Some times ago I encountered with question: How to get function execution time in Haskell program? I asked this question at StackOverflow, and got some useful answers. Here i will try to describe how to do it. For example we have simple haskell program which will calculate sum of prime numbers which are between 0 and 10000. Something like this:
module TimingTest where main :: IO() main = do putStrLn "Start" putStrLn ("Result: " ++ show primesSum) putStrLn "Done" -- -- Returns True if `n` is prime -- isPrime :: Int -> Bool isPrime n = null [ x | x <- [2..n - 1], n `mod` x == 0] primesSum :: Int primesSum = sum [x | x <- [2..10000], isPrime x == True]
Yes, it’s not the best implementation of prime numbers, but it’s not important at the current moment. Let’s see what we have for checking execution time.
First of all, the simplest method to get execution time is time command. Compile our source code and execute:
$ time ./TimingTest
We must get something like this:
real 0m3.503s user 0m3.492s sys 0m0.004s
The second method is just add
:set +s in
ghci before the function execution. Of course it’s not the best method, because functions run much slower in `
The third method is to use TimeIt haskell library by Lennart Augustsson. Very little, but useful library with simple API. It consist only from two functions:
timeIt :: IO a -> IO a -- | Wrap an IO computation so that it prints out the execution time
timeItT :: IO a -> IO (Double, a)Source -- | Wrap an IO computation so that it returns execution time is seconds as well as the real value.
Let’s remake our main function as:
import System.TimeIt main :: IO() main = do putStrLn "Start" timeIt $ putStrLn ("Result: " ++ show primesSum) putStrLn "End"
and will get something like this:
Start Result: 5736396 CPU time: 8.22s
Criterion library provides a powerful but simple way to measure software performance by Bryan O’Sullivan. For using it, will remake again our main function as:
module Main where import Criterion.Main main :: IO() main = defaultMain [ bgroup "Prime numbers." [bench "prime numbers benchmark" $ whnfIO (putStrLn $ show primesSum)] ]
and as a result we will get:
estimating clock resolution... mean is 3.760062 us (160001 iterations) found 3006 outliers among 159999 samples (1.9%) 2461 (1.5%) high severe estimating cost of a clock call... mean is 98.08812 ns (28 iterations) found 4 outliers among 28 samples (14.3%) 4 (14.3%) low severe