Speaker
Description
A good number of applications in physics (and not only) rely on
random number generation for Monte Carlo purposed (or key distribution,
and other tasks).
Most at hand random servers are hash function based, with carefully
studied and tuned algorithms. Depending on complexity and quality of
the samples produced, they can be very good quality like RANLUX (with
10^171 period), or faster, like the Mersenne Twister (x40 faster).
I present the implementation of a true-random multiplier, a code that
relies on a finite set of true-random numbers from a physical source
(in this case atmospheric noise, set of 0.2 M in the 0 ... 9999 range).
The code produces new numbers by combining any 2 random numbers in the
list, at random distance between their list positions. The random offset
relies on a shift register structure involving both the rand() hash
and numbers from the list itself, thereby producing "non-repetitive
repetitions" - i.e. the multiplier has no known period.
The tests of the multiplier are presented and they show good quality.
