1 // Copyright 2015 Google Inc. All rights reserved.
2 //
3 // Licensed under the Apache License, Version 2.0 (the "License");
4 // you may not use this file except in compliance with the License.
5 // You may obtain a copy of the License at
6 //
7 //     http://www.apache.org/licenses/LICENSE-2.0
8 //
9 // Unless required by applicable law or agreed to in writing, software
10 // distributed under the License is distributed on an "AS IS" BASIS,
11 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 // See the License for the specific language governing permissions and
13 // limitations under the License.
14 
15 #include "benchmark/benchmark.h"
16 #include "benchmark_api_internal.h"
17 #include "internal_macros.h"
18 
19 #ifndef BENCHMARK_OS_WINDOWS
20 #include <sys/resource.h>
21 #include <sys/time.h>
22 #include <unistd.h>
23 #endif
24 
25 #include <algorithm>
26 #include <atomic>
27 #include <condition_variable>
28 #include <cstdio>
29 #include <cstdlib>
30 #include <cstring>
31 #include <fstream>
32 #include <iostream>
33 #include <memory>
34 #include <thread>
35 
36 #include "check.h"
37 #include "commandlineflags.h"
38 #include "complexity.h"
39 #include "log.h"
40 #include "mutex.h"
41 #include "re.h"
42 #include "stat.h"
43 #include "string_util.h"
44 #include "sysinfo.h"
45 #include "timers.h"
46 
47 namespace benchmark {
48 
49 namespace {
50 // For non-dense Range, intermediate values are powers of kRangeMultiplier.
51 static const int kRangeMultiplier = 8;
52 // The size of a benchmark family determines is the number of inputs to repeat
53 // the benchmark on. If this is "large" then warn the user during configuration.
54 static const size_t kMaxFamilySize = 100;
55 }  // end namespace
56 
57 namespace internal {
58 
59 //=============================================================================//
60 //                         BenchmarkFamilies
61 //=============================================================================//
62 
63 // Class for managing registered benchmarks.  Note that each registered
64 // benchmark identifies a family of related benchmarks to run.
65 class BenchmarkFamilies {
66  public:
67   static BenchmarkFamilies* GetInstance();
68 
69   // Registers a benchmark family and returns the index assigned to it.
70   size_t AddBenchmark(std::unique_ptr<Benchmark> family);
71 
72   // Extract the list of benchmark instances that match the specified
73   // regular expression.
74   bool FindBenchmarks(const std::string& re,
75                       std::vector<Benchmark::Instance>* benchmarks,
76                       std::ostream* Err);
77 
78  private:
BenchmarkFamilies()79   BenchmarkFamilies() {}
80 
81   std::vector<std::unique_ptr<Benchmark>> families_;
82   Mutex mutex_;
83 };
84 
GetInstance()85 BenchmarkFamilies* BenchmarkFamilies::GetInstance() {
86   static BenchmarkFamilies instance;
87   return &instance;
88 }
89 
AddBenchmark(std::unique_ptr<Benchmark> family)90 size_t BenchmarkFamilies::AddBenchmark(std::unique_ptr<Benchmark> family) {
91   MutexLock l(mutex_);
92   size_t index = families_.size();
93   families_.push_back(std::move(family));
94   return index;
95 }
96 
FindBenchmarks(const std::string & spec,std::vector<Benchmark::Instance> * benchmarks,std::ostream * ErrStream)97 bool BenchmarkFamilies::FindBenchmarks(
98     const std::string& spec, std::vector<Benchmark::Instance>* benchmarks,
99     std::ostream* ErrStream) {
100   CHECK(ErrStream);
101   auto& Err = *ErrStream;
102   // Make regular expression out of command-line flag
103   std::string error_msg;
104   Regex re;
105   if (!re.Init(spec, &error_msg)) {
106     Err << "Could not compile benchmark re: " << error_msg << std::endl;
107     return false;
108   }
109 
110   // Special list of thread counts to use when none are specified
111   const std::vector<int> one_thread = {1};
112 
113   MutexLock l(mutex_);
114   for (std::unique_ptr<Benchmark>& family : families_) {
115     // Family was deleted or benchmark doesn't match
116     if (!family) continue;
117 
118     if (family->ArgsCnt() == -1) {
119       family->Args({});
120     }
121     const std::vector<int>* thread_counts =
122         (family->thread_counts_.empty()
123              ? &one_thread
124              : &static_cast<const std::vector<int>&>(family->thread_counts_));
125     const size_t family_size = family->args_.size() * thread_counts->size();
126     // The benchmark will be run at least 'family_size' different inputs.
