summaryrefslogtreecommitdiff
path: root/tools/perf/builtin-stat.c
blob: e45449938b8007084e81881a4ac5d557a311f70e (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
 * builtin-stat.c
 *
 * Builtin stat command: Give a precise performance counters summary
 * overview about any workload, CPU or specific PID.
 *
 * Sample output:

   $ perf stat ~/hackbench 10
   Time: 0.104

    Performance counter stats for '/home/mingo/hackbench':

       1255.538611  task clock ticks     #      10.143 CPU utilization factor
             54011  context switches     #       0.043 M/sec
               385  CPU migrations       #       0.000 M/sec
             17755  pagefaults           #       0.014 M/sec
        3808323185  CPU cycles           #    3033.219 M/sec
        1575111190  instructions         #    1254.530 M/sec
          17367895  cache references     #      13.833 M/sec
           7674421  cache misses         #       6.112 M/sec

    Wall-clock time elapsed:   123.786620 msecs

 *
 * Copyright (C) 2008, Red Hat Inc, Ingo Molnar <mingo@redhat.com>
 *
 * Improvements and fixes by:
 *
 *   Arjan van de Ven <arjan@linux.intel.com>
 *   Yanmin Zhang <yanmin.zhang@intel.com>
 *   Wu Fengguang <fengguang.wu@intel.com>
 *   Mike Galbraith <efault@gmx.de>
 *   Paul Mackerras <paulus@samba.org>
 *   Jaswinder Singh Rajput <jaswinder@kernel.org>
 *
 * Released under the GPL v2. (and only v2, not any later version)
 */

#include "perf.h"
#include "builtin.h"
#include "util/util.h"
#include "util/parse-options.h"
#include "util/parse-events.h"
#include "util/event.h"
#include "util/evlist.h"
#include "util/evsel.h"
#include "util/debug.h"
#include "util/color.h"
#include "util/header.h"
#include "util/cpumap.h"
#include "util/thread.h"
#include "util/thread_map.h"

#include <sys/prctl.h>
#include <math.h>
#include <locale.h>

#define DEFAULT_SEPARATOR	" "

static struct perf_event_attr default_attrs[] = {

  { .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		},
  { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_FRONTEND	},
  { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_BACKEND	},
  { .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:
 */
static struct perf_event_attr detailed_attrs[] = {

  { .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		},
  { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_FRONTEND	},
  { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_BACKEND	},
  { .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		},

  { .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)				},
};

struct perf_evlist		*evsel_list;

static bool			system_wide			=  false;
static int			run_idx				=  0;

static int			run_count			=  1;
static bool			no_inherit			= false;
static bool			scale				=  true;
static bool			no_aggr				= false;
static pid_t			target_pid			= -1;
static pid_t			target_tid			= -1;
static pid_t			child_pid			= -1;
static bool			null_run			=  false;
static bool			detailed_run			=  false;
static bool			sync_run			=  false;
static bool			big_num				=  true;
static int			big_num_opt			=  -1;
static const char		*cpu_list;
static const char		*csv_sep			= NULL;
static bool			csv_output			= false;

static volatile int done = 0;

struct stats
{
	double n, mean, M2;
};

struct perf_stat {
	struct stats	  res_stats[3];
};

static int perf_evsel__alloc_stat_priv(struct perf_evsel *evsel)
{
	evsel->priv = zalloc(sizeof(struct perf_stat));
	return evsel->priv == NULL ? -ENOMEM : 0;
}

static void perf_evsel__free_stat_priv(struct perf_evsel *evsel)
{
	free(evsel->priv);
	evsel->priv = NULL;
}

static void update_stats(struct stats *stats, u64 val)
{
	double delta;

	stats->n++;
	delta = val - stats->mean;
	stats->mean += delta / stats->n;
	stats->M2 += delta*(val - stats->mean);
}

static double avg_stats(struct stats *stats)
{
	return stats->mean;
}

/*
 * http://en.wikipedia.org/wiki/Algorithms_for_calculating_variance
 *
 *       (\Sum n_i^2) - ((\Sum n_i)^2)/n
 * s^2 = -------------------------------
 *                  n - 1
 *
 * http://en.wikipedia.org/wiki/Stddev
 *
 * The std dev of the mean is related to the std dev by:
 *
 *             s
 * s_mean = -------
 *          sqrt(n)
 *
 */
static double stddev_stats(struct stats *stats)
{
	double variance = stats->M2 / (stats->n - 1);
	double variance_mean = variance / stats->n;

