At 03:54 PM 2/10/00 +0800, "Dave Mullen" <[EMAIL PROTECTED]> wrote:
>
>1) For the LL
>test using FFT multiplys, we have to Transform the data, Multiply the
>matrices, Inverse Transform the data, then subtract 2. I am assuming that
>the only reason we do the Forward and Inverse Transform each time is so we
>can subtract the 2 ? Please correct me here if I'm wrong !
I found this in the Lucas Wiman FAQ. ([EMAIL PROTECTED]). Can't remember
the link though.
My only question would be, "Can we not figure out a way to stay in the
frequency domain and correct the errors internally?"
Dan-Somebody-MilesDaniel
>From the Lucas Wiman FAQ.:
Q:6.4: In Q2.6, you say that it doesn't make sense to use a probable prime
test. But wouldn't it make sense to do this, since you could stay in
frequency space throughout the whole test, and save the time consuming
FFT's and IFFT's?
A: Well, first of all, you could still do this with the LL test
(since x^2-2 has a Fourier transform, or by subtracting the convolution
product of ...00001*...00002). However, the FFT's and IFFT's make certain
that errors don't occur (by transforming back into time space, we can make
certain that errors don't accumulate, remember this is an irrational base).
This also allows "quick checks" e.g. the sum of the squares of the
digits=sum of digits in the square (this takes almost no time at all since
we have to manipulate the number anyway.)
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers