Revision history [back]

I'm doing some largish experiments involving randomness, and I think random number generation may be the bottleneck. Is there a way to make the random number generator significantly faster without sacrificing the validity of results? I don't especially care that the RNG be cryptographically secure or anything like that.