simdjson 0.2.0 → 0.3.0

data/vendor/simdjson/dependencies/json/benchmarks/thirdparty/benchmark/cmake/llvm-toolchain.cmake

@@ -0,0 +1,8 @@
+find_package(LLVMAr REQUIRED)
+set(CMAKE_AR "${LLVMAR_EXECUTABLE}" CACHE FILEPATH "" FORCE)
+
+find_package(LLVMNm REQUIRED)
+set(CMAKE_NM "${LLVMNM_EXECUTABLE}" CACHE FILEPATH "" FORCE)
+
+find_package(LLVMRanLib REQUIRED)
+set(CMAKE_RANLIB "${LLVMRANLIB_EXECUTABLE}" CACHE FILEPATH "" FORCE)

data/vendor/simdjson/dependencies/json/benchmarks/thirdparty/benchmark/cmake/posix_regex.cpp

@@ -0,0 +1,14 @@
+#include <regex.h>
+#include <string>
+int main() {
+  std::string str = "test0159";
+  regex_t re;
+  int ec = regcomp(&re, "^[a-z]+[0-9]+$", REG_EXTENDED | REG_NOSUB);
+  if (ec != 0) {
+    return ec;
+  }
+  int ret = regexec(&re, str.c_str(), 0, nullptr, 0) ? -1 : 0;
+  regfree(&re);
+  return ret;
+}
+

data/vendor/simdjson/dependencies/json/benchmarks/thirdparty/benchmark/cmake/split_list.cmake

@@ -0,0 +1,3 @@
+macro(split_list listname)
+  string(REPLACE ";" " " ${listname} "${${listname}}")
+endmacro()

data/vendor/simdjson/dependencies/json/benchmarks/thirdparty/benchmark/cmake/std_regex.cpp

@@ -0,0 +1,10 @@
+#include <regex>
+#include <string>
+int main() {
+  const std::string str = "test0159";
+  std::regex re;
+  re = std::regex("^[a-z]+[0-9]+$",
+       std::regex_constants::extended | std::regex_constants::nosubs);
+  return std::regex_search(str, re) ? 0 : -1;
+}
+

data/vendor/simdjson/dependencies/json/benchmarks/thirdparty/benchmark/cmake/steady_clock.cpp

@@ -0,0 +1,7 @@
+#include <chrono>
+
+int main() {
+    typedef std::chrono::steady_clock Clock;
+    Clock::time_point tp = Clock::now();
+    ((void)tp);
+}

data/vendor/simdjson/dependencies/json/benchmarks/thirdparty/benchmark/cmake/thread_safety_attributes.cpp

@@ -0,0 +1,4 @@
+#define HAVE_THREAD_SAFETY_ATTRIBUTES
+#include "../src/mutex.h"
+
+int main() {}

data/vendor/simdjson/dependencies/json/benchmarks/thirdparty/benchmark/docs/AssemblyTests.md

@@ -0,0 +1,147 @@
+# Assembly Tests
+
+The Benchmark library provides a number of functions whose primary
+purpose in to affect assembly generation, including `DoNotOptimize`
+and `ClobberMemory`. In addition there are other functions,
+such as `KeepRunning`, for which generating good assembly is paramount.
+
+For these functions it's important to have tests that verify the
+correctness and quality of the implementation. This requires testing
+the code generated by the compiler.
+
+This document describes how the Benchmark library tests compiler output,
+as well as how to properly write new tests.
+
+
+## Anatomy of a Test
+
+Writing a test has two steps:
+
+* Write the code you want to generate assembly for.
+* Add `// CHECK` lines to match against the verified assembly.
+
+Example:
+```c++
+
+// CHECK-LABEL: test_add:
+extern "C" int test_add() {
+    extern int ExternInt;
+    return ExternInt + 1;
+
+    // CHECK: movl ExternInt(%rip), %eax
+    // CHECK: addl %eax
+    // CHECK: ret
+}
+
+```
+
+#### LLVM Filecheck
+
+[LLVM's Filecheck](https://llvm.org/docs/CommandGuide/FileCheck.html)
+is used to test the generated assembly against the `// CHECK` lines
+specified in the tests source file. Please see the documentation
+linked above for information on how to write `CHECK` directives.
+
+#### Tips and Tricks:
+
+* Tests should match the minimal amount of output required to establish
+correctness. `CHECK` directives don't have to match on the exact next line
+after the previous match, so tests should omit checks for unimportant
+bits of assembly. ([`CHECK-NEXT`](https://llvm.org/docs/CommandGuide/FileCheck.html#the-check-next-directive)
+can be used to ensure a match occurs exactly after the previous match).
+
+* The tests are compiled with `-O3 -g0`. So we're only testing the
+optimized output.
+
+* The assembly output is further cleaned up using `tools/strip_asm.py`.
+This removes comments, assembler directives, and unused labels before
+the test is run.
+
+* The generated and stripped assembly file for a test is output under
+`<build-directory>/test/<test-name>.s`
+
+* Filecheck supports using [`CHECK` prefixes](https://llvm.org/docs/CommandGuide/FileCheck.html#cmdoption-check-prefixes)
+to specify lines that should only match in certain situations.
+The Benchmark tests use `CHECK-CLANG` and `CHECK-GNU` for lines that
+are only expected to match Clang or GCC's output respectively. Normal
+`CHECK` lines match against all compilers. (Note: `CHECK-NOT` and
+`CHECK-LABEL` are NOT prefixes. They are versions of non-prefixed
+`CHECK` lines)
+
+* Use `extern "C"` to disable name mangling for specific functions. This
+makes them easier to name in the `CHECK` lines.
+
+
+## Problems Writing Portable Tests
+
+Writing tests which check the code generated by a compiler are
+inherently non-portable. Different compilers and even different compiler
+versions may generate entirely different code. The Benchmark tests
+must tolerate this.
+
+LLVM Filecheck provides a number of mechanisms to help write
+"more portable" tests; including [matching using regular expressions](https://llvm.org/docs/CommandGuide/FileCheck.html#filecheck-pattern-matching-syntax),
+allowing the creation of [named variables](https://llvm.org/docs/CommandGuide/FileCheck.html#filecheck-variables)
+for later matching, and [checking non-sequential matches](https://llvm.org/docs/CommandGuide/FileCheck.html#the-check-dag-directive).
+
+#### Capturing Variables
+
+For example, say GCC stores a variable in a register but Clang stores
+it in memory. To write a test that tolerates both cases we "capture"
+the destination of the store, and then use the captured expression
+to write the remainder of the test.
+
+```c++
+// CHECK-LABEL: test_div_no_op_into_shr:
+extern "C" void test_div_no_op_into_shr(int value) {
+    int divisor = 2;
+    benchmark::DoNotOptimize(divisor); // hide the value from the optimizer
+    return value / divisor;
+
+    // CHECK: movl $2, [[DEST:.*]]
+    // CHECK: idivl [[DEST]]
+    // CHECK: ret
+}
+```
+
+#### Using Regular Expressions to Match Differing Output
+
+Often tests require testing assembly lines which may subtly differ
+between compilers or compiler versions. A common example of this
+is matching stack frame addresses. In this case regular expressions
+can be used to match the differing bits of output. For example:
+
+```c++
+int ExternInt;
+struct Point { int x, y, z; };
+
+// CHECK-LABEL: test_store_point:
+extern "C" void test_store_point() {
+    Point p{ExternInt, ExternInt, ExternInt};
+    benchmark::DoNotOptimize(p);
+
+    // CHECK: movl ExternInt(%rip), %eax
+    // CHECK: movl %eax, -{{[0-9]+}}(%rsp)
+    // CHECK: movl %eax, -{{[0-9]+}}(%rsp)
+    // CHECK: movl %eax, -{{[0-9]+}}(%rsp)
+    // CHECK: ret
+}
+```
+
+## Current Requirements and Limitations
+
+The tests require Filecheck to be installed along the `PATH` of the
+build machine. Otherwise the tests will be disabled.
+
+Additionally, as mentioned in the previous section, codegen tests are
+inherently non-portable. Currently the tests are limited to:
+
+* x86_64 targets.
+* Compiled with GCC or Clang
+
+Further work could be done, at least on a limited basis, to extend the
+tests to other architectures and compilers (using `CHECK` prefixes).
+
+Furthermore, the tests fail for builds which specify additional flags
+that modify code generation, including `--coverage` or `-fsanitize=`.
+

data/vendor/simdjson/dependencies/json/benchmarks/thirdparty/benchmark/docs/tools.md

