1# Benchmark Tools 2 3## compare_bench.py 4 5The `compare_bench.py` utility which can be used to compare the result of benchmarks. 6The program is invoked like: 7 8``` bash 9$ compare_bench.py <old-benchmark> <new-benchmark> [benchmark options]... 10``` 11 12Where `<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. 13 14`[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. 15 16The sample output using the JSON test files under `Inputs/` gives: 17 18``` bash 19$ ./compare_bench.py ./gbench/Inputs/test1_run1.json ./gbench/Inputs/test1_run2.json 20Comparing ./gbench/Inputs/test1_run1.json to ./gbench/Inputs/test1_run2.json 21Benchmark Time CPU Time Old Time New CPU Old CPU New 22------------------------------------------------------------------------------------------------------------- 23BM_SameTimes +0.0000 +0.0000 10 10 10 10 24BM_2xFaster -0.5000 -0.5000 50 25 50 25 25BM_2xSlower +1.0000 +1.0000 50 100 50 100 26BM_1PercentFaster -0.0100 -0.0100 100 99 100 99 27BM_1PercentSlower +0.0100 +0.0100 100 101 100 101 28BM_10PercentFaster -0.1000 -0.1000 100 90 100 90 29BM_10PercentSlower +0.1000 +0.1000 100 110 100 110 30BM_100xSlower +99.0000 +99.0000 100 10000 100 10000 31BM_100xFaster -0.9900 -0.9900 10000 100 10000 100 32BM_10PercentCPUToTime +0.1000 -0.1000 100 110 100 90 33BM_ThirdFaster -0.3333 -0.3334 100 67 100 67 34BM_BadTimeUnit -0.9000 +0.2000 0 0 0 1 35``` 36 37As you can note, the values in `Time` and `CPU` columns are calculated as `(new - old) / |old|`. 38 39When 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: 40 41``` 42./compare_bench.py test/basic_test test/basic_test --benchmark_filter=BM_empty.* 43RUNNING: test/basic_test --benchmark_filter=BM_empty.* --benchmark_out=/tmp/tmpN7LF3a 44Run on (8 X 4000 MHz CPU s) 452017-11-07 23:28:36 46--------------------------------------------------------------------- 47Benchmark Time CPU Iterations 48--------------------------------------------------------------------- 49BM_empty 4 ns 4 ns 170178757 50BM_empty/threads:8 1 ns 7 ns 103868920 51BM_empty_stop_start 0 ns 0 ns 1000000000 52BM_empty_stop_start/threads:8 0 ns 0 ns 1403031720 53RUNNING: /test/basic_test --benchmark_filter=BM_empty.* --benchmark_out=/tmp/tmplvrIp8 54Run on (8 X 4000 MHz CPU s) 552017-11-07 23:28:38 56--------------------------------------------------------------------- 57Benchmark Time CPU Iterations 58--------------------------------------------------------------------- 59BM_empty 4 ns 4 ns 169534855 60BM_empty/threads:8 1 ns 7 ns 104188776 61BM_empty_stop_start 0 ns 0 ns 1000000000 62BM_empty_stop_start/threads:8 0 ns 0 ns 1404159424 63Comparing ../build/test/basic_test to ../build/test/basic_test 64Benchmark Time CPU Time Old Time New CPU Old CPU New 65--------------------------------------------------------------------------------------------------------------------- 66BM_empty -0.0048 -0.0049 4 4 4 4 67BM_empty/threads:8 -0.0123 -0.0054 1 1 7 7 68BM_empty_stop_start -0.0000 -0.0000 0 0 0 0 69BM_empty_stop_start/threads:8 -0.0029 +0.0001 0 0 0 0 70 71``` 72 73As you can note, the values in `Time` and `CPU` columns are calculated as `(new - old) / |old|`. 74Obviously 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. 75 76## compare.py 77 78The `compare.py` can be used to compare the result of benchmarks. 79There are three modes of operation: 80 811. Just compare two benchmarks, what `compare_bench.py` did. 82The program is invoked like: 83 84``` bash 85$ compare.py benchmarks <benchmark_baseline> <benchmark_contender> [benchmark options]... 86``` 87Where `<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. 88 89`[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. 