	return sqrt(variance_mean);
}

struct stats			runtime_nsecs_stats[MAX_NR_CPUS];
struct stats			runtime_cycles_stats[MAX_NR_CPUS];
struct stats			runtime_stalled_cycles_front_stats[MAX_NR_CPUS];
struct stats			runtime_stalled_cycles_back_stats[MAX_NR_CPUS];
struct stats			runtime_branches_stats[MAX_NR_CPUS];
struct stats			runtime_cacherefs_stats[MAX_NR_CPUS];
struct stats			runtime_l1_dcache_stats[MAX_NR_CPUS];
struct stats			walltime_nsecs_stats;

static int create_perf_stat_counter(struct perf_evsel *evsel)
{
	struct perf_event_attr *attr = &evsel->attr;

	if (scale)
		attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED |
				    PERF_FORMAT_TOTAL_TIME_RUNNING;

	attr->inherit = !no_inherit;

	if (system_wide)
		return perf_evsel__open_per_cpu(evsel, evsel_list->cpus, false);

	if (target_pid == -1 && target_tid == -1) {
		attr->disabled = 1;
		attr->enable_on_exec = 1;
	}

	return perf_evsel__open_per_thread(evsel, evsel_list->threads, false);
}

/*
 * Does the counter have nsecs as a unit?
 */
static inline int nsec_counter(struct perf_evsel *evsel)
{
	if (perf_evsel__match(evsel, SOFTWARE, SW_CPU_CLOCK) ||
	    perf_evsel__match(evsel, SOFTWARE, SW_TASK_CLOCK))
		return 1;

	return 0;
}

/*
 * Update various tracking values we maintain to print
 * more semantic information such as miss/hit ratios,
 * instruction rates, etc:
 */
static void update_shadow_stats(struct perf_evsel *counter, u64 *count)
{
	if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
		update_stats(&runtime_nsecs_stats[0], count[0]);
	else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
		update_stats(&runtime_cycles_stats[0], count[0]);
	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_FRONTEND))
		update_stats(&runtime_stalled_cycles_front_stats[0], count[0]);
	else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND))
		update_stats(&runtime_stalled_cycles_back_stats[0], count[0]);
	else if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
		update_stats(&runtime_branches_stats[0], count[0]);
	else if (perf_evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
		update_stats(&runtime_cacherefs_stats[0], count[0]);
	else if (perf_evsel__match(counter, HW_CACHE, HW_CACHE_L1D))
		update_stats(&runtime_l1_dcache_stats[0], count[0]);
}

/*
 * Read out the results of a single counter:
 * aggregate counts across CPUs in system-wide mode
 */
static int read_counter_aggr(struct perf_evsel *counter)
{
	struct perf_stat *ps = counter->priv;
	u64 *count = counter->counts->aggr.values;
	int i;

	if (__perf_evsel__read(counter, evsel_list->cpus->nr,
			       evsel_list->threads->nr, scale) < 0)
		return -1;

	for (i = 0; i < 3; i++)
		update_stats(&ps->res_stats[i], count[i]);

	if (verbose) {
		fprintf(stderr, "%s: %" PRIu64 " %" PRIu64 " %" PRIu64 "\n",
			event_name(counter), count[0], count[1], count[2]);
	}

	/*
	 * Save the full runtime - to allow normalization during printout:
	 */
	update_shadow_stats(counter, count);

	return 0;
}

/*
 * Read out the results of a single counter:
 * do not aggregate counts across CPUs in system-wide mode
 */
static int read_counter(struct perf_evsel *counter)
{
	u64 *count;
	int cpu;

	for (cpu = 0; cpu < evsel_list->cpus->nr; cpu++) {
		if (__perf_evsel__read_on_cpu(counter, cpu, 0, scale) < 0)
			return -1;

		count = counter->counts->cpu[cpu].values;

		update_shadow_stats(counter, count);
	}

	return 0;
}

static int run_perf_stat(int argc __used, const char **argv)
{
	unsigned long long t0, t1;
	struct perf_evsel *counter;
	int status = 0;
	int child_ready_pipe[2], go_pipe[2];
	const bool forks = (argc > 0);
	char buf;

	if (forks && (pipe(child_ready_pipe) < 0 || pipe(go_pipe) < 0)) {
		perror("failed to create pipes");
		exit(1);
	}

	if (forks) {
		if ((child_pid = fork()) < 0)
			perror("failed to fork");

		if (!child_pid) {
			close(child_ready_pipe[0]);
			close(go_pipe[1]);
			fcntl(go_pipe[0], F_SETFD, FD_CLOEXEC);

			/*
			 * Do a dummy execvp to get the PLT entry resolved,
			 * so we avoid the resolver overhead on the real
			 * execvp call.
			 */
			execvp("", (char **)argv);

			/*
			 * Tell the parent we're ready to go
			 */
			close(child_ready_pipe[1]);

			/*
			 * Wait until the parent tells us to go.
			 */
			if (read(go_pipe[0], &buf, 1) == -1)
				perror("unable to read pipe");

			execvp(argv[0], (char **)argv);

			perror(argv[0]);
			exit(-1);
		}

		if (target_tid == -1 && target_pid == -1 && !system_wide)
			evsel_list->threads->map[0] = child_pid;

		/*
		 * Wait for the child to be ready to exec.
		 */
		close(child_ready_pipe[1]);
		close(go_pipe[0]);
		if (read(child_ready_pipe[0], &buf, 1) == -1)
			perror("unable to read pipe");
		close(child_ready_pipe[0]);
	}

	list_for_each_entry(counter, &evsel_list->entries, node) {
		if (create_perf_stat_counter(counter) < 0) {
			if (errno == EINVAL || errno == ENOSYS)
				continue;

			if (errno == EPERM || errno == EACCES) {
				error("You may not have permission to collect %sstats.\n"
				      "\t Consider tweaking"
				      " /proc/sys/kernel/perf_event_paranoid or running as root.",
				      system_wide ? "system-wide " : "");
			} else if (errno == ENOENT) {
				error("%s event is not supported. ", event_name(counter));
			} else {
				error("open_counter returned with %d (%s). "
				      "/bin/dmesg may provide additional information.\n",
				       errno, strerror(errno));
			}
			if (child_pid != -1)
				kill(child_pid, SIGTERM);
			die("Not all events could be opened.\n");
			return -1;
		}
	}

	if (perf_evlist__set_filters(evsel_list)) {
		error("failed to set filter with %d (%s)\n", errno,
			strerror(errno));
		return -1;
	}

	/*
	 * Enable counters and exec the command:
	 */
	t0 = rdclock();

	if (forks) {
		close(go_pipe[1]);
		wait(&status);
	} else {
		while(!done) sleep(1);
	}

	t1 = rdclock();

	update_stats(&walltime_nsecs_stats, t1 - t0);

	if (no_aggr) {
		list_for_each_entry(counter, &evsel_list->entries, node) {
			read_counter(counter);
			perf_evsel__close_fd(counter, evsel_list->cpus->nr, 1);
		}
	} else {
		list_for_each_entry(counter, &evsel_list->entries, node) {
			read_counter_aggr(counter);
			perf_evsel__close_fd(counter, evsel_list->cpus->nr,
					     evsel_list->threads->nr);
		}
	}

	return WEXITSTATUS(status);
}

static void print_noise_pct(double total, double avg)
{
	double pct = 0.0;

	if (avg)
		pct = 100.0*total/avg;

	fprintf(stderr, "  ( +-%6.2f%% )", pct);
}

static void print_noise(struct perf_evsel *evsel, double avg)
{
	struct perf_stat *ps;

	if (run_count == 1)
		return;

	ps = evsel->priv;
	print_noise_pct(stddev_stats(&ps->res_stats[0]), avg);
}

static void nsec_printout(int cpu, struct perf_evsel *evsel, double avg)
{
	double msecs = avg / 1e6;
	char cpustr[16] = { '\0', };
	const char *fmt = csv_output ? "%s%.6f%s%s" : "%s%18.6f%s%-24s";

	if (no_aggr)
		sprintf(cpustr, "CPU%*d%s",
			csv_output ? 0 : -4,
			evsel_list->cpus->map[cpu], csv_sep);

	fprintf(stderr, fmt, cpustr, msecs, csv_sep, event_name(evsel));

	if (evsel->cgrp)
		fprintf(stderr, "%s%s", csv_sep, evsel->cgrp->name);

	if (csv_output)
		return;

	if (perf_evsel__match(evsel, SOFTWARE, SW_TASK_CLOCK))
		fprintf(stderr, " # %8.3f CPUs utilized          ", avg / avg_stats(&walltime_nsecs_stats));
}

static void print_stalled_cycles_frontend(int cpu, struct perf_evsel *evsel __used, double avg)
{
	double total, ratio = 0.0;
	const char *color;

	total = avg_stats(&runtime_cycles_stats[cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = PERF_COLOR_NORMAL;
	if (ratio > 75.0)
		color = PERF_COLOR_RED;
	else if (ratio > 50.0)
		color = PERF_COLOR_MAGENTA;
	else if (ratio > 20.0)
		color = PERF_COLOR_YELLOW;

	fprintf(stderr, " #   ");
	color_fprintf(stderr, color, "%5.2f%%", ratio);
	fprintf(stderr, " frontend cycles idle   ");
}

static void print_stalled_cycles_backend(int cpu, struct perf_evsel *evsel __used, double avg)
{
	double total, ratio = 0.0;
	const char *color;

	total = avg_stats(&runtime_cycles_stats[cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = PERF_COLOR_NORMAL;
	if (ratio > 75.0)
		color = PERF_COLOR_RED;
	else if (ratio > 50.0)
		color = PERF_COLOR_MAGENTA;
	else if (ratio > 25.0)
		color = PERF_COLOR_YELLOW;

	fprintf(stderr, " #   ");
	color_fprintf(stderr, color, "%5.2f%%", ratio);
	fprintf(stderr, "  backend cycles idle   ");
}

static void print_branch_misses(int cpu, struct perf_evsel *evsel __used, double avg)
{
	double total, ratio = 0.0;
	const char *color;

	total = avg_stats(&runtime_branches_stats[cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = PERF_COLOR_NORMAL;
	if (ratio > 20.0)
		color = PERF_COLOR_RED;
	else if (ratio > 10.0)
		color = PERF_COLOR_MAGENTA;
	else if (ratio > 5.0)
		color = PERF_COLOR_YELLOW;

	fprintf(stderr, " #   ");
	color_fprintf(stderr, color, "%5.2f%%", ratio);
	fprintf(stderr, " of all branches        ");
}

static void print_l1_dcache_misses(int cpu, struct perf_evsel *evsel __used, double avg)
{
	double total, ratio = 0.0;
	const char *color;

	total = avg_stats(&runtime_l1_dcache_stats[cpu]);

	if (total)
		ratio = avg / total * 100.0;

	color = PERF_COLOR_NORMAL;
	if (ratio > 20.0)
		color = PERF_COLOR_RED;
	else if (ratio > 10.0)
		color = PERF_COLOR_MAGENTA;
	else if (ratio > 5.0)
		color = PERF_COLOR_YELLOW;

	fprintf(stderr, " #   ");
	color_fprintf(stderr, color, "%5.2f%%", ratio);
	fprintf(stderr, " of all L1-dcache hits  ");
}

static void abs_printout(int cpu, struct perf_evsel *evsel, double avg)
{
	double total, ratio = 0.0;
	char cpustr[16] = { '\0', };
	const char *fmt;

	if (csv_output)
		fmt = "%s%.0f%s%s";
	else if (big_num)
		fmt = "%s%'18.0f%s%-24s";
	else
		fmt = "%s%18.0f%s%-24s";

	if (no_aggr)
		sprintf(cpustr, "CPU%*d%s",
			csv_output ? 0 : -4,
			evsel_list->cpus->map[cpu], csv_sep);
	else
		cpu = 0;

	fprintf(stderr, fmt, cpustr, avg, csv_sep, event_name(evsel));

	if (evsel->cgrp)
		fprintf(stderr, "%s%s", csv_sep, evsel->cgrp->name);

	if (csv_output)
		return;

	if (perf_evsel__match(evsel, HARDWARE, HW_INSTRUCTIONS)) {
		total = avg_stats(&runtime_cycles_stats[cpu]);

		if (total)
			ratio = avg / total;

		fprintf(stderr, " #    %4.2f  insns per cycle        ", ratio);

		total = avg_stats(&runtime_stalled_cycles_front_stats[cpu]);
		total = max(total, avg_stats(&runtime_stalled_cycles_back_stats[cpu]));

		if (total && avg) {
			ratio = total / avg;
			fprintf(stderr, "\n                                            #    %4.2f  stalled cycles per insn", ratio);
		}

	} else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES) &&
			runtime_branches_stats[cpu].n != 0) {
		print_branch_misses(cpu, evsel, avg);
	} else if (
		evsel->attr.type == PERF_TYPE_HW_CACHE &&
		evsel->attr.config ==  ( PERF_COUNT_HW_CACHE_L1D |
					((PERF_COUNT_HW_CACHE_OP_READ) << 8) |
					((PERF_COUNT_HW_CACHE_RESULT_MISS) << 16)) &&
			runtime_l1_dcache_stats[cpu].n != 0) {
		print_l1_dcache_misses(cpu, evsel, avg);
	} else if (perf_evsel__match(evsel, HARDWARE, HW_CACHE_MISSES) &&
			runtime_cacherefs_stats[cpu].n != 0) {
		total = avg_stats(&runtime_cacherefs_stats[cpu]);

		if (total)
			ratio = avg * 100 / total;

		fprintf(stderr, " # %8.3f %% of all cache refs    ", ratio);

	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_FRONTEND)) {
		print_stalled_cycles_frontend(cpu, evsel, avg);
	} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
		print_stalled_cycles_backend(cpu, evsel, avg);
	} else if (perf_evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
		total = avg_stats(&runtime_nsecs_stats[cpu]);

		if (total)
			ratio = 1.0 * avg / total;

		fprintf(stderr, " # %8.3f GHz                    ", ratio);
	} else if (runtime_nsecs_stats[cpu].n != 0) {
		total = avg_stats(&runtime_nsecs_stats[cpu]);

		if (total)
			ratio = 1000.0 * avg / total;

		fprintf(stderr, " # %8.3f M/sec                  ", ratio);
	} else {
		fprintf(stderr, "                                   ");
	}
}

/*
 * Print out the results of a single counter:
 * aggregated counts in system-wide mode
 */
static void print_counter_aggr(struct perf_evsel *counter)
{
	struct perf_stat *ps = counter->priv;
	double avg = avg_stats(&ps->res_stats[0]);
	int scaled = counter->counts->scaled;

	if (scaled == -1) {
		fprintf(stderr, "%*s%s%*s",
			csv_output ? 0 : 18,
			"<not counted>",
			csv_sep,
			csv_output ? 0 : -24,
			event_name(counter));

		if (counter->cgrp)
			fprintf(stderr, "%s%s", csv_sep, counter->cgrp->name);

		fputc('\n', stderr);
		return;
	}

	if (nsec_counter(counter))
		nsec_printout(-1, counter, avg);
	else
		abs_printout(-1, counter, avg);

	if (csv_output) {
		fputc('\n', stderr);
		return;
	}

	print_noise(counter, avg);

	if (scaled) {
		double avg_enabled, avg_running;

		avg_enabled = avg_stats(&ps->res_stats[1]);
		avg_running = avg_stats(&ps->res_stats[2]);

		fprintf(stderr, "  (%.2f%%)", 100 * avg_running / avg_enabled);
	}
	fprintf(stderr, "\n");
}

/*
 * Print out the results of a single counter:
 * does not use aggregated count in system-wide
 */
static void print_counter(struct perf_evsel *counter)
{
	u64 ena, run, val;
	int cpu;

	for (cpu = 0; cpu < evsel_list->cpus->nr; cpu++) {
		val = counter->counts->cpu[cpu].val;
		ena = counter->counts->cpu[cpu].ena;
		run = counter->counts->cpu[cpu].run;
		if (run == 0 || ena == 0) {
			fprintf(stderr, "CPU%*d%s%*s%s%*s",
				csv_output ? 0 : -4,
				evsel_list->cpus->map[cpu], csv_sep,
				csv_output ? 0 : 18,
				"<not counted>", csv_sep,
				csv_output ? 0 : -24,
				event_name(counter));

			if (counter->cgrp)
				fprintf(stderr, "%s%s", csv_sep, counter->cgrp->name);

			fputc('\n', stderr);
			continue;
		}

		if (nsec_counter(counter))
			nsec_printout(cpu, counter, val);
		else
			abs_printout(cpu, counter, val);

		if (!csv_output) {
			print_noise(counter, 1.0);

			if (run != ena)
				fprintf(stderr, "  (%.2f%%)", 100.0 * run / ena);
		}
		fputc('\n', stderr);
	}
}

static void print_stat(int argc, const char **argv)
{
	struct perf_evsel *counter;
	int i;

	fflush(stdout);

	if (!csv_output) {
		fprintf(stderr, "\n");
		fprintf(stderr, " Performance counter stats for ");
		if(target_pid == -1 && target_tid == -1) {
			fprintf(stderr, "\'%s", argv[0]);
			for (i = 1; i < argc; i++)
				fprintf(stderr, " %s", argv[i]);
		} else if (target_pid != -1)
			fprintf(stderr, "process id \'%d", target_pid);
		else
			fprintf(stderr, "thread id \'%d", target_tid);

		fprintf(stderr, "\'");
		if (run_count > 1)
			fprintf(stderr, " (%d runs)", run_count);
		fprintf(stderr, ":\n\n");
	}

	if (no_aggr) {
		list_for_each_entry(counter, &evsel_list->entries, node)
			print_counter(counter);
	} else {
		list_for_each_entry(counter, &evsel_list->entries, node)
			print_counter_aggr(counter);
	}

	if (!csv_output) {
		fprintf(stderr, "\n");
		fprintf(stderr, " %18.9f  seconds time elapsed",
				avg_stats(&walltime_nsecs_stats)/1e9);
		if (run_count > 1) {
			print_noise_pct(stddev_stats(&walltime_nsecs_stats),
					avg_stats(&walltime_nsecs_stats));
		}
		fprintf(stderr, "\n\n");
	}
}

static volatile int signr = -1;

static void skip_signal(int signo)
{
	if(child_pid == -1)
		done = 1;

	signr = signo;
}

static void sig_atexit(void)
{
	if (child_pid != -1)
		kill(child_pid, SIGTERM);

	if (signr == -1)
		return;

	signal(signr, SIG_DFL);
	kill(getpid(), signr);
}

static const char * const stat_usage[] = {
	"perf stat [<options>] [<command>]",
	NULL
};

static int stat__set_big_num(const struct option *opt __used,
			     const char *s __used, int unset)
{
	big_num_opt = unset ? 0 : 1;
	return 0;
}

static const struct option options[] = {
	OPT_CALLBACK('e', "event", &evsel_list, "event",
		     "event selector. use 'perf list' to list available events",
		     parse_events),
	OPT_CALLBACK(0, "filter", &evsel_list, "filter",
		     "event filter", parse_filter),
	OPT_BOOLEAN('i', "no-inherit", &no_inherit,
		    "child tasks do not inherit counters"),
	OPT_INTEGER('p', "pid", &target_pid,
		    "stat events on existing process id"),
	OPT_INTEGER('t', "tid", &target_tid,
		    "stat events on existing thread id"),
	OPT_BOOLEAN('a', "all-cpus", &system_wide,
		    "system-wide collection from all CPUs"),
	OPT_BOOLEAN('c', "scale", &scale,
		    "scale/normalize counters"),
	OPT_INCR('v', "verbose", &verbose,
		    "be more verbose (show counter open errors, etc)"),
	OPT_INTEGER('r', "repeat", &run_count,
		    "repeat command and print average + stddev (max: 100)"),
	OPT_BOOLEAN('n', "null", &null_run,
		    "null run - dont start any counters"),
	OPT_BOOLEAN('d', "detailed", &detailed_run,
		    "detailed run - start a lot of events"),
	OPT_BOOLEAN('S', "sync", &sync_run,
		    "call sync() before starting a run"),
	OPT_CALLBACK_NOOPT('B', "big-num", NULL, NULL, 
			   "print large numbers with thousands\' separators",
			   stat__set_big_num),
	OPT_STRING('C', "cpu", &cpu_list, "cpu",
		    "list of cpus to monitor in system-wide"),
	OPT_BOOLEAN('A', "no-aggr", &no_aggr,
		    "disable CPU count aggregation"),
	OPT_STRING('x', "field-separator", &csv_sep, "separator",
		   "print counts with custom separator"),
	OPT_CALLBACK('G', "cgroup", &evsel_list, "name",
		     "monitor event in cgroup name only",
		     parse_cgroups),
	OPT_END()
};

int cmd_stat(int argc, const char **argv, const char *prefix __used)
{
	struct perf_evsel *pos;
	int status = -ENOMEM;

	setlocale(LC_ALL, "");

	evsel_list = perf_evlist__new(NULL, NULL);
	if (evsel_list == NULL)
		return -ENOMEM;

	argc = parse_options(argc, argv, options, stat_usage,
		PARSE_OPT_STOP_AT_NON_OPTION);

	if (csv_sep)
		csv_output = true;
	else
		csv_sep = DEFAULT_SEPARATOR;

