The_Metatron wrote:I can’t figure out why you aren’t simply treating the output you want like a bucket. A bucket in the shape of zero point shitloads_of_random_numbers.

“0.<AS_MANY_BITS_OF_RANDOM_NUMBERS_HERE_AS_YOU_LIKE>”.

Fill the bucket, use it as your desired number between zero and one. You have the hardware RNG to give you all the random numbers you want to fill that bucket.

Why are you bothering to sign it, get its absolute value then invert it?

scott1328 wrote:

I know it seems stupid, but this function "RANDOM()" is intended to be a drop-in replacement for the builtin random function that is a really poor PRN. It returns a double (64 bit floating point number with a 53 bit mantissa) in the range between 0 and 1.

Most of the applications that make use of this function convert the return value back to a signed four byte integer.

The remainder of the applications only consider the first 2 decimal digits of the return value.

I think the implementation we have is good enough. To avert the overflow error, all that is needed is to flip the most significant bit to 0 instead of using the Absolute value function. Presuming this doesn't break the randomness in any significant way.