summaryrefslogtreecommitdiffstats
path: root/src/testing/benchmark.go
blob: a8f75e9712370adb7f87b6825c74af7d9008a118 (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
// Copyright 2009 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

package testing

import (
	"flag"
	"fmt"
	"internal/race"
	"internal/sysinfo"
	"io"
	"math"
	"os"
	"runtime"
	"sort"
	"strconv"
	"strings"
	"sync"
	"sync/atomic"
	"time"
	"unicode"
)

func initBenchmarkFlags() {
	matchBenchmarks = flag.String("test.bench", "", "run only benchmarks matching `regexp`")
	benchmarkMemory = flag.Bool("test.benchmem", false, "print memory allocations for benchmarks")
	flag.Var(&benchTime, "test.benchtime", "run each benchmark for duration `d`")
}

var (
	matchBenchmarks *string
	benchmarkMemory *bool

	benchTime = benchTimeFlag{d: 1 * time.Second} // changed during test of testing package
)

type benchTimeFlag struct {
	d time.Duration
	n int
}

func (f *benchTimeFlag) String() string {
	if f.n > 0 {
		return fmt.Sprintf("%dx", f.n)
	}
	return time.Duration(f.d).String()
}

func (f *benchTimeFlag) Set(s string) error {
	if strings.HasSuffix(s, "x") {
		n, err := strconv.ParseInt(s[:len(s)-1], 10, 0)
		if err != nil || n <= 0 {
			return fmt.Errorf("invalid count")
		}
		*f = benchTimeFlag{n: int(n)}
		return nil
	}
	d, err := time.ParseDuration(s)
	if err != nil || d <= 0 {
		return fmt.Errorf("invalid duration")
	}
	*f = benchTimeFlag{d: d}
	return nil
}

// Global lock to ensure only one benchmark runs at a time.
var benchmarkLock sync.Mutex

// Used for every benchmark for measuring memory.
var memStats runtime.MemStats

// InternalBenchmark is an internal type but exported because it is cross-package;
// it is part of the implementation of the "go test" command.
type InternalBenchmark struct {
	Name string
	F    func(b *B)
}

// B is a type passed to Benchmark functions to manage benchmark
// timing and to specify the number of iterations to run.
//
// A benchmark ends when its Benchmark function returns or calls any of the methods
// FailNow, Fatal, Fatalf, SkipNow, Skip, or Skipf. Those methods must be called
// only from the goroutine running the Benchmark function.
// The other reporting methods, such as the variations of Log and Error,
// may be called simultaneously from multiple goroutines.
//
// Like in tests, benchmark logs are accumulated during execution
// and dumped to standard output when done. Unlike in tests, benchmark logs
// are always printed, so as not to hide output whose existence may be
// affecting benchmark results.
type B struct {
	common
	importPath       string // import path of the package containing the benchmark
	context          *benchContext
	N                int
	previousN        int           // number of iterations in the previous run
	previousDuration time.Duration // total duration of the previous run
	benchFunc        func(b *B)
	benchTime        benchTimeFlag
	bytes            int64
	missingBytes     bool // one of the subbenchmarks does not have bytes set.
	timerOn          bool
	showAllocResult  bool
	result           BenchmarkResult
	parallelism      int // RunParallel creates parallelism*GOMAXPROCS goroutines
	// The initial states of memStats.Mallocs and memStats.TotalAlloc.
	startAllocs uint64
	startBytes  uint64
	// The net total of this test after being run.
	netAllocs uint64
	netBytes  uint64
	// Extra metrics collected by ReportMetric.
	extra map[string]float64
}

// StartTimer starts timing a test. This function is called automatically
// before a benchmark starts, but it can also be used to resume timing after
// a call to StopTimer.
func (b *B) StartTimer() {
	if !b.timerOn {
		runtime.ReadMemStats(&memStats)
		b.startAllocs = memStats.Mallocs
		b.startBytes = memStats.TotalAlloc
		b.start = time.Now()
		b.timerOn = true
	}
}

// StopTimer stops timing a test. This can be used to pause the timer
// while performing complex initialization that you don't
// want to measure.
func (b *B) StopTimer() {
	if b.timerOn {
		b.duration += time.Since(b.start)
		runtime.ReadMemStats(&memStats)
		b.netAllocs += memStats.Mallocs - b.startAllocs
		b.netBytes += memStats.TotalAlloc - b.startBytes
		b.timerOn = false
	}
}

