-
Notifications
You must be signed in to change notification settings - Fork 20
/
picobench.hpp
1559 lines (1354 loc) · 45.3 KB
/
picobench.hpp
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
// picobench v2.05
// https://github.com/iboB/picobench
//
// A micro microbenchmarking library in a single header file
//
// SPDX-License-Identifier: MIT
//
// MIT License
//
// Copyright(c) 2017-2023 Borislav Stanimirov
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files(the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and / or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions :
//
// The above copyright notice and this permission notice shall be included in all
// copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
// SOFTWARE.
//
//
// VERSION HISTORY
//
// 2.05 (2023-04-26) Fixed MinGW build
// 2.04 (2023-04-12) Added CLI args to run specific benchmarks or suites
// 2.03 (2023-03-21) * Added PICOBENCH_UNIQUE_SYM_SUFFIX
// * Fixed several warnings
// 2.02 (2023-02-16) * Fixed same-func warning if user data is different
// * Macro PICOBENCH_NAMESPACE to change namespace
// * Changed marking of baseline in human-readable reports
// * Minor internal changes in strings
// 2.01 (2019-03-03) * Fixed android build when binding to a signle core
// * Minor doc fixes
// 2.00 (2018-10-30) * Breaking change! runner::run_benchmarks doesn't return
// a report anymore. The report is generated by
// runner::generate_report instead
// * Breaking change! report_output_format doesn't accept
// output streams as arguments. Use set_output_streams.
// * Potentially breaking change (gcc and clang)! Always set
// thread affinity to first core. Macro to turn this off.
// * Added runner::run which performs a full execution
// * Added benchmark results and results comparison
// * Added error enum
// * Macro option to allow a std::function as a benchmark
// * Macros for default iterations and samples
// * Allowing local registration of benchmarks in a runner
// * Added local_runner which doesn't consume registry
// * More force-inline functions in states
// * Fixed some potential compilation warnings
// * Removed tests from header
// * Anonymous namespace for impl-only classes and funcs
// * Added setters and getters for every config option
// 1.05 (2018-07-17) * Counting iterations of state
// * Optionally set thread affinity when running benchmarks
// so as not to miss cpu cycles with the high res clock
// 1.04 (2018-02-06) * User data for benchmarks, which can be seen from states
// * `add_custom_duration` to states so the user can modify time
// * Text table format fixes
// * Custom cmd opts in runner
// * --version CLI command
// 1.03 (2018-01-05) Added helper methods for easier browsing of reports
// 1.02 (2018-01-04) Added parsing of command line
// 1.01 (2018-01-03) * Only taking the fastest sample into account
// * Set default number of samples to 2
// * Added CSV output
// 1.00 (2018-01-01) Initial release
// 0.01 (2017-12-28) Initial prototype release
//
//
// EXAMPLE
//
// void my_function(); // the function you want to benchmark
//
// // write your benchmarking code in a function like this
// static void benchmark_my_function(picobench::state& state)
// {
// // use the state in a range-based for loop to call your code
// for (auto _ : state)
// my_function();
// }
// // create a picobench with your benchmarking code
// PICOBENCH(benchmark_my_function);
//
//
// BASIC DOCUMENTATION
//
// A very brief usage guide follows. For more detailed documentation see the
// README here: https://github.com/iboB/picobench/blob/master/README.md
//
// Simply include this file wherever you need.
// You need to define PICOBENCH_IMPLEMENT_WITH_MAIN (or PICOBENCH_IMPLEMENT if
// you want to write your own main function) in one compilation unit to have
// the implementation compiled there.
//
// The benchmark code must be a `void (picobench::state&)` function which
// you have written. Benchmarks are registered using the `PICOBENCH` macro
// where the only argument is the function's name.
//
// You can have multiple benchmarks in multiple files. All will be run when the
// executable starts.
//
// Typically a benchmark has a loop. To run the loop use the state argument in
// a range-based for loop in your function. The time spent looping is measured
// for the benchmark. You can have initialization/deinitialization code outside
// of the loop and it won't be measured.
