there are functions which return 8, 16, and 32-bit wide integers though they are as fast as 64-bit ones and they make no sense, because result could be MOD'ed with a constant and range functions work all the same way. Same for fp32 functions: one could easily round result of fp64 functions if this is really necessary.
instead of this nonsense it would be better to have normal distribution for fp64, for example.
there are functions which return 8, 16, and 32-bit wide integers though they are as fast as 64-bit ones and they make no sense, because result could be MOD'ed with a constant and range functions work all the same way. Same for fp32 functions: one could easily round result of fp64 functions if this is really necessary.
instead of this nonsense it would be better to have normal distribution for fp64, for example.