Excellent news! How did you test it? One way would be to make it callable by C and then use testu01 but perhaps you have already done this or similar? See here for a bit more detail: https://idontgetoutmuch.wordpress.com/2017/01/14/calling-haskell-from-c/ https://idontgetoutmuch.wordpress.com/2017/01/14/calling-haskell-from-c/. Testing the split function might need thinking about though.
I recently made a SplitMix implentation: http://hackage.haskell.org/package/splitmix http://hackage.haskell.org/package/splitmix
When generating `Word64`s it's noticeable faster than `tf-random` (and `mwc-random` and obviously `random). I didn't benchmark `Double` generation, as `tf-random` own `Random` class doesn't support it [1], and using `random`'s class would be very unfair.
I haven't announced it (well, mentioned on Twitter), but please give it a try!
- Oleg
Dominic Steinitz dominic@steinitz.org http://idontgetoutmuch.wordpress.com