GGNFS or something better?

Zeta-Flux · 2007-08-07, 19:09

I just got a new computer, and want to do a couple of factorizations using SNFS. I could go through the rigomorole of finding a pre-compiled version of GGNFS, and eventually make it work on my machine, but I thought I'd ask the following question here first:

Is GGNFS the best way to go, or is there another program available, which is better for not-too-large numbers. If so, please give (complete) directions for its installation. Thanks.

fivemack · 2007-08-07, 22:40

It depends what you mean by not-too-large. Up to about 170 digits snfs or 130 digits gnfs - say, runs that take less than a CPU-week on a K8/2300 - ggnfs and the factLat.pl script are adequate. Above that, I prefer to use the GNFS polynomial selection and sieving code, to keep the relations in text format, and to process with msieve ... it makes it easier to sieve on multiple processors, ggnfs corrupts the binary relation files worryingly often for my liking and it's not straightforward to recover once the files are corrupted.

2007-08-07, 19:09	#1
Zeta-Flux May 2003 7×13×17 Posts	GGNFS or something better? I just got a new computer, and want to do a couple of factorizations using SNFS. I could go through the rigomorole of finding a pre-compiled version of GGNFS, and eventually make it work on my machine, but I thought I'd ask the following question here first: Is GGNFS the best way to go, or is there another program available, which is better for not-too-large numbers. If so, please give (complete) directions for its installation. Thanks.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
ggnfs not working	sega24	Msieve	3	2016-03-07 23:35
GGNFS not being developed?	jux	Factoring	4	2015-12-27 03:05
GGNFS won't compile?	pakaran	Factoring	14	2015-09-05 23:36
GGNFS & 512 BIT i can puke.	silentfrog	Factoring	1	2009-09-10 08:57
ggnfs	ATH	Factoring	3	2006-08-12 22:50

2007-08-07, 22:40	#2
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 2×7×461 Posts	It depends what you mean by not-too-large. Up to about 170 digits snfs or 130 digits gnfs - say, runs that take less than a CPU-week on a K8/2300 - ggnfs and the factLat.pl script are adequate. Above that, I prefer to use the GNFS polynomial selection and sieving code, to keep the relations in text format, and to process with msieve ... it makes it easier to sieve on multiple processors, ggnfs corrupts the binary relation files worryingly often for my liking and it's not straightforward to recover once the files are corrupted.