BSD-3-Clause licensed by Florian Knupfer
Maintained by [email protected]
This version can be pinned in stack with:chronos-bench-,1961

Module documentation for


chronos performs lazy benchmarking of shell commands with continuous feedback and improving precision.

Please look at the documentation of the module for an overview of the api: Chronos.Bench

How does it work?

chronos will benchmark the specified commands until you abort it with ctrl-c or it reaches some termination criterion specified by the command line. It updates every iteration the measurements of the benchmarks so you can terminate it when you’re satisfied with the presented precision. It will intersperse all benchmarks for fastest possible overview and to distribute any external load over all benchmarks to improve precision.



chronos presents rich command line options.


You can as well use autocomplete of your shell with following command:

source <(chronos --bash-completion-script `which chronos`)

Normally, the output of --bash-completion-script should be coppied in the appropriate directory.


You can use chronos as a haskell library to benchmark pure or impure functions or shell commands.

module Main where

import Chronos.Bench

main :: IO ()
main = defaultMain
  [ bench "fib 1" fib 1
  , bench "fib 2" fib 2
  , bench "fib 4" fib 4
  , bench "fib 8" fib 8

fib :: Int -> Int
fib 1 = 1
fib 2 = 1
fib n = fib (n-1) + fib (n-2)
module Main where

import Chronos.Bench

import Control.Concurrent
import Data.IORef

main :: IO ()
main = defaultMain
  [ benchIO "ioref" (newIORef True)
  , benchIO "mvar"  (newMVar True)
  , benchIO "qsem"  (newQSem 5)
  , benchIO "chan"  (newChan :: IO (Chan Int))
module Main where

import Chronos.Bench

main :: IO ()
main = defaultMain
  [ benchShell "sleep is slow"  "sleep 0"
  , benchShell "echo is fast"   "echo"
  , benchShell "true is faster" "true"


Comparing chronos to bench and hyperfine:


  • intersperses all benchmarks, therefore allowing you fast comparisons
  • uses scientific notation
  • is more robust to external loads because ouf the interspersing (it will affect all benchmarks and not only some)
  • uses a number of significant digits according to the current standard error (so with time more digits are presented and not always 4 which is often the wrong thing to do)
  • uses bars with confidence intervals on the command line for easy comparison

Comparing chronos to criterion:


  • has got a much simpler api
  • can be used in testsuites
  • has got much simpler internals
  • has got less dependencies
  • doesn’t measure allocations
  • doesn’t measure cpu time


Revision history for chronos – 2019-03-10

  • use chronos instead of time
  • performance improvements
  • change module hierarchy – 2019-03-09

  • First version.