mersenneforum.org  

Go Back   mersenneforum.org > Factoring Projects > GMP-ECM

Reply
 
Thread Tools
Old 2011-09-12, 12:41   #1
R.D. Silverman
 
R.D. Silverman's Avatar
 
Nov 2003

3×13×191 Posts
Default Slow Down?

I find it interesting that no new factor has been added to the 'top 10'
since June. Has ECM work slackened off?

I've been running 1000 curves on the 2LM tables. I am up to C248. So far,
no luck. (but I only have ~20 cores, most of which are 32 bits).

I am also surprised at how relatively few factors EPFL found in their
search of the 2+ tables. It seems that kudos should be given to (especially
to Bruce) prior efforts.
R.D. Silverman is offline   Reply With Quote
Old 2011-09-14, 13:07   #2
akruppa
 
akruppa's Avatar
 
"Nancy"
Aug 2002
Alexandria

2,467 Posts
Default

I think it's two reasons mostly, 1. a lot of ECM has been done on Cunningham numbers before so the easy pickings are long gone, and 2. using NFS has become so easy and so well-organized that there are a lot more people devoting their cpu cycles to NFS rather than ECM than there used to be – gotta blame Jason, the GGNFS and NFS@HOME guys.
akruppa is offline   Reply With Quote
Old 2011-09-14, 14:03   #3
debrouxl
 
debrouxl's Avatar
 
Sep 2009

977 Posts
Default

The yoyo@home BOINC grid is doing significant ECM work for several other projects, e.g. OddPerfect, XYYXF and Near-repdigit, and therefore the RSALS BOINC grid
debrouxl is offline   Reply With Quote
Old 2011-09-20, 14:26   #4
bdodson
 
bdodson's Avatar
 
Jun 2005
lehigh.edu

210 Posts
Default

Quote:
Originally Posted by R.D. Silverman View Post
I find it interesting that no new factor has been added to the 'top 10'
since June. Has ECM work slackened off?

I've been running 1000 curves on the 2LM tables. I am up to C248. So far,
no luck. ...
No slacking off over here. The new cluster of 8-core Opterons
recently past 1M curves with B1 = 900M (that's 1003058 curves, not
counting the current pass). The pc/grid is still running just when the
public sites are closed, but that's off-set by the speed of the 64-bit
windows binaries, and 4-curves-at-once on the icore3's. Much/most
of c185-c233 is near/at t60. Likewise, most of the first-five-holes
above c233 are at 3t55. The last of the undertested ranges is now
above c270. In particular, c234-c270 has gotten 2t55. Uhm, that
includes 2LM's, 10800 curves with B1 = 400M or 7600 curves with
B1 = 600M, through c270. The exception being 2- and 2+, where I
don't know which of these has PS3 runs, and haven't attempted any
new curves. The new NFS@Home numbers got t60 (first), which
does include 2p1000.

Also, I'm still working from appc311, plus Sam's few new additions to
the official lists (filling in first holes). Sam and PaulZ appear to have
done quite well on the rest of the new 3- and 3+ extensions, which I
haven't been running.

So with p55's "removed" to 80% with 2t55, I expect few-to-very-few
p53/p54's in the regular list, c185-c270. I'm less optimistic about
p58-p61's, anywhere, and especially above c233. On the range
p63-and-above (the current top10), I still consider this to be in the
"black swan" range. Yes, we do see one, once-in-a-while; but so
unlikely as to have (nearly) zero expectation. In particular, the new
PS3 range B1=1e9 and my local approximation at B1=900e6, while
theoretically p65-optimal - given sufficiently many curves - appears
to perform more like the t60s, at B1=260M, than like the B1=3e9's of
the first PS3 runs. -Bruce

@Alex ("AKruppa")- Both Batlov-Dodson and the part of NFS@Home
running the 15e siever are nearly out of snfs candidates from the
regular Cunningham lists. Likewise, if you've scanned the status of
"c197 Womack+mersenneforum gnfs", mersenneforum appears to have
retired from extended runs with the 64-bit 16e siever.
bdodson is offline   Reply With Quote
Old 2011-09-20, 14:46   #5
fivemack
(loop (#_fork))
 
fivemack's Avatar
 
Feb 2006
Cambridge, England

2×52×127 Posts
Default

The rather high resource usage and rather low rate of relation production for doing a G197 clearly put people off. We're getting there, but three CPU-weeks with 600MB of memory usage per million relations, knowing we'll need six hundred million, turns out not to be attractive.
fivemack is offline   Reply With Quote
Old 2011-09-20, 17:11   #6
Andi47
 
Andi47's Avatar
 
Oct 2004
Austria

2·17·73 Posts
Default

Quote:
Originally Posted by fivemack View Post
The rather high resource usage and rather low rate of relation production for doing a G197 clearly put people off. We're getting there, but three CPU-weeks with 600MB of memory usage per million relations, knowing we'll need six hundred million, turns out not to be attractive.
is the use of the 16e siever feasible on a windows 7 64 bit intel i7 system? Which one would be better to use? the 32 or the 64 bit binary?
Andi47 is offline   Reply With Quote
Old 2011-09-20, 18:43   #7
fivemack
(loop (#_fork))
 
fivemack's Avatar
 
Feb 2006
Cambridge, England

143168 Posts
Default

I don't know - I haven't got one to try with - but I expect it would work. The 64-bit binary would be the one to use; copy down the polynomial from the G197 thread, try to sieve 100M .. 100M+1k, and let's see if it segfaults and how long it takes.

