1HYPERFINE(1) General Commands Manual HYPERFINE(1)
2
6 hyperfine - command-line benchmarking tool
7
9 hyperfine [-ihV] [-w warmupruns] [-r runs] [-p cmd...] [-c cmd] [-s
10 style] [cmd...]
11
13 A command-line benchmarking tool which includes:
14 * Statistical analysis across multiple runs
16 * Support for arbitrary shell commands
17 * Constant feedback about the benchmark progress and current es‐
18 timates
19 * Warmup runs can be executed before the actual benchmark
20 * Cache-clearing commands can be set up before each timing run
21 * Statistical outlier detection to detect interference from
22 other programs and caching effects
23 * Export results to various formats: CSV, JSON, Markdown, Asci‐
24 iDoc
25 * Parameterized benchmarks (e.g. vary the number of threads)
26
28 -w, --warmup warmupruns
29 Perform warmupruns (number) before the actual benchmark. This
31 can be used to fill (disk) caches for I/O-heavy programs.
32 -m, --min-runs minruns
34 Perform at least minruns (number) runs for each command. De‐
36 fault: 10.
37 -M, --max-runs maxruns
39 Perform at most maxruns (number) runs for each command. Default:
41 no limit.
42 -r, --runs runs
44 Perform exactly runs (number) runs for each command. If this op‐
46 tion is not specified, hyperfine automatically determines the
47 number of runs.
48 -s, --setup cmd
50 Execute cmd once before each set of timing runs. This is useful
52 for compiling software or doing other one-off setup. The
53 --setup option can only be specified once.
54 -p, --prepare cmd...
56 Execute cmd before each timing run. This is useful for clearing
58 disk caches, for example. The --prepare option can be specified
59 once for all commands or multiple times, once for each command.
60 In the latter case, each preparation command will be run prior
61 to the corresponding benchmark command.
62 -c, --cleanup cmd
64 Execute cmd after the completion of all benchmarking runs for
66 each individual command to be benchmarked. This is useful if the
67 commands to be benchmarked produce artifacts that need to be
68 cleaned up.
69 -P, --parameter-scan var min max
71 Perform benchmark runs for each value in the range min..max. Re‐
73 places the string '{var}' in each command by the current parame‐
74 ter value.
75 Example:
77 hyperfine -P threads 1 8 'make -j {threads}'
78 This performs benchmarks for 'make -j 1', 'make -j 2', ...,
80 'make -j 8'.
81 -D, --parameter-step-size delta
83 This argument requires --parameter-scan to be specified as well.
85 Traverse the range min..max in steps of delta.
86 Example:
88 hyperfine -P delay 0.3 0.7 -D 0.2 'sleep {delay}'
89 This performs benchmarks for 'sleep 0.3', 'sleep 0.5' and 'sleep
91 0.7'.
92 -L, --parameter-list var values
94 Perform benchmark runs for each value in the comma-separated
96 list of values. Replaces the string '{var}' in each command by
97 the current parameter value.
98 Example:
100 hyperfine -L compiler gcc,clang '{compiler} -O2 main.cpp'
101 This performs benchmarks for 'gcc -O2 main.cpp' and 'clang -O2
103 main.cpp'.
104 --style type
106 Set output style type (default: auto). Set this to 'basic' to
108 disable output coloring and interactive elements. Set it to
109 'full' to enable all effects even if no interactive terminal was
110 detected. Set this to 'nocolor' to keep the interactive output
111 without any colors. Set this to 'color' to keep the colors with‐
112 out any interactive output. Set this to 'none' to disable all
113 the output of the tool. In hyperfine versions v0.4.0..v1.12.0
114 this option took the -s, short option, which is now used by
115 --setup.
116 -S, --shell shell
118 Set the shell to use for executing benchmarked commands.
120 -i, --ignore-failure
122 Ignore non-zero exit codes of the benchmarked programs.
124 -u, --time-unit unit
126 Set the time unit to be used. Default: second. Possible values:
128 millisecond, second.
129 --export-asciidoc file
131 Export the timing summary statistics as an AsciiDoc table to the
133 given file.
134 --export-csv file
136 Export the timing summary statistics as CSV to the given file.
138 If you need the timing results for each individual run, use the
139 JSON export format.
140 --export-json file
142 Export the timing summary statistics and timings of individual
144 runs as JSON to the given file.
145 --export-markdown file
147 Export the timing summary statistics as a Markdown table to the
149 given file.
150 --show-output
152 Print the stdout and stderr of the benchmark instead of sup‐
154 pressing it. This will increase the time it takes for benchmarks
155 to run, so it should only be used for debugging purposes or when
156 trying to benchmark output speed.
157 -n, --command-name name
159 Identify a command with the given name. Commands and names are
161 paired in the same order: the first command executed gets the
162 first name passed as option.
163 -h, --help
165 Print help message.
167 -V, --version
169 Show version information.
171
173 Basic benchmark of 'find . -name todo.txt':
174 hyperfine 'find . -name todo.txt'
175 Perform benchmarks for 'sleep 0.2' and 'sleep 3.2' with a minimum 5
177 runs each:
178 hyperfine --min-runs 5 'sleep 0.2' 'sleep 3.2'
179 Perform a benchmark of 'grep' with a warm disk cache by executing 3
181 runs up front that are not part of the measurement:
182 hyperfine --warmup 3 'grep -R TODO *'
183 Export the results of a parameter scan benchmark to a markdown table:
185 hyperfine --export-markdown output.md --parameter-scan time 1 5 'sleep {time}'
186 Demonstrate when each of --setup, --prepare, cmd and --cleanup will
188 run:
189 hyperfine -L n 1,2 -r 2 --show-output \
190 --setup 'echo setup n={n}' \
191 --prepare 'echo prepare={n}' \
192 --cleanup 'echo cleanup n={n}' \
193 'echo command n={n}'
194
196 David Peter (sharkdp)
197 Source, bug tracker, and additional information can be found on GitHub
199 at: https://github.com/sharkdp/hyperfine
200 HYPERFINE(1)