90 91Example output: 92``` 93$ ./compare.py benchmarks ./a.out ./a.out 94RUNNING: ./a.out --benchmark_out=/tmp/tmprBT5nW 95Run on (8 X 4000 MHz CPU s) 962017-11-07 21:16:44 97------------------------------------------------------ 98Benchmark Time CPU Iterations 99------------------------------------------------------ 100BM_memcpy/8 36 ns 36 ns 19101577 211.669MB/s 101BM_memcpy/64 76 ns 76 ns 9412571 800.199MB/s 102BM_memcpy/512 84 ns 84 ns 8249070 5.64771GB/s 103BM_memcpy/1024 116 ns 116 ns 6181763 8.19505GB/s 104BM_memcpy/8192 643 ns 643 ns 1062855 11.8636GB/s 105BM_copy/8 222 ns 222 ns 3137987 34.3772MB/s 106BM_copy/64 1608 ns 1608 ns 432758 37.9501MB/s 107BM_copy/512 12589 ns 12589 ns 54806 38.7867MB/s 108BM_copy/1024 25169 ns 25169 ns 27713 38.8003MB/s 109BM_copy/8192 201165 ns 201112 ns 3486 38.8466MB/s 110RUNNING: ./a.out --benchmark_out=/tmp/tmpt1wwG_ 111Run on (8 X 4000 MHz CPU s) 1122017-11-07 21:16:53 113------------------------------------------------------ 114Benchmark Time CPU Iterations 115------------------------------------------------------ 116BM_memcpy/8 36 ns 36 ns 19397903 211.255MB/s 117BM_memcpy/64 73 ns 73 ns 9691174 839.635MB/s 118BM_memcpy/512 85 ns 85 ns 8312329 5.60101GB/s 119BM_memcpy/1024 118 ns 118 ns 6438774 8.11608GB/s 120BM_memcpy/8192 656 ns 656 ns 1068644 11.6277GB/s 121BM_copy/8 223 ns 223 ns 3146977 34.2338MB/s 122BM_copy/64 1611 ns 1611 ns 435340 37.8751MB/s 123BM_copy/512 12622 ns 12622 ns 54818 38.6844MB/s 124BM_copy/1024 25257 ns 25239 ns 27779 38.6927MB/s 125BM_copy/8192 205013 ns 205010 ns 3479 38.108MB/s 126Comparing ./a.out to ./a.out 127Benchmark Time CPU Time Old Time New CPU Old CPU New 128------------------------------------------------------------------------------------------------------ 129BM_memcpy/8 +0.0020 +0.0020 36 36 36 36 130BM_memcpy/64 -0.0468 -0.0470 76 73 76 73 131BM_memcpy/512 +0.0081 +0.0083 84 85 84 85 132BM_memcpy/1024 +0.0098 +0.0097 116 118 116 118 133BM_memcpy/8192 +0.0200 +0.0203 643 656 643 656 134BM_copy/8 +0.0046 +0.0042 222 223 222 223 135BM_copy/64 +0.0020 +0.0020 1608 1611 1608 1611 136BM_copy/512 +0.0027 +0.0026 12589 12622 12589 12622 137BM_copy/1024 +0.0035 +0.0028 25169 25257 25169 25239 138BM_copy/8192 +0.0191 +0.0194 201165 205013 201112 205010 139``` 140 141What 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. 142As you can note, the values in `Time` and `CPU` columns are calculated as `(new - old) / |old|`. 143 1442. Compare two different filters of one benchmark 145The program is invoked like: 146 147``` bash 148$ compare.py filters <benchmark> <filter_baseline> <filter_contender> [benchmark options]... 149``` 150Where `<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. 151 152Where `<filter_baseline>` and `<filter_contender>` are the same regex filters that you would pass to the `[--benchmark_filter=<regex>]` parameter of the benchmark binary. 153 154`[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. 155 156Example output: 157``` 158$ ./compare.py filters ./a.out BM_memcpy BM_copy 159RUNNING: ./a.out --benchmark_filter=BM_memcpy --benchmark_out=/tmp/tmpBWKk0k 160Run on (8 X 4000 MHz CPU s) 1612017-11-07 21:37:28 162------------------------------------------------------ 163Benchmark Time CPU Iterations 164------------------------------------------------------ 165BM_memcpy/8 36 ns 36 ns 17891491 211.215MB/s 166BM_memcpy/64 74 ns 74 ns 9400999 825.646MB/s 167BM_memcpy/512 87 ns 87 ns 8027453 5.46126GB/s 168BM_memcpy/1024 111 ns 111 ns 6116853 8.5648GB/s 169BM_memcpy/8192 657 ns 656 ns 1064679 11.6247GB/s 170RUNNING: ./a.out --benchmark_filter=BM_copy --benchmark_out=/tmp/tmpAvWcOM 171Run on (8 X 4000 MHz CPU s) 1722017-11-07 21:37:33 