	/*
	 * let the spreadsheet do the pretty-printing
	 */
	if (csv_output) {
		/* User explicitely passed -B? */
		if (big_num_opt == 1) {
			fprintf(stderr, "-B option not supported with -x\n");
			usage_with_options(stat_usage, options);
		} else /* Nope, so disable big number formatting */
			big_num = false;
	} else if (big_num_opt == 0) /* User passed --no-big-num */
		big_num = false;

	if (!argc && target_pid == -1 && target_tid == -1)
		usage_with_options(stat_usage, options);
	if (run_count <= 0)
		usage_with_options(stat_usage, options);

	/* no_aggr, cgroup are for system-wide only */
	if ((no_aggr || nr_cgroups) && !system_wide) {
		fprintf(stderr, "both cgroup and no-aggregation "
			"modes only available in system-wide mode\n");

		usage_with_options(stat_usage, options);
	}

	/* Set attrs and nr_counters if no event is selected and !null_run */
	if (detailed_run) {
		size_t c;

		for (c = 0; c < ARRAY_SIZE(detailed_attrs); ++c) {
			pos = perf_evsel__new(&detailed_attrs[c], c);
			if (pos == NULL)
				goto out;
			perf_evlist__add(evsel_list, pos);
		}
	}
	/* Set attrs and nr_counters if no event is selected and !null_run */
	if (!detailed_run && !null_run && !evsel_list->nr_entries) {
		size_t c;

		for (c = 0; c < ARRAY_SIZE(default_attrs); ++c) {
			pos = perf_evsel__new(&default_attrs[c], c);
			if (pos == NULL)
				goto out;
			perf_evlist__add(evsel_list, pos);
		}
	}

	if (target_pid != -1)
		target_tid = target_pid;

	evsel_list->threads = thread_map__new(target_pid, target_tid);
	if (evsel_list->threads == NULL) {
		pr_err("Problems finding threads of monitor\n");
		usage_with_options(stat_usage, options);
	}

	if (system_wide)
		evsel_list->cpus = cpu_map__new(cpu_list);
	else
		evsel_list->cpus = cpu_map__dummy_new();

	if (evsel_list->cpus == NULL) {
		perror("failed to parse CPUs map");
		usage_with_options(stat_usage, options);
		return -1;
	}

	list_for_each_entry(pos, &evsel_list->entries, node) {
		if (perf_evsel__alloc_stat_priv(pos) < 0 ||
		    perf_evsel__alloc_counts(pos, evsel_list->cpus->nr) < 0 ||
		    perf_evsel__alloc_fd(pos, evsel_list->cpus->nr, evsel_list->threads->nr) < 0)
			goto out_free_fd;
	}

	/*
	 * We dont want to block the signals - that would cause
	 * child tasks to inherit that and Ctrl-C would not work.
	 * What we want is for Ctrl-C to work in the exec()-ed
	 * task, but being ignored by perf stat itself:
	 */
	atexit(sig_atexit);
	signal(SIGINT,  skip_signal);
	signal(SIGALRM, skip_signal);
	signal(SIGABRT, skip_signal);

	status = 0;
	for (run_idx = 0; run_idx < run_count; run_idx++) {
		if (run_count != 1 && verbose)
			fprintf(stderr, "[ perf stat: executing run #%d ... ]\n", run_idx + 1);

		if (sync_run)
			sync();

		status = run_perf_stat(argc, argv);
	}

	if (status != -1)
		print_stat(argc, argv);
out_free_fd:
	list_for_each_entry(pos, &evsel_list->entries, node)
		perf_evsel__free_stat_priv(pos);
	perf_evlist__delete_maps(evsel_list);
out:
	perf_evlist__delete(evsel_list);
	return status;
}