Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Riccardo Mancini <rickyman7@gmail.com>
To:     Arnaldo Carvalho de Melo <acme@kernel.org>
Cc:     Ian Rogers <irogers@google.com>,
        Namhyung Kim <namhyung@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        Mark Rutland <mark.rutland@arm.com>,
        Jiri Olsa <jolsa@redhat.com>, linux-kernel@vger.kernel.org,
        linux-perf-users@vger.kernel.org,
        Riccardo Mancini <rickyman7@gmail.com>
Subject: [RFC PATCH v1 34/37] perf record: apply multithreading in init and fini phases
Date:   Sat, 21 Aug 2021 11:19:40 +0200
Message-Id: <f73e485eb0d29539722da396555d9fcbd0267642.1629490974.git.rickyman7@gmail.com>
In-Reply-To: <cover.1629490974.git.rickyman7@gmail.com>
References: <cover.1629490974.git.rickyman7@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

This patch grows the multithreaded portion of perf-record, marked by the
perf_set_multithreaded and perf_set_singlethreaded functions to the whole
init and fini part of __cmd_record.
By doing so, perf-record can take advantage of the parallelized evlist
operations (open, enable, disable, close).

This patch also needs to handle the case in which evlist and synthesis
multithreading are not enabled at the same time. Therefore, in
record__synthesize multithreading is enabled/disabled and then
disabled/renabled if needed.

Signed-off-by: Riccardo Mancini <rickyman7@gmail.com>
---
 tools/perf/builtin-record.c | 22 +++++++++++++++++++++-
 1 file changed, 21 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index e2d2445e05d7c07a..db9ec08db672f994 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -1411,6 +1411,7 @@ static int record__synthesize(struct record *rec, bool tail)
 	struct perf_tool *tool = &rec->tool;
 	int err = 0;
 	event_op f = process_synthesized_event;
+	bool perf_was_singlethreaded = perf_singlethreaded;
 
 	if (rec->opts.tail_synthesize != tail)
 		return 0;
@@ -1499,12 +1500,16 @@ static int record__synthesize(struct record *rec, bool tail)
 	if (rec->opts.multithreaded_synthesis) {
 		perf_set_multithreaded();
 		f = process_locked_synthesized_event;
+	} else {
+		perf_set_singlethreaded();
 	}
 
 	err = __machine__synthesize_threads(machine, tool, &opts->target, rec->evlist->core.threads,
 					    f, opts->sample_address);
 
-	if (rec->opts.multithreaded_synthesis)
+	if (!perf_was_singlethreaded && perf_singlethreaded)
+		perf_set_multithreaded();
+	if (perf_was_singlethreaded && !perf_singlethreaded)
 		perf_set_singlethreaded();
 
 out:
@@ -1735,6 +1740,9 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)
 
 	record__uniquify_name(rec);
 
+	if (rec->opts.multithreaded_evlist)
+		perf_set_multithreaded();
+
 	if (record__open(rec) != 0) {
 		err = -1;
 		goto out_child;
@@ -1877,6 +1885,10 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)
 		}
 	}
 
+	// disable locks in the main thread since there is no multithreading
+	if (rec->opts.multithreaded_evlist)
+		perf_set_singlethreaded();
+
 	trigger_ready(&auxtrace_snapshot_trigger);
 	trigger_ready(&switch_output_trigger);
 	perf_hooks__invoke_record_start();
@@ -1998,6 +2010,10 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)
 		}
 	}
 
+	// reenable multithreading for evlist
+	if (rec->opts.multithreaded_evlist)
+		perf_set_multithreaded();
+
 	trigger_off(&auxtrace_snapshot_trigger);
 	trigger_off(&switch_output_trigger);
 
@@ -2099,6 +2115,10 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)
 
 	if (!opts->no_bpf_event)
 		evlist__stop_sb_thread(rec->sb_evlist);
+
+	// disable multithreaded mode on exit
+	if (rec->opts.multithreaded_evlist)
+		perf_set_singlethreaded();
 	return status;
 }
 
-- 
2.31.1