173---------------------------------------------------- 174Benchmark Time CPU Iterations 175---------------------------------------------------- 176BM_copy/8 227 ns 227 ns 3038700 33.6264MB/s 177BM_copy/64 1640 ns 1640 ns 426893 37.2154MB/s 178BM_copy/512 12804 ns 12801 ns 55417 38.1444MB/s 179BM_copy/1024 25409 ns 25407 ns 27516 38.4365MB/s 180BM_copy/8192 202986 ns 202990 ns 3454 38.4871MB/s 181Comparing BM_memcpy to BM_copy (from ./a.out) 182Benchmark Time CPU Time Old Time New CPU Old CPU New 183-------------------------------------------------------------------------------------------------------------------- 184[BM_memcpy vs. BM_copy]/8 +5.2829 +5.2812 36 227 36 227 185[BM_memcpy vs. BM_copy]/64 +21.1719 +21.1856 74 1640 74 1640 186[BM_memcpy vs. BM_copy]/512 +145.6487 +145.6097 87 12804 87 12801 187[BM_memcpy vs. BM_copy]/1024 +227.1860 +227.1776 111 25409 111 25407 188[BM_memcpy vs. BM_copy]/8192 +308.1664 +308.2898 657 202986 656 202990 189``` 190 191As 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. 192As you can note, the values in `Time` and `CPU` columns are calculated as `(new - old) / |old|`. 193 1943. Compare filter one from benchmark one to filter two from benchmark two: 195The program is invoked like: 196 197``` bash 198$ compare.py filters <benchmark_baseline> <filter_baseline> <benchmark_contender> <filter_contender> [benchmark options]... 199``` 200 201Where `<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. 202 203Where `<filter_baseline>` and `<filter_contender>` are the same regex filters that you would pass to the `[--benchmark_filter=<regex>]` parameter of the benchmark binary. 204 205`[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. 206 207Example output: 208``` 209$ ./compare.py benchmarksfiltered ./a.out BM_memcpy ./a.out BM_copy 210RUNNING: ./a.out --benchmark_filter=BM_memcpy --benchmark_out=/tmp/tmp_FvbYg 211Run on (8 X 4000 MHz CPU s) 2122017-11-07 21:38:27 213------------------------------------------------------ 214Benchmark Time CPU Iterations 215------------------------------------------------------ 216BM_memcpy/8 37 ns 37 ns 18953482 204.118MB/s 217BM_memcpy/64 74 ns 74 ns 9206578 828.245MB/s 218BM_memcpy/512 91 ns 91 ns 8086195 5.25476GB/s 219BM_memcpy/1024 120 ns 120 ns 5804513 7.95662GB/s 220BM_memcpy/8192 664 ns 664 ns 1028363 11.4948GB/s 221RUNNING: ./a.out --benchmark_filter=BM_copy --benchmark_out=/tmp/tmpDfL5iE 222Run on (8 X 4000 MHz CPU s) 2232017-11-07 21:38:32 224---------------------------------------------------- 225Benchmark Time CPU Iterations 226---------------------------------------------------- 227BM_copy/8 230 ns 230 ns 2985909 33.1161MB/s 228BM_copy/64 1654 ns 1653 ns 419408 36.9137MB/s 229BM_copy/512 13122 ns 13120 ns 53403 37.2156MB/s 230BM_copy/1024 26679 ns 26666 ns 26575 36.6218MB/s 231BM_copy/8192 215068 ns 215053 ns 3221 36.3283MB/s 232Comparing BM_memcpy (from ./a.out) to BM_copy (from ./a.out) 233Benchmark Time CPU Time Old Time New CPU Old CPU New 234-------------------------------------------------------------------------------------------------------------------- 235[BM_memcpy vs. BM_copy]/8 +5.1649 +5.1637 37 230 37 230 236[BM_memcpy vs. BM_copy]/64 +21.4352 +21.4374 74 1654 74 1653 237[BM_memcpy vs. BM_copy]/512 +143.6022 +143.5865 91 13122 91 13120 238[BM_memcpy vs. BM_copy]/1024 +221.5903 +221.4790 120 26679 120 26666 239[BM_memcpy vs. BM_copy]/8192 +322.9059 +323.0096 664 215068 664 215053 240``` 241This is a mix of the previous two modes, two (potentially different) benchmark binaries are run, and a different filter is applied to each one. 242As you can note, the values in `Time` and `CPU` columns are calculated as `(new - old) / |old|`. 243