Mersenne: ECM memory usage

[Steve Phipps]

I've been carrying out ECM for some time now on small exponents (under 
40,000) and I'm curious about the amount of memory that it uses.

According to readme.txt, the minimum memory required is 192 times the FFT
size. For the exponents I'm looking at, I suspect that the FFT size is of
order a kilobyte (BTW, can I look the FFT sizes and breakpoints up
somewhere?) and so the minimum memory required is less than 1MB.

However, I use mprime with the available memory set to 24MB and I've
noticed that ECM uses almost all of this. I'd like to know why ECM is 
using so much memory - does it enable it to run faster? And how much
memory would ECM 'like' to use if it was available?

Regards, 
Steve

_
Unsubscribe  list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ  -- http://www.tasam.com/~lrwiman/FAQ-mers

Re: Mersenne: ECM memory usage

[George Woltman]

Hi,

At 10:08 AM 3/29/2001 +1000, Steve Phipps wrote:
For the exponents I'm looking at, I suspect that the FFT size is of
order a kilobyte (BTW, can I look the FFT sizes and breakpoints up
somewhere?) and so the minimum memory required is less than 1MB.

A size 1024 FFT can handle exponents up to 22599.  For rough estimating
purposes, divide the exponent by 21 and round up to the next FFT size.

However, I use mprime with the available memory set to 24MB and I've
noticed that ECM uses almost all of this. I'd like to know why ECM is
using so much memory - does it enable it to run faster?

Yes, more memory can be used to save some operations.

  And how much
memory would ECM 'like' to use if it was available?

I've never studied this.  I suspect you quickly reach the point
of diminishing returns.  Try mprime with 8MB and I'll bet you'll
notice little difference.

Regards,
George

_
Unsubscribe  list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ  -- http://www.tasam.com/~lrwiman/FAQ-mers