…module for this.
…rmance regression that warrants further investigation.
…^5000). This represents a genuine problem, but for now I'm commenting out the relevant tests to get things passing. This problem and several others are fixed on the new_api branch.
…dom methods to generate a coefficient.
… comparison against other RNGs on hackage.
… checkin. next is not a sufficient substitute for randomBounded in the Int instance. It doesn't have sufficient genRange.
…y it was slower than other types using randomBounded, but this patch circumvents the problem by using next directly.
…tional tests of randomR alongside random.