diff --git a/README.md b/README.md index a94925c223..19cf93036e 100644 --- a/README.md +++ b/README.md @@ -180,7 +180,7 @@ BENCHMARK_MAIN(); ``` To run the benchmark, compile and link against the `benchmark` library -(libbenchmark.a/.so). If you followed the build steps above, this library will +(libbenchmark.a/.so). If you followed the build steps above, this library will be under the build directory you created. ```bash @@ -300,6 +300,8 @@ too (`-lkstat`). [Setting the Time Unit](#setting-the-time-unit) +[Random Interleaving](docs/random_interleaving.md) + [User-Requested Performance Counters](docs/perf_counters.md) [Preventing Optimization](#preventing-optimization) @@ -400,8 +402,8 @@ Write benchmark results to a file with the `--benchmark_out=` option (or set `BENCHMARK_OUT`). Specify the output format with `--benchmark_out_format={json|console|csv}` (or set `BENCHMARK_OUT_FORMAT={json|console|csv}`). Note that the 'csv' reporter is -deprecated and the saved `.csv` file -[is not parsable](https://github.com/google/benchmark/issues/794) by csv +deprecated and the saved `.csv` file +[is not parsable](https://github.com/google/benchmark/issues/794) by csv parsers. Specifying `--benchmark_out` does not suppress the console output. diff --git a/docs/random_interleaving.md b/docs/random_interleaving.md new file mode 100644 index 0000000000..2471b46bb0 --- /dev/null +++ b/docs/random_interleaving.md @@ -0,0 +1,26 @@ + + +# Random Interleaving + +[Random Interleaving](https://github.com/google/benchmark/issues/1051) is a +technique to lower run-to-run variance. It breaks the execution of a +microbenchmark into multiple chunks and randomly interleaves them with chunks +from other microbenchmarks in the same benchmark test. Data shows it is able to +lower run-to-run variance by +[40%](https://github.com/google/benchmark/issues/1051) on average. + +To use, set `--benchmark_enable_random_interleaving=true`. + +It's a known issue that random interleaving may increase the benchmark execution +time, if: + +1. A benchmark has costly setup and / or teardown. Random interleaving will run + setup and teardown many times and may increase test execution time + significantly. +2. The time to run a single benchmark iteration is larger than the desired time + per repetition (i.e., `benchmark_min_time / benchmark_repetitions`). + +The overhead of random interleaving can be controlled by +`--benchmark_random_interleaving_max_overhead`. The default value is 0.4 meaning +the total execution time under random interlaving is limited by 1.4 x original +total execution time. Set it to `inf` for unlimited overhead. diff --git a/src/benchmark.cc b/src/benchmark.cc index d205232b8a..272794e147 100644 --- a/src/benchmark.cc +++ b/src/benchmark.cc @@ -33,8 +33,10 @@ #include #include #include +#include #include #include +#include #include #include #include @@ -54,6 +56,18 @@ #include "thread_manager.h" #include "thread_timer.h" +// Each benchmark can be repeated a number of times, and within each +// *repetition*, we run the user-defined benchmark function a number of +// *iterations*. The number of repetitions is determined based on flags +// (--benchmark_repetitions). +namespace { + +// Attempt to make each repetition run for at least this much of time. +constexpr double kDefaultMinTimeTotalSecs = 0.5; +constexpr int kRandomInterleavingDefaultRepetitions = 12; + +} // namespace + // Print a list of benchmarks. This option overrides all other options. DEFINE_bool(benchmark_list_tests, false); @@ -62,16 +76,39 @@ DEFINE_bool(benchmark_list_tests, false); // linked into the binary are run. DEFINE_string(benchmark_filter, "."); -// Minimum number of seconds we should run benchmark before results are -// considered significant. For cpu-time based tests, this is the lower bound -// on the total cpu time used by all threads that make up the test. For -// real-time based tests, this is the lower bound on the elapsed time of the -// benchmark execution, regardless of number of threads. -DEFINE_double(benchmark_min_time, 0.5); +// Do NOT read these flags directly. Use Get*() to read them. +namespace do_not_read_flag_directly { + +// Minimum number of seconds we should run benchmark per repetition before +// results are considered significant. For cpu-time based tests, this is the +// lower bound on the total cpu time used by all threads that make up the test. +// For real-time based tests, this is the lower bound on the elapsed time of the +// benchmark execution, regardless of number of threads. If left unset, will use +// kDefaultMinTimeTotalSecs / FLAGS_benchmark_repetitions, if random +// interleaving is enabled. Otherwise, will use kDefaultMinTimeTotalSecs. +// Do NOT read this flag directly. Use GetMinTime() to read this flag. +DEFINE_double(benchmark_min_time, -1.0); // The number of runs of each benchmark. If greater than 1, the mean and -// standard deviation of the runs will be reported. -DEFINE_int32(benchmark_repetitions, 1); +// standard deviation of the runs will be reported. By default, the number of +// repetitions is 1 if random interleaving is disabled, and up to +// kDefaultRepetitions if random interleaving is enabled. (Read the +// documentation for random