On Fri, 2009-12-04 at 22:51 +1100, Patrick Caldon wrote:
I'm looking for the "right" concurrency library/semantics for what should be a reasonably simple problem.
I have a little simulator:
runWorldSim :: MTGen -> SimState -> IO SimState
it takes about a second to run on a PC. It's functional except it whacks the rng, which needs IO.
Wait! This is not going to work! You cannot use the MTGen from the mersenne-random in a concurrent IO program because the C code uses a single global mutable RNG state. Your "independent" simulations would not be independent and you would not get reproducible results. Indeed you could get incorrect results or segfaults because the C code does not expect to be called from multiple threads simultaneously (there is no locking). Personally I would attack this by eliminating the IO. There's no justification for a random number generator being in IO. And look at the problems it causes! There are other MT implementations that do not use C code which assumes it's ok to use one single global mutable RNG state for an entire process. There are pure-Haskell MT impls that use mutable variables in ST but give an overall pure lazy list of random numbers. If you don't need MT specifically then there are other fast RNGs too. Duncan