@@ -0,0 +1,242 @@
+# Benchmark Tools
+
+## compare_bench.py
+
+The `compare_bench.py` utility which can be used to compare the result of benchmarks.
+The program is invoked like:
+
+``` bash
+$ compare_bench.py <old-benchmark> <new-benchmark> [benchmark options]...
+```
+
+Where `<old-benchmark>` and `<new-benchmark>` either specify a benchmark executable file, or a JSON output file. The type of the input file is automatically detected. If a benchmark executable is specified then the benchmark is run to obtain the results. Otherwise the results are simply loaded from the output file.
+
+`[benchmark options]` will be passed to the benchmarks invocations. They can be anything that binary accepts, be it either normal `--benchmark_*` parameters, or some custom parameters your binary takes.
+
+The sample output using the JSON test files under `Inputs/` gives:
+
+``` bash
+$ ./compare_bench.py ./gbench/Inputs/test1_run1.json ./gbench/Inputs/test1_run2.json
+Comparing ./gbench/Inputs/test1_run1.json to ./gbench/Inputs/test1_run2.json
+Benchmark                        Time             CPU      Time Old      Time New       CPU Old       CPU New
+-------------------------------------------------------------------------------------------------------------
+BM_SameTimes                  +0.0000         +0.0000            10            10            10            10
+BM_2xFaster                   -0.5000         -0.5000            50            25            50            25
+BM_2xSlower                   +1.0000         +1.0000            50           100            50           100
+BM_1PercentFaster             -0.0100         -0.0100           100            99           100            99
+BM_1PercentSlower             +0.0100         +0.0100           100           101           100           101
+BM_10PercentFaster            -0.1000         -0.1000           100            90           100            90
+BM_10PercentSlower            +0.1000         +0.1000           100           110           100           110
+BM_100xSlower                +99.0000        +99.0000           100         10000           100         10000
+BM_100xFaster                 -0.9900         -0.9900         10000           100         10000           100
+BM_10PercentCPUToTime         +0.1000         -0.1000           100           110           100            90
+BM_ThirdFaster                -0.3333         -0.3334           100            67           100            67
+BM_BadTimeUnit                -0.9000         +0.2000             0             0             0             1
+```
+
+As you can note, the values in `Time` and `CPU` columns are calculated as `(new - old) / |old|`.
+
+When a benchmark executable is run, the raw output from the benchmark is printed in real time to stdout. The sample output using `benchmark/basic_test` for both arguments looks like:
+
+```
+./compare_bench.py  test/basic_test test/basic_test  --benchmark_filter=BM_empty.*
+RUNNING: test/basic_test --benchmark_filter=BM_empty.* --benchmark_out=/tmp/tmpN7LF3a
+Run on (8 X 4000 MHz CPU s)
+2017-11-07 23:28:36
+---------------------------------------------------------------------
+Benchmark                              Time           CPU Iterations
+---------------------------------------------------------------------
+BM_empty                               4 ns          4 ns  170178757
+BM_empty/threads:8                     1 ns          7 ns  103868920
+BM_empty_stop_start                    0 ns          0 ns 1000000000
+BM_empty_stop_start/threads:8          0 ns          0 ns 1403031720
+RUNNING: /test/basic_test --benchmark_filter=BM_empty.* --benchmark_out=/tmp/tmplvrIp8
+Run on (8 X 4000 MHz CPU s)
+2017-11-07 23:28:38
+---------------------------------------------------------------------
+Benchmark                              Time           CPU Iterations
+---------------------------------------------------------------------
+BM_empty                               4 ns          4 ns  169534855
+BM_empty/threads:8                     1 ns          7 ns  104188776
+BM_empty_stop_start                    0 ns          0 ns 1000000000
+BM_empty_stop_start/threads:8          0 ns          0 ns 1404159424
+Comparing ../build/test/basic_test to ../build/test/basic_test
+Benchmark                                Time             CPU      Time Old      Time New       CPU Old       CPU New
+---------------------------------------------------------------------------------------------------------------------
+BM_empty                              -0.0048         -0.0049             4             4             4             4
+BM_empty/threads:8                    -0.0123         -0.0054             1             1             7             7
+BM_empty_stop_start                   -0.0000         -0.0000             0             0             0             0
+BM_empty_stop_start/threads:8         -0.0029         +0.0001             0             0             0             0
+
+```
+
+As you can note, the values in `Time` and `CPU` columns are calculated as `(new - old) / |old|`.
+Obviously this example doesn't give any useful output, but it's intended to show the output format when 'compare_bench.py' needs to run benchmarks.
+
+## compare.py
+
+The `compare.py` can be used to compare the result of benchmarks.
+There are three modes of operation:
+
+1. Just compare two benchmarks, what `compare_bench.py` did.
+The program is invoked like:
+
+``` bash
+$ compare.py benchmarks <benchmark_baseline> <benchmark_contender> [benchmark options]...
+```
+Where `<benchmark_baseline>` and `<benchmark_contender>` either specify a benchmark executable file, or a JSON output file. The type of the input file is automatically detected. If a benchmark executable is specified then the benchmark is run to obtain the results. Otherwise the results are simply loaded from the output file.
+
+`[benchmark options]` will be passed to the benchmarks invocations. They can be anything that binary accepts, be it either normal `--benchmark_*` parameters, or some custom parameters your binary takes.
+
+Example output:
+```
+$ ./compare.py benchmarks ./a.out ./a.out
+RUNNING: ./a.out --benchmark_out=/tmp/tmprBT5nW
+Run on (8 X 4000 MHz CPU s)
+2017-11-07 21:16:44
+------------------------------------------------------
+Benchmark               Time           CPU Iterations
+------------------------------------------------------
+BM_memcpy/8            36 ns         36 ns   19101577   211.669MB/s
+BM_memcpy/64           76 ns         76 ns    9412571   800.199MB/s
+BM_memcpy/512          84 ns         84 ns    8249070   5.64771GB/s
+BM_memcpy/1024        116 ns        116 ns    6181763   8.19505GB/s
+BM_memcpy/8192        643 ns        643 ns    1062855   11.8636GB/s
+BM_copy/8             222 ns        222 ns    3137987   34.3772MB/s
+BM_copy/64           1608 ns       1608 ns     432758   37.9501MB/s
+BM_copy/512         12589 ns      12589 ns      54806   38.7867MB/s
+BM_copy/1024        25169 ns      25169 ns      27713   38.8003MB/s
+BM_copy/8192       201165 ns     201112 ns       3486   38.8466MB/s
+RUNNING: ./a.out --benchmark_out=/tmp/tmpt1wwG_
+Run on (8 X 4000 MHz CPU s)
+2017-11-07 21:16:53
+------------------------------------------------------
+Benchmark               Time           CPU Iterations
+------------------------------------------------------
+BM_memcpy/8            36 ns         36 ns   19397903   211.255MB/s
+BM_memcpy/64           73 ns         73 ns    9691174   839.635MB/s
+BM_memcpy/512          85 ns         85 ns    8312329   5.60101GB/s
+BM_memcpy/1024        118 ns        118 ns    6438774   8.11608GB/s
+BM_memcpy/8192        656 ns        656 ns    1068644   11.6277GB/s
+BM_copy/8             223 ns        223 ns    3146977   34.2338MB/s
+BM_copy/64           1611 ns       1611 ns     435340   37.8751MB/s
+BM_copy/512         12622 ns      12622 ns      54818   38.6844MB/s
+BM_copy/1024        25257 ns      25239 ns      27779   38.6927MB/s
+BM_copy/8192       205013 ns     205010 ns       3479    38.108MB/s
+Comparing ./a.out to ./a.out
+Benchmark                 Time             CPU      Time Old      Time New       CPU Old       CPU New
+------------------------------------------------------------------------------------------------------
+BM_memcpy/8            +0.0020         +0.0020            36            36            36            36
+BM_memcpy/64           -0.0468         -0.0470            76            73            76            73
+BM_memcpy/512          +0.0081         +0.0083            84            85            84            85
+BM_memcpy/1024         +0.0098         +0.0097           116           118           116           118
+BM_memcpy/8192         +0.0200         +0.0203           643           656           643           656
+BM_copy/8              +0.0046         +0.0042           222           223           222           223
+BM_copy/64             +0.0020         +0.0020          1608          1611          1608          1611
+BM_copy/512            +0.0027         +0.0026         12589         12622         12589         12622
+BM_copy/1024           +0.0035         +0.0028         25169         25257         25169         25239
+BM_copy/8192           +0.0191         +0.0194        201165        205013        201112        205010
+```
+
+What it does is for the every benchmark from the first run it looks for the benchmark with exactly the same name in the second run, and then compares the results. If the names differ, the benchmark is omitted from the diff.
+As you can note, the values in `Time` and `CPU` columns are calculated as `(new - old) / |old|`.
+
+2. Compare two different filters of one benchmark
+The program is invoked like:
+
+``` bash
+$ compare.py filters <benchmark> <filter_baseline> <filter_contender> [benchmark options]...
+```
+Where `<benchmark>` either specify a benchmark executable file, or a JSON output file. The type of the input file is automatically detected. If a benchmark executable is specified then the benchmark is run to obtain the results. Otherwise the results are simply loaded from the output file.
+
+Where `<filter_baseline>` and `<filter_contender>` are the same regex filters that you would pass to the `[--benchmark_filter=<regex>]` parameter of the benchmark binary.
+
+`[benchmark options]` will be passed to the benchmarks invocations. They can be anything that binary accepts, be it either normal `--benchmark_*` parameters, or some custom parameters your binary takes.
+
+Example output:
+```
+$ ./compare.py filters ./a.out BM_memcpy BM_copy
+RUNNING: ./a.out --benchmark_filter=BM_memcpy --benchmark_out=/tmp/tmpBWKk0k
+Run on (8 X 4000 MHz CPU s)
+2017-11-07 21:37:28
+------------------------------------------------------
+Benchmark               Time           CPU Iterations
+------------------------------------------------------
+BM_memcpy/8            36 ns         36 ns   17891491   211.215MB/s
+BM_memcpy/64           74 ns         74 ns    9400999   825.646MB/s
+BM_memcpy/512          87 ns         87 ns    8027453   5.46126GB/s
+BM_memcpy/1024        111 ns        111 ns    6116853    8.5648GB/s
+BM_memcpy/8192        657 ns        656 ns    1064679   11.6247GB/s
+RUNNING: ./a.out --benchmark_filter=BM_copy --benchmark_out=/tmp/tmpAvWcOM
+Run on (8 X 4000 MHz CPU s)
+2017-11-07 21:37:33
+----------------------------------------------------
+Benchmark             Time           CPU Iterations
+----------------------------------------------------
+BM_copy/8           227 ns        227 ns    3038700   33.6264MB/s
+BM_copy/64         1640 ns       1640 ns     426893   37.2154MB/s
+BM_copy/512       12804 ns      12801 ns      55417   38.1444MB/s
+BM_copy/1024      25409 ns      25407 ns      27516   38.4365MB/s
+BM_copy/8192     202986 ns     202990 ns       3454   38.4871MB/s
+Comparing BM_memcpy to BM_copy (from ./a.out)
+Benchmark                               Time             CPU      Time Old      Time New       CPU Old       CPU New
+--------------------------------------------------------------------------------------------------------------------
+[BM_memcpy vs. BM_copy]/8            +5.2829         +5.2812            36           227            36           227
+[BM_memcpy vs. BM_copy]/64          +21.1719        +21.1856            74          1640            74          1640
+[BM_memcpy vs. BM_copy]/512        +145.6487       +145.6097            87         12804            87         12801
+[BM_memcpy vs. BM_copy]/1024       +227.1860       +227.1776           111         25409           111         25407
+[BM_memcpy vs. BM_copy]/8192       +308.1664       +308.2898           657        202986           656        202990
+```
+
+As you can see, it applies filter to the benchmarks, both when running the benchmark, and before doing the diff. And to make the diff work, the matches are replaced with some common string. Thus, you can compare two different benchmark families within one benchmark binary.
+As you can note, the values in `Time` and `CPU` columns are calculated as `(new - old) / |old|`.
+
+3. Compare filter one from benchmark one to filter two from benchmark two:
+The program is invoked like:
+
+``` bash
+$ compare.py filters <benchmark_baseline> <filter_baseline> <benchmark_contender> <filter_contender> [benchmark options]...
+```
+
+Where `<benchmark_baseline>` and `<benchmark_contender>` either specify a benchmark executable file, or a JSON output file. The type of the input file is automatically detected. If a benchmark executable is specified then the benchmark is run to obtain the results. Otherwise the results are simply loaded from the output file.
+
+Where `<filter_baseline>` and `<filter_contender>` are the same regex filters that you would pass to the `[--benchmark_filter=<regex>]` parameter of the benchmark binary.
+
+`[benchmark options]` will be passed to the benchmarks invocations. They can be anything that binary accepts, be it either normal `--benchmark_*` parameters, or some custom parameters your binary takes.
+
+Example output:
+```
+$ ./compare.py benchmarksfiltered ./a.out BM_memcpy ./a.out BM_copy
+RUNNING: ./a.out --benchmark_filter=BM_memcpy --benchmark_out=/tmp/tmp_FvbYg
+Run on (8 X 4000 MHz CPU s)
+2017-11-07 21:38:27
+------------------------------------------------------
+Benchmark               Time           CPU Iterations
+------------------------------------------------------
+BM_memcpy/8            37 ns         37 ns   18953482   204.118MB/s
+BM_memcpy/64           74 ns         74 ns    9206578   828.245MB/s
+BM_memcpy/512          91 ns         91 ns    8086195   5.25476GB/s
+BM_memcpy/1024        120 ns        120 ns    5804513   7.95662GB/s
+BM_memcpy/8192        664 ns        664 ns    1028363   11.4948GB/s
+RUNNING: ./a.out --benchmark_filter=BM_copy --benchmark_out=/tmp/tmpDfL5iE
+Run on (8 X 4000 MHz CPU s)
+2017-11-07 21:38:32
+----------------------------------------------------
+Benchmark             Time           CPU Iterations
+----------------------------------------------------
+BM_copy/8           230 ns        230 ns    2985909   33.1161MB/s
+BM_copy/64         1654 ns       1653 ns     419408   36.9137MB/s
+BM_copy/512       13122 ns      13120 ns      53403   37.2156MB/s
+BM_copy/1024      26679 ns      26666 ns      26575   36.6218MB/s
+BM_copy/8192     215068 ns     215053 ns       3221   36.3283MB/s
+Comparing BM_memcpy (from ./a.out) to BM_copy (from ./a.out)
+Benchmark                               Time             CPU      Time Old      Time New       CPU Old       CPU New
+--------------------------------------------------------------------------------------------------------------------
+[BM_memcpy vs. BM_copy]/8            +5.1649         +5.1637            37           230            37           230
+[BM_memcpy vs. BM_copy]/64          +21.4352        +21.4374            74          1654            74          1653
+[BM_memcpy vs. BM_copy]/512        +143.6022       +143.5865            91         13122            91         13120
+[BM_memcpy vs. BM_copy]/1024       +221.5903       +221.4790           120         26679           120         26666
+[BM_memcpy vs. BM_copy]/8192       +322.9059       +323.0096           664        215068           664        215053
+```
+This is a mix of the previous two modes, two (potentially different) benchmark binaries are run, and a different filter is applied to each one.
+As you can note, the values in `Time` and `CPU` columns are calculated as `(new - old) / |old|`.

