perf stat: Support overall statistics for interval mode
cover perf stat: Support overall statistics for interval mode
Currently perf-stat supports to print counts at regular interval (-I), but it's not very easy for user to get the overall statistics.
With this patchset, it supports to report the summary at the end of interval output.
For example,
root@kbl-ppc:~# perf stat -e cycles -I1000 --interval-count 2
# time counts unit events
1.000412064 2,281,114 cycles
2.001383658 2,547,880 cyclesPerformance counter stats for 'system wide':
4,828,994 cycles
2.002860349 seconds time elapsed root@kbl-ppc:~# perf stat -e cycles,instructions -I1000 --interval-count 2
# time counts unit events
1.000389902 1,536,093 cycles
1.000389902 420,226 instructions # 0.27 insn per cycle
2.001433453 2,213,952 cycles
2.001433453 735,465 instructions # 0.33 insn per cyclePerformance counter stats for 'system wide':
3,750,045 cycles
1,155,691 instructions # 0.31 insn per cycle
2.003023361 seconds time elapsed root@kbl-ppc:~# perf stat -M CPI,IPC -I1000 --interval-count 2
# time counts unit events
1.000435121 905,303 inst_retired.any # 2.9 CPI
1.000435121 2,663,333 cycles
1.000435121 914,702 inst_retired.any # 0.3 IPC
1.000435121 2,676,559 cpu_clk_unhalted.thread
2.001615941 1,951,092 inst_retired.any # 1.8 CPI
2.001615941 3,551,357 cycles
2.001615941 1,950,837 inst_retired.any # 0.5 IPC
2.001615941 3,551,044 cpu_clk_unhalted.threadPerformance counter stats for 'system wide':
2,856,395 inst_retired.any # 2.2 CPI
6,214,690 cycles
2,865,539 inst_retired.any # 0.5 IPC
6,227,603 cpu_clk_unhalted.thread2.003403078 seconds time elapsed
Jin Yao (2):
perf evsel: Create counts for collecting summary data perf stat: Report summary for interval mode
tools/perf/builtin-stat.c | 14 ++++++- tools/perf/util/evsel.c | 10 ++++- tools/perf/util/evsel.h | 1 + tools/perf/util/stat.c | 77 +++++++++++++++++++++++++++++++++++++++ tools/perf/util/stat.h | 5 +++ 5 files changed, 103 insertions(+), 4 deletions(-)
Signature · 1 line
2.17.1
PATCH 1/2 perf evsel: Create counts for collecting summary data
It would be useful to support the overall statistics for perf-stat interval mode. For example, report the summary at the end of "perf-stat -I" output.
But since perf-stat can support many aggregation modes, such as --per-thread, --per-socket, -M and etc, we need a solution which doesn't bring much complexity.
The idea is to create new 'evsel->summary_counts' which sums up the counts delta per interval. Before reporting the summary, we copy the data from evsel->summary_counts to evsel->counts, and next we just follow current code.
--- tools/perf/util/evsel.c | 10 ++++++++-- tools/perf/util/evsel.h | 1 + tools/perf/util/stat.c | 20 ++++++++++++++++++++ 3 files changed, 29 insertions(+), 2 deletions(-)
tools/perf/util/evsel.c +8 -2
@@ -1286,22 +1286,28 @@ void evsel__delete(struct evsel *evsel)void perf_evsel__compute_deltas (struct evsel *evsel,int cpu,int thread,struct perf_counts_values *count) { -struct perf_counts_values tmp; +struct perf_counts_values tmp, *summary; -if (!evsel->prev_raw_counts) +if (!evsel->prev_raw_counts || !evsel->summary_counts)return ;if (cpu ==-1 ) { tmp = evsel->prev_raw_counts->aggr; evsel->prev_raw_counts->aggr = *count; + summary = &evsel->summary_counts->aggr; }else { tmp = *perf_counts (evsel->prev_raw_counts, cpu, thread); *perf_counts (evsel->prev_raw_counts, cpu, thread) = *count; + summary =perf_counts (evsel->summary_counts, cpu, thread); } count->val = count->val - tmp.val; count->ena = count->ena - tmp.ena; count->run = count->run - tmp.run; + + summary->val += count->val; + summary->ena += count->ena; + summary->run += count->run; } void perf_counts_values__scale(struct perf_counts_values *count,
tools/perf/util/evsel.h +1 -0
@@ -46,6 +46,7 @@ struct evsel {char *filter;struct perf_counts *counts;struct perf_counts *prev_raw_counts; +struct perf_counts *summary_counts;int idx;unsigned long max_events;unsigned long nr_events_printed;
tools/perf/util/stat.c +20 -0
@@ -171,6 +171,24 @@ static void perf_evsel__reset_prev_raw_counts(struct evsel *evsel) } } +static int perf_evsel__alloc_summary_counts (struct evsel *evsel, +int ncpus,int nthreads) +{ +struct perf_counts *counts; + + counts =perf_counts__new (ncpus, nthreads); +if (counts) + evsel->summary_counts = counts; + +return counts ?0 : -ENOMEM ; +} + +static void perf_evsel__free_summary_counts (struct evsel *evsel) +{ +perf_counts__delete (evsel->summary_counts); + evsel->summary_counts =NULL ; +} +static int perf_evsel__alloc_stats (struct evsel *evsel,bool alloc_raw) {int ncpus =perf_evsel__nr_cpus (evsel);
@@ -178,6 +196,7 @@ static int perf_evsel__alloc_stats(struct evsel *evsel, bool alloc_raw)if (perf_evsel__alloc_stat_priv (evsel) <0 ||perf_evsel__alloc_counts (evsel, ncpus, nthreads) <0 || +perf_evsel__alloc_summary_counts (evsel, ncpus, nthreads) <0 || (alloc_raw &&perf_evsel__alloc_prev_raw_counts (evsel, ncpus, nthreads) <0 ))return -ENOMEM ;
@@ -208,6 +227,7 @@ void perf_evlist__free_stats(struct evlist *evlist)perf_evsel__free_stat_priv (evsel);perf_evsel__free_counts (evsel);perf_evsel__free_prev_raw_counts (evsel); +perf_evsel__free_summary_counts (evsel); } }
Signature · 1 line
2.17.1
PATCH 2/2 perf stat: Report summary for interval mode
In reply to Jin YaoCurrently perf-stat supports to print counts at regular interval (-I), but it's not very easy for user to get the overall statistics.
The patch uses 'evsel->summary_counts' to sum up the per interval counts and copy the counts to 'evsel->counts' after printing the interval results. Next, we just follow the non-interval processing.
Let's see some examples,
root@kbl-ppc:~# perf stat -e cycles -I1000 --interval-count 2
# time counts unit events
1.000412064 2,281,114 cycles
2.001383658 2,547,880 cyclesPerformance counter stats for 'system wide':
4,828,994 cycles
2.002860349 seconds time elapsed root@kbl-ppc:~# perf stat -e cycles,instructions -I1000 --interval-count 2
# time counts unit events
1.000389902 1,536,093 cycles
1.000389902 420,226 instructions # 0.27 insn per cycle
2.001433453 2,213,952 cycles
2.001433453 735,465 instructions # 0.33 insn per cyclePerformance counter stats for 'system wide':
3,750,045 cycles
1,155,691 instructions # 0.31 insn per cycle
2.003023361 seconds time elapsed root@kbl-ppc:~# perf stat -M CPI,IPC -I1000 --interval-count 2
# time counts unit events
1.000435121 905,303 inst_retired.any # 2.9 CPI
1.000435121 2,663,333 cycles
1.000435121 914,702 inst_retired.any # 0.3 IPC
1.000435121 2,676,559 cpu_clk_unhalted.thread
2.001615941 1,951,092 inst_retired.any # 1.8 CPI
2.001615941 3,551,357 cycles
2.001615941 1,950,837 inst_retired.any # 0.5 IPC
2.001615941 3,551,044 cpu_clk_unhalted.threadPerformance counter stats for 'system wide':
2,856,395 inst_retired.any # 2.2 CPI
6,214,690 cycles
