Portability | CPP, FFI |
---|---|

Stability | experimental |

Maintainer | Don Stewart <dons@galois.com> |

Tested with: GHC 6.8.2

Generate pseudo-random numbers using the SIMD-oriented Fast Mersenne Twister(SFMT)
pseudorandom number generator. This is a *much* faster generator than
the default `System.Random`

generator for Haskell (~50x faster
generation for Doubles, on a core 2 duo), however, it is not
nearly as flexible.

This library may be compiled with the '-f use_sse2' or '-f use_altivec' flags to configure, on intel and powerpc machines respectively, to enable high performance vector instructions to be used. This typically results in a 2-3x speedup in generation time.

This will work for newer intel chips such as Pentium 4s, and Core, Core2* chips.

# The random number generator

A single, global SIMD fast mersenne twister random number generator This generator is evidence that you have initialised the generator,

## Initialising the generator

newMTGen :: Maybe Word32 -> IO MTGenSource

Return an initialised SIMD Fast Mersenne Twister. The generator is initialised based on the clock time, if Nothing is passed as a seed. For deterministic behaviour, pass an explicit seed.

Due to the current SFMT library being vastly impure, currently only a single generator is allowed per-program. Attempts to reinitialise it will fail.

# Random values of various types

Instances MTRandom for Word, Word64, Word32, Word16, Word8 all return, quickly, a random inhabintant of that type, in its full range. Similarly for Int types.

Int and Word will be 32 bits on a 32 bit machine, and 64 on a 64 bit machine. The double precision will be 32 bits on a 32 bit machine, and 53 on a 64 bit machine.

The MTRandom instance for Double returns a Double in the interval [0,1). The Bool instance takes the lower bit off a random word.

Given an initialised SFMT generator, the MTRandom allows the programmer to extract values of a variety of types.

Minimal complete definition: `random`

.

# Miscellaneous

Returns the identification string for the SMFT version. The string shows the word size, the Mersenne exponent, and all parameters of this generator.

An example, calculation of pi via a monte carlo method:

import System.Random.Mersenne import System.Environment

We'll roll the dice `lim`

times,

main = do [lim] <- mapM readIO =<< getArgs

Now, define a loop that runs this many times, plotting a `x`

and `y`

position, then working out if its in and outside the circle.
The ratio of inside/total points at then gives us an approximation
of pi.

let go :: Int -> Int -> IO Double go throws ins | throws >= lim = return ((4 * fromIntegral ins) / (fromIntegral throws)) | otherwise = do x <- random g :: IO Double y <- random g :: IO Double if x * x + y * y < 1 then go (throws+1) $! ins + 1 else go (throws+1) ins

Compiling this, '-fexcess-precision', for accurate Doubles,

$ ghc -fexcess-precision -fvia-C pi.hs -o pi $ ./pi 10000000 3.1417304