|
| 1 | +// (C) Copyright Matt Borland 2022. |
| 2 | +// Use, modification and distribution are subject to the |
| 3 | +// Boost Software License, Version 1.0. (See accompanying file |
| 4 | +// LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt) |
| 5 | + |
| 6 | +#include <boost/math/special_functions/next.hpp> |
| 7 | +#include <benchmark/benchmark.h> |
| 8 | + |
| 9 | +template <typename T> |
| 10 | +void float_distance(benchmark::State& state) |
| 11 | +{ |
| 12 | + const auto difference = static_cast<int>(state.range(0)); |
| 13 | + T left = 2; |
| 14 | + T right = boost::math::float_advance(left, difference); |
| 15 | + |
| 16 | + for (auto _ : state) |
| 17 | + { |
| 18 | + benchmark::DoNotOptimize(boost::math::float_distance(left, right)); |
| 19 | + } |
| 20 | + state.SetComplexityN(state.range(0)); |
| 21 | +} |
| 22 | + |
| 23 | +BENCHMARK_TEMPLATE(float_distance, float)->RangeMultiplier(2)->Range(1 << 1, 1 << 14)->Complexity()->UseRealTime(); |
| 24 | +BENCHMARK_TEMPLATE(float_distance, double)->RangeMultiplier(2)->Range(1 << 1, 1 << 14)->Complexity()->UseRealTime(); |
| 25 | + |
| 26 | +BENCHMARK_MAIN(); |
| 27 | + |
| 28 | +/* |
| 29 | +Run on Apple M1 Pro Arch using Apple Clang 14.0.0 (15OCT22) |
| 30 | +
|
| 31 | +Original performance (Boost 1.80.0): |
| 32 | +
|
| 33 | +Unable to determine clock rate from sysctl: hw.cpufrequency: No such file or directory |
| 34 | +This does not affect benchmark measurements, only the metadata output. |
| 35 | +2022-10-15T15:24:07-07:00 |
| 36 | +Running ./new_next_performance |
| 37 | +Run on (10 X 24.0916 MHz CPU s) |
| 38 | +CPU Caches: |
| 39 | + L1 Data 64 KiB |
| 40 | + L1 Instruction 128 KiB |
| 41 | + L2 Unified 4096 KiB (x10) |
| 42 | +Load Average: 1.86, 2.53, 5.83 |
| 43 | +--------------------------------------------------------------------------------- |
| 44 | +Benchmark Time CPU Iterations |
| 45 | +--------------------------------------------------------------------------------- |
| 46 | +float_distance<float>/2/real_time 61.4 ns 61.4 ns 9074469 |
| 47 | +float_distance<float>/4/real_time 61.7 ns 61.7 ns 11384150 |
| 48 | +float_distance<float>/8/real_time 61.4 ns 61.4 ns 10814604 |
| 49 | +float_distance<float>/16/real_time 61.7 ns 61.7 ns 11348376 |
| 50 | +float_distance<float>/32/real_time 61.4 ns 61.4 ns 11387167 |
| 51 | +float_distance<float>/64/real_time 61.6 ns 61.6 ns 11131932 |
| 52 | +float_distance<float>/128/real_time 61.4 ns 61.4 ns 11382029 |
| 53 | +float_distance<float>/256/real_time 61.4 ns 61.4 ns 11307649 |
| 54 | +float_distance<float>/512/real_time 61.4 ns 61.4 ns 11376048 |
| 55 | +float_distance<float>/1024/real_time 61.4 ns 61.4 ns 11355748 |
| 56 | +float_distance<float>/2048/real_time 61.8 ns 61.8 ns 11373776 |
| 57 | +float_distance<float>/4096/real_time 61.4 ns 61.4 ns 11382368 |
| 58 | +float_distance<float>/8192/real_time 61.4 ns 61.4 ns 11353453 |
| 59 | +float_distance<float>/16384/real_time 61.4 ns 61.4 ns 11378298 |
| 60 | +float_distance<float>/real_time_BigO 61.48 (1) 61.47 (1) |
| 61 | +float_distance<float>/real_time_RMS 0 % 0 % |
| 62 | +float_distance<double>/2/real_time 55.6 ns 55.6 ns 12580218 |
| 63 | +float_distance<double>/4/real_time 55.6 ns 55.6 ns 12577835 |
| 64 | +float_distance<double>/8/real_time 55.6 ns 55.6 ns 12564909 |
| 65 | +float_distance<double>/16/real_time 56.2 ns 56.2 ns 12554909 |
| 66 | +float_distance<double>/32/real_time 56.0 ns 56.0 ns 12544381 |
| 67 | +float_distance<double>/64/real_time 55.6 ns 55.6 ns 12566488 |
| 68 | +float_distance<double>/128/real_time 55.6 ns 55.6 ns 12499581 |