data/vendor/simdjson/dependencies/json/benchmarks/thirdparty/benchmark/include/benchmark/benchmark.h

@@ -0,0 +1,1456 @@
+// 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.
+
+// Support for registering benchmarks for functions.
+
+/* Example usage:
+// Define a function that executes the code to be measured a
+// specified number of times:
+static void BM_StringCreation(benchmark::State& state) {
+  for (auto _ : state)
+    std::string empty_string;
+}
+
+// Register the function as a benchmark
+BENCHMARK(BM_StringCreation);
+
+// Define another benchmark
+static void BM_StringCopy(benchmark::State& state) {
+  std::string x = "hello";
+  for (auto _ : state)
+    std::string copy(x);
+}
+BENCHMARK(BM_StringCopy);
+
+// Augment the main() program to invoke benchmarks if specified
+// via the --benchmarks command line flag.  E.g.,
+//       my_unittest --benchmark_filter=all
+//       my_unittest --benchmark_filter=BM_StringCreation
+//       my_unittest --benchmark_filter=String
+//       my_unittest --benchmark_filter='Copy|Creation'
+int main(int argc, char** argv) {
+  benchmark::Initialize(&argc, argv);
+  benchmark::RunSpecifiedBenchmarks();
+  return 0;
+}
+
+// Sometimes a family of microbenchmarks can be implemented with
+// just one routine that takes an extra argument to specify which
+// one of the family of benchmarks to run.  For example, the following
+// code defines a family of microbenchmarks for measuring the speed
+// of memcpy() calls of different lengths:
+
+static void BM_memcpy(benchmark::State& state) {
+  char* src = new char[state.range(0)]; char* dst = new char[state.range(0)];
+  memset(src, 'x', state.range(0));
+  for (auto _ : state)
+    memcpy(dst, src, state.range(0));
+  state.SetBytesProcessed(int64_t(state.iterations()) *
+                          int64_t(state.range(0)));
+  delete[] src; delete[] dst;
+}
+BENCHMARK(BM_memcpy)->Arg(8)->Arg(64)->Arg(512)->Arg(1<<10)->Arg(8<<10);
+
+// The preceding code is quite repetitive, and can be replaced with the
+// following short-hand.  The following invocation will pick a few
+// appropriate arguments in the specified range and will generate a
+// microbenchmark for each such argument.
+BENCHMARK(BM_memcpy)->Range(8, 8<<10);
+
+// You might have a microbenchmark that depends on two inputs.  For
+// example, the following code defines a family of microbenchmarks for
+// measuring the speed of set insertion.
+static void BM_SetInsert(benchmark::State& state) {
+  set<int> data;
+  for (auto _ : state) {
+    state.PauseTiming();
+    data = ConstructRandomSet(state.range(0));
+    state.ResumeTiming();
+    for (int j = 0; j < state.range(1); ++j)
+      data.insert(RandomNumber());
+  }
+}
+BENCHMARK(BM_SetInsert)
+   ->Args({1<<10, 128})
+   ->Args({2<<10, 128})
+   ->Args({4<<10, 128})
+   ->Args({8<<10, 128})
+   ->Args({1<<10, 512})
+   ->Args({2<<10, 512})
+   ->Args({4<<10, 512})
+   ->Args({8<<10, 512});
+
+// The preceding code is quite repetitive, and can be replaced with
+// the following short-hand.  The following macro will pick a few
+// appropriate arguments in the product of the two specified ranges
+// and will generate a microbenchmark for each such pair.
+BENCHMARK(BM_SetInsert)->Ranges({{1<<10, 8<<10}, {128, 512}});
+
+// For more complex patterns of inputs, passing a custom function
+// to Apply allows programmatic specification of an
+// arbitrary set of arguments to run the microbenchmark on.
+// The following example enumerates a dense range on
+// one parameter, and a sparse range on the second.
+static void CustomArguments(benchmark::internal::Benchmark* b) {
+  for (int i = 0; i <= 10; ++i)
+    for (int j = 32; j <= 1024*1024; j *= 8)
+      b->Args({i, j});
+}
+BENCHMARK(BM_SetInsert)->Apply(CustomArguments);
+
+// Templated microbenchmarks work the same way:
+// Produce then consume 'size' messages 'iters' times
+// Measures throughput in the absence of multiprogramming.
+template <class Q> int BM_Sequential(benchmark::State& state) {
+  Q q;
+  typename Q::value_type v;
+  for (auto _ : state) {
+    for (int i = state.range(0); i--; )
+      q.push(v);
+    for (int e = state.range(0); e--; )
+      q.Wait(&v);
+  }
+  // actually messages, not bytes:
+  state.SetBytesProcessed(
+      static_cast<int64_t>(state.iterations())*state.range(0));
+}
+BENCHMARK_TEMPLATE(BM_Sequential, WaitQueue<int>)->Range(1<<0, 1<<10);
+
+Use `Benchmark::MinTime(double t)` to set the minimum time used to run the
+benchmark. This option overrides the `benchmark_min_time` flag.
+
+void BM_test(benchmark::State& state) {
+ ... body ...
+}
+BENCHMARK(BM_test)->MinTime(2.0); // Run for at least 2 seconds.
+
+In a multithreaded test, it is guaranteed that none of the threads will start
+until all have reached the loop start, and all will have finished before any
+thread exits the loop body. As such, any global setup or teardown you want to
+do can be wrapped in a check against the thread index:
+
+static void BM_MultiThreaded(benchmark::State& state) {
+  if (state.thread_index == 0) {
+    // Setup code here.
+  }
+  for (auto _ : state) {
+    // Run the test as normal.
+  }
+  if (state.thread_index == 0) {
+    // Teardown code here.
+  }
+}
+BENCHMARK(BM_MultiThreaded)->Threads(4);
+
+
+If a benchmark runs a few milliseconds it may be hard to visually compare the
+measured times, since the output data is given in nanoseconds per default. In
+order to manually set the time unit, you can specify it manually:
+
+BENCHMARK(BM_test)->Unit(benchmark::kMillisecond);
+*/
+
+#ifndef BENCHMARK_BENCHMARK_H_
+#define BENCHMARK_BENCHMARK_H_
+
+
+// The _MSVC_LANG check should detect Visual Studio 2015 Update 3 and newer.
+#if __cplusplus >= 201103L || (defined(_MSVC_LANG) && _MSVC_LANG >= 201103L)
+#define BENCHMARK_HAS_CXX11
+#endif
+
+#include <stdint.h>
+
+#include <algorithm>
+#include <cassert>
+#include <cstddef>
+#include <iosfwd>
+#include <string>
+#include <vector>
+#include <map>
+#include <set>
+
+#if defined(BENCHMARK_HAS_CXX11)
+#include <type_traits>
+#include <initializer_list>
+#include <utility>
+#endif
+
+#if defined(_MSC_VER)
+#include <intrin.h> // for _ReadWriteBarrier
+#endif
+
+#ifndef BENCHMARK_HAS_CXX11
+#define BENCHMARK_DISALLOW_COPY_AND_ASSIGN(TypeName) \
+  TypeName(const TypeName&);                         \
+  TypeName& operator=(const TypeName&)
+#else
+#define BENCHMARK_DISALLOW_COPY_AND_ASSIGN(TypeName) \
+  TypeName(const TypeName&) = delete;                \
+  TypeName& operator=(const TypeName&) = delete
+#endif
+
+#if defined(__GNUC__)
+#define BENCHMARK_UNUSED __attribute__((unused))
+#define BENCHMARK_ALWAYS_INLINE __attribute__((always_inline))
+#define BENCHMARK_NOEXCEPT noexcept
+#define BENCHMARK_NOEXCEPT_OP(x) noexcept(x)
+#elif defined(_MSC_VER) && !defined(__clang__)
+#define BENCHMARK_UNUSED
+#define BENCHMARK_ALWAYS_INLINE __forceinline
+#if _MSC_VER >= 1900
+#define BENCHMARK_NOEXCEPT noexcept
+#define BENCHMARK_NOEXCEPT_OP(x) noexcept(x)
+#else
+#define BENCHMARK_NOEXCEPT
+#define BENCHMARK_NOEXCEPT_OP(x)
+#endif
+#define __func__ __FUNCTION__
+#else
+#define BENCHMARK_UNUSED
+#define BENCHMARK_ALWAYS_INLINE
+#define BENCHMARK_NOEXCEPT
+#define BENCHMARK_NOEXCEPT_OP(x)
+#endif
+
+#define BENCHMARK_INTERNAL_TOSTRING2(x) #x
+#define BENCHMARK_INTERNAL_TOSTRING(x) BENCHMARK_INTERNAL_TOSTRING2(x)
+
+#if defined(__GNUC__)
+#define BENCHMARK_BUILTIN_EXPECT(x, y) __builtin_expect(x, y)
+#define BENCHMARK_DEPRECATED_MSG(msg) __attribute__((deprecated(msg)))
+#else
+#define BENCHMARK_BUILTIN_EXPECT(x, y) x
+#define BENCHMARK_DEPRECATED_MSG(msg)
+#define BENCHMARK_WARNING_MSG(msg) __pragma(message(__FILE__ "(" BENCHMARK_INTERNAL_TOSTRING(__LINE__) ") : warning note: " msg))
+#endif
+
+#if defined(__GNUC__) && !defined(__clang__)
+#define BENCHMARK_GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
+#endif
+
+namespace benchmark {
+class BenchmarkReporter;
+
+void Initialize(int* argc, char** argv);
+
+// Report to stdout all arguments in 'argv' as unrecognized except the first.
+// Returns true there is at least on unrecognized argument (i.e. 'argc' > 1).
+bool ReportUnrecognizedArguments(int argc, char** argv);
+
+// Generate a list of benchmarks matching the specified --benchmark_filter flag
+// and if --benchmark_list_tests is specified return after printing the name
+// of each matching benchmark. Otherwise run each matching benchmark and
+// report the results.
+//
+// The second and third overload use the specified 'console_reporter' and
+//  'file_reporter' respectively. 'file_reporter' will write to the file
+//  specified
+//   by '--benchmark_output'. If '--benchmark_output' is not given the