127     // If 'family_size' is very large warn the user.
128     if (family_size > kMaxFamilySize) {
129       Err << "The number of inputs is very large. " << family->name_
130           << " will be repeated at least " << family_size << " times.\n";
131     }
132     // reserve in the special case the regex ".", since we know the final
133     // family size.
134     if (spec == ".") benchmarks->reserve(family_size);
135 
136     for (auto const& args : family->args_) {
137       for (int num_threads : *thread_counts) {
138         Benchmark::Instance instance;
139         instance.name = family->name_;
140         instance.benchmark = family.get();
141         instance.report_mode = family->report_mode_;
142         instance.arg = args;
143         instance.time_unit = family->time_unit_;
144         instance.range_multiplier = family->range_multiplier_;
145         instance.min_time = family->min_time_;
146         instance.repetitions = family->repetitions_;
147         instance.use_real_time = family->use_real_time_;
148         instance.use_manual_time = family->use_manual_time_;
149         instance.complexity = family->complexity_;
150         instance.complexity_lambda = family->complexity_lambda_;
151         instance.threads = num_threads;
152 
153         // Add arguments to instance name
154         size_t arg_i = 0;
155         for (auto const& arg : args) {
156           instance.name += "/";
157 
158           if (arg_i < family->arg_names_.size()) {
159             const auto& arg_name = family->arg_names_[arg_i];
160             if (!arg_name.empty()) {
161               instance.name +=
162                   StringPrintF("%s:", family->arg_names_[arg_i].c_str());
163             }
164           }
165 
166           AppendHumanReadable(arg, &instance.name);
167           ++arg_i;
168         }
169 
170         if (!IsZero(family->min_time_)) {
171           instance.name += StringPrintF("/min_time:%0.3f", family->min_time_);
172         }
173         if (family->repetitions_ != 0) {
174           instance.name += StringPrintF("/repeats:%d", family->repetitions_);
175         }
176         if (family->use_manual_time_) {
177           instance.name += "/manual_time";
178         } else if (family->use_real_time_) {
179           instance.name += "/real_time";
180         }
181 
182         // Add the number of threads used to the name
183         if (!family->thread_counts_.empty()) {
184           instance.name += StringPrintF("/threads:%d", instance.threads);
185         }
186 
187         if (re.Match(instance.name)) {
188           instance.last_benchmark_instance = (&args == &family->args_.back());
189           benchmarks->push_back(std::move(instance));
190         }
191       }
192     }
193   }
194   return true;
195 }
196 
RegisterBenchmarkInternal(Benchmark * bench)197 Benchmark* RegisterBenchmarkInternal(Benchmark* bench) {
198   std::unique_ptr<Benchmark> bench_ptr(bench);
199   BenchmarkFamilies* families = BenchmarkFamilies::GetInstance();
200   families->AddBenchmark(std::move(bench_ptr));
201   return bench;
202 }
203 
204 // FIXME: This function is a hack so that benchmark.cc can access
205 // `BenchmarkFamilies`
FindBenchmarksInternal(const std::string & re,std::vector<Benchmark::Instance> * benchmarks,std::ostream * Err)206 bool FindBenchmarksInternal(const std::string& re,
207                             std::vector<Benchmark::Instance>* benchmarks,
208                             std::ostream* Err) {
209   return BenchmarkFamilies::GetInstance()->FindBenchmarks(re, benchmarks, Err);
210 }
211 
212 //=============================================================================//
