hol/bench is a C++ benchmarking library. You can use it in any project by including benc.h, but if your project is using Bazel you'll get a little more out of it. The project relies on C++17, but could probably be modified to use an earlier version if needed.
Check out example/bench/example-benchmarks.cpp for some code snippets to learn from.
To use with your Bazel project, copy the directory structure of this project as-is. You'll want to copy /bench into your project as a package, and then use the build target of /example as a guide for how you can add benchmarks to any of your packages.
bazel run //example:benchmark
This will run all benchmarks defined by the specified target and output the results to stdout. (There are optional command line arguments to outputting to a file, but that's not fully implemented yet.)
bazel run bench
This will run every target in your project of the form //*:benchmark and append their results to /bench/bench_results.csv. Running on the example, your output will look something like this:
The CSV is intended to accumulate multiple runs of the same tests over time. This is useful so you can track how changes to your code have impacted performance across commits. A powerful way to do this is to copy-paste the CSV into a Google Sheets, split into columns, and then create a pivot table of mean values for test name vs datetime. You can also do neat things like graph performance across successive runs over time.
Here is a partial list of things left to implement:
- Implement proper argument parsing. Individual benchmarks currently don't handle datetime and/or appending to an output file well.
- Implement commandline filters for functions, bench sets, etc.
- Find a better naming convention than simply
//*:benchmarkso that packages can have multiple benchmark targets - Clean up display code