interleaving to see why it might be less than +// kDefaultRepetitions.) +// Do NOT read this flag directly, Use GetRepetitions() to access this flag. +DEFINE_int32(benchmark_repetitions, -1); + +} // namespace do_not_read_flag_directly + +// The maximum overhead allowed for random interleaving. A value X means total +// execution time under random interleaving is limited by +// (1 + X) * original total execution time. Set to 'inf' to allow infinite +// overhead. +DEFINE_double(benchmark_random_interleaving_max_overhead, 0.4); + +// If set, enable random interleaving. See +// http://github.com/google/benchmark/issues/1051 for details. +DEFINE_bool(benchmark_enable_random_interleaving, false); // Report the result of each benchmark repetitions. When 'true' is specified // only the mean, standard deviation, and other statistics are reported for @@ -122,6 +159,30 @@ DEFINE_kvpairs(benchmark_context, {}); std::map* global_context = nullptr; +// Performance measurements always come with random variances. Defines a +// factor by which the required number of iterations is overestimated in order +// to reduce the probability that the minimum time requirement will not be met. +const double kSafetyMultiplier = 1.4; + +// Wraps --benchmark_min_time and returns valid default values if not supplied. +double GetMinTime() { + const double default_min_time = kDefaultMinTimeTotalSecs / GetRepetitions(); + const double flag_min_time = + do_not_read_flag_directly::FLAGS_benchmark_min_time; + return flag_min_time >= 0.0 ? flag_min_time : default_min_time; +} + +// Wraps --benchmark_repetitions and return valid default value if not supplied. +int GetRepetitions() { + const int default_repetitions = + FLAGS_benchmark_enable_random_interleaving + ? kRandomInterleavingDefaultRepetitions + : 1; + const int flag_repetitions = + do_not_read_flag_directly::FLAGS_benchmark_repetitions; + return flag_repetitions >= 0 ? flag_repetitions : default_repetitions; +} + // FIXME: wouldn't LTO mess this up? void UseCharPointer(char const volatile*) {} @@ -241,6 +302,39 @@ void State::FinishKeepRunning() { namespace internal { namespace { +// Flushes streams after invoking reporter methods that write to them. This +// ensures users get timely updates even when streams are not line-buffered. +void FlushStreams(BenchmarkReporter* reporter) { + if (!reporter) return; + std::flush(reporter->GetOutputStream()); + std::flush(reporter->GetErrorStream()); +}; + +// Reports in both display and file reporters. +void Report(BenchmarkReporter* display_reporter, + BenchmarkReporter* file_reporter, const RunResults& run_results) { + auto report_one = [](BenchmarkReporter* reporter, + bool aggregates_only, + const RunResults& results) { + assert(reporter); + // If there are no aggregates, do output non-aggregates. + aggregates_only &= !results.aggregates_only.empty(); + if (!aggregates_only) + reporter->ReportRuns(results.non_aggregates); + if (!results.aggregates_only.empty()) + reporter->ReportRuns(results.aggregates_only); + }; + + report_one(display_reporter, run_results.display_report_aggregates_only, + run_results); + if (file_reporter) + report_one(file_reporter, run_results.file_report_aggregates_only, + run_results); + + FlushStreams(display_reporter); + FlushStreams(file_reporter); +}; + void RunBenchmarks(const std::vector& benchmarks, BenchmarkReporter* display_reporter, BenchmarkReporter* file_reporter) { @@ -248,7 +342,7 @@ void RunBenchmarks(const std::vector& benchmarks, CHECK(display_reporter != nullptr); // Determine the width of the name field using a minimum width of 10. - bool might_have_aggregates = FLAGS_benchmark_repetitions > 1; + bool might_have_aggregates = GetRepetitions() > 1; size_t name_field_width = 10; size_t stat_field_width = 0; for (const BenchmarkInstance& benchmark : benchmarks) { @@ -256,8 +350,9 @@ void RunBenchmarks(const std::vector& benchmarks, std::max(name_field_width, benchmark.name().str().size()); might_have_aggregates |= benchmark.repetitions() > 1; - for (const auto& Stat : benchmark.statistics()) + for (const auto& Stat : benchmark.statistics()) { stat_field_width = std::max(stat_field_width, Stat.name_.size()); + } } if (might_have_aggregates) name_field_width += 1 + stat_field_width; @@ -268,45 +363,61 @@ void RunBenchmarks(const std::vector& benchmarks, // Keep track of running times of all instances of current benchmark std::vector complexity_reports; - // We flush streams after invoking reporter methods that write to them. This - // ensures users get timely updates even when streams are not line-buffered. - auto flushStreams = [](BenchmarkReporter* reporter) { - if (!reporter) return; - std::flush(reporter->GetOutputStream()); - std::flush(reporter->GetErrorStream()); - }; - if (display_reporter->ReportContext(context) && (!file_reporter || file_reporter->ReportContext(context))) { - flushStreams(display_reporter); - flushStreams(file_reporter); - - for (const auto& benchmark : benchmarks) { - RunResults run_results = RunBenchmark(benchmark, &complexity_reports); - - auto report = [&run_results](BenchmarkReporter* reporter, - bool report_aggregates_only) { - assert(reporter); - // If there are no aggregates, do output