2,865,539 inst_retired.any # 0.5 IPC
6,227,603 cpu_clk_unhalted.thread2.003403078 seconds time elapsed
--- tools/perf/builtin-stat.c | 14 ++++++++-- tools/perf/util/stat.c | 57 +++++++++++++++++++++++++++++++++++++++ tools/perf/util/stat.h | 5 ++++ 3 files changed, 74 insertions(+), 2 deletions(-)
tools/perf/builtin-stat.c +12 -2
@@ -355,6 +355,7 @@ static void read_counters(struct timespec *rs)static void process_interval (void ) {struct timespec ts, rs; +struct stats walltime_nsecs_stats_bak;clock_gettime (CLOCK_MONOTONIC , &ts);diff_timespec (&rs, &ts, &ref_time);
@@ -367,9 +368,11 @@ static void process_interval(void)pr_err ("failed to write stat round event\n" ); } + walltime_nsecs_stats_bak = walltime_nsecs_stats;init_stats (&walltime_nsecs_stats);update_stats (&walltime_nsecs_stats, stat_config.interval *1000000 );print_counters (&rs,0 ,NULL ); + walltime_nsecs_stats = walltime_nsecs_stats_bak; }static void enable_counters (void )
@@ -732,7 +735,14 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)* avoid arbitrary skew, we must read all counters before closing any * group leaders. */ -read_counters (&(struct timespec ) { .tv_nsec = t1-t0 }); +if (!interval) +read_counters (&(struct timespec ) { .tv_nsec = t1-t0 }); +else { + stat_config.interval =0 ; + stat_config.summary = true; +perf_evlist__copy_summary_counts (evsel_list); +perf_evlist__process_summary_counts (&stat_config, evsel_list); + }/* * We need to keep evsel_list alive, because it's processed
@@ -2149,7 +2159,7 @@ int cmd_stat(int argc, const char **argv) } } -if (!forever && status !=-1 && !interval) +if (!forever && status !=-1 && (!interval || stat_config.summary))print_counters (NULL , argc, argv);if (STAT_RECORD ) {
tools/perf/util/stat.c +57 -0
63 lines not on this page · raw message
tools/perf/util/stat.h +5 -0
@@ -110,6 +110,7 @@ struct perf_stat_config {bool all_kernel;bool all_user;bool percore_show_thread; +bool summary;FILE *output;unsigned int interval;unsigned int timeout;
@@ -199,6 +200,10 @@ void perf_evlist__free_stats(struct evlist *evlist);void perf_evlist__reset_stats (struct evlist *evlist);void perf_evlist__reset_prev_raw_counts (struct evlist *evlist); +void perf_evlist__copy_summary_counts (struct evlist *evlist); +void perf_evlist__process_summary_counts (struct perf_stat_config *config, +struct evlist *evlist); +int perf_stat_process_counter (struct perf_stat_config *config,struct evsel *counter);struct perf_tool ;
Signature · 1 line
2.17.1
Quoted 51 lines · Jin Yao
Currently perf-stat supports to print counts at regular interval (-I), but it's not very easy for user to get the overall statistics.
With this patchset, it supports to report the summary at the end of interval output.
For example,
root@kbl-ppc:~# perf stat -e cycles -I1000 --interval-count 2 # time counts unit events 1.000412064 2,281,114 cycles 2.001383658 2,547,880 cyclesPerformance counter stats for 'system wide':
4,828,994 cycles 2.002860349 seconds time elapsedroot@kbl-ppc:~# perf stat -e cycles,instructions -I1000 --interval-count 2 # time counts unit events 1.000389902 1,536,093 cycles 1.000389902 420,226 instructions # 0.27 insn per cycle 2.001433453 2,213,952 cycles 2.001433453 735,465 instructions # 0.33 insn per cyclePerformance counter stats for 'system wide':
3,750,045 cycles 1,155,691 instructions # 0.31 insn per cycle 2.003023361 seconds time elapsedroot@kbl-ppc:~# perf stat -M CPI,IPC -I1000 --interval-count 2 # time counts unit events 1.000435121 905,303 inst_retired.any # 2.9 CPI 1.000435121 2,663,333 cycles 1.000435121 914,702 inst_retired.any # 0.3 IPC 1.000435121 2,676,559 cpu_clk_unhalted.thread 2.001615941 1,951,092 inst_retired.any # 1.8 CPI 2.001615941 3,551,357 cycles 2.001615941 1,950,837 inst_retired.any # 0.5 IPC 2.001615941 3,551,044 cpu_clk_unhalted.threadPerformance counter stats for 'system wide':
2,856,395 inst_retired.any # 2.2 CPI 6,214,690 cycles 2,865,539 inst_retired.any # 0.5 IPC 6,227,603 cpu_clk_unhalted.thread
Quoted 5 lines
2.003403078 seconds time elapsedJin Yao (2):
perf evsel: Create counts for collecting summary data perf stat: Report summary for interval mode
good idea, but I can't apply this on latest Arnaldo's perf/core, could you please rebase?