//
#if !defined(PICOBENCH_HPP_INCLUDED)
#define PICOBENCH_HPP_INCLUDED
#include <cstdint>
#include <chrono>
#include <vector>
#if defined(PICOBENCH_STD_FUNCTION_BENCHMARKS)
# include <functional>
#endif
#define PICOBENCH_VERSION 2.05
#define PICOBENCH_VERSION_STR "2.05"
#if defined(PICOBENCH_DEBUG)
# include <cassert>
# define I_PICOBENCH_ASSERT assert
#else
# define I_PICOBENCH_ASSERT(...)
#endif
#if defined(__GNUC__)
# define PICOBENCH_INLINE __attribute__((always_inline))
#elif defined(_MSC_VER)
# define PICOBENCH_INLINE __forceinline
#else
# define PICOBENCH_INLINE inline
#endif
#if !defined(PICOBENCH_NAMESPACE)
# define PICOBENCH_NAMESPACE picobench
#endif
namespace PICOBENCH_NAMESPACE
{
#if defined(_MSC_VER) || defined(__MINGW32__) || defined(PICOBENCH_TEST)
struct high_res_clock
{
typedef long long rep;
typedef std::nano period;
typedef std::chrono::duration<rep, period> duration;
typedef std::chrono::time_point<high_res_clock> time_point;
static const bool is_steady = true;
static time_point now();
};
#else
using high_res_clock = std::chrono::high_resolution_clock;
#endif
using result_t = intptr_t;
class state
{
public:
explicit state(int num_iterations, uintptr_t user_data = 0)
: _user_data(user_data)
, _iterations(num_iterations)
{
I_PICOBENCH_ASSERT(_iterations > 0);
}
int iterations() const { return _iterations; }
int64_t duration_ns() const { return _duration_ns; }
void add_custom_duration(int64_t duration_ns) { _duration_ns += duration_ns; }
uintptr_t user_data() const { return _user_data; }
// optionally set result of benchmark
// this can be used as a value sync to prevent optimizations
// or a way to check whether benchmarks produce the same results
void set_result(uintptr_t data) { _result = data; }
result_t result() const { return _result; }
PICOBENCH_INLINE
void start_timer()
{
_start = high_res_clock::now();
}
PICOBENCH_INLINE
void stop_timer()
{
auto duration = high_res_clock::now() - _start;
_duration_ns = std::chrono::duration_cast<std::chrono::nanoseconds>(duration).count();
}
struct iterator
{
PICOBENCH_INLINE
iterator(state* parent)
: _counter(0)
, _lim(parent->iterations())
, _state(parent)
{
I_PICOBENCH_ASSERT(_counter < _lim);
}
PICOBENCH_INLINE
iterator()
: _counter(0)
, _lim(0)
, _state(nullptr)
{}
PICOBENCH_INLINE
iterator& operator++()
{
I_PICOBENCH_ASSERT(_counter < _lim);
++_counter;
return *this;
}
PICOBENCH_INLINE
bool operator!=(const iterator&) const
{
if (_counter < _lim) return true;
_state->stop_timer();
return false;
}
PICOBENCH_INLINE
int operator*() const
{
return _counter;
}
private:
int _counter;
const int _lim;
state* _state;
};
PICOBENCH_INLINE
iterator begin()
{
start_timer();
return iterator(this);
}
PICOBENCH_INLINE
iterator end()
{
return iterator();
}
private:
high_res_clock::time_point _start;
int64_t _duration_ns = 0;
uintptr_t _user_data;
int _iterations;
result_t _result = 0;
};
// this can be used for manual measurement
class scope
{
public:
PICOBENCH_INLINE
scope(state& s)
: _state(s)
{
_state.start_timer();
}
PICOBENCH_INLINE
~scope()
{