non-aggregates. - report_aggregates_only &= !run_results.aggregates_only.empty(); - if (!report_aggregates_only) - reporter->ReportRuns(run_results.non_aggregates); - if (!run_results.aggregates_only.empty()) - reporter->ReportRuns(run_results.aggregates_only); - }; - - report(display_reporter, run_results.display_report_aggregates_only); - if (file_reporter) - report(file_reporter, run_results.file_report_aggregates_only); - - flushStreams(display_reporter); - flushStreams(file_reporter); + FlushStreams(display_reporter); + FlushStreams(file_reporter); + + // Without random interleaving, benchmarks are executed in the order of: + // A, A, ..., A, B, B, ..., B, C, C, ..., C, ... + // That is, repetition is within RunBenchmark(), hence the name + // inner_repetitions. + // With random interleaving, benchmarks are executed in the order of: + // {Random order of A, B, C, ...}, {Random order of A, B, C, ...}, ... + // That is, repetitions is outside of RunBenchmark(), hence the name + // outer_repetitions. + int inner_repetitions = + FLAGS_benchmark_enable_random_interleaving ? 1 : GetRepetitions(); + int outer_repetitions = + FLAGS_benchmark_enable_random_interleaving ? GetRepetitions() : 1; + std::vector benchmark_indices(benchmarks.size()); + for (size_t i = 0; i < benchmarks.size(); ++i) { + benchmark_indices[i] = i; + } + + std::random_device rd; + std::mt19937 g(rd()); + // 'run_results_vector' and 'benchmarks' are parallel arrays. + std::vector run_results_vector(benchmarks.size()); + for (int i = 0; i < outer_repetitions; i++) { + if (FLAGS_benchmark_enable_random_interleaving) { + std::shuffle(benchmark_indices.begin(), benchmark_indices.end(), g); + } + for (size_t j : benchmark_indices) { + // Repetitions will be automatically adjusted under random interleaving. + if (!FLAGS_benchmark_enable_random_interleaving || + i < benchmarks[j].RandomInterleavingRepetitions()) { + RunBenchmark(benchmarks[j], outer_repetitions, inner_repetitions, + &complexity_reports, &run_results_vector[j]); + if (!FLAGS_benchmark_enable_random_interleaving) { + // Print out reports as they come in. + Report(display_reporter, file_reporter, run_results_vector.at(j)); + } + } + } + } + + if (FLAGS_benchmark_enable_random_interleaving) { + // Print out all reports at the end of the test. + for (const RunResults& run_results : run_results_vector) { + Report(display_reporter, file_reporter, run_results); + } } } display_reporter->Finalize(); if (file_reporter) file_reporter->Finalize(); - flushStreams(display_reporter); - flushStreams(file_reporter); + FlushStreams(display_reporter); + FlushStreams(file_reporter); } // Disable deprecated warnings temporarily because we need to reference @@ -456,6 +567,7 @@ void PrintUsageAndExit() { " [--benchmark_filter=]\n" " [--benchmark_min_time=]\n" " [--benchmark_repetitions=]\n" + " [--benchmark_enable_random_interleaving={true|false}]\n" " [--benchmark_report_aggregates_only={true|false}]\n" " [--benchmark_display_aggregates_only={true|false}]\n" " [--benchmark_format=]\n" @@ -476,10 +588,16 @@ void ParseCommandLineFlags(int* argc, char** argv) { if (ParseBoolFlag(argv[i], "benchmark_list_tests", &FLAGS_benchmark_list_tests) || ParseStringFlag(argv[i], "benchmark_filter", &FLAGS_benchmark_filter) || - ParseDoubleFlag(argv[i], "benchmark_min_time", - &FLAGS_benchmark_min_time) || - ParseInt32Flag(argv[i], "benchmark_repetitions", - &FLAGS_benchmark_repetitions) || + ParseDoubleFlag( + argv[i], "benchmark_min_time", + &do_not_read_flag_directly::FLAGS_benchmark_min_time) || + ParseInt32Flag( + argv[i], "benchmark_repetitions", + &do_not_read_flag_directly::FLAGS_benchmark_repetitions) || + ParseBoolFlag(argv[i], "benchmark_enable_random_interleaving", + &FLAGS_benchmark_enable_random_interleaving) || + ParseDoubleFlag(argv[i], "benchmark_random_interleaving_max_overhead", + &FLAGS_benchmark_random_interleaving_max_overhead) || ParseBoolFlag(argv[i], "benchmark_report_aggregates_only", &FLAGS_benchmark_report_aggregates_only) || ParseBoolFlag(argv[i], "benchmark_display_aggregates_only", diff --git a/src/benchmark_adjust_repetitions.cc b/src/benchmark_adjust_repetitions.cc new file mode 100644 index 0000000000..2847927628 --- /dev/null +++ b/src/benchmark_adjust_repetitions.cc @@ -0,0 +1,125 @@ +// Copyright 2015 Google Inc. All rights reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "benchmark_adjust_repetitions.h" + +#include "benchmark_api_internal.h" +#include "log.h" + +namespace benchmark { +namespace internal { + +namespace { + +constexpr double kNanosecondInSecond = 1e-9; + +} // namespace + +int ComputeRandomInterleavingRepetitions( + InternalRandomInterleavingRepetitionsInput input) { + // Find the repetitions such that total overhead is bounded. Let + // n = desired number of repetitions, i.e., the output of this method. + // t = total real execution time per repetition including overhead, + // (input.total_execution_time_per_repetition). + // o = maximum allowed increase in total real execution time due to random + // interleaving, measured as a fraction (input.max_overhead). + // e = estimated total execution time without Random Interleaving + // We want + // t * n / e <= 1 + o + // I.e., + // n <= (1 + o) * e / t + // + // Let + // h = overhead per repetition, which include all setup / teardown time and + // also the execution time of preliminary trials used to search for the + // correct number of iterations. + // r = real execution time per repetition not including overhead + // (input.real_accumulated_time_per_repetition). + // s = measured execution time per repetition not including overhead, + // which can be either real or CPU time + // (input.accumulated_time_per_repetition). + // We have + // h = t - r + // + // Let + // m = total minimum measured execution time for all repetitions + // (input.min_time_per_repetition * input.max_repetitions). + // Let + // f = m / s + // f is the scale factor between m and s, and will be used to estimate + // l, the total real execution time for all repetitions excluding the + // overhead. It's reasonable to assume that the real execution time excluding + // the overhead is proportional to the measured time. Hence we expect to see + // l / r to be equal to m / s. That is, l / r = f, thus, l = r * f. Then the + // total execution time e can be estimated by h + l, which is h + r * f. + // e = h + r * f + // Note that this might be an underestimation. If number of repetitions is + // reduced, we may need to run more iterations per repetition, and that may + // increase the number of preliminary trials needed to find the correct + // number of iterations. + + double h = std::max(0.0, input.total_execution_time_per_repetition - + input.real_time_used_per_repetition); + double r = + std::max(input.real_time_used_per_repetition, kNanosecondInSecond); + double s = + std::max(input.time_used_per_repetition, kNanosecondInSecond); + double m = input.min_time_per_repetition * input.max_repetitions; + + // f = m / s + // RunBenchmark() always overshoot the iteration count by kSafetyMultiplier. + // Apply the same factor here. + // f = kSafetyMultiplier * m / s + // Also we want to make sure 1 <= f <= input.max_repetitions. Note that we + // may not be able to reach m because the total iters per repetition is + // upper bounded by --benchmark_max_iters. This behavior is preserved in + // Random Interleaving, as we won't run repetitions more than + // input.max_repetitions to reach m. + + double f = kSafetyMultiplier * m / s; + f = std::min(std::max(f, 1.0), static_cast(input.max_repetitions)); + + double e = h + r * f; + // n <= (1 + o) * e / t = (1 + o) * e / (h + r) + // Also we want to make sure 1 <= n <= input.max_repetition, and (h + r) > 0. + double n = (1 + input.max_overhead) * e / (h + r); + n = std::min(std::max(n, 1.0), static_cast(input.max_repetitions)); + + int n_int = static_cast(n); + + VLOG(2) << "Computed random interleaving repetitions" + << "\n input.total_execution_time_per_repetition: " + << input.total_execution_time_per_repetition + << "\n input.time_used_per_repetition: " + << input.time_used_per_repetition + << "\n input.real_time_used_per_repetition: " + << input.real_time_used_per_repetition + << "\n input.min_time_per_repetitions: " + << input.min_time_per_repetition + << "\n input.max_repetitions: " << input.max_repetitions + << "\n input.max_overhead: " << input.max_overhead + << "\n h: " << h + << "\n r: " << r + << "\n s: " << s + << "\n f: " << f + << "\n m: " << m + << "\n e: " << e + << "\n n: " << n + << "\n n_int: " << n_int; + + return n_int; +} + +} // internal +} // benchmark diff --git a/src/benchmark_adjust_repetitions.h b/src/benchmark_adjust_repetitions.h new file mode 100644 index 0000000000..21a666afe0 --- /dev/null +++ b/src/benchmark_adjust_repetitions.h @@ -0,0 +1,42 @@ +// Copyright 2015 Google Inc. All rights reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#ifndef BENCHMARK_ADJUST_REPETITIONS_H +#define BENCHMARK_ADJUST_REPETITIONS_H + +#include "benchmark/benchmark.h" +#include "commandlineflags.h" + +namespace benchmark { +namespace internal { + +// Defines the input tuple to ComputeRandomInterleavingRepetitions(). +struct InternalRandomInterleavingRepetitionsInput { + double total_execution_time_per_repetition; + double time_used_per_repetition; + double real_time_used_per_repetition; + double min_time_per_repetition; + double max_overhead; + int max_repetitions; +}; + +// Should be called right after the first repetition is completed to estimate +// the number of iterations. +int ComputeRandomInterleavingRepetitions( + InternalRandomInterleavingRepetitionsInput input); + +} // end namespace internal +} // end namespace benchmark + +#endif // BENCHMARK_ADJUST_REPETITIONS_H diff --git a/src/benchmark_api_internal.cc b/src/benchmark_api_internal.cc index 553ff44e5c..ddd46bee63 100644 --- a/src/benchmark_api_internal.cc +++ b/src/benchmark_api_internal.cc @@ -2,8 +2,11 @@ #include +#include "check.h" #include "string_util.h" +DECLARE_bool(benchmark_enable_random_interleaving); + namespace benchmark { namespace internal { @@ -21,9 +24,12 @@ BenchmarkInstance::BenchmarkInstance(Benchmark* benchmark, complexity_lambda_(benchmark_.complexity_lambda_), statistics_(benchmark_.statistics_), repetitions_(benchmark_.repetitions_), - min_time_(benchmark_.min_time_), + min_time_(!IsZero(benchmark_.min_time_) ? benchmark_.min_time_ + : GetMinTime()), iterations_(benchmark_.iterations_), threads_(thread_count) { + CHECK(!IsZero(min_time_)) << "min_time must be non-zero."; + name_.function_name = benchmark_.name_; size_t arg_i = 0; @@ -77,6 +83,32 @@ BenchmarkInstance::BenchmarkInstance(Benchmark* benchmark, } } +double BenchmarkInstance::MinTime() const { + if (FLAGS_benchmark_enable_random_interleaving) { + // Random Interleaving will automatically adjust + // random_interleaving_repetitions(). Dividing + // total execution time by random_interleaving_repetitions() gives + // the adjusted min_time per repetition. + return min_time_ * GetRepetitions() / RandomInterleavingRepetitions(); + } + return min_time_; +} + +int BenchmarkInstance::RandomInterleavingRepetitions() const { + return random_interleaving_repetitions_ < 0 + ? GetRepetitions() + : random_interleaving_repetitions_; +} + +bool BenchmarkInstance::RandomInterleavingRepetitionsInitialized() const { + return random_interleaving_repetitions_ >= 0; +} + +void BenchmarkInstance::InitRandomInterleavingRepetitions( + int reps) const { + random_interleaving_repetitions_ = reps; +} + State BenchmarkInstance::Run( IterationCount iters, int thread_id, internal::ThreadTimer* timer, internal::ThreadManager* manager, diff --git a/src/benchmark_api_internal.h b/src/benchmark_api_internal.h index b0595e5a12..0ff8dafbe6 100644 --- a/src/benchmark_api_internal.h +++ b/src/benchmark_api_internal.h @@ -1,6 +1,10 @@ #ifndef BENCHMARK_API_INTERNAL_H #define BENCHMARK_API_INTERNAL_H +#include "benchmark/benchmark.h" +#include "commandlineflags.h" + +#include #include #include #include @@ -14,12 +18,25 @@ namespace benchmark { namespace internal { +extern const double kSafetyMultiplier; + // Information kept per benchmark we may want to run class BenchmarkInstance { public: BenchmarkInstance(Benchmark* benchmark, const std::vector& args, int threads); + // Returns number of repetitions for Random Interleaving. This will be + // initialized later once we finish the first repetition, if Random + // Interleaving is enabled. See also ComputeRandominterleavingrepetitions(). + int RandomInterleavingRepetitions() const; + + // Returns true if repetitions for Random Interleaving is initialized. + bool RandomInterleavingRepetitionsInitialized() const; + + // Initializes number of repetitions for random interleaving. + void InitRandomInterleavingRepetitions(int reps) const; + const BenchmarkName& name() const { return name_; } AggregationReportMode aggregation_report_mode() const { return aggregation_report_mode_; @@ -32,7 +49,7 @@ class BenchmarkInstance { BigOFunc& complexity_lambda() const { return *complexity_lambda_; } const std::vector& statistics() const { return statistics_; } int repetitions() const { return repetitions_; } - double min_time() const { return min_time_; } + double MinTime() const; IterationCount iterations() const { return iterations_; } int threads() const { return threads_; } @@ -53,12 +70,13 @@ class BenchmarkInstance { bool use_manual_time_; BigO complexity_; BigOFunc* complexity_lambda_; - UserCounters counters_; - const std::vector& statistics_; + std::vector statistics_; int repetitions_; double min_time_; IterationCount iterations_; - int threads_; // Number of concurrent threads to us + int threads_; + UserCounters counters_; + mutable int random_interleaving_repetitions_ = -1; }; bool FindBenchmarksInternal(const std::string& re, @@ -69,6 +87,10 @@ bool IsZero(double n); ConsoleReporter::OutputOptions GetOutputOptions(bool force_no_color = false); +double GetMinTime(); + +int GetRepetitions(); + } // end namespace internal } // end namespace benchmark diff --git a/src/benchmark_runner.cc b/src/benchmark_runner.cc index 3cffbbfad5..330cb4494b 100644 --- a/src/benchmark_runner.cc +++ b/src/benchmark_runner.cc @@ -15,6 +15,7 @@ #include "benchmark_runner.h" #include "benchmark/benchmark.h" #include "benchmark_api_internal.h" +#include "benchmark_adjust_repetitions.h" #include "internal_macros.h" #ifndef BENCHMARK_OS_WINDOWS @@ -52,6 +53,9 @@ #include "thread_manager.h" #include "thread_timer.h" +DECLARE_bool(benchmark_enable_random_interleaving); +DECLARE_double(benchmark_random_interleaving_max_overhead); + namespace benchmark { namespace internal { @@ -67,7 +71,7 @@ BenchmarkReporter::Run CreateRunReport( const internal::ThreadManager::Result& results, IterationCount memory_iterations, const MemoryManager::Result& memory_result, double seconds, - int64_t repetition_index) { + int repetition_index) { // Create report about this benchmark run. BenchmarkReporter::Run report; @@ -138,12 +142,16 @@ void RunInThread(const BenchmarkInstance* b, IterationCount iters, class BenchmarkRunner { public: BenchmarkRunner(const benchmark::internal::BenchmarkInstance& b_, - std::vector* complexity_reports_) + int outer_repetitions_, + int inner_repetitions_, + std::vector* complexity_reports_, + RunResults* run_results_) : b(b_), complexity_reports(*complexity_reports_), - min_time(!IsZero(b.min_time()) ? b.min_time() : FLAGS_benchmark_min_time), - repeats(b.repetitions() != 0 ? b.repetitions() - : FLAGS_benchmark_repetitions), + run_results(*run_results_), + outer_repetitions(outer_repetitions_), + inner_repetitions(inner_repetitions_), + repeats(b.repetitions() != 