Last fiddled with by fivemack on 2011-09-20 at 18:43
fivemack is offline   Reply With Quote
Old 2011-09-20, 21:03   #8
axn
 
axn's Avatar
 
Jun 2003

23×3×193 Posts
Default

Quote:
Originally Posted by fivemack View Post
I don't know - I haven't got one to try with - but I expect it would work. The 64-bit binary would be the one to use; copy down the polynomial from the G197 thread, try to sieve 100M .. 100M+1k, and let's see if it segfaults and how long it takes.
Unless the 64 bit asm has been ported to windows, 32-bit binary will give better performance under Win 7.

Last fiddled with by axn on 2011-09-20 at 21:03 Reason: getter -> better
axn is online now   Reply With Quote
Old 2011-09-20, 22:15   #9
R.D. Silverman
 
R.D. Silverman's Avatar
 
Nov 2003

1D1916 Posts
Default

Quote:
Originally Posted by fivemack View Post
The rather high resource usage and rather low rate of relation production for doing a G197 clearly put people off. We're getting there, but three CPU-weeks with 600MB of memory usage per million relations, knowing we'll need six hundred million, turns out not to be attractive.
The IGG strikes again.
R.D. Silverman is offline   Reply With Quote
Old 2011-09-21, 01:58   #10
bdodson
 
bdodson's Avatar
 
Jun 2005
lehigh.edu

210 Posts
Default Current Curve counts (RE: Slow Down?)

Quote:
Originally Posted by R.D. Silverman View Post
The IGG strikes again.
I appear to have hijacked my own post to this thread on ECMNET, and
in particular, my reply to Bob's question. Here's a repost.
Quote:
Originally Posted by bdodson
No slacking off over here. The new cluster of 8-core Opterons
recently past 1M curves with B1 = 900M (that's 1003058 curves, not
counting the current pass). The pc/grid is still running just when the
public sites are closed, but that's off-set by the speed of the 64-bit
windows binaries, and 4-curves-at-once on the icore3's. Much/most
of c185-c233 is near/at t60. Likewise, most of the first-five-holes
above c233 are at 3t55. The last of the undertested ranges is now
above c270. In particular, c234-c270 has gotten 2t55. Uhm, that
includes 2LM's, 10800 curves with B1 = 400M or 7600 curves with
B1 = 600M, through c270. The exception being 2- and 2+, where I
don't know which of these has PS3 runs, and haven't attempted any
new curves. The new NFS@Home numbers got t60 (first), which
does include 2p1000.

Also, I'm still working from appc311, plus Sam's few new additions to
the official lists (filling in first holes). Sam and PaulZ appear to have
done quite well on the rest of the new 3- and 3+ extensions, which I
haven't been running.

So with p55's "removed" to 80% with 2t55, I expect few-to-very-few
p53/p54's in the regular list, c185-c270. I'm less optimistic about
p58-p61's, anywhere, and especially above c233. On the range
p63-and-above (the current top10), I still consider this to be in the
"black swan" range. Yes, we do see one, once-in-a-while; but so
unlikely as to have (nearly) zero expectation. In particular, the new
PS3 range B1=1e9 and my local approximation at B1=900e6, while
theoretically p65-optimal - given sufficiently many curves - appears
to perform more like the t60s, at B1=260M, than like the B1=3e9's of
the first PS3 runs. -Bruce
bdodson is offline   Reply With Quote
Old 2011-09-21, 08:28   #11
fivemack
(loop (#_fork))
 
fivemack's Avatar
 
Feb 2006
Cambridge, England

2·52·127 Posts
Default

Quote:
Originally Posted by R.D. Silverman View Post
The IGG strikes again.
Please be clearer
fivemack is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Slow progress! Xyzzy Miscellaneous Math 62 2017-03-02 00:35
How hot is too hot? Slow is too slow? petrw1 Hardware 13 2008-11-10 23:25
Slow computer Housemouse Hardware 7 2008-02-15 18:18
Celeron 2.40 too slow? rudi_m Hardware 14 2005-10-11 03:31
Web server slow HackMaster PrimeNet 28 2005-06-21 21:01

All times are UTC. The time now is 02:30.

Thu Jul 9 02:30:06 UTC 2020 up 106 days, 3 mins, 0 users, load averages: 1.84, 1.91, 1.76

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.