ` [PATCH v5 5/7] perf evsel: Remove pmu_name James Clark
@ 2024-08-28 14:07 ` James Clark
2024-08-28 14:07 ` [PATCH v5 7/7] perf test: Add a test for default perf stat command James Clark
2024-08-28 17:02 ` [PATCH v5 0/7] Event parsing fixes Andi Kleen
7 siblings, 0 replies; 14+ messages in thread
From: James Clark @ 2024-08-28 14:07 UTC (permalink / raw)
To: irogers, linux-perf-users, kan.liang, ak
Cc: James Clark, John Garry, Will Deacon, Mike Leach, Leo Yan,
Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo,
Namhyung Kim, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Adrian Hunter, Weilin Wang, Dominique Martinet, Athira Rajeev,
Colin Ian King, Yang Jihong, Ze Gao, Yanteng Si, Jing Zhang,
Sun Haiyong, Yicong Yang, linux-arm-kernel, linux-kernel
PMUs aren't listed in /sys/devices/ on DT devices, so change the search
directory to /sys/bus/event_source/devices which works everywhere. Also
add armv8_cortex_* as a known PMU type to search for to make the test
run on more devices.
Signed-off-by: James Clark <james.clark@linaro.org>
---
tools/perf/tests/shell/stat.sh | 12 +++++++-----
1 file changed, 7 insertions(+), 5 deletions(-)
diff --git a/tools/perf/tests/shell/stat.sh b/tools/perf/tests/shell/stat.sh
index 3f1e67795490..525d0c44fdc6 100755
--- a/tools/perf/tests/shell/stat.sh
+++ b/tools/perf/tests/shell/stat.sh
@@ -117,16 +117,18 @@ test_cputype() {
# Find a known PMU for cputype.
pmu=""
- for i in cpu cpu_atom armv8_pmuv3_0
+ devs="/sys/bus/event_source/devices"
+ for i in $devs/cpu $devs/cpu_atom $devs/armv8_pmuv3_0 $devs/armv8_cortex_*
do
- if test -d "/sys/devices/$i"
+ i_base=$(basename "$i")
+ if test -d "$i"
then
- pmu="$i"
+ pmu="$i_base"
break
fi
- if perf stat -e "$i/instructions/" true > /dev/null 2>&1
+ if perf stat -e "$i_base/instructions/" true > /dev/null 2>&1
then
- pmu="$i"
+ pmu="$i_base"
break
fi
done
--
2.34.1
^ permalink raw reply related [flat|nested] 14+ messages in thread* [PATCH v5 7/7] perf test: Add a test for default perf stat command
2024-08-28 14:07 [PATCH v5 0/7] Event parsing fixes James Clark
` (5 preceding siblings ...)
2024-08-28 14:07 ` [PATCH v5 6/7] perf test: Make stat test work on DT devices James Clark
@ 2024-08-28 14:07 ` James Clark
2024-08-30 6:32 ` Namhyung Kim
2024-08-28 17:02 ` [PATCH v5 0/7] Event parsing fixes Andi Kleen
7 siblings, 1 reply; 14+ messages in thread
From: James Clark @ 2024-08-28 14:07 UTC (permalink / raw)
To: irogers, linux-perf-users, kan.liang, ak
Cc: James Clark, John Garry, Will Deacon, Mike Leach, Leo Yan,
Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo,
Namhyung Kim, Mark Rutland, Alexander Shishkin, Jiri Olsa,
Adrian Hunter, Weilin Wang, Dominique Martinet, Athira Rajeev,
Colin Ian King, Yang Jihong, Ze Gao, Sun Haiyong, Jing Zhang,
Yicong Yang, linux-arm-kernel, linux-kernel
Test that one cycles event is opened for each core PMU when "perf stat"
is run without arguments.
The event line can either be output as "pmu/cycles/" or just "cycles" if
there is only one PMU. Include 2 spaces for padding in the one PMU case
to avoid matching when the word cycles is included in metric
descriptions.
Signed-off-by: James Clark <james.clark@linaro.org>
---
tools/perf/tests/shell/stat.sh | 21 +++++++++++++++++++++
1 file changed, 21 insertions(+)
diff --git a/tools/perf/tests/shell/stat.sh b/tools/perf/tests/shell/stat.sh
index 525d0c44fdc6..24ace1de71cc 100755
--- a/tools/perf/tests/shell/stat.sh
+++ b/tools/perf/tests/shell/stat.sh
@@ -148,6 +148,26 @@ test_cputype() {
echo "cputype test [Success]"
}
+test_hybrid() {
+ # Test the default stat command on hybrid devices opens one cycles event for
+ # each CPU type.
+ echo "hybrid test"
+
+ # Count the number of core PMUs
+ pmus=$(ls /sys/bus/event_source/devices/*/cpus 2>/dev/null | wc -l)
+
+ # Run default Perf stat
+ cycles_events=$(perf stat -- true 2>&1 | grep -E "/cycles/| cycles " | wc -l)
+
+ if [ "$pmus" -ne "$cycles_events" ]
+ then
+ echo "hybrid test [Found $pmus PMUs but $cycles_events cycles events. Failed]"
+ err=1
+ return
+ fi
+ echo "hybrid test [Success]"
+}
+
test_default_stat
test_stat_record_report
test_stat_record_script
@@ -155,4 +175,5 @@ test_stat_repeat_weak_groups
test_topdown_groups
test_topdown_weak_groups
test_cputype
+test_hybrid
exit $err
--
2.34.1
^ permalink raw reply related [flat|nested] 14+ messages in thread* Re: [PATCH v5 7/7] perf test: Add a test for default perf stat command
2024-08-28 14:07 ` [PATCH v5 7/7] perf test: Add a test for default perf stat command James Clark
@ 2024-08-30 6:32 ` Namhyung Kim
2024-08-30 8:29 ` James Clark
0 siblings, 1 reply; 14+ messages in thread
From: Namhyung Kim @ 2024-08-30 6:32 UTC (permalink / raw)
To: James Clark
Cc: irogers, linux-perf-users, kan.liang, ak, John Garry, Will Deacon,
Mike Leach, Leo Yan, Peter Zijlstra, Ingo Molnar,
Arnaldo Carvalho de Melo, Mark Rutland, Alexander Shishkin,
Jiri Olsa, Adrian Hunter, Weilin Wang, Dominique Martinet,
Athira Rajeev, Colin Ian King, Yang Jihong, Ze Gao, Sun Haiyong,
Jing Zhang, Yicong Yang, linux-arm-kernel, linux-kernel
Hello,
On Wed, Aug 28, 2024 at 03:07:21PM +0100, James Clark wrote:
> Test that one cycles event is opened for each core PMU when "perf stat"
> is run without arguments.
>
> The event line can either be output as "pmu/cycles/" or just "cycles" if
> there is only one PMU. Include 2 spaces for padding in the one PMU case
> to avoid matching when the word cycles is included in metric
> descriptions.
>
> Signed-off-by: James Clark <james.clark@linaro.org>
> ---
> tools/perf/tests/shell/stat.sh | 21 +++++++++++++++++++++
> 1 file changed, 21 insertions(+)
>
> diff --git a/tools/perf/tests/shell/stat.sh b/tools/perf/tests/shell/stat.sh
> index 525d0c44fdc6..24ace1de71cc 100755
> --- a/tools/perf/tests/shell/stat.sh
> +++ b/tools/perf/tests/shell/stat.sh
> @@ -148,6 +148,26 @@ test_cputype() {
> echo "cputype test [Success]"
> }
>
> +test_hybrid() {
> + # Test the default stat command on hybrid devices opens one cycles event for
> + # each CPU type.
> + echo "hybrid test"
> +
> + # Count the number of core PMUs
> + pmus=$(ls /sys/bus/event_source/devices/*/cpus 2>/dev/null | wc -l)
Is it working on non-hybrid systems? I don't think they have cpus file
in the core PMU.
Thanks,
Namhyung
> +
> + # Run default Perf stat
> + cycles_events=$(perf stat -- true 2>&1 | grep -E "/cycles/| cycles " | wc -l)
> +
> + if [ "$pmus" -ne "$cycles_events" ]
> + then
> + echo "hybrid test [Found $pmus PMUs but $cycles_events cycles events. Failed]"
> + err=1
> + return
> + fi
> + echo "hybrid test [Success]"
> +}
> +
> test_default_stat
> test_stat_record_report
> test_stat_record_script
> @@ -155,4 +175,5 @@ test_stat_repeat_weak_groups
> test_topdown_groups
> test_topdown_weak_groups
> test_cputype
> +test_hybrid
> exit $err
> --
> 2.34.1
>
^ permalink raw reply [flat|nested] 14+ messages in thread* Re: [PATCH v5 7/7] perf test: Add a test for default perf stat command
2024-08-30 6:32 ` Namhyung Kim
@ 2024-08-30 8:29 ` James Clark
2024-08-30 8:45 ` James Clark
0 siblings, 1 reply; 14+ messages in thread
From: James Clark @ 2024-08-30 8:29 UTC (permalink / raw)
To: Namhyung Kim
Cc: irogers, linux-perf-users, kan.liang, ak, John Garry, Will Deacon,
Mike Leach, Leo Yan, Peter Zijlstra, Ingo Molnar,
Arnaldo Carvalho de Melo, Mark Rutland, Alexander Shishkin,
Jiri Olsa, Adrian Hunter, Weilin Wang, Dominique Martinet,
Athira Rajeev, Colin Ian King, Yang Jihong, Ze Gao, Sun Haiyong,
Jing Zhang, Yicong Yang, linux-arm-kernel, linux-kernel
On 30/08/2024 7:32 am, Namhyung Kim wrote:
> Hello,
>
> On Wed, Aug 28, 2024 at 03:07:21PM +0100, James Clark wrote:
>> Test that one cycles event is opened for each core PMU when "perf stat"
>> is run without arguments.
>>
>> The event line can either be output as "pmu/cycles/" or just "cycles" if
>> there is only one PMU. Include 2 spaces for padding in the one PMU case
>> to avoid matching when the word cycles is included in metric
>> descriptions.
>>
>> Signed-off-by: James Clark <james.clark@linaro.org>
>> ---
>> tools/perf/tests/shell/stat.sh | 21 +++++++++++++++++++++
>> 1 file changed, 21 insertions(+)
>>
>> diff --git a/tools/perf/tests/shell/stat.sh b/tools/perf/tests/shell/stat.sh
>> index 525d0c44fdc6..24ace1de71cc 100755
>> --- a/tools/perf/tests/shell/stat.sh
>> +++ b/tools/perf/tests/shell/stat.sh
>> @@ -148,6 +148,26 @@ test_cputype() {
>> echo "cputype test [Success]"
>> }
>>
>> +test_hybrid() {
>> + # Test the default stat command on hybrid devices opens one cycles event for
>> + # each CPU type.
>> + echo "hybrid test"
>> +
>> + # Count the number of core PMUs
>> + pmus=$(ls /sys/bus/event_source/devices/*/cpus 2>/dev/null | wc -l)
>
> Is it working on non-hybrid systems? I don't think they have cpus file
> in the core PMU.
>
> Thanks,
> Namhyung
>
Good point I only tested on Arm non-hybrid. I can change it to assume 1
PMU for no cpus files?
^ permalink raw reply [flat|nested] 14+ messages in thread* Re: [PATCH v5 7/7] perf test: Add a test for default perf stat command
2024-08-30 8:29 ` James Clark
@ 2024-08-30 8:45 ` James Clark
2024-09-02 17:43 ` Namhyung Kim
0 siblings, 1 reply; 14+ messages in thread
From: James Clark @ 2024-08-30 8:45 UTC (permalink / raw)
To: Namhyung Kim
Cc: irogers, linux-perf-users, kan.liang, ak, John Garry, Will Deacon,
Mike Leach, Leo Yan, Peter Zijlstra, Ingo Molnar,
Arnaldo Carvalho de Melo, Mark Rutland, Alexander Shishkin,
Jiri Olsa, Adrian Hunter, Weilin Wang, Dominique Martinet,
Athira Rajeev, Colin Ian King, Yang Jihong, Ze Gao, Sun Haiyong,
Jing Zhang, Yicong Yang, linux-arm-kernel, linux-kernel
On 30/08/2024 9:29 am, James Clark wrote:
>
>
> On 30/08/2024 7:32 am, Namhyung Kim wrote:
>> Hello,
>>
>> On Wed, Aug 28, 2024 at 03:07:21PM +0100, James Clark wrote:
>>> Test that one cycles event is opened for each core PMU when "perf stat"
>>> is run without arguments.
>>>
>>> The event line can either be output as "pmu/cycles/" or just "cycles" if
>>> there is only one PMU. Include 2 spaces for padding in the one PMU case
>>> to avoid matching when the word cycles is included in metric
>>> descriptions.
>>>
>>> Signed-off-by: James Clark <james.clark@linaro.org>
>>> ---
>>> tools/perf/tests/shell/stat.sh | 21 +++++++++++++++++++++
>>> 1 file changed, 21 insertions(+)
>>>
>>> diff --git a/tools/perf/tests/shell/stat.sh
>>> b/tools/perf/tests/shell/stat.sh
>>> index 525d0c44fdc6..24ace1de71cc 100755
>>> --- a/tools/perf/tests/shell/stat.sh
>>> +++ b/tools/perf/tests/shell/stat.sh
>>> @@ -148,6 +148,26 @@ test_cputype() {
>>> echo "cputype test [Success]"
>>> }
>>> +test_hybrid() {
>>> + # Test the default stat command on hybrid devices opens one cycles
>>> event for
>>> + # each CPU type.
>>> + echo "hybrid test"
>>> +
>>> + # Count the number of core PMUs
>>> + pmus=$(ls /sys/bus/event_source/devices/*/cpus 2>/dev/null | wc -l)
>>
>> Is it working on non-hybrid systems? I don't think they have cpus file
>> in the core PMU.
>>
>> Thanks,
>> Namhyung
>>
>
> Good point I only tested on Arm non-hybrid. I can change it to assume 1
> PMU for no cpus files?
Or maybe assume 1 if a /sys/bus/event_source/devices/cpu folder exists?
Not sure which is best but either will work.
^ permalink raw reply [flat|nested] 14+ messages in thread* Re: [PATCH v5 7/7] perf test: Add a test for default perf stat command
2024-08-30 8:45 ` James Clark
@ 2024-09-02 17:43 ` Namhyung Kim
0 siblings, 0 replies; 14+ messages in thread
From: Namhyung Kim @ 2024-09-02 17:43 UTC (permalink / raw)
To: James Clark
Cc: irogers, linux-perf-users, kan.liang, ak, John Garry, Will Deacon,
Mike Leach, Leo Yan, Peter Zijlstra, Ingo Molnar,
Arnaldo Carvalho de Melo, Mark Rutland, Alexander Shishkin,
Jiri Olsa, Adrian Hunter, Weilin Wang, Dominique Martinet,
Athira Rajeev, Colin Ian King, Yang Jihong, Ze Gao, Sun Haiyong,
Jing Zhang, Yicong Yang, linux-arm-kernel, linux-kernel
On Fri, Aug 30, 2024 at 09:45:11AM +0100, James Clark wrote:
>
>
> On 30/08/2024 9:29 am, James Clark wrote:
> >
> >
> > On 30/08/2024 7:32 am, Namhyung Kim wrote:
> > > Hello,
> > >
> > > On Wed, Aug 28, 2024 at 03:07:21PM +0100, James Clark wrote:
> > > > Test that one cycles event is opened for each core PMU when "perf stat"
> > > > is run without arguments.
> > > >
> > > > The event line can either be output as "pmu/cycles/" or just "cycles" if
> > > > there is only one PMU. Include 2 spaces for padding in the one PMU case
> > > > to avoid matching when the word cycles is included in metric
> > > > descriptions.
> > > >
> > > > Signed-off-by: James Clark <james.clark@linaro.org>
> > > > ---
> > > > tools/perf/tests/shell/stat.sh | 21 +++++++++++++++++++++
> > > > 1 file changed, 21 insertions(+)
> > > >
> > > > diff --git a/tools/perf/tests/shell/stat.sh
> > > > b/tools/perf/tests/shell/stat.sh
> > > > index 525d0c44fdc6..24ace1de71cc 100755
> > > > --- a/tools/perf/tests/shell/stat.sh
> > > > +++ b/tools/perf/tests/shell/stat.sh
> > > > @@ -148,6 +148,26 @@ test_cputype() {
> > > > echo "cputype test [Success]"
> > > > }
> > > > +test_hybrid() {
> > > > + # Test the default stat command on hybrid devices opens one
> > > > cycles event for
> > > > + # each CPU type.
> > > > + echo "hybrid test"
> > > > +
> > > > + # Count the number of core PMUs
> > > > + pmus=$(ls /sys/bus/event_source/devices/*/cpus 2>/dev/null | wc -l)
> > >
> > > Is it working on non-hybrid systems? I don't think they have cpus file
> > > in the core PMU.
> > >
> > > Thanks,
> > > Namhyung
> > >
> >
> > Good point I only tested on Arm non-hybrid. I can change it to assume 1
> > PMU for no cpus files?
>
> Or maybe assume 1 if a /sys/bus/event_source/devices/cpu folder exists? Not
> sure which is best but either will work.
Some arch might not have a cpu PMU, I think we can assume 1 and update
only if it finds cpus files.
