[Haskell-cafe] [ANN] tasty-bench-0.2.5: featherlight benchmark framework

Andrew Lelechenko andrew.lelechenko at gmail.com
Sun Apr 11 19:41:54 UTC 2021

I'm happy to announce a release of tasty-bench-0.2.5.
It is a featherlight benchmark framework with API mimicking criterion and gauge.


* It supports GHCs from 7.0 to 9.2 alpha.
  Running benchmarks on a bleeding edge GHC is important to identify
  potential performance degradation before a compiler is released.

* A build on a clean machine is 16 times faster than criterion
  and 4 times faster than gauge. A build without dependencies
  is 6 times faster than criterion and 8 times faster than gauge.
  One can build benchmarks on CI without wasting excessive resources.

* Benchmarks can be written together with tests, and can itself be
  turned into a performance regression test suite.
  In fact our benchmarks are just regular tasty tests.

* Comparisons between benchmarks are readily available out of the box:
  no external tools (like criterion-compare or bench-show) are required.
  One can compare results between runs and between individual benchmarks
  instead of squinting at absolute numbers.

Best regards,

More information about the Haskell-Cafe mailing list