
1 Aug
2009
1 Aug
'09
3:40 p.m.
Sterling Clover wrote:
For other random numbers, with different properties (faster, but with tradeoffs in robustness, or ability to split, or both), you can check hackage for at least mersenne-random and gsl-random.
gsl-random is definitely worth a try for performance. I recently did a rough benchmark (not rigorous) of Control.Monad.Random, which uses System.Random, and Control.Monad.MC, which uses GSL.Random. Control.Monad.Random took about 3.6 times longer to generate a few million random numbers compared to Control.Monad.MC, and used five times as much heap. The GSL-based version generated about 340,000 random numbers per second on a 1.8GHz notebook. Anton