mersenneforum.org  

Go Back   mersenneforum.org > Factoring Projects > Aliquot Sequences

Reply
 
Thread Tools
Old 2015-06-05, 09:06   #177
fivemack
(loop (#_fork))
 
fivemack's Avatar
 
Feb 2006
Cambridge, England

13×491 Posts
Default

msieve -v -nc1 filter_maxrels=200000000

If you're doing experiments on this relation set, I would be interested also to see what the minimum relation count is for target_density=120 to work; my suspicion is that, with 32-bit large primes, it's worth doing enough oversieving to get a higher target density even if you're not doing the sieving and linear algebra on the same hardware.

At least at one point you had to put
Code:
msieve -v -nc1 "filter_maxrels=200000000 target_density=120"
(note the quotes) to get the arguments parsed correctly; not sure if this has now been fixed.

Last fiddled with by fivemack on 2015-06-05 at 09:08
fivemack is offline   Reply With Quote
Old 2015-06-08, 16:01   #178
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

2·2,339 Posts
Default

3408:1589 is factored, p65*p97. 1590 was easy. 1591 has a C157 which has survived a t40 and 600 curves @ 11M; I am running a standard t45 on it now, and welcome assistance for whatever part of a t50 is proper.

I used -s flag to update factordb for the sequence; it appears that since aliqueit is managing the ECM presently, the p25 already found in 1591 wasn't uploaded. Sorry 'bout that!

I'll tinker with the filtering step and report my findings- there aren't many 32-bit runs on C162s out yet.
VBCurtis is offline   Reply With Quote
Old 2015-06-10, 23:20   #179
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

2·2,339 Posts
Default

A t45 is complete on the C157 at i1591. I have completed 400 curves at B1 =43M, and will finish 1000 more.

I will NFS this C157 in July if nobody claims it before then.

Findings on the C162 with 32LP:
233M relations could not build a matrix.
236M relations produced a matrix size 9.53M.
240M relations with default target-density, the actual run, was a matrix of size 8.78M. This took 320 thread-hours to solve. The last 4.4M relations took 48 thread-hours to sieve on the same machine.
Target-density 80 produced 8.25M matrix.
Density 90 produced 7.83M matrix.
Density 96 produced 7.62M matrix.
Density 100 could not build a matrix.
Making a broad assumption that matrix solve time scales with the square of the size of the matrix, a 9.53M matrix would have taken 45-50 extra thread-hours to solve, so my "oversieving" cost me no net time even at default density. If I'd set target-density to 90 or 96, I could have saved ~25% of the matrix step, or 70-80 thread-hours (plus the ~50 saved already from 236M rels vs 240M rels at default density!). 50 hours extra sieve, 100+ hours saved in matrix step.

My data matches fivemack's opinion, that for 32LP the extra sieve time is more than made up in shorter matrix steps; I will set target-density to 96 or higher for my next 32-bit project.

I once edited the factmsieve script to add target-density in, but have lost those changes. Can someone remind me where to place the flag? I suppose it'll take just a couple attempts to get the edit syntax correct anyway..
VBCurtis is offline   Reply With Quote
Old 2015-06-11, 01:59   #180
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

32×17×61 Posts
Default

Quote:
Originally Posted by VBCurtis View Post
I once edited the factmsieve script to add target-density in, but have lost those changes. Can someone remind me where to place the flag? I suppose it'll take just a couple attempts to get the edit syntax correct anyway..
You put it in quotes behind the -nc1, like
msieve {...other params...} -nc1 "target_density=96"
When in doubt, run
msieve -h
Batalov is offline   Reply With Quote
Old 2015-06-11, 06:13   #181
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

111068 Posts
Default

My doubt was about the python script, rather than the msieve invocation. Makes sense it is as simple as searching for "-nc1" in the script and adding the flag as if it were command line.

Thanks!
VBCurtis is offline   Reply With Quote
Old 2015-06-14, 03:19   #182
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

2×2,339 Posts
Default

A total of 1550 curves at B1 = 43M are complete on the current c157.
If nobody takes this number on, I will find resources in July to perform NFS.
VBCurtis is offline   Reply With Quote
Old 2015-10-20, 23:35   #183
swellman
 
swellman's Avatar
 
Jun 2012

22·3·13·19 Posts
Default

Did the C157 ever get the full t50? I can't take on the NFS job right now but I can help with ECM (if needed).
swellman is online now   Reply With Quote
Old 2015-10-21, 02:27   #184
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

2×2,339 Posts
Default

I never did the NFS, as my cuda-enabled msieve rig failed in June so I lost my poly-searching setup. I'll be happy to do the NFS if the usual suspects can procure a suitable poly. Perhaps 500-1000 more curves first?
VBCurtis is offline   Reply With Quote
Old 2015-10-21, 09:29   #185
swellman
 
swellman's Avatar
 
Jun 2012

22·3·13·19 Posts
Default

Ok I will run some additional ECM pending a poly being found. I don't have a GPU rig to aid with the poly select either or I'd help with that.

Should this thread become sticky once again?
swellman is online now   Reply With Quote
Old 2015-10-24, 03:34   #186
RichD
 
RichD's Avatar
 
Sep 2008
Kansas

1100110101102 Posts
Default C157 @ i1591

I ran A5 to 5M and the two best are:
Code:
N: 9622573172947172690216533113544901989270930284548712969416487159589518659037713792560995347053599943835805911303098595474551471251812082141957434773586035021
# expecting poly E from 1.91e-12 to > 2.20e-12
R0: -1226437530675169727727847599102
R1: 4489146410702701
A0: 1085620711869412190796638577436472845
A1: -13524714026065933306465543047704
A2: 2354089978283392085793977
A3: 22197682057227374702
A4: 2053627590300
A5: 3467880
# skew 1073123.83, size 2.686e-15, alpha -7.096, combined = 2.044e-12 rroots = 3
Code:
N: 9622573172947172690216533113544901989270930284548712969416487159589518659037713792560995347053599943835805911303098595474551471251812082141957434773586035021
# expecting poly E from 1.91e-12 to > 2.20e-12
R0: -3726397299028232594516190800757
R1: 228399129977819
A0: -275716441178826782149000795012629753184
A1: 374337043347538241259074213027768
A2: 35851573769457152514399928
A3: -3163034830851676700
A4: -195302199429
A5: 13392
# skew 12398157.51, size 2.645e-15, alpha -6.725, combined = 2.039e-12 rroots = 1
RichD is offline   Reply With Quote
Old 2015-10-24, 05:46   #187
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

2×2,339 Posts
Default

I'll test-sieve these two and pick one to run, then I'l wait for Swellman's last ECM run before commencing NFS.
VBCurtis is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Reserved for MF - Sequence 3366 RichD Aliquot Sequences 468 2021-01-27 01:16
Reserved for MF - Sequence 4788 schickel Aliquot Sequences 2934 2021-01-07 18:52
Reserved for MF - Sequence 276 kar_bon Aliquot Sequences 127 2020-12-17 10:05
Team Sieve #37: 3408:i1287 RichD Aliquot Sequences 14 2013-08-02 17:02
80M to 64 bits ... but not really reserved petrw1 Lone Mersenne Hunters 82 2010-01-11 01:57

All times are UTC. The time now is 13:11.

Mon Mar 8 13:11:30 UTC 2021 up 95 days, 9:22, 0 users, load averages: 2.09, 2.44, 2.44

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, 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.