block wiedemann and block lanczos

ravlyuchenko · 2011-05-02, 09:22

I would like to make a comparison between these two algorithms. Has anyone grafted block lanczos into CADO-NFS or Block Wiedemann into ggnfs+msieve? Or is that complicated?

jasonp · 2011-05-02, 10:47

That's quite complex; the two suites can exchange data but it would be very difficult for them to exchange code. Both the workflow and the parallel architecture used by each suite are quite different.

It is straightforward to give the same set of relations to both suites, though; both use the GGNFS relation format by default.

ravlyuchenko · 2011-05-09, 01:51

hello jasonp,

i'm trying to exchange data, and i found .purged in CADO might be the counterpart to .dat in msieve and .small.bin = dat.mat, is that true? and what's .small.cw.bin and .small. rw.bin corresponding to? or what is .dat.cyc in msieve corresponding to? Could you give me a suggestion in comparison between these two linalg alg.

thanks : )

Quote:

Originally Posted by jasonp

That's quite complex; the two suites can exchange data but it would be very difficult for them to exchange code. Both the workflow and the parallel architecture used by each suite are quite different.

It is straightforward to give the same set of relations to both suites, though; both use the GGNFS relation format by default.

jasonp · 2011-05-09, 10:47

By exchanging data I meant that the CADO suite includes a script that can do the postprocessing with all the CADO tools given a file of NFS relations from some other package. I actually not that familiar with the intermediate files that the postprocessing generates.

Christenson · 2011-05-09, 10:52

Hi jason:
Is there a reference to format of the NFS relations files? The example might help my studying.
E.

jasonp · 2011-05-09, 13:16

For Msieve, the updated Readme.nfs has all the formats that are amenable to being read/written by other tools (this is a work in progress right now). The CADO suite defaults to the GGNFS format for relations, and other stuff it needs is in a parameters file. The source has a directory full of such files, along with a script called factor_from_rels that looks promising for jumpstarting the postprocessing.

2011-05-02, 09:22	#1
ravlyuchenko May 2011 3 Posts	block wiedemann and block lanczos I would like to make a comparison between these two algorithms. Has anyone grafted block lanczos into CADO-NFS or Block Wiedemann into ggnfs+msieve? Or is that complicated?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Block Wiedemann for beginners	paul0	Factoring	7	2015-11-16 17:09
Msieve with MPI block Lanczos	jasonp	Msieve	104	2013-06-17 11:28
Block Lanczos with a reordering pass	jasonp	Msieve	18	2010-02-07 08:33
Yet another Block Wiedemann thread	pstach	Factoring	25	2009-01-04 12:46
P-1: Block of 130: 19.8-19.9M	dave_0273	Completed Missions	3	2005-07-11 20:22

2011-05-02, 10:47	#2
jasonp Tribal Bullet Oct 2004 3565₁₀ Posts	That's quite complex; the two suites can exchange data but it would be very difficult for them to exchange code. Both the workflow and the parallel architecture used by each suite are quite different. It is straightforward to give the same set of relations to both suites, though; both use the GGNFS relation format by default.

2011-05-09, 10:47	#4
jasonp Tribal Bullet Oct 2004 5×23×31 Posts	By exchanging data I meant that the CADO suite includes a script that can do the postprocessing with all the CADO tools given a file of NFS relations from some other package. I actually not that familiar with the intermediate files that the postprocessing generates.

2011-05-09, 10:52	#5
Christenson Dec 2010 Monticello 11100000011₂ Posts	Hi jason: Is there a reference to format of the NFS relations files? The example might help my studying. E.

2011-05-09, 13:16	#6
jasonp Tribal Bullet Oct 2004 5·23·31 Posts	For Msieve, the updated Readme.nfs has all the formats that are amenable to being read/written by other tools (this is a work in progress right now). The CADO suite defaults to the GGNFS format for relations, and other stuff it needs is in a parameters file. The source has a directory full of such files, along with a script called factor_from_rels that looks promising for jumpstarting the postprocessing.