Just a Curiosity - Page 2

VBCurtis · 2020-06-16, 15:45

That is an excellent and superior viewpoint. As a non-programmer, I tend to think in terms of "can this be improved for little effort?", and running CADO as a black-box that receives input and returns factors seems much easier to implement than breaking out individual tasks and coding individual calls for those things. Your scripts already do those individual things, so you know from experience they can be implemented; I do not, so I advocated for the simplest change that adds speed (CADO as black box).

bsquared · 2020-06-16, 16:08

The CADO-as-a-black-box approach might not be too bad to add to yafu... I really don't want to have to deal with param files, managing different stages, etc. But if someone already has good params set up and I can just use a system call to "cado.py -n 12489...123971", then wait to read and parse an output file, that seems doable.

Would have to figure out how to run tune without running entire jobs though. Anyone know if that can be done (i.e., request to only collect N relations)?

EdH · 2020-06-16, 16:37

Quote:

Originally Posted by bsquared

The CADO-as-a-black-box approach might not be too bad to add to yafu... I really don't want to have to deal with param files, managing different stages, etc. But if someone already has good params set up and I can just use a system call to "cado.py -n 12489...123971", then wait to read and parse an output file, that seems doable.

Would have to figure out how to run tune without running entire jobs though. Anyone know if that can be done (i.e., request to only collect N relations)?

Actually, I can give more details later, but once you set up CADO-NFS, it is basically run as simply as:

Code:

go to cado-nfs directory
 ,/cado-nfs.py <number>
go get factors

During the setup and periodically afterwords, the param files can be updated by simply replacing the existing ones. And, yes, you can tell it the number of relations wanted and to stop before LA, etc. I can give you more details later. Or, maybe VBCurtis can.

EdH · 2020-06-16, 17:24

A couple more details to consider:

CADO-NFS, on its own, uses a filename base from its parameter files and places all data in a randomly named folder in the /tmp directory. Both of these need to be addressed (easily done) in order to make retrieving factors easier. They can both be addressed simply by setting the working directory and filename on the command line. In fact, all (or, close to all) parameters can be overriden by command line values

If the /tmp directory is used, a computer reset will remove all work, so a more permanent location should be used for anything of any length.

Starting a new CADO-NFS number will fail if prior number data is encountered, so any previous data should be removed once factors are retrieved. The random directory naming keeps this from happening, but makes factor retrieval more complex.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Results Curiosity	Gordon	PrimeNet	1	2015-07-30 06:49
I see this as a bit of a mathematical curiosity...	petrw1	Math	4	2015-07-19 02:33
A Curiosity	R.D. Silverman	GMP-ECM	4	2012-05-10 20:00
P-1 question of curiosity	JuanTutors	Math	3	2004-10-20 04:37
Curiosity	ET_	Math	7	2004-02-13 17:36

2020-06-16, 15:45	#12
VBCurtis "Curtis" Feb 2005 Riverside, CA 4,621 Posts	That is an excellent and superior viewpoint. As a non-programmer, I tend to think in terms of "can this be improved for little effort?", and running CADO as a black-box that receives input and returns factors seems much easier to implement than breaking out individual tasks and coding individual calls for those things. Your scripts already do those individual things, so you know from experience they can be implemented; I do not, so I advocated for the simplest change that adds speed (CADO as black box).

2020-06-16, 16:08	#13
bsquared "Ben" Feb 2007 D21₁₆ Posts	The CADO-as-a-black-box approach might not be too bad to add to yafu... I really don't want to have to deal with param files, managing different stages, etc. But if someone already has good params set up and I can just use a system call to "cado.py -n 12489...123971", then wait to read and parse an output file, that seems doable. Would have to figure out how to run tune without running entire jobs though. Anyone know if that can be done (i.e., request to only collect N relations)?

2020-06-16, 17:24	#15
EdH "Ed Hall" Dec 2009 Adirondack Mtns DD9₁₆ Posts	A couple more details to consider: CADO-NFS, on its own, uses a filename base from its parameter files and places all data in a randomly named folder in the /tmp directory. Both of these need to be addressed (easily done) in order to make retrieving factors easier. They can both be addressed simply by setting the working directory and filename on the command line. In fact, all (or, close to all) parameters can be overriden by command line values If the /tmp directory is used, a computer reset will remove all work, so a more permanent location should be used for anything of any length. Starting a new CADO-NFS number will fail if prior number data is encountered, so any previous data should be removed once factors are retrieved. The random directory naming keeps this from happening, but makes factor retrieval more complex.