// ResetTimer zeroes the elapsed benchmark time and memory allocation counters
// and deletes user-reported metrics.
// It does not affect whether the timer is running.
func (b *B) ResetTimer() {
	if b.extra == nil {
		// Allocate the extra map before reading memory stats.
		// Pre-size it to make more allocation unlikely.
		b.extra = make(map[string]float64, 16)
	} else {
		for k := range b.extra {
			delete(b.extra, k)
		}
	}
	if b.timerOn {
		runtime.ReadMemStats(&memStats)
		b.startAllocs = memStats.Mallocs
		b.startBytes = memStats.TotalAlloc
		b.start = time.Now()
	}
	b.duration = 0
	b.netAllocs = 0
	b.netBytes = 0
}

// SetBytes records the number of bytes processed in a single operation.
// If this is called, the benchmark will report ns/op and MB/s.
func (b *B) SetBytes(n int64) { b.bytes = n }

// ReportAllocs enables malloc statistics for this benchmark.
// It is equivalent to setting -test.benchmem, but it only affects the
// benchmark function that calls ReportAllocs.
func (b *B) ReportAllocs() {
	b.showAllocResult = true
}

// runN runs a single benchmark for the specified number of iterations.
func (b *B) runN(n int) {
	benchmarkLock.Lock()
	defer benchmarkLock.Unlock()
	defer b.runCleanup(normalPanic)
	// Try to get a comparable environment for each run
	// by clearing garbage from previous runs.
	runtime.GC()
	b.raceErrors = -race.Errors()
	b.N = n
	b.parallelism = 1
	b.ResetTimer()
	b.StartTimer()
	b.benchFunc(b)
	b.StopTimer()
	b.previousN = n
	b.previousDuration = b.duration
	b.raceErrors += race.Errors()
	if b.raceErrors > 0 {
		b.Errorf("race detected during execution of benchmark")
	}
}

func min(x, y int64) int64 {
	if x > y {
		return y
	}
	return x
}

func max(x, y int64) int64 {
	if x < y {
		return y
	}
	return x
}

// run1 runs the first iteration of benchFunc. It reports whether more
// iterations of this benchmarks should be run.
func (b *B) run1() bool {
	if ctx := b.context; ctx != nil {
		// Extend maxLen, if needed.
		if n := len(b.name) + ctx.extLen + 1; n > ctx.maxLen {
			ctx.maxLen = n + 8 // Add additional slack to avoid too many jumps in size.
		}
	}
	go func() {
		// Signal that we're done whether we return normally
		// or by FailNow's runtime.Goexit.
		defer func() {
			b.signal <- true
		}()

		b.runN(1)
	}()
	<-b.signal
	if b.failed {
		fmt.Fprintf(b.w, "--- FAIL: %s\n%s", b.name, b.output)
		return false
	}
	// Only print the output if we know we are not going to proceed.
	// Otherwise it is printed in processBench.
	if atomic.LoadInt32(&b.hasSub) != 0 || b.finished {
		tag := "BENCH"
		if b.skipped {
			tag = "SKIP"
		}
		if b.chatty != nil && (len(b.output) > 0 || b.finished) {
			b.trimOutput()
			fmt.Fprintf(b.w, "--- %s: %s\n%s", tag, b.name, b.output)
		}
		return false
	}
	return true
}

var labelsOnce sync.Once

// run executes the benchmark in a separate goroutine, including all of its
// subbenchmarks. b must not have subbenchmarks.
func (b *B) run() {
	labelsOnce.Do(func() {
		fmt.Fprintf(b.w, "goos: %s\n", runtime.GOOS)
		fmt.Fprintf(b.w, "goarch: %s\n", runtime.GOARCH)
		if b.importPath != "" {
			fmt.Fprintf(b.w, "pkg: %s\n", b.importPath)
		}
		if cpu := sysinfo.CPU.Name(); cpu != "" {
			fmt.Fprintf(b.w, "cpu: %s\n", cpu)
		}
	})
	if b.context != nil {
		// Running go test --test.bench
		b.context.processBench(b) // Must call doBench.
	} else {
		// Running func Benchmark.
		b.doBench()
	}
}

func (b *B) doBench() BenchmarkResult {
	go b.launch()
	<-b.signal
	return b.result
}

// launch launches the benchmark function. It gradually increases the number
// of benchmark iterations until the benchmark runs for the requested benchtime.
// launch is run by the doBench function as a separate goroutine.
// run1 must have been called on b.
func (b *B) launch() {
	// Signal that we're done whether we return normally
	// or by FailNow's runtime.Goexit.
	defer func() {
		b.signal <- true
	}()

	// Run the benchmark for at least the specified amount of time.
	if b.benchTime.n > 0 {
		b.runN(b.benchTime.n)
	} else {
		d := b.benchTime.d
		for n := int64(1); !b.failed && b.duration < d && n < 1e9; {
			last := n
			// Predict required iterations.
			goalns := d.Nanoseconds()
			prevIters := int64(b.N)
			prevns := b.duration.Nanoseconds()
			if prevns <= 0 {
				// Round up, to avoid div by zero.
				prevns = 1
			}
			// Order of operations matters.
			// For very fast benchmarks, prevIters ~= prevns.
			// If you divide first, you get 0 or 1,
			// which can hide an order of magnitude in execution time.
			// So multiply first, then divide.
			n = goalns * prevIters / prevns
			// Run more iterations than we think we'll need (1.2x).
			n += n / 5
			// Don't grow too fast in case we had timing errors previously.
			n = min(n, 100*last)
			// Be sure to run at least one more than last time.
			n = max(n, last+1)
			// Don't run more than 1e9 times. (This also keeps n in int range on 32 bit platforms.)
			n = min(n, 1e9)
			b.runN(int(n))
		}
	}
	b.result = BenchmarkResult{b.N, b.duration, b.bytes, b.netAllocs, b.netBytes, b.extra}
}

// ReportMetric adds "n unit" to the reported benchmark results.
// If the metric is per-iteration, the caller should divide by b.N,
// and by convention units should end in "/op".
// ReportMetric overrides any previously reported value for the same unit.
// ReportMetric panics if unit is the empty string or if unit contains
// any whitespace.
// If unit is a unit normally reported by the benchmark framework itself
// (such as "allocs/op"), ReportMetric will override that metric.
// Setting "ns/op" to 0 will suppress that built-in metric.
func (b *B) ReportMetric(n float64, unit string) {
	if unit == "" {
		panic("metric unit must not be empty")
	}
	if strings.IndexFunc(unit, unicode.IsSpace) >= 0 {
		panic("metric unit must not contain whitespace")
	}
	b.extra[unit] = n
}

// BenchmarkResult contains the results of a benchmark run.
type BenchmarkResult struct {
	N         int           // The number of iterations.
	T         time.Duration // The total time taken.
	Bytes     int64         // Bytes processed in one iteration.
	MemAllocs uint64        // The total number of memory allocations.
	MemBytes  uint64        // The total number of bytes allocated.

	// Extra records additional metrics reported by ReportMetric.
	Extra map[string]float64
}

// NsPerOp returns the "ns/op" metric.
func (r BenchmarkResult) NsPerOp() int64 {
	if v, ok := r.Extra["ns/op"]; ok {
		return int64(v)
	}
	if r.N <= 0 {
		return 0
	}
	return r.T.Nanoseconds() / int64(r.N)
}

// mbPerSec returns the "MB/s" metric.
func (r BenchmarkResult) mbPerSec() float64 {
	if v, ok := r.Extra["MB/s"]; ok {
		return v
	}
	if r.Bytes <= 0 || r.T <= 0 || r.N <= 0 {
		return 0
	}
	return (float64(r.Bytes) * float64(r.N) / 1e6) / r.T.Seconds()
}

// AllocsPerOp returns the "allocs/op" metric,
// which is calculated as r.MemAllocs / r.N.
func (r BenchmarkResult) AllocsPerOp() int64 {
	if v, ok := r.Extra["allocs/op"]; ok {
		return int64(v)
	}
	if r.N <= 0 {
		return 0
	}
	return int64(r.MemAllocs) / int64(r.N)
}

// AllocedBytesPerOp returns the "B/op" metric,
// which is calculated as r.MemBytes / r.N.
func (r BenchmarkResult) AllocedBytesPerOp() int64 {
	if v, ok := r.Extra["B/op"]; ok {
		return int64(v)
	}
	if r.N <= 0 {
		return 0
	}
	return int64(r.MemBytes) / int64(r.N)
}

// String returns a summary of the benchmark results.
// It follows the benchmark result line format from
// https://golang.org/design/14313-benchmark-format, not including the
// benchmark name.
// Extra metrics override built-in metrics of the same name.
// String does not include allocs/op or B/op, since those are reported
// by MemString.
func (r BenchmarkResult) String() string {
	buf := new(strings.Builder)
	fmt.Fprintf(buf, "%8d", r.N)

	// Get ns/op as a float.
	ns, ok := r.Extra["ns/op"]
	if !ok {
		ns = float64(r.T.Nanoseconds()) / float64(r.N)
	}
	if ns != 0 {
		buf.WriteByte('\t')
		prettyPrint(buf, ns, "ns/op")
	}

	if mbs := r.mbPerSec(); mbs != 0 {
		fmt.Fprintf(buf, "\t%7.2f MB/s", mbs)
	}

	// Print extra metrics that aren't represented in the standard
	// metrics.
	var extraKeys []string
	for k := range r.Extra {
		switch k {
		case "ns/op", "MB/s", "B/op", "allocs/op":
			// Built-in metrics reported elsewhere.
			continue
		}
		extraKeys = append(extraKeys, k)
	}
	sort.Strings(extraKeys)
	for _, k := range extraKeys {
		buf.WriteByte('\t')
		prettyPrint(buf, r.Extra[k], k)
	}
	return buf.String()
}

func prettyPrint(w io.Writer, x float64, unit string) {
	// Print all numbers with 10 places before the decimal point
	// and small numbers with four sig figs. Field widths are
	// chosen to fit the whole part in 10 places while aligning
	// the decimal point of all fractional formats.
	var format string
	switch y := math.Abs(x); {
	case y == 0 || y >= 999.95:
		format = "%10.0f %s"
	case y >= 99.995:
		format = "%12.1f %s"
	case y >= 9.9995:
		format = "%13.2f %s"
	case y >= 0.99995:
		format = "%14.3f %s"
	case y >= 0.099995:
		format = "%15.4f %s"
	case y >= 0.0099995:
		format = "%16.5f %s"
	case y >= 0.00099995:
		format = "%17.6f %s"
	default:
		format = "%18.7f %s"
	}
	fmt.Fprintf(w, format, x, unit)
}

// MemString returns r.AllocedBytesPerOp and r.AllocsPerOp in the same format as 'go test'.
func (r BenchmarkResult) MemString() string {
	return fmt.Sprintf("%8d B/op\t%8d allocs/op",
		r.AllocedBytesPerOp(), r.AllocsPerOp())
}

// benchmarkName returns full name of benchmark including procs suffix.
func benchmarkName(name string, n int) string {
	if n != 1 {
		return fmt.Sprintf("%s-%d", name, n)
	}
	return name
}

type benchContext struct {
	match *matcher

	maxLen int // The largest recorded benchmark name.
	extLen int // Maximum extension length.
}

// RunBenchmarks is an internal function but exported because it is cross-package;
// it is part of the implementation of the "go test" command.
func RunBenchmarks(matchString func(pat, str string) (bool, error), benchmarks []InternalBenchmark) {
	runBenchmarks("", matchString, benchmarks)
}

func runBenchmarks(importPath string, matchString func(pat, str string) (bool, error), benchmarks []InternalBenchmark) bool {
	// If no flag was specified, don't run benchmarks.
	if len(*matchBenchmarks) == 0 {
		return true
	}
	// Collect matching benchmarks and determine longest name.
	maxprocs := 1
	for _, procs := range cpuList {
		if procs > maxprocs {
			maxprocs = procs
		}
	}
	ctx := &benchContext{
		match:  newMatcher(matchString, *matchBenchmarks, "-test.bench"),
		extLen: len(benchmarkName("", maxprocs)),
	}
	var bs []InternalBenchmark
	for _, Benchmark := range benchmarks {
		if _, matched, _ := ctx.match.fullName(nil, Benchmark.Name); matched {
			bs = append(bs, Benchmark)
			benchName := benchmarkName(Benchmark.Name, maxprocs)
			if l := len(benchName) + ctx.extLen + 1; l > ctx.maxLen {
				ctx.maxLen = l
			}
		}
	}
	main := &B{
		common: common{
			name:  "Main",
			w:     os.Stdout,
			bench: true,
		},
		importPath: importPath,
		benchFunc: func(b *B) {
			for _, Benchmark := range bs {
				b.Run(Benchmark.Name, Benchmark.F)
			}
		},
		benchTime: benchTime,
		context:   ctx,
	}
	if Verbose() {
		main.chatty = newChattyPrinter(main.w)
	}
	main.runN(1)
	return !main.failed
}

// processBench runs bench b for the configured CPU counts and prints the results.
func (ctx *benchContext) processBench(b *B) {
	for i, procs := range cpuList {
		for j := uint(0); j < *count; j++ {
			runtime.GOMAXPROCS(procs)
			benchName := benchmarkName(b.name, procs)

			// If it's chatty, we've already printed this information.
			if b.chatty == nil {
				fmt.Fprintf(b.w, "%-*s\t", ctx.maxLen, benchName)
			}
			// Recompute the running time for all but the first iteration.
			if i > 0 || j > 0 {
				b = &B{
					common: common{
						signal: make(chan bool),
						name:   b.name,
						w:      b.w,
						chatty: b.chatty,
						bench:  true,
					},
					benchFunc: b.benchFunc,
					benchTime: b.benchTime,
				}
				b.run1()
			}
			r := b.doBench()
			if b.failed {
				// The output could be very long here, but probably isn't.
				// We print it all, regardless, because we don't want to trim the reason
				// the benchmark failed.
				fmt.Fprintf(b.w, "--- FAIL: %s\n%s", benchName, b.output)
				continue
			}
			results := r.String()
			if b.chatty != nil {
				fmt.Fprintf(b.w, "%-*s\t", ctx.maxLen, benchName)
			}
			if *benchmarkMemory || b.showAllocResult {
				results += "\t" + r.MemString()
			}
			fmt.Fprintln(b.w, results)
			// Unlike with tests, we ignore the -chatty flag and always print output for
			// benchmarks since the output generation time will skew the results.
			if len(b.output) > 0 {
				b.trimOutput()
				fmt.Fprintf(b.w, "--- BENCH: %s\n%s", benchName, b.output)
			}
			if p := runtime.GOMAXPROCS(-1); p != procs {
				fmt.Fprintf(os.Stderr, "testing: %s left GOMAXPROCS set to %d\n", benchName, p)
			}
		}
	}
}

// Run benchmarks f as a subbenchmark with the given name. It reports
// whether there were any failures.
//
// A subbenchmark is like any other benchmark. A benchmark that calls Run at
// least once will not be measured itself and will be called once with N=1.
func (b *B) Run(name string, f func(b *B)) bool {
	// Since b has subbenchmarks, we will no longer run it as a benchmark itself.
	// Release the lock and acquire it on exit to ensure locks stay paired.
	atomic.StoreInt32(&b.hasSub, 1)
	benchmarkLock.Unlock()
	defer benchmarkLock.Lock()

	benchName, ok, partial := b.name, true, false
	if b.context != nil {
		benchName, ok, partial = b.context.match.fullName(&b.common, name)
	}
	if !ok {
		return true
	}
	var pc [maxStackLen]uintptr
	n := runtime.Callers(2, pc[:])
	sub := &B{
		common: common{
			signal:  make(chan bool),
			name:    benchName,
			parent:  &b.common,
			level:   b.level + 1,
			creator: pc[:n],
			w:       b.w,
			chatty:  b.chatty,
			bench:   true,
		},
		importPath: b.importPath,
		benchFunc:  f,
		benchTime:  b.benchTime,
		context:    b.context,
	}
	if partial {
		// Partial name match, like -bench=X/Y matching BenchmarkX.
		// Only process sub-benchmarks, if any.
		atomic.StoreInt32(&sub.hasSub, 1)
	}

	if b.chatty != nil {
		labelsOnce.Do(func() {
			fmt.Printf("goos: %s\n", runtime.GOOS)
			fmt.Printf("goarch: %s\n", runtime.GOARCH)
			if b.importPath != "" {
				fmt.Printf("pkg: %s\n", b.importPath)
			}
			if cpu := sysinfo.CPU.Name(); cpu != "" {
				fmt.Printf("cpu: %s\n", cpu)
			}
		})

		fmt.Println(benchName)
	}

	if sub.run1() {
		sub.run()
	}
	b.add(sub.result)
	return !sub.failed
}

// add simulates running benchmarks in sequence in a single iteration. It is
// used to give some meaningful results in case func Benchmark is used in
// combination with Run.
func (b *B) add(other BenchmarkResult) {
	r := &b.result
	// The aggregated BenchmarkResults resemble running all subbenchmarks as
	// in sequence in a single benchmark.
	r.N = 1
	r.T += time.Duration(other.NsPerOp())
	if other.Bytes == 0 {
		// Summing Bytes is meaningless in aggregate if not all subbenchmarks
		// set it.
		b.missingBytes = true
		r.Bytes = 0
	}
	if !b.missingBytes {
		r.Bytes += other.Bytes
	}
	r.MemAllocs += uint64(other.AllocsPerOp())
	r.MemBytes += uint64(other.AllocedBytesPerOp())
}

// trimOutput shortens the output from a benchmark, which can be very long.
func (b *B) trimOutput() {
	// The output is likely to appear multiple times because the benchmark
	// is run multiple times, but at least it will be seen. This is not a big deal
	// because benchmarks rarely print, but just in case, we trim it if it's too long.
	const maxNewlines = 10
	for nlCount, j := 0, 0; j < len(b.output); j++ {
		if b.output[j] == '\n' {
			nlCount++
			if nlCount >= maxNewlines {
				b.output = append(b.output[:j], "\n\t... [output truncated]\n"...)
				break
			}
		}
	}
}

// A PB is used by RunParallel for running parallel benchmarks.
type PB struct {
	globalN *uint64 // shared between all worker goroutines iteration counter
	grain   uint64  // acquire that many iterations from globalN at once
	cache   uint64  // local cache of acquired iterations
	bN      uint64  // total number of iterations to execute (b.N)
}

// Next reports whether there are more iterations to execute.
func (pb *PB) Next() bool {
	if pb.cache == 0 {
		n := atomic.AddUint64(pb.globalN, pb.grain)
		if n <= pb.bN {
			pb.cache = pb.grain
		} else if n < pb.bN+pb.grain {
			pb.cache = pb.bN + pb.grain - n
		} else {
			return false
		}
	}
	pb.cache--
	return true
}

// RunParallel runs a benchmark in parallel.
// It creates multiple goroutines and distributes b.N iterations among them.
// The number of goroutines defaults to GOMAXPROCS. To increase parallelism for
// non-CPU-bound benchmarks, call SetParallelism before RunParallel.
// RunParallel is usually used with the go test -cpu flag.
//
// The body function will be run in each goroutine. It should set up any
// goroutine-local state and then iterate until pb.Next returns false.
// It should not use the StartTimer, StopTimer, or ResetTimer functions,
// because they have global effect. It should also not call Run.
func (b *B) RunParallel(body func(*PB)) {
	if b.N == 0 {
		return // Nothing to do when probing.
	}
	// Calculate grain size as number of iterations that take ~100µs.
	// 100µs is enough to amortize the overhead and provide sufficient
	// dynamic load balancing.
	grain := uint64(0)
	if b.previousN > 0 && b.previousDuration > 0 {
		grain = 1e5 * uint64(b.previousN) / uint64(b.previousDuration)
	}
	if grain < 1 {
		grain = 1
	}
	// We expect the inner loop and function call to take at least 10ns,
	// so do not do more than 100µs/10ns=1e4 iterations.
	if grain > 1e4 {
		grain = 1e4
	}

	n := uint64(0)
	numProcs := b.parallelism * runtime.GOMAXPROCS(0)
	var wg sync.WaitGroup
	wg.Add(numProcs)
	for p := 0; p < numProcs; p++ {
		go func() {
			defer wg.Done()
			pb := &PB{
				globalN: &n,
				grain:   grain,
				bN:      uint64(b.N),
			}
			body(pb)
		}()
	}
	wg.Wait()
	if n <= uint64(b.N) && !b.Failed() {
		b.Fatal("RunParallel: body exited without pb.Next() == false")
	}
}

// SetParallelism sets the number of goroutines used by RunParallel to p*GOMAXPROCS.
// There is usually no need to call SetParallelism for CPU-bound benchmarks.
// If p is less than 1, this call will have no effect.
func (b *B) SetParallelism(p int) {
	if p >= 1 {
		b.parallelism = p
	}
}

// Benchmark benchmarks a single function. It is useful for creating
// custom benchmarks that do not use the "go test" command.
//
// If f depends on testing flags, then Init must be used to register
// those flags before calling Benchmark and before calling flag.Parse.
//
// If f calls Run, the result will be an estimate of running all its
// subbenchmarks that don't call Run in sequence in a single benchmark.
func Benchmark(f func(b *B)) BenchmarkResult {
	b := &B{
		common: common{
			signal: make(chan bool),
			w:      discard{},
		},
		benchFunc: f,
		benchTime: benchTime,
	}
	if b.run1() {
		b.run()
	}
	return b.result
}

type discard struct{}

func (discard) Write(b []byte) (n int, err error) { return len(b), nil }