213 //                               Benchmark
214 //=============================================================================//
215 
Benchmark(const char * name)216 Benchmark::Benchmark(const char* name)
217     : name_(name),
218       report_mode_(RM_Unspecified),
219       time_unit_(kNanosecond),
220       range_multiplier_(kRangeMultiplier),
221       min_time_(0),
222       repetitions_(0),
223       use_real_time_(false),
224       use_manual_time_(false),
225       complexity_(oNone),
226       complexity_lambda_(nullptr) {}
227 
~Benchmark()228 Benchmark::~Benchmark() {}
229 
AddRange(std::vector<int> * dst,int lo,int hi,int mult)230 void Benchmark::AddRange(std::vector<int>* dst, int lo, int hi, int mult) {
231   CHECK_GE(lo, 0);
232   CHECK_GE(hi, lo);
233   CHECK_GE(mult, 2);
234 
235   // Add "lo"
236   dst->push_back(lo);
237 
238   static const int kint32max = std::numeric_limits<int32_t>::max();
239 
240   // Now space out the benchmarks in multiples of "mult"
241   for (int32_t i = 1; i < kint32max / mult; i *= mult) {
242     if (i >= hi) break;
243     if (i > lo) {
244       dst->push_back(i);
245     }
246   }
247   // Add "hi" (if different from "lo")
248   if (hi != lo) {
249     dst->push_back(hi);
250   }
251 }
252 
Arg(int x)253 Benchmark* Benchmark::Arg(int x) {
254   CHECK(ArgsCnt() == -1 || ArgsCnt() == 1);
255   args_.push_back({x});
256   return this;
257 }
258 
Unit(TimeUnit unit)259 Benchmark* Benchmark::Unit(TimeUnit unit) {
260   time_unit_ = unit;
261   return this;
262 }
263 
Range(int start,int limit)264 Benchmark* Benchmark::Range(int start, int limit) {
265   CHECK(ArgsCnt() == -1 || ArgsCnt() == 1);
266   std::vector<int> arglist;
267   AddRange(&arglist, start, limit, range_multiplier_);
268 
269   for (int i : arglist) {
270     args_.push_back({i});
271   }
272   return this;
273 }
274 
Ranges(const std::vector<std::pair<int,int>> & ranges)275 Benchmark* Benchmark::Ranges(const std::vector<std::pair<int, int>>& ranges) {
276   CHECK(ArgsCnt() == -1 || ArgsCnt() == static_cast<int>(ranges.size()));
277   std::vector<std::vector<int>> arglists(ranges.size());
278   std::size_t total = 1;
279   for (std::size_t i = 0; i < ranges.size(); i++) {
280     AddRange(&arglists[i], ranges[i].first, ranges[i].second,
281              range_multiplier_);
282     total *= arglists[i].size();
283   }
284 
285   std::vector<std::size_t> ctr(arglists.size(), 0);
286 
287   for (std::size_t i = 0; i < total; i++) {
288     std::vector<int> tmp;
289     tmp.reserve(arglists.size());
290 
291     for (std::size_t j = 0; j < arglists.size(); j++) {
292       tmp.push_back(arglists[j].at(ctr[j]));
293     }
294 
295     args_.push_back(std::move(tmp));
296 
297     for (std::size_t j = 0; j < arglists.size(); j++) {
298       if (ctr[j] + 1 < arglists[j].size()) {
299         ++ctr[j];
300         break;
301       }
302       ctr[j] = 0;
303     }
304   }
305   return this;
306 }
307 
ArgName(const std::string & name)308 Benchmark* Benchmark::ArgName(const std::string& name) {
309   CHECK(ArgsCnt() == -1 || ArgsCnt() == 1);
310   arg_names_ = {name};
311   return this;
312 }
313 
ArgNames(const std::vector<std::string> & names)314 Benchmark* Benchmark::ArgNames(const std::vector<std::string>& names) {
315   CHECK(ArgsCnt() == -1 || ArgsCnt() == static_cast<int>(names.size()));
316   arg_names_ = names;
317   return this;
318 }
319 
DenseRange(int start,int limit,int step)320 Benchmark* Benchmark::DenseRange(int start, int limit, int step) {
321   CHECK(ArgsCnt() == -1 || ArgsCnt() == 1);
322   CHECK_GE(start, 0);
323   CHECK_LE(start, limit);
324   for (int arg = start; arg <= limit; arg += step) {
325     args_.push_back({arg});