_state.stop_timer();
}
private:
state& _state;
};
#if defined(PICOBENCH_STD_FUNCTION_BENCHMARKS)
using benchmark_proc = std::function<void(state&)>;
#else
using benchmark_proc = void(*)(state&);
#endif
class benchmark
{
public:
const char* name() const { return _name; }
benchmark& iterations(std::vector<int> data) { _state_iterations = std::move(data); return *this; }
benchmark& samples(int n) { _samples = n; return *this; }
benchmark& label(const char* label) { _name = label; return *this; }
benchmark& baseline(bool b = true) { _baseline = b; return *this; }
benchmark& user_data(uintptr_t data) { _user_data = data; return *this; }
protected:
friend class runner;
benchmark(const char* name, benchmark_proc proc);
const char* _name;
const benchmark_proc _proc;
bool _baseline = false;
uintptr_t _user_data = 0;
std::vector<int> _state_iterations;
int _samples = 0;
};
// used for globally functions
// note that you can instantiate a runner and register local benchmarks for it alone
class global_registry
{
public:
static int set_bench_suite(const char* name);
static benchmark& new_benchmark(const char* name, benchmark_proc proc);
};
}
// Optionally define PICOBENCH_UNIQUE_SYM_SUFFIX to replace __LINE__ with something
// non standard like __COUNTER__ in case you need multiple PICOBENCH macros in a
// macro of yours
#if !defined(PICOBENCH_UNIQUE_SYM_SUFFIX)
#define PICOBENCH_UNIQUE_SYM_SUFFIX __LINE__
#endif
#define I_PICOBENCH_PP_CAT(a, b) I_PICOBENCH_PP_INTERNAL_CAT(a, b)
#define I_PICOBENCH_PP_INTERNAL_CAT(a, b) a##b
#define PICOBENCH_SUITE(name) \
static int I_PICOBENCH_PP_CAT(picobench_suite, PICOBENCH_UNIQUE_SYM_SUFFIX) = \
PICOBENCH_NAMESPACE::global_registry::set_bench_suite(name)
#define PICOBENCH(func) \
static auto& I_PICOBENCH_PP_CAT(picobench, PICOBENCH_UNIQUE_SYM_SUFFIX) = \
PICOBENCH_NAMESPACE::global_registry::new_benchmark(#func, func)
#if defined(PICOBENCH_IMPLEMENT_WITH_MAIN)
# define PICOBENCH_IMPLEMENT
# define PICOBENCH_IMPLEMENT_MAIN
#endif
#endif // PICOBENCH_HPP_INCLUDED
#if defined(PICOBENCH_IMPLEMENT)
#include <random>
#include <iostream>
#include <fstream>
#include <iomanip>
#include <map>
#include <memory>
#include <cstring>
#include <cstdlib>
#include <algorithm>
#if defined(_WIN32)
# define WIN32_LEAN_AND_MEAN
# include <Windows.h>
#else
# if !defined(PICOBENCH_DONT_BIND_TO_ONE_CORE)
# if defined(__APPLE__)
# include <mach/mach.h>
# else
# include <sched.h>
# endif
# endif
#endif
namespace PICOBENCH_NAMESPACE
{
// namespace
// {
enum error_t
{
no_error,
error_bad_cmd_line_argument, // ill-formed command-line argument
error_unknown_cmd_line_argument, // command argument looks like a picobench one, but isn't
error_sample_compare, // benchmark produced different results across samples
error_benchmark_compare, // two benchmarks of the same suite and dimension produced different results
};
class report
{
public:
struct benchmark_problem_space
{
int dimension; // number of iterations for the problem space
int samples; // number of samples taken
int64_t total_time_ns; // fastest sample!!!
result_t result; // result of fastest sample
};
struct benchmark
{
const char* name;
bool is_baseline;
std::vector<benchmark_problem_space> data;
};
struct suite
{
const char* name;
std::vector<benchmark> benchmarks; // benchmark view
const benchmark* find_benchmark(const char* bname) const
{
for (auto& b : benchmarks)
{
if (strcmp(b.name, bname) == 0)
return &b;
}
return nullptr;
}
const benchmark* find_baseline() const
{
for (auto& b : benchmarks)
{
if (b.is_baseline)
return &b;
}
return nullptr;
}
};
std::vector<suite> suites;
error_t error = no_error;
const suite* find_suite(const char* name) const
{
for (auto& s : suites)
{
if (strcmp(s.name, name) == 0)
return &s;
}
return nullptr;
}
void to_text(std::ostream& out) const
{
using namespace std;
for (auto& suite : suites)
{
if (suite.name)
{
out << "## " << suite.name << ":\n";
}
out.put('\n');
out <<
" Name (* = baseline) | Dim | Total ms | ns/op |Baseline| Ops/second\n";
out <<
"--------------------------|--------:|----------:|--------:|-------:|----------:\n";
auto problem_space_view = get_problem_space_view(suite);
for (auto& ps : problem_space_view)
{
const problem_space_benchmark* baseline = nullptr;
for (auto& bm : ps.second)
{
if (bm.is_baseline)
{
baseline = &bm;
break;
}
}
for (auto& bm : ps.second)
{
out << ' ' << bm.name;
auto pad = 24 - int(strlen(bm.name));
if (bm.is_baseline)
{
out << " *";
pad -= 2;
}
for (int i = 0; i < pad; ++i) {
out.put(' ');
}
out << " |"
<< setw(8) << ps.first << " |"
<< setw(10) << fixed << setprecision(3) << double(bm.total_time_ns) / 1000000.0 << " |";
auto ns_op = (bm.total_time_ns / ps.first);
if (ns_op > 99999999)
{
int e = 0;
while (ns_op > 999999)
{
++e;
ns_op /= 10;
}
out << ns_op << 'e' << e;
}
else
{
out << setw(8) << ns_op;
}
out << " |";
if (baseline == &bm)
{
out << " - |";
}
else if (baseline)
{
out << setw(7) << fixed << setprecision(3)
<< double(bm.total_time_ns) / double(baseline->total_time_ns) << " |";
}
else
{
// no baseline to compare to
out << " ??? |";
}
auto ops_per_sec = ps.first * (1000000000.0 / double(bm.total_time_ns));
out << setw(11) << fixed << setprecision(1) << ops_per_sec << "\n";
}
}
out.put('\n');
}
}
void to_text_concise(std::ostream& out)
{
using namespace std;
for (auto& suite : suites)
{
if (suite.name)
{
out << "## " << suite.name << ":\n";
}
out.put('\n');
out <<
" Name (* = baseline) | ns/op | Baseline | Ops/second\n";
out <<
"--------------------------|--------:|---------:|-----------:\n";
const benchmark* baseline = nullptr;
for (auto& bm : suite.benchmarks)
{
if (bm.is_baseline)
{
baseline = &bm;
break;
}
}
I_PICOBENCH_ASSERT(baseline);
int64_t baseline_total_time = 0;
int baseline_total_iterations = 0;
for (auto& d : baseline->data)
{
baseline_total_time += d.total_time_ns;
baseline_total_iterations += d.dimension;
}
int64_t baseline_ns_per_op = baseline_total_time / baseline_total_iterations;
for (auto& bm : suite.benchmarks)
{
out << ' ' << bm.name;
auto pad = 24 - int(strlen(bm.name));
if (bm.is_baseline)
{
out << " *";
pad -= 2;
}
for (int i = 0; i < pad; ++i) {
out.put(' ');
}
int64_t total_time = 0;
int total_iterations = 0;
for (auto& d : bm.data)
{
total_time += d.total_time_ns;
total_iterations += d.dimension;
}
int64_t ns_per_op = total_time / total_iterations;
out << " |" << setw(8) << ns_per_op << " |";
if (&bm == baseline)
{
out << " - |";
}
else
{
out << setw(9) << fixed << setprecision(3)
<< double(ns_per_op) / double(baseline_ns_per_op) << " |";
}
auto ops_per_sec = total_iterations * (1000000000.0 / double(total_time));
out << setw(12) << fixed << setprecision(1) << ops_per_sec << "\n";
}
out.put('\n');
}
}
void to_csv(std::ostream& out, bool header = true) const
{
using namespace std;
if (header)
{
out << "Suite,Benchmark,b,D,S,\"Total ns\",Result,\"ns/op\",Baseline\n";
}
for (auto& suite : suites)
{
const benchmark* baseline = nullptr;
for (auto& bm : suite.benchmarks)
{
if (bm.is_baseline)
{
baseline = &bm;
break;
}
}
I_PICOBENCH_ASSERT(baseline);
for (auto& bm : suite.benchmarks)
{
for (auto& d : bm.data)
{
if (suite.name)
{
out << '"' << suite.name << '"';;
}
out << ",\"" << bm.name << "\",";
if (&bm == baseline)
{
out << '*';
}
out << ','
<< d.dimension << ','
<< d.samples << ','
<< d.total_time_ns << ','
<< d.result << ','
<< (d.total_time_ns / d.dimension) << ',';
if (baseline)
{
for (auto& bd : baseline->data)
{
if (bd.dimension == d.dimension)
{
out << fixed << setprecision(3) << (double(d.total_time_ns) / double(bd.total_time_ns));
}
}
}
out << '\n';
}
}
}
}
struct problem_space_benchmark
{
const char* name;
bool is_baseline;
int64_t total_time_ns; // fastest sample!!!
result_t result; // result of fastest sample
};
static std::map<int, std::vector<problem_space_benchmark>> get_problem_space_view(const suite& s)
{
std::map<int, std::vector<problem_space_benchmark>> res;
for (auto& bm : s.benchmarks)
{
for (auto& d : bm.data)
{
auto& pvbs = res[d.dimension];
pvbs.push_back({ bm.name, bm.is_baseline, d.total_time_ns, d.result });
}
}
return res;
}
private:
};
class benchmark_impl : public benchmark
{
public:
benchmark_impl(const char* name, benchmark_proc proc)
: benchmark(name, proc)
{}
private:
friend class runner;
// state
std::vector<state> _states; // length is _samples * _state_iterations.size()
std::vector<state>::iterator _istate;
};
class picostring
{
public:
picostring() = default;
explicit picostring(const char* text)
: str(text)
, len(int(strlen(text)))
{}
picostring(const char* text, int len)
: str(text)
, len(len)
{}
const char* str;
int len = 0;
// checks whether other begins with this string
bool is_start_of(const char* other) const
{
return strncmp(str, other, size_t(len)) == 0;
}
bool operator==(const picostring& other) const
{
if (len != other.len) return false;
return strncmp(str, other.str, size_t(len)) == 0;
}
bool operator==(const char* other) const
{
return operator==(picostring(other));
}
};
class null_streambuf : public std::streambuf
{
public:
virtual int overflow(int c) override { return c; }
};
struct null_stream : public std::ostream
{
null_stream() : std::ostream(&_buf) {}
private:
null_streambuf _buf;
} cnull;
enum class report_output_format
{
text,
concise_text,
csv,
};
#if !defined(PICOBENCH_DEFAULT_ITERATIONS)
# define PICOBENCH_DEFAULT_ITERATIONS { 8, 64, 512, 4096, 8192 }
#endif
#if !defined(PICOBENCH_DEFAULT_SAMPLES)
# define PICOBENCH_DEFAULT_SAMPLES 2
#endif
using benchmarks_vector = std::vector<std::unique_ptr<benchmark_impl>>;
struct rsuite
{
const char* name;
benchmarks_vector benchmarks;
};
class registry
{
public:
benchmark& add_benchmark(const char* name, benchmark_proc proc)
{
auto b = new benchmark_impl(name, proc);
benchmarks_for_current_suite().emplace_back(b);
return *b;
}
void set_suite(const char* name)
{
_current_suite_name = name;
}
const char*& current_suite_name()
{
return _current_suite_name;
}
benchmarks_vector& benchmarks_for_current_suite()
{
for (auto& s : _suites)
{
if (s.name == _current_suite_name)
return s.benchmarks;
if (s.name && _current_suite_name && strcmp(s.name, _current_suite_name) == 0)
return s.benchmarks;
}
_suites.push_back({ _current_suite_name, {} });
return _suites.back().benchmarks;
}
protected:
friend class runner;
const char* _current_suite_name = nullptr;
std::vector<rsuite> _suites;
};
registry& g_registry()
{
static registry r;
return r;
}
class runner : public registry
{
public:
runner(bool local = false)
: _default_state_iterations(PICOBENCH_DEFAULT_ITERATIONS)
, _default_samples(PICOBENCH_DEFAULT_SAMPLES)
{
if (!local)
{
_suites = std::move(g_registry()._suites);
}
}
int run(int benchmark_random_seed = -1)
{
if (should_run())
{
run_benchmarks(benchmark_random_seed);
auto report = generate_report();
std::ostream* out = _stdout;
std::ofstream fout;
