CADO NFS - Page 42

EdH · 2020-01-22, 15:42

I thought this has been referenced somewhere, but I can't find it.

I have several machines, some of which may have older versions. Is there a way I can tell how old an install is?

Dylan14 · 2020-01-22, 17:06

Quote:

Originally Posted by EdH

I thought this has been referenced somewhere, but I can't find it.

I have several machines, some of which may have older versions. Is there a way I can tell how old an install is?

Try typing in a terminal that is accessing the root of the cado directory the following:

Code:

git log

this will yield all the revisions. The one that says (HEAD -> master, origin/master, origin/coverage, origin/HEAD) is the the one you're on. For example, I am on rev f65aab7f3ccca673cb9086d67f2526f1fde10a91, dated Oct 10, 2019.

EdH · 2020-01-22, 21:43

Thanks! I really must study git more. . .

Of two machines checked, one had the Head, etc. info and the other didn't. It just gave four entries with their dates and a brief description.

fivemack · 2020-01-27, 19:57

I had thought I could use -admin X -incr 420 and -admin {X+210} -incr 420 on two machines to search a wide c5 range efficiently. But in fact polyselect adjusts admin to the next multiple of incr, and is deterministic, so all 36 thread-days run on the second computer was wasted.

EdH · 2020-02-02, 20:01

This is quite disappointing!

The last three CADO-NFS runs I've worked have had server failures like the one that were plaguing the 2,2330L team project. The server quits handing out WUs. This last one has stopped twice in one day. Restarting isn't really an issue, but the loss of all the hours when it happens overnight or even during the day, if it isn't noticed, is quite annoying.

I don't think I ever had this issue prior to the last time I installed a current revision. After this run finishes, I will be moving back to a prior one.

henryzz · 2020-02-26, 16:17

What is the sublat option on las? As far as I can tell it sieves a subset of the sieve region and should reduce memory usage(presumably with a speed penalty). On the version, I am using it just crashes.
Is this option worth anything? Does it work on the latest source?

EdH · 2020-03-09, 20:08

I have a few issues that warrant further study:

In all my SNFS runs, I need to explicitly invoke local machine clients. I don't need to do this for GNFS. I don't see (recognize) any setting(s) in the params.cxx files.

In my latest GNFS run, I noticed that during LA, only 50% of my CPU was in use. I'm running an i7 2600 with four cores and two threads/core. Is CADO-NFS limiting the use to core count? Can/should I override it to use thread count?

In my latest GNFS runs, using a recent revision of CADO-NFS, I have noticed the server stopping the issuing of WUs, at which time I need to stop/restart the server. It still accepts WUs, but doesn't issue any. (This appears to be the same as we had with the 2,2330L project.) I never experienced this in earlier GNFS/SNFS runs.

RichD · 2020-03-11, 04:35

Quote:

Originally Posted by EdH

In my latest GNFS run, I noticed that during LA, only 50% of my CPU was in use. I'm running an i7 2600 with four cores and two threads/core. Is CADO-NFS limiting the use to core count? Can/should I override it to use thread count?

I have a Core i7-2600K standalone running a GNFS job. (No network clients.) It finally got to the "krylov" stage and I am noticing the same thing. Recalling several years ago Msieve did not benefit with HT in Block Lanczos. Perhaps there is something similar to their implementation of Block Weidemann, which may be intentional.

EdH · 2020-03-11, 13:24

Quote:

Originally Posted by RichD

I have a Core i7-2600K standalone running a GNFS job. (No network clients.) It finally got to the "krylov" stage and I am noticing the same thing. Recalling several years ago Msieve did not benefit with HT in Block Lanczos. Perhaps there is something similar to their implementation of Block Weidemann, which may be intentional.

I've never looked at that with msieve. But, msieve runs LA faster and when I'm not lazy, I use it for the LA portion of larger jobs. I suppose this warrants more study on my part. . .

EdH · 2020-04-01, 15:29

I am working on a Colab Instance of CADO-NFS to pick up the 100k-150k section of relations of a current snfs run. (I hope to eventually document my adventure in my "How I . . ." threads.)

I will not be accessing the server, but will, instead, initiate the Colab session as a separate sever with all the parameters of the first except the qmin setting. I am concerned about the addition of the new relations to the original set, since the original sever has no record of issuing the work area. Is there a method of adding the alternate relations to the existing ones?

I'm expecting that my msieve side process, that I'm also going to run, will have no trouble with them, but I'm expecting the CADO-NFS process to have issues.

Thoughts from anyone?

VBCurtis · 2020-04-01, 15:57

I have the same question/doubt, and I agree that msieve won't care. When I've done this myself, I've just cat'ed together the relations files from the side run, then fed the main-run relations and the side-run file together to remdups and then msieve.

I don't know how to tell CADO to assimilate those extra relations; it would be nice if there was a feature similar to factmsieve's spairs.add where we could manually drop new relations in and CADO would pick them up via an occasional check for a ".add" file.

2020-01-22, 15:42	#452
EdH "Ed Hall" Dec 2009 Adirondack Mtns 2×1,811 Posts	Is there a way to find the version? I thought this has been referenced somewhere, but I can't find it. I have several machines, some of which may have older versions. Is there a way I can tell how old an install is?

2020-01-27, 19:57	#455
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 14356₈ Posts	well that was silly I had thought I could use -admin X -incr 420 and -admin {X+210} -incr 420 on two machines to search a wide c5 range efficiently. But in fact polyselect adjusts admin to the next multiple of incr, and is deterministic, so all 36 thread-days run on the second computer was wasted. Last fiddled with by fivemack on 2020-01-27 at 19:57

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
CADO-NFS on windows	jux	CADO-NFS	22	2019-11-12 12:08
CADO help	henryzz	CADO-NFS	4	2017-11-20 15:14
CADO and WinBlows	akruppa	Programming	22	2015-12-31 08:37
CADO-NFS	skan	Information & Answers	1	2013-10-22 07:00
CADO	R.D. Silverman	Factoring	4	2008-11-06 12:35

2020-01-22, 21:43	#454
EdH "Ed Hall" Dec 2009 Adirondack Mtns 111000100110₂ Posts	Thanks! I really must study git more. . . Of two machines checked, one had the Head, etc. info and the other didn't. It just gave four entries with their dates and a brief description.

2020-02-02, 20:01	#456
EdH "Ed Hall" Dec 2009 Adirondack Mtns 111000100110₂ Posts	This is quite disappointing! The last three CADO-NFS runs I've worked have had server failures like the one that were plaguing the 2,2330L team project. The server quits handing out WUs. This last one has stopped twice in one day. Restarting isn't really an issue, but the loss of all the hours when it happens overnight or even during the day, if it isn't noticed, is quite annoying. I don't think I ever had this issue prior to the last time I installed a current revision. After this run finishes, I will be moving back to a prior one.

2020-02-26, 16:17	#457
henryzz Just call me Henry "David" Sep 2007 Cambridge (GMT/BST) 13272₈ Posts	What is the sublat option on las? As far as I can tell it sieves a subset of the sieve region and should reduce memory usage(presumably with a speed penalty). On the version, I am using it just crashes. Is this option worth anything? Does it work on the latest source?

2020-03-09, 20:08	#458
EdH "Ed Hall" Dec 2009 Adirondack Mtns 2×1,811 Posts	I have a few issues that warrant further study: In all my SNFS runs, I need to explicitly invoke local machine clients. I don't need to do this for GNFS. I don't see (recognize) any setting(s) in the params.cxx files. In my latest GNFS run, I noticed that during LA, only 50% of my CPU was in use. I'm running an i7 2600 with four cores and two threads/core. Is CADO-NFS limiting the use to core count? Can/should I override it to use thread count? In my latest GNFS runs, using a recent revision of CADO-NFS, I have noticed the server stopping the issuing of WUs, at which time I need to stop/restart the server. It still accepts WUs, but doesn't issue any. (This appears to be the same as we had with the 2,2330L project.) I never experienced this in earlier GNFS/SNFS runs.

2020-04-01, 15:29	#461
EdH "Ed Hall" Dec 2009 Adirondack Mtns 2×1,811 Posts	I am working on a Colab Instance of CADO-NFS to pick up the 100k-150k section of relations of a current snfs run. (I hope to eventually document my adventure in my "How I . . ." threads.) I will not be accessing the server, but will, instead, initiate the Colab session as a separate sever with all the parameters of the first except the qmin setting. I am concerned about the addition of the new relations to the original set, since the original sever has no record of issuing the work area. Is there a method of adding the alternate relations to the existing ones? I'm expecting that my msieve side process, that I'm also going to run, will have no trouble with them, but I'm expecting the CADO-NFS process to have issues. Thoughts from anyone?

2020-04-01, 15:57	#462
VBCurtis "Curtis" Feb 2005 Riverside, CA 4,673 Posts	I have the same question/doubt, and I agree that msieve won't care. When I've done this myself, I've just cat'ed together the relations files from the side run, then fed the main-run relations and the side-run file together to remdups and then msieve. I don't know how to tell CADO to assimilate those extra relations; it would be nice if there was a feature similar to factmsieve's spairs.add where we could manually drop new relations in and CADO would pick them up via an occasional check for a ".add" file.