Re: Prng Alvo
- From: Phoenix <ribeiroalvo@xxxxxxxxx>
- Date: Sat, 18 Apr 2009 10:07:55 -0700 (PDT)
On 18 Abr, 15:27, rossum <rossu...@xxxxxxxxxxxx> wrote:
I read the paper, but I think the Diehard tests and Ent program, are
already quite strong.
However, I compared, whith the same size, many of the binary files
generated by PRNG with many binary files whith truly random numbers,
from Random.org, Hotbits and Lavarand and came to nice conclusions.
The Diehard test, have already a large number tests (All 15) including
Runs tests and outhers.
In tests with the Diehard I have not get p-values, and a lot of
results are matched with the teoric expected value.
I have made tests whith outher algoritms, (LCGs) and the resukts are
bad whit a lot of p-values.
Anouther ones are very slow and heavy.
The beauty on this formula is the simplicity, speed, and, I think, a
high quality randomness.
I asck. Mersenne Twister algorithm, always pass all this tests?
Include the Auto-correlation?