326   }
327   return this;
328 }
329 
Args(const std::vector<int> & args)330 Benchmark* Benchmark::Args(const std::vector<int>& args) {
331   CHECK(ArgsCnt() == -1 || ArgsCnt() == static_cast<int>(args.size()));
332   args_.push_back(args);
333   return this;
334 }
335 
Apply(void (* custom_arguments)(Benchmark * benchmark))336 Benchmark* Benchmark::Apply(void (*custom_arguments)(Benchmark* benchmark)) {
337   custom_arguments(this);
338   return this;
339 }
340 
RangeMultiplier(int multiplier)341 Benchmark* Benchmark::RangeMultiplier(int multiplier) {
342   CHECK(multiplier > 1);
343   range_multiplier_ = multiplier;
344   return this;
345 }
346 
Repetitions(int n)347 Benchmark* Benchmark::Repetitions(int n) {
348   CHECK(n > 0);
349   repetitions_ = n;
350   return this;
351 }
352 
ReportAggregatesOnly(bool value)353 Benchmark* Benchmark::ReportAggregatesOnly(bool value) {
354   report_mode_ = value ? RM_ReportAggregatesOnly : RM_Default;
355   return this;
356 }
357 
MinTime(double t)358 Benchmark* Benchmark::MinTime(double t) {
359   CHECK(t > 0.0);
360   min_time_ = t;
361   return this;
362 }
363 
UseRealTime()364 Benchmark* Benchmark::UseRealTime() {
365   CHECK(!use_manual_time_)
366       << "Cannot set UseRealTime and UseManualTime simultaneously.";
367   use_real_time_ = true;
368   return this;
369 }
370 
UseManualTime()371 Benchmark* Benchmark::UseManualTime() {
372   CHECK(!use_real_time_)
373       << "Cannot set UseRealTime and UseManualTime simultaneously.";
374   use_manual_time_ = true;
375   return this;
376 }
377 
Complexity(BigO complexity)378 Benchmark* Benchmark::Complexity(BigO complexity) {
379   complexity_ = complexity;
380   return this;
381 }
382 
Complexity(BigOFunc * complexity)383 Benchmark* Benchmark::Complexity(BigOFunc* complexity) {
384   complexity_lambda_ = complexity;
385   complexity_ = oLambda;
386   return this;
387 }
388 
Threads(int t)389 Benchmark* Benchmark::Threads(int t) {
390   CHECK_GT(t, 0);
391   thread_counts_.push_back(t);
392   return this;
393 }
394 
ThreadRange(int min_threads,int max_threads)395 Benchmark* Benchmark::ThreadRange(int min_threads, int max_threads) {
396   CHECK_GT(min_threads, 0);
397   CHECK_GE(max_threads, min_threads);
398 
399   AddRange(&thread_counts_, min_threads, max_threads, 2);
400   return this;
401 }
402 
DenseThreadRange(int min_threads,int max_threads,int stride)403 Benchmark* Benchmark::DenseThreadRange(int min_threads, int max_threads,
404                                        int stride) {
405   CHECK_GT(min_threads, 0);
406   CHECK_GE(max_threads, min_threads);
407   CHECK_GE(stride, 1);
408 
409   for (auto i = min_threads; i < max_threads; i += stride) {
410     thread_counts_.push_back(i);
411   }
412   thread_counts_.push_back(max_threads);
413   return this;
414 }
415 
ThreadPerCpu()416 Benchmark* Benchmark::ThreadPerCpu() {
417   static int num_cpus = NumCPUs();
418   thread_counts_.push_back(num_cpus);
419   return this;
420 }
421 
SetName(const char * name)422 void Benchmark::SetName(const char* name) { name_ = name; }
423 
ArgsCnt() const424 int Benchmark::ArgsCnt() const {
425   if (args_.empty()) {
426     if (arg_names_.empty()) return -1;
427     return static_cast<int>(arg_names_.size());
428   }
429   return static_cast<int>(args_.front().size());
430 }
431 
432 //=============================================================================//
433 //                            FunctionBenchmark
434 //=============================================================================//
435 
Run(State & st)436 void FunctionBenchmark::Run(State& st) { func_(st); }
437 
438 }  // end namespace internal
439 }  // end namespace benchmark
440