+//  'file_reporter' is ignored.
+//
+// RETURNS: The number of matching benchmarks.
+size_t RunSpecifiedBenchmarks();
+size_t RunSpecifiedBenchmarks(BenchmarkReporter* console_reporter);
+size_t RunSpecifiedBenchmarks(BenchmarkReporter* console_reporter,
+                              BenchmarkReporter* file_reporter);
+
+// If this routine is called, peak memory allocation past this point in the
+// benchmark is reported at the end of the benchmark report line. (It is
+// computed by running the benchmark once with a single iteration and a memory
+// tracer.)
+// TODO(dominic)
+// void MemoryUsage();
+
+namespace internal {
+class Benchmark;
+class BenchmarkImp;
+class BenchmarkFamilies;
+
+void UseCharPointer(char const volatile*);
+
+// Take ownership of the pointer and register the benchmark. Return the
+// registered benchmark.
+Benchmark* RegisterBenchmarkInternal(Benchmark*);
+
+// Ensure that the standard streams are properly initialized in every TU.
+int InitializeStreams();
+BENCHMARK_UNUSED static int stream_init_anchor = InitializeStreams();
+
+}  // namespace internal
+
+
+#if (!defined(__GNUC__) && !defined(__clang__)) || defined(__pnacl__) || \
+    defined(__EMSCRIPTEN__)
+# define BENCHMARK_HAS_NO_INLINE_ASSEMBLY
+#endif
+
+
+// The DoNotOptimize(...) function can be used to prevent a value or
+// expression from being optimized away by the compiler. This function is
+// intended to add little to no overhead.
+// See: https://youtu.be/nXaxk27zwlk?t=2441
+#ifndef BENCHMARK_HAS_NO_INLINE_ASSEMBLY
+template <class Tp>
+inline BENCHMARK_ALWAYS_INLINE
+void DoNotOptimize(Tp const& value) {
+    asm volatile("" : : "r,m"(value) : "memory");
+}
+
+template <class Tp>
+inline BENCHMARK_ALWAYS_INLINE void DoNotOptimize(Tp& value) {
+#if defined(__clang__)
+  asm volatile("" : "+r,m"(value) : : "memory");
+#else
+  asm volatile("" : "+m,r"(value) : : "memory");
+#endif
+}
+
+// Force the compiler to flush pending writes to global memory. Acts as an
+// effective read/write barrier
+inline BENCHMARK_ALWAYS_INLINE void ClobberMemory() {
+  asm volatile("" : : : "memory");
+}
+#elif defined(_MSC_VER)
+template <class Tp>
+inline BENCHMARK_ALWAYS_INLINE void DoNotOptimize(Tp const& value) {
+  internal::UseCharPointer(&reinterpret_cast<char const volatile&>(value));
+  _ReadWriteBarrier();
+}
+
+inline BENCHMARK_ALWAYS_INLINE void ClobberMemory() {
+  _ReadWriteBarrier();
+}
+#else
+template <class Tp>
+inline BENCHMARK_ALWAYS_INLINE void DoNotOptimize(Tp const& value) {
+  internal::UseCharPointer(&reinterpret_cast<char const volatile&>(value));
+}
+// FIXME Add ClobberMemory() for non-gnu and non-msvc compilers
+#endif
+
+
+
+// This class is used for user-defined counters.
+class Counter {
+public:
+
+  enum Flags {
+    kDefaults   = 0,
+    // Mark the counter as a rate. It will be presented divided
+    // by the duration of the benchmark.
+    kIsRate     = 1,
+    // Mark the counter as a thread-average quantity. It will be
+    // presented divided by the number of threads.
+    kAvgThreads = 2,
+    // Mark the counter as a thread-average rate. See above.
+    kAvgThreadsRate = kIsRate|kAvgThreads
+  };
+
+  double value;
+  Flags  flags;
+
+  BENCHMARK_ALWAYS_INLINE
+  Counter(double v = 0., Flags f = kDefaults) : value(v), flags(f) {}
+
+  BENCHMARK_ALWAYS_INLINE operator double const& () const { return value; }
+  BENCHMARK_ALWAYS_INLINE operator double      & ()       { return value; }
+
+};
+
+// This is the container for the user-defined counters.
+typedef std::map<std::string, Counter> UserCounters;
+
+
+// TimeUnit is passed to a benchmark in order to specify the order of magnitude
+// for the measured time.
+enum TimeUnit { kNanosecond, kMicrosecond, kMillisecond };
+
+// BigO is passed to a benchmark in order to specify the asymptotic
+// computational
+// complexity for the benchmark. In case oAuto is selected, complexity will be
+// calculated automatically to the best fit.
+enum BigO { oNone, o1, oN, oNSquared, oNCubed, oLogN, oNLogN, oAuto, oLambda };
+
+// BigOFunc is passed to a benchmark in order to specify the asymptotic
+// computational complexity for the benchmark.
+typedef double(BigOFunc)(int64_t);
+
+// StatisticsFunc is passed to a benchmark in order to compute some descriptive
+// statistics over all the measurements of some type
+typedef double(StatisticsFunc)(const std::vector<double>&);
+
+struct Statistics {
+  std::string name_;
+  StatisticsFunc* compute_;
+
+  Statistics(std::string name, StatisticsFunc* compute)
+    : name_(name), compute_(compute) {}
+};
+
+namespace internal {
+class ThreadTimer;
+class ThreadManager;
+
+enum ReportMode
+#if defined(BENCHMARK_HAS_CXX11)
+  : unsigned
+#else
+#endif
+  {
+  RM_Unspecified,  // The mode has not been manually specified
+  RM_Default,      // The mode is user-specified as default.
+  RM_ReportAggregatesOnly
+};
+}  // namespace internal
+
+// State is passed to a running Benchmark and contains state for the
+// benchmark to use.
+class State {
+ public:
+  struct StateIterator;
+  friend struct StateIterator;
+
+  // Returns iterators used to run each iteration of a benchmark using a
+  // C++11 ranged-based for loop. These functions should not be called directly.
+  //
+  // REQUIRES: The benchmark has not started running yet. Neither begin nor end
+  // have been called previously.
+  //
+  // NOTE: KeepRunning may not be used after calling either of these functions.
+  BENCHMARK_ALWAYS_INLINE StateIterator begin();
+  BENCHMARK_ALWAYS_INLINE StateIterator end();
+
+  // Returns true if the benchmark should continue through another iteration.
+  // NOTE: A benchmark may not return from the test until KeepRunning() has
+  // returned false.
+  bool KeepRunning();
+
+  // Returns true iff the benchmark should run n more iterations.
+  // REQUIRES: 'n' > 0.
+  // NOTE: A benchmark must not return from the test until KeepRunningBatch()
+  // has returned false.
+  // NOTE: KeepRunningBatch() may overshoot by up to 'n' iterations.
+  //
+  // Intended usage:
+  //   while (state.KeepRunningBatch(1000)) {
+  //     // process 1000 elements
+  //   }
+  bool KeepRunningBatch(size_t n);
+
+  // REQUIRES: timer is running and 'SkipWithError(...)' has not been called
+  //           by the current thread.
+  // Stop the benchmark timer.  If not called, the timer will be
+  // automatically stopped after the last iteration of the benchmark loop.
+  //
+  // For threaded benchmarks the PauseTiming() function only pauses the timing
+  // for the current thread.
+  //
+  // NOTE: The "real time" measurement is per-thread. If different threads
+  // report different measurements the largest one is reported.
+  //
+  // NOTE: PauseTiming()/ResumeTiming() are relatively
+  // heavyweight, and so their use should generally be avoided
+  // within each benchmark iteration, if possible.
+  void PauseTiming();
+
+  // REQUIRES: timer is not running and 'SkipWithError(...)' has not been called
+  //           by the current thread.
+  // Start the benchmark timer.  The timer is NOT running on entrance to the
+  // benchmark function. It begins running after control flow enters the
+  // benchmark loop.
+  //
+  // NOTE: PauseTiming()/ResumeTiming() are relatively
+  // heavyweight, and so their use should generally be avoided
+  // within each benchmark iteration, if possible.
+  void ResumeTiming();
+
+  // REQUIRES: 'SkipWithError(...)' has not been called previously by the
+  //            current thread.
+  // Report the benchmark as resulting in an error with the specified 'msg'.
+  // After this call the user may explicitly 'return' from the benchmark.
+  //
+  // If the ranged-for style of benchmark loop is used, the user must explicitly
+  // break from the loop, otherwise all future iterations will be run.
+  // If the 'KeepRunning()' loop is used the current thread will automatically
+  // exit the loop at the end of the current iteration.
+  //
+  // For threaded benchmarks only the current thread stops executing and future
+  // calls to `KeepRunning()` will block until all threads have completed
+  // the `KeepRunning()` loop. If multiple threads report an error only the
+  // first error message is used.
+  //
+  // NOTE: Calling 'SkipWithError(...)' does not cause the benchmark to exit
+  // the current scope immediately. If the function is called from within
+  // the 'KeepRunning()' loop the current iteration will finish. It is the users
+  // responsibility to exit the scope as needed.
+  void SkipWithError(const char* msg);
+
+  // REQUIRES: called exactly once per iteration of the benchmarking loop.