0 ? b.repetitions() : inner_repetitions), has_explicit_iteration_count(b.iterations() != 0), pool(b.threads() - 1), iters(has_explicit_iteration_count ? b.iterations() : 1), @@ -163,6 +171,7 @@ class BenchmarkRunner { internal::ARM_DisplayReportAggregatesOnly); run_results.file_report_aggregates_only = (b.aggregation_report_mode() & internal::ARM_FileReportAggregatesOnly); + CHECK(FLAGS_benchmark_perf_counters.empty() || perf_counters_measurement.IsValid()) << "Perf counters were requested but could not be set up."; @@ -179,21 +188,20 @@ class BenchmarkRunner { if ((b.complexity() != oNone) && b.last_benchmark_instance) { auto additional_run_stats = ComputeBigO(complexity_reports); run_results.aggregates_only.insert(run_results.aggregates_only.end(), - additional_run_stats.begin(), - additional_run_stats.end()); + additional_run_stats.begin(), + additional_run_stats.end()); complexity_reports.clear(); } } - RunResults&& get_results() { return std::move(run_results); } - private: - RunResults run_results; - const benchmark::internal::BenchmarkInstance& b; std::vector& complexity_reports; - const double min_time; + RunResults& run_results; + + const int outer_repetitions; + const int inner_repetitions; const int repeats; const bool has_explicit_iteration_count; @@ -268,13 +276,14 @@ class BenchmarkRunner { IterationCount PredictNumItersNeeded(const IterationResults& i) const { // See how much iterations should be increased by. // Note: Avoid division by zero with max(seconds, 1ns). - double multiplier = min_time * 1.4 / std::max(i.seconds, 1e-9); + double multiplier = + b.MinTime() * kSafetyMultiplier / std::max(i.seconds, 1e-9); // If our last run was at least 10% of FLAGS_benchmark_min_time then we // use the multiplier directly. // Otherwise we use at most 10 times expansion. // NOTE: When the last run was at least 10% of the min time the max // expansion should be 14x. - bool is_significant = (i.seconds / min_time) > 0.1; + bool is_significant = (i.seconds / b.MinTime()) > 0.1; multiplier = is_significant ? multiplier : std::min(10.0, multiplier); if (multiplier <= 1.0) multiplier = 2.0; @@ -295,14 +304,17 @@ class BenchmarkRunner { // or because an error was reported. return i.results.has_error_ || i.iters >= kMaxIterations || // Too many iterations already. - i.seconds >= min_time || // The elapsed time is large enough. - // CPU time is specified but the elapsed real time greatly exceeds - // the minimum time. - // Note that user provided timers are except from this sanity check. - ((i.results.real_time_used >= 5 * min_time) && !b.use_manual_time()); + i.seconds >= b.MinTime() || // The elapsed time is large enough. + // CPU time is specified but the + // elapsed real time greatly exceeds + // the minimum time. Note that user + // provided timers are except from this + // sanity check. + ((i.results.real_time_used >= 5 * b.MinTime()) && + !b.use_manual_time()); } - void DoOneRepetition(int64_t repetition_index) { + void DoOneRepetition(int repetition_index) { const bool is_the_first_repetition = repetition_index == 0; IterationResults i; @@ -312,8 +324,10 @@ class BenchmarkRunner { // Please do note that the if there are repetitions, the iteration count // is *only* calculated for the *first* repetition, and other repetitions // simply use that precomputed iteration count. + const auto exec_start = benchmark::ChronoClockNow(); for (;;) { i = DoNIterations(); + const auto exec_end = benchmark::ChronoClockNow(); // Do we consider the results to be significant? // If we are doing repetitions, and the first repetition was already done, @@ -324,7 +338,38 @@ class BenchmarkRunner { has_explicit_iteration_count || ShouldReportIterationResults(i); - if (results_are_significant) break; // Good, let's report them! + if (results_are_significant) { + // The number of repetitions for random interleaving may be reduced + // to limit the increase in benchmark execution time. When this happens + // the target execution time for each repetition is increased. We may + // need to rerun trials to calculate iters according to the increased + // target execution time. + bool rerun_trial = false; + // If random interleaving is enabled and the repetitions is not + // initialized, do it now. + if (FLAGS_benchmark_enable_random_interleaving && + !b.RandomInterleavingRepetitionsInitialized()) { + InternalRandomInterleavingRepetitionsInput input; + input.total_execution_time_per_repetition = exec_end - exec_start; + input.time_used_per_repetition = i.seconds; + input.real_time_used_per_repetition = i.results.real_time_used; + input.min_time_per_repetition = GetMinTime(); + input.max_overhead = FLAGS_benchmark_random_interleaving_max_overhead; + input.max_repetitions = GetRepetitions(); + b.InitRandomInterleavingRepetitions( + ComputeRandomInterleavingRepetitions(input)); + // If the number of repetitions changed, need to rerun the last trial + // because iters may also change. Note that we only need to do this + // if accumulated_time < b.MinTime(), i.e., the iterations we have + // run is not enough for the already adjusted b.MinTime(). + // Otherwise, we will still skip the rerun. + rerun_trial = + b.RandomInterleavingRepetitions() < GetRepetitions() && + i.seconds < b.MinTime() && !has_explicit_iteration_count; + } + + if (!rerun_trial) break; // Good, let's report them! + } // Nope, bad iteration. Let's re-estimate the hopefully-sufficient // iteration count, and run the benchmark again... @@ -341,7 +386,8 @@ class BenchmarkRunner { if (memory_manager != nullptr) { // Only run a few iterations to reduce the impact of one-time // allocations in benchmarks that are not properly managed. - memory_iterations = std::min(16, iters); + memory_iterations = std::min( + 16 / outer_repetitions + (16 % outer_repetitions != 0), iters); memory_manager->Start(); std::unique_ptr manager; manager.reset(new internal::ThreadManager(1)); @@ -367,11 +413,12 @@ class BenchmarkRunner { } // end namespace -RunResults RunBenchmark( - const benchmark::internal::BenchmarkInstance& b, - std::vector* complexity_reports) { - internal::BenchmarkRunner r(b, complexity_reports); - return r.get_results(); +void RunBenchmark(const benchmark::internal::BenchmarkInstance& b, + const int outer_repetitions, const int inner_repetitions, + std::vector* complexity_reports, + RunResults* run_results) { + internal::BenchmarkRunner r(b, outer_repetitions, inner_repetitions, + complexity_reports, run_results); } } // end namespace internal diff --git a/src/benchmark_runner.h b/src/benchmark_runner.h index 9b0cf2a64e..e29aa32306 100644 --- a/src/benchmark_runner.h +++ b/src/benchmark_runner.h @@ -42,9 +42,10 @@ struct RunResults { bool file_report_aggregates_only = false; }; -RunResults RunBenchmark( - const benchmark::internal::BenchmarkInstance& b, - std::vector* complexity_reports); +void RunBenchmark(const benchmark::internal::BenchmarkInstance& b, + int outer_repetitions, int inner_repetitions, + std::vector* complexity_reports, + RunResults* run_results); } // namespace internal diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt index 1e7b6829f2..8945f5005f 100644 --- a/test/CMakeLists.txt +++ b/test/CMakeLists.txt @@ -196,6 +196,7 @@ if (BENCHMARK_ENABLE_GTEST_TESTS) add_gtest(benchmark_gtest) add_gtest(benchmark_name_gtest) + add_gtest(benchmark_random_interleaving_gtest) add_gtest(commandlineflags_gtest) add_gtest(statistics_gtest) add_gtest(string_util_gtest) diff --git a/test/benchmark_random_interleaving_gtest.cc b/test/benchmark_random_interleaving_gtest.cc new file mode 100644 index 0000000000..5e8329a4e6 --- /dev/null +++ b/test/benchmark_random_interleaving_gtest.cc @@ -0,0 +1,271 @@ +#include +#include +#include + +#include "../src/benchmark_adjust_repetitions.h" +#include "../src/string_util.h" +#include "benchmark/benchmark.h" +#include "gmock/gmock.h" +#include "gtest/gtest.h" + +DECLARE_bool(benchmark_enable_random_interleaving); +DECLARE_string(benchmark_filter); +DECLARE_double(benchmark_random_interleaving_max_overhead); + +namespace do_not_read_flag_directly { +DECLARE_int32(benchmark_repetitions); +} // namespace do_not_read_flag_directly + +namespace benchmark { +namespace internal { +namespace { + +class EventQueue : public std::queue { + public: + void Put(const std::string& event) { + push(event); + } + + void Clear() { + while (!empty()) { + pop(); + } + } + + std::string Get() { + std::string event = front(); + pop(); + return event; + } +}; + +static EventQueue* queue = new EventQueue; + +class NullReporter : public BenchmarkReporter { + public: + bool ReportContext(const Context& /*context*/) override { + return true; + } + void ReportRuns(const std::vector& /* report */) override {} +}; + +class BenchmarkTest : public testing::Test { + public: + static void SetupHook(int /* num_threads */) { queue->push("Setup"); } + + static void TeardownHook(int /* num_threads */) { queue->push("Teardown"); } + + void Execute(const std::string& pattern) { + queue->Clear(); + + BenchmarkReporter* reporter = new NullReporter; + FLAGS_benchmark_filter = pattern; + RunSpecifiedBenchmarks(reporter); + delete reporter; + + queue->Put("DONE"); // End marker + } +}; + +static void BM_Match1(benchmark::State& state) { + const int64_t arg = state.range(0); + + for (auto _ : state) {} + queue->Put(StrFormat("BM_Match1/%d", static_cast(arg))); +} +BENCHMARK(BM_Match1) + ->Iterations(100) + ->Arg(1) + ->Arg(2) + ->Arg(3) + ->Range(10, 80) + ->Args({90}) + ->Args({100}); + +static void BM_MatchOverhead(benchmark::State& state) { + const int64_t arg = state.range(0); + + for (auto _ : state) {} + queue->Put(StrFormat("BM_MatchOverhead/%d", static_cast(arg))); +} +BENCHMARK(BM_MatchOverhead) + ->Iterations(100) + ->Arg(64) + ->Arg(80); + +TEST_F(BenchmarkTest, Match1) { + Execute("BM_Match1"); + ASSERT_EQ("BM_Match1/1", queue->Get()); + ASSERT_EQ("BM_Match1/2", queue->Get()); + ASSERT_EQ("BM_Match1/3", queue->Get()); + ASSERT_EQ("BM_Match1/10", queue->Get()); + ASSERT_EQ("BM_Match1/64", queue->Get()); + ASSERT_EQ("BM_Match1/80", queue->Get()); + ASSERT_EQ("BM_Match1/90", queue->Get()); + ASSERT_EQ("BM_Match1/100", queue->Get()); + ASSERT_EQ("DONE", queue->Get()); +} + +TEST_F(BenchmarkTest, Match1WithRepetition) { + do_not_read_flag_directly::FLAGS_benchmark_repetitions = 2; + + Execute("BM_Match1/(64|80)"); + ASSERT_EQ("BM_Match1/64", queue->Get()); + ASSERT_EQ("BM_Match1/64", queue->Get()); + ASSERT_EQ("BM_Match1/80", queue->Get()); + ASSERT_EQ("BM_Match1/80", queue->Get()); + ASSERT_EQ("DONE", queue->Get()); +} + +TEST_F(BenchmarkTest, Match1WithRandomInterleaving) { + FLAGS_benchmark_enable_random_interleaving = true; + do_not_read_flag_directly::FLAGS_benchmark_repetitions = 100; + FLAGS_benchmark_random_interleaving_max_overhead = + std::numeric_limits::infinity(); + + std::vector expected({"BM_Match1/64", "BM_Match1/80"}); + std::map interleaving_count; + Execute("BM_Match1/(64|80)"); + for (int i = 0; i < 100; ++i) { + std::vector interleaving; + interleaving.push_back(queue->Get()); + interleaving.push_back(queue->Get()); + EXPECT_THAT(interleaving, testing::UnorderedElementsAreArray(expected)); + interleaving_count[StrFormat("%s,%s", interleaving[0].c_str(), + interleaving[1].c_str())]++; + } + EXPECT_GE(interleaving_count.size(), 2) << "Interleaving was not randomized."; + ASSERT_EQ("DONE", queue->Get()); +} + +TEST_F(BenchmarkTest, Match1WithRandomInterleavingAndZeroOverhead) { + FLAGS_benchmark_enable_random_interleaving = true; + do_not_read_flag_directly::FLAGS_benchmark_repetitions = 100; + FLAGS_benchmark_random_interleaving_max_overhead = 0; + + // ComputeRandomInterleavingRepetitions() will kick in and rerun each + // benchmark once with increased iterations. Then number of repetitions will + // be reduced to < 100. The first 4 executions should be + // 2 x BM_MatchOverhead/64 and 2 x BM_MatchOverhead/80. + std::vector expected( + {"BM_MatchOverhead/64", "BM_MatchOverhead/80", "BM_MatchOverhead/64", + "BM_MatchOverhead/80"}); + std::map interleaving_count; + Execute("BM_MatchOverhead/(64|80)"); + std::vector interleaving; + interleaving.push_back(queue->Get()); + interleaving.push_back(queue->Get()); + interleaving.push_back(queue->Get()); + interleaving.push_back(queue->Get()); + EXPECT_THAT(interleaving, testing::UnorderedElementsAreArray(expected)); + ASSERT_LT(queue->size(), 100) << "# Repetitions was not reduced to < 100."; +} + +InternalRandomInterleavingRepetitionsInput CreateInput( + double total, double time, double real_time, double min_time, + double overhead, int repetitions) { + InternalRandomInterleavingRepetitionsInput input; + input.total_execution_time_per_repetition = total; + input.time_used_per_repetition = time; + input.real_time_used_per_repetition = real_time; + input.min_time_per_repetition = min_time; + input.max_overhead = overhead; + input.max_repetitions = repetitions; + return input; +} + +TEST(Benchmark, ComputeRandomInterleavingRepetitions) { + // On wall clock time. + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.05, 0.05, 0.05, 0.05, 0.0, 10)), + 10); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.05, 0.05, 0.05, 0.05, 0.4, 10)), + 10); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.06, 0.05, 0.05, 0.05, 0.0, 10)), + 8); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.06, 0.05, 0.05, 0.05, 0.4, 10)), + 10); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.08, 0.05, 0.05, 0.05, 0.0, 10)), + 6); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.08, 0.05, 0.05, 0.05, 0.4, 10)), + 9); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.26, 0.25, 0.25, 0.05, 0.0, 10)), + 2); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.25, 0.25, 0.25, 0.05, 0.4, 10)), + 3); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.26, 0.25, 0.25, 0.05, 0.0, 10)), + 2); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.26, 0.25, 0.25, 0.05, 0.4, 10)), + 3); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.38, 0.25, 0.25, 0.05, 0.0, 10)), + 2); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.38, 0.25, 0.25, 0.05, 0.4, 10)), + 3); + + // On CPU time. + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.1, 0.05, 0.1, 0.05, 0.0, 10)), + 10); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.1, 0.05, 0.1, 0.05, 0.4, 10)), + 10); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.11, 0.05, 0.1, 0.05, 0.0, 10)), + 9); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.11, 0.05, 0.1, 0.05, 0.4, 10)), + 10); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.15, 0.05, 0.1, 0.05, 0.0, 10)), + 7); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.15, 0.05, 0.1, 0.05, 0.4, 10)), + 9); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.5, 0.25, 0.5, 0.05, 0.0, 10)), + 2); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.5, 0.25, 0.5, 0.05, 0.4, 10)), + 3); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.51, 0.25, 0.5, 0.05, 0.0, 10)), + 2); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.51, 0.25, 0.5, 0.05, 0.4, 10)), + 3); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.8, 0.25, 0.5, 0.05, 0.0, 10)), + 2); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.8, 0.25, 0.5, 0.05, 0.4, 10)), + 2); + + // Corner cases. + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.0, 0.25, 0.5, 0.05, 0.4, 10)), + 3); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.8, 0.0, 0.5, 0.05, 0.4, 10)), + 9); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.8, 0.25, 0.0, 0.05, 0.4, 10)), + 1); + EXPECT_EQ(ComputeRandomInterleavingRepetitions( + CreateInput(0.8, 0.25, 0.5, 0.0, 0.4, 10)), + 1); +} + +} // namespace +} // namespace internal +} // namespace benchmark