public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Dapeng Mi <dapeng1.mi@linux.intel.com>
To: Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@redhat.com>,
	Arnaldo Carvalho de Melo <acme@kernel.org>,
	Namhyung Kim <namhyung@kernel.org>,
	Ian Rogers <irogers@google.com>,
	Adrian Hunter <adrian.hunter@intel.com>,
	Alexander Shishkin <alexander.shishkin@linux.intel.com>,
	Kan Liang <kan.liang@linux.intel.com>
Cc: linux-perf-users@vger.kernel.org, linux-kernel@vger.kernel.org,
	Yanfei Xu <yanfei.xu@intel.com>, Dapeng Mi <dapeng1.mi@intel.com>,
	Dapeng Mi <dapeng1.mi@linux.intel.com>
Subject: [PATCH 3/4] perf parse-events: Don't move slots event when no topdwon metrics event
Date: Wed,  3 Jul 2024 06:40:36 +0800	[thread overview]
Message-ID: <20240702224037.343958-4-dapeng1.mi@linux.intel.com> (raw)
In-Reply-To: <20240702224037.343958-1-dapeng1.mi@linux.intel.com>

Currently the helper arch_evlist__cmp() would unconditionally move slots
event to be the leader event even though there is no topdown metrics
event in the group.

perf stat -e "{instructions,slots}" -C 0 sleep 1
WARNING: events were regrouped to match PMUs

 Performance counter stats for 'CPU(s) 0':

        27,581,148      slots
         8,390,827      instructions

       1.045546595 seconds time elapsed

This is an overkill. It's not necessary to move slots event as the leader
event if there is no topdown metrics event.

Thus only regroup events when there are both topdown slots and metrics
events in a group.

Signed-off-by: Dapeng Mi <dapeng1.mi@linux.intel.com>
---
 tools/perf/arch/x86/util/evlist.c |  6 ++++--
 tools/perf/util/evlist.h          |  7 ++++++-
 tools/perf/util/parse-events.c    | 35 ++++++++++++++++++-------------
 3 files changed, 31 insertions(+), 17 deletions(-)

diff --git a/tools/perf/arch/x86/util/evlist.c b/tools/perf/arch/x86/util/evlist.c
index 7215c7c7b435..a1e78be6ebd1 100644
--- a/tools/perf/arch/x86/util/evlist.c
+++ b/tools/perf/arch/x86/util/evlist.c
@@ -73,9 +73,11 @@ int arch_evlist__add_default_attrs(struct evlist *evlist,
 	return ___evlist__add_default_attrs(evlist, attrs, nr_attrs);
 }
 
-int arch_evlist__cmp(const struct evsel *lhs, const struct evsel *rhs)
+int arch_evlist__cmp(const struct evsel *lhs, const struct evsel *rhs, void *priv)
 {
-	if (topdown_sys_has_perf_metrics() &&
+	struct sort_priv *_priv = priv;
+
+	if (topdown_sys_has_perf_metrics() && _priv->topdown_metrics_in_group &&
 	    (arch_evsel__must_be_in_group(lhs) || arch_evsel__must_be_in_group(rhs))) {
 		/* Ensure the topdown slots comes first. */
 		if (strcasestr(lhs->name, "slots") && !strcasestr(lhs->name, "uops_retired.slots"))
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index cb91dc9117a2..14c858dcf5a2 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -89,6 +89,11 @@ struct evsel_str_handler {
 	void	   *handler;
 };
 
+struct sort_priv {
+	int force_grouped_idx;
+	bool topdown_metrics_in_group;
+};
+
 struct evlist *evlist__new(void);
 struct evlist *evlist__new_default(void);
 struct evlist *evlist__new_dummy(void);
@@ -112,7 +117,7 @@ int arch_evlist__add_default_attrs(struct evlist *evlist,
 #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 arch_evlist__cmp(const struct evsel *lhs, const struct evsel *rhs, void *priv);
 
 int evlist__add_dummy(struct evlist *evlist);
 struct evsel *evlist__add_aux_dummy(struct evlist *evlist, bool system_wide);
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 6ed0f9c5581d..a3f7173a7ae2 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -1962,19 +1962,21 @@ static int evsel__compute_group_pmu_name(struct evsel *evsel,
 	return evsel->group_pmu_name ? 0 : -ENOMEM;
 }
 
-__weak int arch_evlist__cmp(const struct evsel *lhs, const struct evsel *rhs)
+__weak int arch_evlist__cmp(const struct evsel *lhs, const struct evsel *rhs,
+			    void *priv __maybe_unused)
 {
 	/* Order by insertion index. */
 	return lhs->core.idx - rhs->core.idx;
 }
 