Thanks,
Namhyung
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: [PATCH v5 0/7] Event parsing fixes
2024-08-28 14:07 [PATCH v5 0/7] Event parsing fixes James Clark
` (6 preceding siblings ...)
2024-08-28 14:07 ` [PATCH v5 7/7] perf test: Add a test for default perf stat command James Clark
@ 2024-08-28 17:02 ` Andi Kleen
2024-08-28 17:14 ` Ian Rogers
7 siblings, 1 reply; 14+ messages in thread
From: Andi Kleen @ 2024-08-28 17:02 UTC (permalink / raw)
To: James Clark
Cc: irogers, linux-perf-users, kan.liang, John Garry, Will Deacon,
Mike Leach, Leo Yan, Peter Zijlstra, Ingo Molnar,
Arnaldo Carvalho de Melo, Namhyung Kim, Mark Rutland,
Alexander Shishkin, Jiri Olsa, Adrian Hunter, Weilin Wang,
Athira Rajeev, Dominique Martinet, Yang Jihong, Colin Ian King,
Ze Gao, Yunseong Kim, Sun Haiyong, Jing Zhang, Yicong Yang,
linux-arm-kernel, linux-kernel
On Wed, Aug 28, 2024 at 03:07:14PM +0100, James Clark wrote:
> I rebased this one and made some other fixes so that I could test it,
> so I thought I'd repost it here in case it's helpful. I also added a
> new test.
>
> But for the testing it all looks ok.
>
> There is one small difference where it now hides _all_ <not supported>
> events, when previously it would only hide some selected subset of
> events like "stalled-cycles-frontend". I think this is now more
> consistent across platforms because, for example, Apple M only has
> cycles and instructions, and the rest of the default events would
> always show as <not supported> there.
I'm not a big fan of hiding all of <unsupported>, when they are explicitly
specified on the command line they should be absolutely shown.
I do have tools that specify events on the command line and expect
the same order/events in the output. This might well cause breakage.
-Andi
^ permalink raw reply [flat|nested] 14+ messages in thread* Re: [PATCH v5 0/7] Event parsing fixes
2024-08-28 17:02 ` [PATCH v5 0/7] Event parsing fixes Andi Kleen
@ 2024-08-28 17:14 ` Ian Rogers
0 siblings, 0 replies; 14+ messages in thread
From: Ian Rogers @ 2024-08-28 17:14 UTC (permalink / raw)
To: Andi Kleen
Cc: James Clark, linux-perf-users, kan.liang, John Garry, Will Deacon,
Mike Leach, Leo Yan, Peter Zijlstra, Ingo Molnar,
Arnaldo Carvalho de Melo, Namhyung Kim, Mark Rutland,
Alexander Shishkin, Jiri Olsa, Adrian Hunter, Weilin Wang,
Athira Rajeev, Dominique Martinet, Yang Jihong, Colin Ian King,
Ze Gao, Yunseong Kim, Sun Haiyong, Jing Zhang, Yicong Yang,
linux-arm-kernel, linux-kernel
On Wed, Aug 28, 2024 at 10:02 AM Andi Kleen <ak@linux.intel.com> wrote:
>
> On Wed, Aug 28, 2024 at 03:07:14PM +0100, James Clark wrote:
> > I rebased this one and made some other fixes so that I could test it,
> > so I thought I'd repost it here in case it's helpful. I also added a
> > new test.
> >
> > But for the testing it all looks ok.
> >
> > There is one small difference where it now hides _all_ <not supported>
> > events, when previously it would only hide some selected subset of
> > events like "stalled-cycles-frontend". I think this is now more
> > consistent across platforms because, for example, Apple M only has
> > cycles and instructions, and the rest of the default events would
> > always show as <not supported> there.
>
> I'm not a big fan of hiding all of <unsupported>, when they are explicitly
> specified on the command line they should be absolutely shown.
>
> I do have tools that specify events on the command line and expect
> the same order/events in the output. This might well cause breakage.
The patch series doesn't do this, it just doesn't display unsupported
events that came from the set of default events - default evsels have
a "skippable" flag set. In patch 3 James has added to
should_skip_zero_counter in stat-display.c:
+ /*
+ * Skip unsupported default events when not verbose. (default events
+ * are all marked 'skippable').
+ */
+ if (verbose == 0 && counter->skippable && !counter->supported)
+ return true;
Thanks,
Ian
^ permalink raw reply [flat|nested] 14+ messages in thread