+  // Set the manually measured time for this benchmark iteration, which
+  // is used instead of automatically measured time if UseManualTime() was
+  // specified.
+  //
+  // For threaded benchmarks the final value will be set to the largest
+  // reported values.
+  void SetIterationTime(double seconds);
+
+  // Set the number of bytes processed by the current benchmark
+  // execution.  This routine is typically called once at the end of a
+  // throughput oriented benchmark.  If this routine is called with a
+  // value > 0, the report is printed in MB/sec instead of nanoseconds
+  // per iteration.
+  //
+  // REQUIRES: a benchmark has exited its benchmarking loop.
+  BENCHMARK_ALWAYS_INLINE
+  void SetBytesProcessed(int64_t bytes) { bytes_processed_ = bytes; }
+
+  BENCHMARK_ALWAYS_INLINE
+  int64_t bytes_processed() const { return bytes_processed_; }
+
+  // If this routine is called with complexity_n > 0 and complexity report is
+  // requested for the
+  // family benchmark, then current benchmark will be part of the computation
+  // and complexity_n will
+  // represent the length of N.
+  BENCHMARK_ALWAYS_INLINE
+  void SetComplexityN(int64_t complexity_n) { complexity_n_ = complexity_n; }
+
+  BENCHMARK_ALWAYS_INLINE
+  int64_t complexity_length_n() { return complexity_n_; }
+
+  // If this routine is called with items > 0, then an items/s
+  // label is printed on the benchmark report line for the currently
+  // executing benchmark. It is typically called at the end of a processing
+  // benchmark where a processing items/second output is desired.
+  //
+  // REQUIRES: a benchmark has exited its benchmarking loop.
+  BENCHMARK_ALWAYS_INLINE
+  void SetItemsProcessed(int64_t items) { items_processed_ = items; }
+
+  BENCHMARK_ALWAYS_INLINE
+  int64_t items_processed() const { return items_processed_; }
+
+  // If this routine is called, the specified label is printed at the
+  // end of the benchmark report line for the currently executing
+  // benchmark.  Example:
+  //  static void BM_Compress(benchmark::State& state) {
+  //    ...
+  //    double compress = input_size / output_size;
+  //    state.SetLabel(StrFormat("compress:%.1f%%", 100.0*compression));
+  //  }
+  // Produces output that looks like:
+  //  BM_Compress   50         50   14115038  compress:27.3%
+  //
+  // REQUIRES: a benchmark has exited its benchmarking loop.
+  void SetLabel(const char* label);
+
+  void BENCHMARK_ALWAYS_INLINE SetLabel(const std::string& str) {
+    this->SetLabel(str.c_str());
+  }
+
+  // Range arguments for this run. CHECKs if the argument has been set.
+  BENCHMARK_ALWAYS_INLINE
+  int64_t range(std::size_t pos = 0) const {
+    assert(range_.size() > pos);
+    return range_[pos];
+  }
+
+  BENCHMARK_DEPRECATED_MSG("use 'range(0)' instead")
+  int64_t range_x() const { return range(0); }
+
+  BENCHMARK_DEPRECATED_MSG("use 'range(1)' instead")
+  int64_t range_y() const { return range(1); }
+
+  BENCHMARK_ALWAYS_INLINE
+  size_t iterations() const {
+    if (BENCHMARK_BUILTIN_EXPECT(!started_, false)) {
+      return 0;
+    }
+    return max_iterations - total_iterations_ + batch_leftover_;
+  }
+
+private: // items we expect on the first cache line (ie 64 bytes of the struct)
+
+  // When total_iterations_ is 0, KeepRunning() and friends will return false.
+  // May be larger than max_iterations.
+  size_t total_iterations_;
+
+  // When using KeepRunningBatch(), batch_leftover_ holds the number of
+  // iterations beyond max_iters that were run. Used to track
+  // completed_iterations_ accurately.
+  size_t batch_leftover_;
+
+public:
+  const size_t max_iterations;
+
+private:
+  bool started_;
+  bool finished_;
+  bool error_occurred_;
+
+private: // items we don't need on the first cache line
+  std::vector<int64_t> range_;
+
+  int64_t bytes_processed_;
+  int64_t items_processed_;
+
+  int64_t complexity_n_;
+
+ public:
+  // Container for user-defined counters.
+  UserCounters counters;
+  // Index of the executing thread. Values from [0, threads).
+  const int thread_index;
+  // Number of threads concurrently executing the benchmark.
+  const int threads;
+
+
+  // TODO(EricWF) make me private
+  State(size_t max_iters, const std::vector<int64_t>& ranges, int thread_i,
+        int n_threads, internal::ThreadTimer* timer,
+        internal::ThreadManager* manager);
+
+ private:
+  void StartKeepRunning();
+  // Implementation of KeepRunning() and KeepRunningBatch().
+  // is_batch must be true unless n is 1.
+  bool KeepRunningInternal(size_t n, bool is_batch);
+  void FinishKeepRunning();
+  internal::ThreadTimer* timer_;
+  internal::ThreadManager* manager_;
+  BENCHMARK_DISALLOW_COPY_AND_ASSIGN(State);
+};
+
+inline BENCHMARK_ALWAYS_INLINE
+bool State::KeepRunning() {
+  return KeepRunningInternal(1, /*is_batch=*/ false);
+}
+
+inline BENCHMARK_ALWAYS_INLINE
+bool State::KeepRunningBatch(size_t n) {
+  return KeepRunningInternal(n, /*is_batch=*/ true);
+}
+
+inline BENCHMARK_ALWAYS_INLINE
+bool State::KeepRunningInternal(size_t n, bool is_batch) {
+  // total_iterations_ is set to 0 by the constructor, and always set to a
+  // nonzero value by StartKepRunning().
+  assert(n > 0);
+  // n must be 1 unless is_batch is true.
+  assert(is_batch || n == 1);
+  if (BENCHMARK_BUILTIN_EXPECT(total_iterations_ >= n, true)) {
+    total_iterations_ -= n;
+    return true;
+  }
+  if (!started_) {
+    StartKeepRunning();
+    if (!error_occurred_ && total_iterations_ >= n) {
+      total_iterations_-= n;
+      return true;
+    }
+  }
+  // For non-batch runs, total_iterations_ must be 0 by now.
+  if (is_batch && total_iterations_ != 0) {
+    batch_leftover_  = n - total_iterations_;
+    total_iterations_ = 0;
+    return true;
+  }
+  FinishKeepRunning();
+  return false;
+}
+
+struct State::StateIterator {
+  struct BENCHMARK_UNUSED Value {};
+  typedef std::forward_iterator_tag iterator_category;
+  typedef Value value_type;
+  typedef Value reference;
+  typedef Value pointer;
+  typedef std::ptrdiff_t difference_type;
+
+ private:
+  friend class State;
+  BENCHMARK_ALWAYS_INLINE
+  StateIterator() : cached_(0), parent_() {}
+
+  BENCHMARK_ALWAYS_INLINE
+  explicit StateIterator(State* st)
+      : cached_(st->error_occurred_ ? 0 : st->max_iterations), parent_(st) {}
+
+ public:
+  BENCHMARK_ALWAYS_INLINE
+  Value operator*() const { return Value(); }
+
+  BENCHMARK_ALWAYS_INLINE
+  StateIterator& operator++() {
+    assert(cached_ > 0);
+    --cached_;
+    return *this;
+  }
+
+  BENCHMARK_ALWAYS_INLINE
+  bool operator!=(StateIterator const&) const {
+    if (BENCHMARK_BUILTIN_EXPECT(cached_ != 0, true)) return true;
+    parent_->FinishKeepRunning();
+    return false;
+  }
+
+ private:
+  size_t cached_;
+  State* const parent_;
+};
+
+inline BENCHMARK_ALWAYS_INLINE State::StateIterator State::begin() {
+  return StateIterator(this);
+}
+inline BENCHMARK_ALWAYS_INLINE State::StateIterator State::end() {
+  StartKeepRunning();
+  return StateIterator();
+}
+
+namespace internal {
+
+typedef void(Function)(State&);
+
+// ------------------------------------------------------
+// Benchmark registration object.  The BENCHMARK() macro expands
+// into an internal::Benchmark* object.  Various methods can
+// be called on this object to change the properties of the benchmark.
+// Each method returns "this" so that multiple method calls can
+// chained into one expression.
+class Benchmark {
+ public:
+  virtual ~Benchmark();
+
+  // Note: the following methods all return "this" so that multiple
+  // method calls can be chained together in one expression.
+
+  // Run this benchmark once with "x" as the extra argument passed
+  // to the function.
+  // REQUIRES: The function passed to the constructor must accept an arg1.
+  Benchmark* Arg(int64_t x);
+
+  // Run this benchmark with the given time unit for the generated output report
+  Benchmark* Unit(TimeUnit unit);
+
+  // Run this benchmark once for a number of values picked from the
+  // range [start..limit].  (start and limit are always picked.)
+  // REQUIRES: The function passed to the constructor must accept an arg1.
+  Benchmark* Range(int64_t start, int64_t limit);
+
+  // Run this benchmark once for all values in the range [start..limit] with
+  // specific step
+  // REQUIRES: The function passed to the constructor must accept an arg1.
+  Benchmark* DenseRange(int64_t start, int64_t limit, int step = 1);
+
+  // Run this benchmark once with "args" as the extra arguments passed
+  // to the function.
+  // REQUIRES: The function passed to the constructor must accept arg1, arg2 ...
+  Benchmark* Args(const std::vector<int64_t>& args);
+
+  // Equivalent to Args({x, y})
+  // NOTE: This is a legacy C++03 interface provided for compatibility only.
+  //   New code should use 'Args'.
+  Benchmark* ArgPair(int64_t x, int64_t y) {
+    std::vector<int64_t> args;
+    args.push_back(x);
+    args.push_back(y);
+    return Args(args);
+  }
+
+  // Run this benchmark once for a number of values picked from the
+  // ranges [start..limit].  (starts and limits are always picked.)
+  // REQUIRES: The function passed to the constructor must accept arg1, arg2 ...
+  Benchmark* Ranges(const std::vector<std::pair<int64_t, int64_t> >& ranges);
+
+  // Equivalent to ArgNames({name})
+  Benchmark* ArgName(const std::string& name);
+
+  // Set the argument names to display in the benchmark name. If not called,
+  // only argument values will be shown.
+  Benchmark* ArgNames(const std::vector<std::string>& names);
+
+  // Equivalent to Ranges({{lo1, hi1}, {lo2, hi2}}).
+  // NOTE: This is a legacy C++03 interface provided for compatibility only.
+  //   New code should use 'Ranges'.
+  Benchmark* RangePair(int64_t lo1, int64_t hi1, int64_t lo2, int64_t hi2) {
+    std::vector<std::pair<int64_t, int64_t> > ranges;
+    ranges.push_back(std::make_pair(lo1, hi1));
+    ranges.push_back(std::make_pair(lo2, hi2));
+    return Ranges(ranges);
+  }
+
+  // Pass this benchmark object to *func, which can customize
+  // the benchmark by calling various methods like Arg, Args,
+  // Threads, etc.
+  Benchmark* Apply(void (*func)(Benchmark* benchmark));
+
+  // Set the range multiplier for non-dense range. If not called, the range
+  // multiplier kRangeMultiplier will be used.
+  Benchmark* RangeMultiplier(int multiplier);
+
+  // Set the minimum amount of time to use when running this benchmark. This
+  // option overrides the `benchmark_min_time` flag.
+  // REQUIRES: `t > 0` and `Iterations` has not been called on this benchmark.
+  Benchmark* MinTime(double t);
+
+  // Specify the amount of iterations that should be run by this benchmark.
+  // REQUIRES: 'n > 0' and `MinTime` has not been called on this benchmark.
+  //
+  // NOTE: This function should only be used when *exact* iteration control is
+  //   needed and never to control or limit how long a benchmark runs, where
+  // `--benchmark_min_time=N` or `MinTime(...)` should be used instead.
+  Benchmark* Iterations(size_t n);
+
+  // Specify the amount of times to repeat this benchmark. This option overrides
+  // the `benchmark_repetitions` flag.
+  // REQUIRES: `n > 0`
+  Benchmark* Repetitions(int n);
+
+  // Specify if each repetition of the benchmark should be reported separately
+  // or if only the final statistics should be reported. If the benchmark
+  // is not repeated then the single result is always reported.
+  Benchmark* ReportAggregatesOnly(bool value = true);
+
+  // If a particular benchmark is I/O bound, runs multiple threads internally or
+  // if for some reason CPU timings are not representative, call this method. If
+  // called, the elapsed time will be used to control how many iterations are
+  // run, and in the printing of items/second or MB/seconds values.  If not
+  // called, the cpu time used by the benchmark will be used.
+  Benchmark* UseRealTime();
+
+  // If a benchmark must measure time manually (e.g. if GPU execution time is
+  // being
+  // measured), call this method. If called, each benchmark iteration should
+  // call
+  // SetIterationTime(seconds) to report the measured time, which will be used
+  // to control how many iterations are run, and in the printing of items/second
+  // or MB/second values.
+  Benchmark* UseManualTime();
+
+  // Set the asymptotic computational complexity for the benchmark. If called
+  // the asymptotic computational complexity will be shown on the output.
+  Benchmark* Complexity(BigO complexity = benchmark::oAuto);
+
+  // Set the asymptotic computational complexity for the benchmark. If called
+  // the asymptotic computational complexity will be shown on the output.
+  Benchmark* Complexity(BigOFunc* complexity);
+
+  // Add this statistics to be computed over all the values of benchmark run
+  Benchmark* ComputeStatistics(std::string name, StatisticsFunc* statistics);
+
+  // Support for running multiple copies of the same benchmark concurrently
+  // in multiple threads.  This may be useful when measuring the scaling
+  // of some piece of code.
+
+  // Run one instance of this benchmark concurrently in t threads.
+  Benchmark* Threads(int t);
+
+  // Pick a set of values T from [min_threads,max_threads].
+  // min_threads and max_threads are always included in T.  Run this
+  // benchmark once for each value in T.  The benchmark run for a
+  // particular value t consists of t threads running the benchmark
+  // function concurrently.  For example, consider:
+  //    BENCHMARK(Foo)->ThreadRange(1,16);
+  // This will run the following benchmarks:
+  //    Foo in 1 thread
+  //    Foo in 2 threads
+  //    Foo in 4 threads
+  //    Foo in 8 threads
+  //    Foo in 16 threads
+  Benchmark* ThreadRange(int min_threads, int max_threads);
+
+  // For each value n in the range, run this benchmark once using n threads.
+  // min_threads and max_threads are always included in the range.
+  // stride specifies the increment. E.g. DenseThreadRange(1, 8, 3) starts
+  // a benchmark with 1, 4, 7 and 8 threads.
+  Benchmark* DenseThreadRange(int min_threads, int max_threads, int stride = 1);
+
+  // Equivalent to ThreadRange(NumCPUs(), NumCPUs())
+  Benchmark* ThreadPerCpu();
+
+  virtual void Run(State& state) = 0;
+
+  // Used inside the benchmark implementation
+  struct Instance;
+
+ protected:
+  explicit Benchmark(const char* name);
+  Benchmark(Benchmark const&);
+  void SetName(const char* name);
+
+  int ArgsCnt() const;
+
+ private:
+  friend class BenchmarkFamilies;
+
+  std::string name_;
+  ReportMode report_mode_;
+  std::vector<std::string> arg_names_;   // Args for all benchmark runs
+  std::vector<std::vector<int64_t> > args_;  // Args for all benchmark runs
+  TimeUnit time_unit_;
+  int range_multiplier_;
+  double min_time_;
+  size_t iterations_;
+  int repetitions_;
+  bool use_real_time_;
+  bool use_manual_time_;
+  BigO complexity_;
+  BigOFunc* complexity_lambda_;
+  std::vector<Statistics> statistics_;
+  std::vector<int> thread_counts_;
+
+  Benchmark& operator=(Benchmark const&);
+};
+
+}  // namespace internal
+
+// Create and register a benchmark with the specified 'name' that invokes
+// the specified functor 'fn'.
+//
+// RETURNS: A pointer to the registered benchmark.
+internal::Benchmark* RegisterBenchmark(const char* name,
+                                       internal::Function* fn);
+
+#if defined(BENCHMARK_HAS_CXX11)
+template <class Lambda>
+internal::Benchmark* RegisterBenchmark(const char* name, Lambda&& fn);
+#endif
+
+// Remove all registered benchmarks. All pointers to previously registered
+// benchmarks are invalidated.
+void ClearRegisteredBenchmarks();
+
+namespace internal {
+// The class used to hold all Benchmarks created from static function.
+// (ie those created using the BENCHMARK(...) macros.
+class FunctionBenchmark : public Benchmark {
+ public:
+  FunctionBenchmark(const char* name, Function* func)
+      : Benchmark(name), func_(func) {}
+
+  virtual void Run(State& st);
+
+ private:
+  Function* func_;
+};
+
+#ifdef BENCHMARK_HAS_CXX11
+template <class Lambda>
+class LambdaBenchmark : public Benchmark {
+ public:
+  virtual void Run(State& st) { lambda_(st); }
+
+ private:
+  template <class OLambda>
+  LambdaBenchmark(const char* name, OLambda&& lam)
+      : Benchmark(name), lambda_(std::forward<OLambda>(lam)) {}
+
+  LambdaBenchmark(LambdaBenchmark const&) = delete;
+
+ private:
+  template <class Lam>
+  friend Benchmark* ::benchmark::RegisterBenchmark(const char*, Lam&&);
+
+  Lambda lambda_;
+};
+#endif
+
+}  // namespace internal
+
+inline internal::Benchmark* RegisterBenchmark(const char* name,
+                                              internal::Function* fn) {
+  return internal::RegisterBenchmarkInternal(
+      ::new internal::FunctionBenchmark(name, fn));
+}
+
+#ifdef BENCHMARK_HAS_CXX11
+template <class Lambda>
+internal::Benchmark* RegisterBenchmark(const char* name, Lambda&& fn) {
+  using BenchType =
+      internal::LambdaBenchmark<typename std::decay<Lambda>::type>;
+  return internal::RegisterBenchmarkInternal(
+      ::new BenchType(name, std::forward<Lambda>(fn)));
+}
+#endif
+
+#if defined(BENCHMARK_HAS_CXX11) && \
+    (!defined(BENCHMARK_GCC_VERSION) || BENCHMARK_GCC_VERSION >= 409)
+template <class Lambda, class... Args>
+internal::Benchmark* RegisterBenchmark(const char* name, Lambda&& fn,
+                                       Args&&... args) {
+  return benchmark::RegisterBenchmark(
+      name, [=](benchmark::State& st) { fn(st, args...); });
+}
+#else
+#define BENCHMARK_HAS_NO_VARIADIC_REGISTER_BENCHMARK
+#endif
+
+// The base class for all fixture tests.
+class Fixture : public internal::Benchmark {
+ public:
+  Fixture() : internal::Benchmark("") {}
+
+  virtual void Run(State& st) {
+    this->SetUp(st);
+    this->BenchmarkCase(st);
+    this->TearDown(st);
+  }
+
+  // These will be deprecated ...
+  virtual void SetUp(const State&) {}
+  virtual void TearDown(const State&) {}
+  // ... In favor of these.
+  virtual void SetUp(State& st) { SetUp(const_cast<const State&>(st)); }
+  virtual void TearDown(State& st) { TearDown(const_cast<const State&>(st)); }
+
+ protected:
+  virtual void BenchmarkCase(State&) = 0;
+};
+
+}  // namespace benchmark
+
+// ------------------------------------------------------
+// Macro to register benchmarks
+
+// Check that __COUNTER__ is defined and that __COUNTER__ increases by 1
+// every time it is expanded. X + 1 == X + 0 is used in case X is defined to be
+// empty. If X is empty the expression becomes (+1 == +0).
+#if defined(__COUNTER__) && (__COUNTER__ + 1 == __COUNTER__ + 0)
+#define BENCHMARK_PRIVATE_UNIQUE_ID __COUNTER__
+#else
+#define BENCHMARK_PRIVATE_UNIQUE_ID __LINE__
+#endif
+
+// Helpers for generating unique variable names
+#define BENCHMARK_PRIVATE_NAME(n) \
+  BENCHMARK_PRIVATE_CONCAT(_benchmark_, BENCHMARK_PRIVATE_UNIQUE_ID, n)
+#define BENCHMARK_PRIVATE_CONCAT(a, b, c) BENCHMARK_PRIVATE_CONCAT2(a, b, c)
+#define BENCHMARK_PRIVATE_CONCAT2(a, b, c) a##b##c
+
+#define BENCHMARK_PRIVATE_DECLARE(n)                                 \
+  static ::benchmark::internal::Benchmark* BENCHMARK_PRIVATE_NAME(n) \
+      BENCHMARK_UNUSED
+
+#define BENCHMARK(n)                                     \
+  BENCHMARK_PRIVATE_DECLARE(n) =                         \
+      (::benchmark::internal::RegisterBenchmarkInternal( \
+          new ::benchmark::internal::FunctionBenchmark(#n, n)))
+
+// Old-style macros
+#define BENCHMARK_WITH_ARG(n, a) BENCHMARK(n)->Arg((a))
+#define BENCHMARK_WITH_ARG2(n, a1, a2) BENCHMARK(n)->Args({(a1), (a2)})
+#define BENCHMARK_WITH_UNIT(n, t) BENCHMARK(n)->Unit((t))
+#define BENCHMARK_RANGE(n, lo, hi) BENCHMARK(n)->Range((lo), (hi))
+#define BENCHMARK_RANGE2(n, l1, h1, l2, h2) \
+  BENCHMARK(n)->RangePair({{(l1), (h1)}, {(l2), (h2)}})
+
+#ifdef BENCHMARK_HAS_CXX11
+
+// Register a benchmark which invokes the function specified by `func`
+// with the additional arguments specified by `...`.
+//
+// For example:
+//
+// template <class ...ExtraArgs>`
+// void BM_takes_args(benchmark::State& state, ExtraArgs&&... extra_args) {
+//  [...]
+//}
+// /* Registers a benchmark named "BM_takes_args/int_string_test` */
+// BENCHMARK_CAPTURE(BM_takes_args, int_string_test, 42, std::string("abc"));
+#define BENCHMARK_CAPTURE(func, test_case_name, ...)     \
+  BENCHMARK_PRIVATE_DECLARE(func) =                      \
+      (::benchmark::internal::RegisterBenchmarkInternal( \
+          new ::benchmark::internal::FunctionBenchmark(  \
+              #func "/" #test_case_name,                 \
+              [](::benchmark::State& st) { func(st, __VA_ARGS__); })))
+
+#endif  // BENCHMARK_HAS_CXX11
+
+// This will register a benchmark for a templatized function.  For example:
+//
+// template<int arg>
+// void BM_Foo(int iters);
+//
+// BENCHMARK_TEMPLATE(BM_Foo, 1);
+//
+// will register BM_Foo<1> as a benchmark.
+#define BENCHMARK_TEMPLATE1(n, a)                        \
+  BENCHMARK_PRIVATE_DECLARE(n) =                         \
+      (::benchmark::internal::RegisterBenchmarkInternal( \
+          new ::benchmark::internal::FunctionBenchmark(#n "<" #a ">", n<a>)))
+
+#define BENCHMARK_TEMPLATE2(n, a, b)                                         \
+  BENCHMARK_PRIVATE_DECLARE(n) =                                             \
+      (::benchmark::internal::RegisterBenchmarkInternal(                     \
+          new ::benchmark::internal::FunctionBenchmark(#n "<" #a "," #b ">", \
+                                                       n<a, b>)))
+
+#ifdef BENCHMARK_HAS_CXX11
+#define BENCHMARK_TEMPLATE(n, ...)                       \
+  BENCHMARK_PRIVATE_DECLARE(n) =                         \
+      (::benchmark::internal::RegisterBenchmarkInternal( \
+          new ::benchmark::internal::FunctionBenchmark(  \
+              #n "<" #__VA_ARGS__ ">", n<__VA_ARGS__>)))
+#else
+#define BENCHMARK_TEMPLATE(n, a) BENCHMARK_TEMPLATE1(n, a)
+#endif
+
+#define BENCHMARK_PRIVATE_DECLARE_F(BaseClass, Method)        \
+  class BaseClass##_##Method##_Benchmark : public BaseClass { \
+   public:                                                    \
+    BaseClass##_##Method##_Benchmark() : BaseClass() {        \
+      this->SetName(#BaseClass "/" #Method);                  \
+    }                                                         \
+                                                              \
+   protected:                                                 \
+    virtual void BenchmarkCase(::benchmark::State&);          \
+  };
+
+#define BENCHMARK_TEMPLATE1_PRIVATE_DECLARE_F(BaseClass, Method, a) \
+  class BaseClass##_##Method##_Benchmark : public BaseClass<a> {    \
+   public:                                                          \
+    BaseClass##_##Method##_Benchmark() : BaseClass<a>() {           \
+      this->SetName(#BaseClass"<" #a ">/" #Method);                 \
+    }                                                               \
+                                                                    \
+   protected:                                                       \
+    virtual void BenchmarkCase(::benchmark::State&);                \
+  };
+
+#define BENCHMARK_TEMPLATE2_PRIVATE_DECLARE_F(BaseClass, Method, a, b) \
+  class BaseClass##_##Method##_Benchmark : public BaseClass<a, b> {    \
+   public:                                                             \
+    BaseClass##_##Method##_Benchmark() : BaseClass<a, b>() {           \
+      this->SetName(#BaseClass"<" #a "," #b ">/" #Method);             \
+    }                                                                  \
+                                                                       \
+   protected:                                                          \
+    virtual void BenchmarkCase(::benchmark::State&);                   \
+  };
+
+#ifdef BENCHMARK_HAS_CXX11
+#define BENCHMARK_TEMPLATE_PRIVATE_DECLARE_F(BaseClass, Method, ...)       \
+  class BaseClass##_##Method##_Benchmark : public BaseClass<__VA_ARGS__> { \
+   public:                                                                 \
+    BaseClass##_##Method##_Benchmark() : BaseClass<__VA_ARGS__>() {        \
+      this->SetName(#BaseClass"<" #__VA_ARGS__ ">/" #Method);              \
+    }                                                                      \
+                                                                           \
+   protected:                                                              \
+    virtual void BenchmarkCase(::benchmark::State&);                       \
+  };
+#else
+#define BENCHMARK_TEMPLATE_PRIVATE_DECLARE_F(n, a) BENCHMARK_TEMPLATE1_PRIVATE_DECLARE_F(n, a)
+#endif
+
+#define BENCHMARK_DEFINE_F(BaseClass, Method)    \
+  BENCHMARK_PRIVATE_DECLARE_F(BaseClass, Method) \
+  void BaseClass##_##Method##_Benchmark::BenchmarkCase
+
+#define BENCHMARK_TEMPLATE1_DEFINE_F(BaseClass, Method, a)    \
+  BENCHMARK_TEMPLATE1_PRIVATE_DECLARE_F(BaseClass, Method, a) \
+  void BaseClass##_##Method##_Benchmark::BenchmarkCase
+
+#define BENCHMARK_TEMPLATE2_DEFINE_F(BaseClass, Method, a, b)    \
+  BENCHMARK_TEMPLATE2_PRIVATE_DECLARE_F(BaseClass, Method, a, b) \
+  void BaseClass##_##Method##_Benchmark::BenchmarkCase
+
+#ifdef BENCHMARK_HAS_CXX11
+#define BENCHMARK_TEMPLATE_DEFINE_F(BaseClass, Method, ...)            \
+  BENCHMARK_TEMPLATE_PRIVATE_DECLARE_F(BaseClass, Method, __VA_ARGS__) \
+  void BaseClass##_##Method##_Benchmark::BenchmarkCase
+#else
+#define BENCHMARK_TEMPLATE_DEFINE_F(BaseClass, Method, a) BENCHMARK_TEMPLATE1_DEFINE_F(BaseClass, Method, a)
+#endif
+
+#define BENCHMARK_REGISTER_F(BaseClass, Method) \
+  BENCHMARK_PRIVATE_REGISTER_F(BaseClass##_##Method##_Benchmark)
+
+#define BENCHMARK_PRIVATE_REGISTER_F(TestName) \
+  BENCHMARK_PRIVATE_DECLARE(TestName) =        \
+      (::benchmark::internal::RegisterBenchmarkInternal(new TestName()))
+
+// This macro will define and register a benchmark within a fixture class.
+#define BENCHMARK_F(BaseClass, Method)           \
+  BENCHMARK_PRIVATE_DECLARE_F(BaseClass, Method) \
+  BENCHMARK_REGISTER_F(BaseClass, Method);       \
+  void BaseClass##_##Method##_Benchmark::BenchmarkCase
+
+#define BENCHMARK_TEMPLATE1_F(BaseClass, Method, a)           \
+  BENCHMARK_TEMPLATE1_PRIVATE_DECLARE_F(BaseClass, Method, a) \
+  BENCHMARK_REGISTER_F(BaseClass, Method);                    \
+  void BaseClass##_##Method##_Benchmark::BenchmarkCase
+
+#define BENCHMARK_TEMPLATE2_F(BaseClass, Method, a, b)           \
+  BENCHMARK_TEMPLATE2_PRIVATE_DECLARE_F(BaseClass, Method, a, b) \
+  BENCHMARK_REGISTER_F(BaseClass, Method);                       \
+  void BaseClass##_##Method##_Benchmark::BenchmarkCase
+
+#ifdef BENCHMARK_HAS_CXX11
+#define BENCHMARK_TEMPLATE_F(BaseClass, Method, ...)           \
+  BENCHMARK_TEMPLATE_PRIVATE_DECLARE_F(BaseClass, Method, __VA_ARGS__) \
+  BENCHMARK_REGISTER_F(BaseClass, Method);                     \
+  void BaseClass##_##Method##_Benchmark::BenchmarkCase
+#else
+#define BENCHMARK_TEMPLATE_F(BaseClass, Method, a) BENCHMARK_TEMPLATE1_F(BaseClass, Method, a)
+#endif
+
+// Helper macro to create a main routine in a test that runs the benchmarks
+#define BENCHMARK_MAIN()                   \
+  int main(int argc, char** argv) {        \
+    ::benchmark::Initialize(&argc, argv);  \
+    if (::benchmark::ReportUnrecognizedArguments(argc, argv)) return 1; \
+    ::benchmark::RunSpecifiedBenchmarks(); \
+  }                                        \
+  int main(int, char**)
+
+
+// ------------------------------------------------------
+// Benchmark Reporters
+
+namespace benchmark {
+
+struct CPUInfo {
+  struct CacheInfo {
+    std::string type;
+    int level;
+    int size;
+    int num_sharing;
+  };
+
+  int num_cpus;
+  double cycles_per_second;
+  std::vector<CacheInfo> caches;
+  bool scaling_enabled;
+
+  static const CPUInfo& Get();
+
+ private:
+  CPUInfo();
+  BENCHMARK_DISALLOW_COPY_AND_ASSIGN(CPUInfo);
+};
+
+// Interface for custom benchmark result printers.
+// By default, benchmark reports are printed to stdout. However an application
+// can control the destination of the reports by calling
+// RunSpecifiedBenchmarks and passing it a custom reporter object.
+// The reporter object must implement the following interface.
+class BenchmarkReporter {
+ public:
+  struct Context {
+    CPUInfo const& cpu_info;
+    // The number of chars in the longest benchmark name.
+    size_t name_field_width;
+    static const char *executable_name;
+    Context();
+  };
+
+  struct Run {
+    Run()
+        : error_occurred(false),
+          iterations(1),
+          time_unit(kNanosecond),
+          real_accumulated_time(0),
+          cpu_accumulated_time(0),
+          bytes_per_second(0),
+          items_per_second(0),
+          max_heapbytes_used(0),
+          complexity(oNone),
+          complexity_lambda(),
+          complexity_n(0),
+          report_big_o(false),
+          report_rms(false),
+          counters() {}
+
+    std::string benchmark_name;
+    std::string report_label;  // Empty if not set by benchmark.
+    bool error_occurred;
+    std::string error_message;
+
+    int64_t iterations;
+    TimeUnit time_unit;
+    double real_accumulated_time;
+    double cpu_accumulated_time;
+
+    // Return a value representing the real time per iteration in the unit
+    // specified by 'time_unit'.
+    // NOTE: If 'iterations' is zero the returned value represents the
+    // accumulated time.
+    double GetAdjustedRealTime() const;
+
+    // Return a value representing the cpu time per iteration in the unit
+    // specified by 'time_unit'.
+    // NOTE: If 'iterations' is zero the returned value represents the
+    // accumulated time.
+    double GetAdjustedCPUTime() const;
+
+    // Zero if not set by benchmark.
+    double bytes_per_second;
+    double items_per_second;
+
+    // This is set to 0.0 if memory tracing is not enabled.
+    double max_heapbytes_used;
+
+    // Keep track of arguments to compute asymptotic complexity
+    BigO complexity;
+    BigOFunc* complexity_lambda;
+    int64_t complexity_n;
+
+    // what statistics to compute from the measurements
+    const std::vector<Statistics>* statistics;
+
+    // Inform print function whether the current run is a complexity report
+    bool report_big_o;
+    bool report_rms;
+
+    UserCounters counters;
+  };
+
+  // Construct a BenchmarkReporter with the output stream set to 'std::cout'
+  // and the error stream set to 'std::cerr'
+  BenchmarkReporter();
+
+  // Called once for every suite of benchmarks run.
+  // The parameter "context" contains information that the
+  // reporter may wish to use when generating its report, for example the
+  // platform under which the benchmarks are running. The benchmark run is
+  // never started if this function returns false, allowing the reporter
+  // to skip runs based on the context information.
+  virtual bool ReportContext(const Context& context) = 0;
+
+  // Called once for each group of benchmark runs, gives information about
+  // cpu-time and heap memory usage during the benchmark run. If the group
+  // of runs contained more than two entries then 'report' contains additional
+  // elements representing the mean and standard deviation of those runs.
+  // Additionally if this group of runs was the last in a family of benchmarks
+  // 'reports' contains additional entries representing the asymptotic
+  // complexity and RMS of that benchmark family.
+  virtual void ReportRuns(const std::vector<Run>& report) = 0;
+
+  // Called once and only once after ever group of benchmarks is run and
+  // reported.
+  virtual void Finalize() {}
+
+  // REQUIRES: The object referenced by 'out' is valid for the lifetime
+  // of the reporter.
+  void SetOutputStream(std::ostream* out) {
+    assert(out);
+    output_stream_ = out;
+  }
+
+  // REQUIRES: The object referenced by 'err' is valid for the lifetime
+  // of the reporter.
+  void SetErrorStream(std::ostream* err) {
+    assert(err);
+    error_stream_ = err;
+  }
+
+  std::ostream& GetOutputStream() const { return *output_stream_; }
+
+  std::ostream& GetErrorStream() const { return *error_stream_; }
+
+  virtual ~BenchmarkReporter();
+
+  // Write a human readable string to 'out' representing the specified
+  // 'context'.
+  // REQUIRES: 'out' is non-null.
+  static void PrintBasicContext(std::ostream* out, Context const& context);
+
+ private:
+  std::ostream* output_stream_;
+  std::ostream* error_stream_;
+};
+
+// Simple reporter that outputs benchmark data to the console. This is the
+// default reporter used by RunSpecifiedBenchmarks().
+class ConsoleReporter : public BenchmarkReporter {
+public:
+  enum OutputOptions {
+    OO_None = 0,
+    OO_Color = 1,
+    OO_Tabular = 2,
+    OO_ColorTabular = OO_Color|OO_Tabular,
+    OO_Defaults = OO_ColorTabular
+  };
+  explicit ConsoleReporter(OutputOptions opts_ = OO_Defaults)
+      : output_options_(opts_), name_field_width_(0),
+        prev_counters_(), printed_header_(false) {}
+
+  virtual bool ReportContext(const Context& context);
+  virtual void ReportRuns(const std::vector<Run>& reports);
+
+ protected:
+  virtual void PrintRunData(const Run& report);
+  virtual void PrintHeader(const Run& report);
+
+  OutputOptions output_options_;
+  size_t name_field_width_;
+  UserCounters prev_counters_;
+  bool printed_header_;
+};
+
+class JSONReporter : public BenchmarkReporter {
+ public:
+  JSONReporter() : first_report_(true) {}
+  virtual bool ReportContext(const Context& context);
+  virtual void ReportRuns(const std::vector<Run>& reports);
+  virtual void Finalize();
+
+ private:
+  void PrintRunData(const Run& report);
+
+  bool first_report_;
+};
+
+class CSVReporter : public BenchmarkReporter {
+ public:
+  CSVReporter() : printed_header_(false) {}
+  virtual bool ReportContext(const Context& context);
+  virtual void ReportRuns(const std::vector<Run>& reports);
+
+ private:
+  void PrintRunData(const Run& report);
+
+  bool printed_header_;
+  std::set< std::string > user_counter_names_;
+};
+
+inline const char* GetTimeUnitString(TimeUnit unit) {
+  switch (unit) {
+    case kMillisecond:
+      return "ms";
+    case kMicrosecond:
+      return "us";
+    case kNanosecond:
+    default:
+      return "ns";
+  }
+}
+
+inline double GetTimeUnitMultiplier(TimeUnit unit) {
+  switch (unit) {
+    case kMillisecond:
+      return 1e3;
+    case kMicrosecond:
+      return 1e6;
+    case kNanosecond:
+    default:
+      return 1e9;
+  }
+}
+
+} // namespace benchmark
+
+#endif  // BENCHMARK_BENCHMARK_H_