-static int evlist__cmp(void *_fg_idx, const struct list_head *l, const struct list_head *r)
+static int evlist__cmp(void *_sort_priv, const struct list_head *l, const struct list_head *r)
 {
 	const struct perf_evsel *lhs_core = container_of(l, struct perf_evsel, node);
 	const struct evsel *lhs = container_of(lhs_core, struct evsel, core);
 	const struct perf_evsel *rhs_core = container_of(r, struct perf_evsel, node);
 	const struct evsel *rhs = container_of(rhs_core, struct evsel, core);
-	int *force_grouped_idx = _fg_idx;
+	struct sort_priv *sort_priv = _sort_priv;
+	int force_grouped_idx = sort_priv->force_grouped_idx;
 	int lhs_sort_idx, rhs_sort_idx, ret;
 	const char *lhs_pmu_name, *rhs_pmu_name;
 	bool lhs_has_group, rhs_has_group;
@@ -1992,8 +1994,8 @@ static int evlist__cmp(void *_fg_idx, const struct list_head *l, const struct li
 		lhs_sort_idx = lhs_core->leader->idx;
 	} else {
 		lhs_has_group = false;
-		lhs_sort_idx = *force_grouped_idx != -1 && arch_evsel__must_be_in_group(lhs)
-			? *force_grouped_idx
+		lhs_sort_idx = force_grouped_idx != -1 && arch_evsel__must_be_in_group(lhs)
+			? force_grouped_idx
 			: lhs_core->idx;
 	}
 	if (rhs_core->leader != rhs_core || rhs_core->nr_members > 1) {
@@ -2001,8 +2003,8 @@ static int evlist__cmp(void *_fg_idx, const struct list_head *l, const struct li
 		rhs_sort_idx = rhs_core->leader->idx;
 	} else {
 		rhs_has_group = false;
-		rhs_sort_idx = *force_grouped_idx != -1 && arch_evsel__must_be_in_group(rhs)
-			? *force_grouped_idx
+		rhs_sort_idx = force_grouped_idx != -1 && arch_evsel__must_be_in_group(rhs)
+			? force_grouped_idx
 			: rhs_core->idx;
 	}
 
@@ -2019,16 +2021,17 @@ static int evlist__cmp(void *_fg_idx, const struct list_head *l, const struct li
 	}
 
 	/* Architecture specific sorting. */
-	return arch_evlist__cmp(lhs, rhs);
+	return arch_evlist__cmp(lhs, rhs, _sort_priv);
 }
 
 static int parse_events__sort_events_and_fix_groups(struct list_head *list)
 {
-	int idx = 0, force_grouped_idx = -1;
 	struct evsel *pos, *cur_leader = NULL;
 	struct perf_evsel *cur_leaders_grp = NULL;
 	bool idx_changed = false, cur_leader_force_grouped = false;
 	int orig_num_leaders = 0, num_leaders = 0;
+	struct sort_priv sort_priv = {-1, false};
+	int idx = 0;
 	int ret;
 
 	/*
@@ -2053,13 +2056,17 @@ static int parse_events__sort_events_and_fix_groups(struct list_head *list)
 		pos->core.idx = idx++;
 
 		/* Remember an index to sort all forced grouped events together to. */
-		if (force_grouped_idx == -1 && pos == pos_leader && pos->core.nr_members < 2 &&
-		    arch_evsel__must_be_in_group(pos))
-			force_grouped_idx = pos->core.idx;
+		if (sort_priv.force_grouped_idx == -1 && pos == pos_leader &&
+		    pos->core.nr_members < 2 && arch_evsel__must_be_in_group(pos))
+			sort_priv.force_grouped_idx = pos->core.idx;
+
+		if (!sort_priv.topdown_metrics_in_group &&
+		    strcasestr(pos->name, "topdown"))
+			sort_priv.topdown_metrics_in_group = true;
 	}
 
 	/* Sort events. */
-	list_sort(&force_grouped_idx, list, evlist__cmp);
+	list_sort(&sort_priv, list, evlist__cmp);
 
 	/*
 	 * Recompute groups, splitting for PMUs and adding groups for events
@@ -2070,7 +2077,7 @@ static int parse_events__sort_events_and_fix_groups(struct list_head *list)
 		const struct evsel *pos_leader = evsel__leader(pos);
 		const char *pos_pmu_name = pos->group_pmu_name;
 		const char *cur_leader_pmu_name;
-		bool pos_force_grouped = force_grouped_idx != -1 &&
+		bool pos_force_grouped = sort_priv.force_grouped_idx != -1 &&
 			arch_evsel__must_be_in_group(pos);
 
 		/* Reset index and nr_members. */
-- 
2.40.1


  parent reply	other threads:[~2024-07-02  6:39 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-07-02 22:40 [PATCH 0/4] Bug fixes on topdown metrics group leader selection Dapeng Mi
2024-07-02 22:40 ` [PATCH 1/4] perf topdown: Correct leader selection with sample_read enabled Dapeng Mi
2024-07-02 16:05   ` Liang, Kan
2024-07-03  2:46     ` Mi, Dapeng
2024-07-03 13:50       ` Liang, Kan
2024-07-02 22:40 ` [PATCH 2/4] perf parse-events: Don't move topdown metrics events when sorting events Dapeng Mi
2024-07-02 22:40 ` Dapeng Mi [this message]
2024-07-02 18:03   ` [PATCH 3/4] perf parse-events: Don't move slots event when no topdwon metrics event Liang, Kan
2024-07-03  2:51     ` Mi, Dapeng
2024-07-02 22:40 ` [PATCH 4/4] perf tests: Add leader sampling test in record tests Dapeng Mi
2024-07-02 18:07   ` Liang, Kan
2024-07-03  2:53     ` Mi, Dapeng

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20240702224037.343958-4-dapeng1.mi@linux.intel.com \
    --to=dapeng1.mi@linux.intel.com \
    --cc=acme@kernel.org \
    --cc=adrian.hunter@intel.com \
    --cc=alexander.shishkin@linux.intel.com \
    --cc=dapeng1.mi@intel.com \
    --cc=irogers@google.com \
    --cc=kan.liang@linux.intel.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-perf-users@vger.kernel.org \
    --cc=mingo@redhat.com \
    --cc=namhyung@kernel.org \
    --cc=peterz@infradead.org \
    --cc=yanfei.xu@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox