Currently all the -p option PID arguments tasks values
get aggregated and printed as single values.
Adding --per-tasks option to print values per task.
$ perf stat -e cycles,instructions --per-task -p 12451,16173
^C
Performance counter stats for process id '12451,16173':
TASK-12451 77,784 cycles
TASK-16173 64,809 cycles
TASK-12451 14,432 instructions
TASK-16173 14,452 instructions
3.854957122 seconds time elapsed
Also woks under interval mode:
$ ./perf stat -e cycles,instructions --per-task -p 16431,16173 -I 1000
# time task counts unit events
1.000085816 TASK-16173 0 cycles
1.000085816 TASK-16431 3,358,360,926 cycles
1.000085816 TASK-16173 0 instructions
1.000085816 TASK-16431 9,062,422,086 instructions
2.000212262 TASK-16173 65,386 cycles
2.000212262 TASK-16431 3,349,355,309 cycles
2.000212262 TASK-16173 12,151 instructions
2.000212262 TASK-16431 9,039,401,422 instructions
3.000333402 TASK-16173 62,797 cycles
3.000333402 TASK-16431 3,357,140,183 cycles
3.000333402 TASK-16173 12,208 instructions
3.000333402 TASK-16431 9,058,080,762 instructions
^C 3.375949851 TASK-16173 0 cycles
3.375949851 TASK-16431 1,264,764,804 cycles
3.375949851 TASK-16173 0 instructions
3.375949851 TASK-16431 3,414,532,317 instructions
Link: http://lkml.kernel.org/n/tip-0v0ixd9k7o9z1u8hqngm1...@git.kernel.org
Signed-off-by: Jiri Olsa <jo...@kernel.org>
---
tools/perf/builtin-stat.c | 57 +++++++++++++++++++++++++++++++++++++++++++++--
tools/perf/util/stat.h | 1 +
2 files changed, 56 insertions(+), 2 deletions(-)
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 4a7cad8debac..293d1029e2ba 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -237,6 +237,7 @@ process_counter_values(struct perf_evsel *evsel, int cpu,
int thread,
count = &zero;
switch (aggr_mode) {
+ case AGGR_TASK:
case AGGR_CORE:
case AGGR_SOCKET:
case AGGR_NONE:
@@ -605,6 +606,11 @@ static void aggr_printout(struct perf_evsel *evsel, int
id, int nr)
csv_output ? 0 : -4,
perf_evsel__cpus(evsel)->map[id], csv_sep);
break;
+ case AGGR_TASK:
+ fprintf(output, "TASK-%*d%s",
+ csv_output ? 0 : -8,
+ evsel->threads->map[id], csv_sep);
+ break;
case AGGR_GLOBAL:
default:
break;
@@ -753,6 +759,40 @@ static void print_aggr(char *prefix)
}
}
+static void print_aggr_task(struct perf_evsel *counter, char *prefix)
+{
+ int nthreads = thread_map__nr(counter->threads);
+ int ncpus = cpu_map__nr(counter->cpus);
+ int cpu, thread;
+ double uval;
+
+ for (thread = 0; thread < nthreads; thread++) {
+ u64 ena = 0, run = 0, val = 0;
+
+ for (cpu = 0; cpu < ncpus; cpu++) {
+ val += perf_counts(counter->counts, cpu, thread)->val;
+ ena += perf_counts(counter->counts, cpu, thread)->ena;
+ run += perf_counts(counter->counts, cpu, thread)->run;
+ }
+
+ if (prefix)
+ fprintf(output, "%s", prefix);
+
+ uval = val * counter->scale;
+
+ if (nsec_counter(counter))
+ nsec_printout(thread, 0, counter, uval);
+ else
+ abs_printout(thread, 0, counter, uval);
+
+ if (!csv_output)
+ print_noise(counter, 1.0);
+
+ print_running(run, ena);
+ fputc('\n', output);
+ }
+}
+
/*
* Print out the results of a single counter:
* aggregated counts in system-wide mode
@@ -879,6 +919,9 @@ static void print_interval(char *prefix, struct timespec
*ts)
case AGGR_NONE:
fprintf(output, "# time CPU
counts %*s events\n", unit_width, "unit");
break;
+ case AGGR_TASK:
+ fprintf(output, "# time task
counts %*s events\n", unit_width, "unit");
+ break;
case AGGR_GLOBAL:
default:
fprintf(output, "# time counts
%*s events\n", unit_width, "unit");
@@ -947,6 +990,10 @@ static void print_counters(struct timespec *ts, int argc,
const char **argv)
case AGGR_SOCKET:
print_aggr(prefix);
break;
+ case AGGR_TASK:
+ evlist__for_each(evsel_list, counter)
+ print_aggr_task(counter, prefix);
+ break;
case AGGR_GLOBAL:
evlist__for_each(evsel_list, counter)
print_counter_aggr(counter, prefix);
@@ -1034,6 +1081,7 @@ static int perf_stat_init_aggr_mode(void)
break;
case AGGR_NONE:
case AGGR_GLOBAL:
+ case AGGR_TASK:
default:
break;
}
@@ -1258,6 +1306,8 @@ int cmd_stat(int argc, const char **argv, const char
*prefix __maybe_unused)
"aggregate counts per processor socket", AGGR_SOCKET),
OPT_SET_UINT(0, "per-core", &aggr_mode,
"aggregate counts per physical processor core", AGGR_CORE),
+ OPT_SET_UINT(0, "per-task", &aggr_mode,
+ "aggregate counts per task", AGGR_TASK),
OPT_UINTEGER('D', "delay", &initial_delay,
"ms to wait before starting measurement after program
start"),
OPT_END()
@@ -1349,8 +1399,11 @@ int cmd_stat(int argc, const char **argv, const char
*prefix __maybe_unused)
run_count = 1;
}
- /* no_aggr, cgroup are for system-wide only */
- if ((aggr_mode != AGGR_GLOBAL || nr_cgroups) &&
+ /*
+ * no_aggr, cgroup are for system-wide only
+ * --per-task is aggregated per task, we dont mix it with cpu mode
+ */
+ if (((aggr_mode != AGGR_GLOBAL && aggr_mode != AGGR_TASK) ||
nr_cgroups) &&
!target__has_cpu(&target)) {
fprintf(stderr, "both cgroup and no-aggregation "
"modes only available in system-wide mode\n");
diff --git a/tools/perf/util/stat.h b/tools/perf/util/stat.h
index 615c779eb42a..e4c616af057d 100644
--- a/tools/perf/util/stat.h
+++ b/tools/perf/util/stat.h
@@ -29,6 +29,7 @@ enum aggr_mode {
AGGR_GLOBAL,
AGGR_SOCKET,
AGGR_CORE,
+ AGGR_TASK,
};
void update_stats(struct stats *stats, u64 val);
--
1.9.3
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
