p: and q: implement Miller-Rabin as modelled by Cliff Reiter.
You can see the model by following the link in the "q: improved"
item in the J5.04 release notes.
http://www.jsoftware.com/books/help/release/qco.htm

If the test says a number is composite then it is composite;
if it says a number is prime then the probability of an error 
is 0.25^100.  Calling 1&p: or q: more than once on the same
argument does not buy you anything.



----- Original Message -----
From: Andrew Nikitin <[EMAIL PROTECTED]>
Date: Saturday, June 17, 2006 6:33 pm
Subject: [Jprogramming] details on 1&p:

> Dictionary says about 1&p:
> "Currently, arguments larger than 2^31 are tested to be prime 
> according to a 
> probabilistic algorithm (Miller-Rabin)."
> 
> Miller Rabin (as described in a book) is supposed to choose a at 
> random, 
> raise it to some power (modulo n) and test the result for certain 
> property. 
> This test process is repeated until desired probability is reached.
> Few questions:
> How J chooses a?
> How many times it performs the test?
> Does it choose different a if 1&p: called on same number again?
> 
> The point of my questions is, if I want to test that number is 
> prime with 
> probability of mistake less then, say 2^-200, does it make sense 
> to call 
> 1&p: more than once, and if yes, how many times (assuming 1/4 per 
> random a).


----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to