thanks,
jirka
Quoted 11 lines
tools/perf/builtin-stat.c | 14 ++++++- tools/perf/util/evsel.c | 10 ++++- tools/perf/util/evsel.h | 1 + tools/perf/util/stat.c | 77 +++++++++++++++++++++++++++++++++++++++ tools/perf/util/stat.h | 5 +++ 5 files changed, 103 insertions(+), 4 deletions(-)Signature · 1 line
2.17.1
Quoted 58 lines · Jiri Olsa
Quoted 53 lines · Jin Yao
Currently perf-stat supports to print counts at regular interval (-I), but it's not very easy for user to get the overall statistics.
With this patchset, it supports to report the summary at the end of interval output.
For example,
root@kbl-ppc:~# perf stat -e cycles -I1000 --interval-count 2 # time counts unit events 1.000412064 2,281,114 cycles 2.001383658 2,547,880 cyclesPerformance counter stats for 'system wide': 4,828,994 cycles 2.002860349 seconds time elapsedroot@kbl-ppc:~# perf stat -e cycles,instructions -I1000 --interval-count 2 # time counts unit events 1.000389902 1,536,093 cycles 1.000389902 420,226 instructions # 0.27 insn per cycle 2.001433453 2,213,952 cycles 2.001433453 735,465 instructions # 0.33 insn per cyclePerformance counter stats for 'system wide':
3,750,045 cycles 1,155,691 instructions # 0.31 insn per cycle 2.003023361 seconds time elapsedroot@kbl-ppc:~# perf stat -M CPI,IPC -I1000 --interval-count 2 # time counts unit events 1.000435121 905,303 inst_retired.any # 2.9 CPI 1.000435121 2,663,333 cycles 1.000435121 914,702 inst_retired.any # 0.3 IPC 1.000435121 2,676,559 cpu_clk_unhalted.thread 2.001615941 1,951,092 inst_retired.any # 1.8 CPI 2.001615941 3,551,357 cycles 2.001615941 1,950,837 inst_retired.any # 0.5 IPC 2.001615941 3,551,044 cpu_clk_unhalted.threadPerformance counter stats for 'system wide': 2,856,395 inst_retired.any # 2.2 CPI 6,214,690 cycles 2,865,539 inst_retired.any # 0.5 IPC 6,227,603 cpu_clk_unhalted.thread 2.003403078 seconds time elapsedQuoted 3 lines
Jin Yao (2):
perf evsel: Create counts for collecting summary data perf stat: Report summary for interval mode
Quoted 6 lines
good idea, but I can't apply this on latest Arnaldo's perf/core, could you please rebase?
thanks,
jirka
Oh, yes, it couldn't be applied on perf/core. I've rebased it.
Thanks
Jin Yao
Quoted 12 lines
Quoted 11 lines
tools/perf/builtin-stat.c | 14 ++++++- tools/perf/util/evsel.c | 10 ++++- tools/perf/util/evsel.h | 1 + tools/perf/util/stat.c | 77 +++++++++++++++++++++++++++++++++++++++ tools/perf/util/stat.h | 5 +++ 5 files changed, 103 insertions(+), 4 deletions(-)Signature · 1 line
2.17.1