| 69 | +float_distance<double>/256/real_time 55.6 ns 55.6 ns 12565661 |
| 70 | +float_distance<double>/512/real_time 56.1 ns 56.1 ns 12550023 |
| 71 | +float_distance<double>/1024/real_time 55.8 ns 55.8 ns 12568603 |
| 72 | +float_distance<double>/2048/real_time 55.6 ns 55.6 ns 12546049 |
| 73 | +float_distance<double>/4096/real_time 55.6 ns 55.6 ns 12528525 |
| 74 | +float_distance<double>/8192/real_time 55.9 ns 55.9 ns 12563030 |
| 75 | +float_distance<double>/16384/real_time 56.0 ns 56.0 ns 12447644 |
| 76 | +float_distance<double>/real_time_BigO 55.78 (1) 55.78 (1) |
| 77 | +float_distance<double>/real_time_RMS 0 % 0 % |
| 78 | +
|
| 79 | +New performance: |
| 80 | +
|
| 81 | +Unable to determine clock rate from sysctl: hw.cpufrequency: No such file or directory |
| 82 | +This does not affect benchmark measurements, only the metadata output. |
| 83 | +2022-10-15T15:31:37-07:00 |
| 84 | +Running ./new_next_performance |
| 85 | +Run on (10 X 24.122 MHz CPU s) |
| 86 | +CPU Caches: |
| 87 | + L1 Data 64 KiB |
| 88 | + L1 Instruction 128 KiB |
| 89 | + L2 Unified 4096 KiB (x10) |
| 90 | +Load Average: 2.12, 2.17, 4.26 |
| 91 | +--------------------------------------------------------------------------------- |
| 92 | +Benchmark Time CPU Iterations |
| 93 | +--------------------------------------------------------------------------------- |
| 94 | +float_distance<float>/2/real_time 15.8 ns 15.8 ns 42162717 |
| 95 | +float_distance<float>/4/real_time 15.9 ns 15.9 ns 44213877 |
| 96 | +float_distance<float>/8/real_time 15.8 ns 15.8 ns 43972542 |
| 97 | +float_distance<float>/16/real_time 15.8 ns 15.8 ns 44209456 |
| 98 | +float_distance<float>/32/real_time 15.8 ns 15.8 ns 44200244 |
| 99 | +float_distance<float>/64/real_time 15.8 ns 15.8 ns 44239293 |
| 100 | +float_distance<float>/128/real_time 15.8 ns 15.8 ns 44171202 |
| 101 | +float_distance<float>/256/real_time 15.8 ns 15.8 ns 44241507 |
| 102 | +float_distance<float>/512/real_time 15.9 ns 15.8 ns 44230034 |
| 103 | +float_distance<float>/1024/real_time 15.8 ns 15.8 ns 44241554 |
| 104 | +float_distance<float>/2048/real_time 15.8 ns 15.8 ns 44220802 |
| 105 | +float_distance<float>/4096/real_time 15.8 ns 15.8 ns 44220441 |
| 106 | +float_distance<float>/8192/real_time 15.9 ns 15.9 ns 44213994 |
| 107 | +float_distance<float>/16384/real_time 15.8 ns 15.8 ns 44215413 |
| 108 | +float_distance<float>/real_time_BigO 15.83 (1) 15.83 (1) |
| 109 | +float_distance<float>/real_time_RMS 0 % 0 % |
| 110 | +float_distance<double>/2/real_time 15.5 ns 15.5 ns 45098165 |
| 111 | +float_distance<double>/4/real_time 15.6 ns 15.6 ns 45065465 |
| 112 | +float_distance<double>/8/real_time 15.5 ns 15.5 ns 45058733 |
| 113 | +float_distance<double>/16/real_time 15.8 ns 15.7 ns 45078404 |
| 114 | +float_distance<double>/32/real_time 15.5 ns 15.5 ns 44832734 |
| 115 | +float_distance<double>/64/real_time 15.5 ns 15.5 ns 45077303 |
| 116 | +float_distance<double>/128/real_time 15.5 ns 15.5 ns 45067255 |
| 117 | +float_distance<double>/256/real_time 15.5 ns 15.5 ns 45073844 |
| 118 | +float_distance<double>/512/real_time 15.6 ns 15.6 ns 45109342 |
| 119 | +float_distance<double>/1024/real_time 15.5 ns 15.5 ns 44845180 |
| 120 | +float_distance<double>/2048/real_time 15.5 ns 15.5 ns 45051846 |
| 121 | +float_distance<double>/4096/real_time 15.5 ns 15.5 ns 45064317 |
| 122 | +float_distance<double>/8192/real_time 15.5 ns 15.5 ns 45115653 |
| 123 | +float_distance<double>/16384/real_time 15.5 ns 15.5 ns 45067642 |
| 124 | +float_distance<double>/real_time_BigO 15.54 (1) 15.54 (1) |
| 125 | +float_distance<double>/real_time_RMS 0 % 0 % |
| 126 | +*/ |
0 commit comments