if (preferred_output_filename())
{
fout.open(preferred_output_filename());
if (!fout.is_open())
{
std::cerr << "Error: Could not open output file `" << preferred_output_filename() << "`\n";
return 1;
}
out = &fout;
}
switch (preferred_output_format())
{
case report_output_format::text:
report.to_text(*out);
break;
case report_output_format::concise_text:
report.to_text_concise(*out);
break;
case report_output_format::csv:
report.to_csv(*out);
break;
}
}
return error();
}
void run_benchmarks(int random_seed = -1)
{
I_PICOBENCH_ASSERT(_error == no_error && _should_run);
if (random_seed == -1)
{
random_seed = int(std::random_device()());
}
std::minstd_rand rnd(random_seed);
// vector of all benchmarks
std::vector<benchmark_impl*> benchmarks;
for (auto& suite : _suites)
{
// also identify a baseline in this loop
// if there is no explicit one, set the first one as a baseline
bool found_baseline = false;
for (auto irb = suite.benchmarks.begin(); irb != suite.benchmarks.end(); ++irb)
{
auto& rb = *irb;
rb->_states.clear(); // clear states so we can safely call run_benchmarks multiple times
benchmarks.push_back(rb.get());
if (rb->_baseline)
{
found_baseline = true;
}
#if !defined(PICOBENCH_STD_FUNCTION_BENCHMARKS)
// check for same func
for (auto ib = irb+1; ib != suite.benchmarks.end(); ++ib)
{
auto& b = *ib;
if (rb->_proc == b->_proc && rb->_user_data == b->_user_data)
{
*_stdwarn << "Warning: " << rb->name() << " and " << b->name()
<< " are benchmarks of the same function.\n";
}
}
#endif
}
if (!found_baseline && !suite.benchmarks.empty())
{
suite.benchmarks.front()->_baseline = true;
}
}
// initialize benchmarks
for (auto b : benchmarks)
{
const std::vector<int>& state_iterations =
b->_state_iterations.empty() ?
_default_state_iterations :
b->_state_iterations;
if (b->_samples == 0)
b->_samples = _default_samples;
b->_states.reserve(state_iterations.size() * size_t(b->_samples));
// fill states while random shuffling them
for (auto iters : state_iterations)
{
for (int i = 0; i < b->_samples; ++i)
{
auto index = rnd() % (b->_states.size() + 1);
auto pos = b->_states.begin() + long(index);
b->_states.emplace(pos, iters, b->_user_data);
}
}
b->_istate = b->_states.begin();
}
#if !defined(PICOBENCH_DONT_BIND_TO_ONE_CORE)
// set thread affinity to first cpu
// so the high resolution clock doesn't miss cycles
{
#if defined(_WIN32)
SetThreadAffinityMask(GetCurrentThread(), 1);
#elif defined(__APPLE__)
thread_affinity_policy_data_t policy = {0};
thread_policy_set(
pthread_mach_thread_np(pthread_self()),
THREAD_AFFINITY_POLICY,
(thread_policy_t)&policy, 1);
#else
cpu_set_t cpuset;
CPU_ZERO(&cpuset);
CPU_SET(0, &cpuset);
sched_setaffinity(0, sizeof(cpu_set_t), &cpuset);
#endif
}
#endif
// we run a random benchmark from it incrementing _istate for each
// when _istate reaches _states.end(), we erase the benchmark
// when the vector becomes empty, we're done
while (!benchmarks.empty())
{
auto i = benchmarks.begin() + long(rnd() % benchmarks.size());
auto& b = *i;
b->_proc(*b->_istate);
++b->_istate;
if (b->_istate == b->_states.end())
{
benchmarks.erase(i);
}
}
}
// function to compare results