Skip to Content.
Sympa Menu

cado-nfs - Re: [Cado-nfs-discuss] Newer Commit Takes Longer to filter (Ed Hall)

Subject: Discussion related to cado-nfs

List archive

Re: [Cado-nfs-discuss] Newer Commit Takes Longer to filter (Ed Hall)


Chronological Thread 
    < Chronological >      < Thread >
  • From: Ed Hall <ed_ka2fwj@yahoo.com>
  • To: Paul Zimmermann <Paul.Zimmermann@inria.fr>
  • Cc: "cado-nfs-discuss@lists.gforge.inria.fr" <cado-nfs-discuss@lists.gforge.inria.fr>
  • Subject: Re: [Cado-nfs-discuss] Newer Commit Takes Longer to filter (Ed Hall)
  • Date: Wed, 16 Sep 2020 14:18:03 +0000 (UTC)
  • Authentication-results: mail3-smtp-sop.national.inria.fr; spf=None smtp.pra=ed_ka2fwj@yahoo.com; spf=Pass smtp.mailfrom=ed_ka2fwj@yahoo.com; spf=None smtp.helo=postmaster@sonic302-2.consmr.mail.bf2.yahoo.com
  • Ironport-phdr: 9a23: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
  • List-archive: <http://lists.gforge.inria.fr/pipermail/cado-nfs-discuss/>
  • List-id: A discussion list for Cado-NFS <cado-nfs-discuss.lists.gforge.inria.fr>
Dear Team,

I have continued to run the May commit and have found that larger composites seem to have actually shorter dup2 filtering times.  Is it possible this is just an issue for smaller composites, which don't really take that much total real time?

Here is a comparison of a couple shorter times for larger composites:

c139:
Filtering - Duplicate Removal, splitting pass: Total cpu/real time for dup1: 612.93/234.107
Filtering - Duplicate Removal, removal pass: Total cpu/real time for dup2: 1907.08/553.812

c143:
Filtering - Duplicate Removal, splitting pass: Total cpu/real time for dup1: 627.07/233.989
Filtering - Duplicate Removal, removal pass: Total cpu/real time for dup2: 1987.68/783.616

I have attached the four dup2 files plus log for the c143 referenced above at: https://www.mersenneforum.org/attachment.php?attachmentid=23338&d=1600265152

I hope I haven't caused you guys to "pursue undomesticated fowl."

I did revert all my farm to the May commit for the recent tests.  Let me know if you still want me to revert them to an earlier commit and whether I should revert the entire farm, or go with the original mix and match farm that produced the shorter times for the earlier c130.  OR, if you would rather I run something in the c15x region on the already set up May commit, I could easily do that.

Thanks for all the work!

Sincerely,
Edwin Hall


On Wednesday, September 16, 2020, 2:37:50 AM EDT, Paul Zimmermann <paul.zimmermann@inria.fr> wrote:


      Dear Ed,

thanks. The dup2 time did not change much between these two versions:

May 2020: Total cpu/real time for dup2: 1634.55/1185.76
Aug 2020: Total cpu/real time for dup2: 1793.81/1259.68

Thus the issue seems to be between the 2019 version and the May 2020 one.

Please can you make a similar tarball for the 2019 version you used?

Best regards,
Paul


Archive powered by MHonArc 2.6.19+.

Top of Page