mersenneforum.org  

Go Back   mersenneforum.org > Factoring Projects > Factoring

Reply
 
Thread Tools
Old 2012-12-02, 10:02   #89
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

1C3516 Posts
Default

I'll do 330-340M. (PS: My website/aliquot pages are now on a dedicated server, and so should be less temperamental than the last week. )

Edit: FWIW, here's the script I used to run this job:
Code:
#! /bin/bash

siever="15e"
start=330000000 # q-range start
len=10000000 # q-range length
cores=4
affin="2 4 8" # Hex, one entry for all but one core (see `man taskset`)
a="1" # Hex affin for the last core (not necessarily literally last)

diff=$(($len/$cores))
echo "diff: $diff"

echo "affin: $a; start: $start"
nice -n 19 taskset $a ./gnfs-lasieve4I$siever snfs -v -R -o rels$a.dat -r -f $start -c $diff &
let start+=$diff

for a in $affin; do # Loops over each space-separated substring of $affin
     echo "affin: $a; start: $start"
     nice -n 19 taskset $a ./gnfs-lasieve4I$siever snfs -v -R -o rels$a.dat -r -f $start -c $diff 2> $(($start/1000)).t &
     start=$(($start+$diff))
done
It prints the output of one of the instances to terminal, while redirecting the other instances to files.

If you had 8 cores, you might do something like
Code:
affin="1 2 4 8 10 40 80" # Hex, one entry for all but one core (see `man taskset`)
a="20" # Hex affin for the last core (not necessarily literally last)
In this case, the 6th instance would be printed to terminal, the rest to file.

Last fiddled with by Dubslow on 2012-12-02 at 10:14
Dubslow is offline   Reply With Quote
Old 2012-12-02, 10:24   #90
fivemack
(loop (#_fork))
 
fivemack's Avatar
 
Feb 2006
Cambridge, England

23×797 Posts
Default

Err, 330-340 includes the range that em99010pepe is running: maybe 334 would be a better place to start?

I should finish 360-385 in the next couple of days, which might even bring enough relations to finish the job; if not, I'll take 340-360.
fivemack is offline   Reply With Quote
Old 2012-12-02, 10:26   #91
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

1C3516 Posts
Default

Quote:
Originally Posted by fivemack View Post
Err, 330-340 includes the range that em99010pepe is running: maybe 334 would be a better place to start?

I should finish 360-385 in the next couple of days, which might even bring enough relations to finish the job; if not, I'll take 340-360.
Derp. I guess I'll do 334-340.

PS Between the four cores, I'm getting around 0.16 sec/rel.

Last fiddled with by Dubslow on 2012-12-02 at 10:35
Dubslow is offline   Reply With Quote
Old 2012-12-02, 19:32   #92
em99010pepe
 
em99010pepe's Avatar
 
Sep 2004

283010 Posts
Default

Quote:
Originally Posted by fivemack View Post
Err, 330-340 includes the range that em99010pepe is running: maybe 334 would be a better place to start?

I should finish 360-385 in the next couple of days, which might even bring enough relations to finish the job; if not, I'll take 340-360.
fivemack, I didn't reserve 330-334 so it is free.
em99010pepe is offline   Reply With Quote
Old 2012-12-07, 19:05   #93
fivemack
(loop (#_fork))
 
fivemack's Avatar
 
Feb 2006
Cambridge, England

23×797 Posts
Default

360-385 finished at 1102 this morning; it didn't give enough relations for a matrix. So I'm starting 340-360 now. It'll probably take almost until Christmas (I'm doing a few C159 GNFS jobs for aliquot sequences in parallel); I hope dubslow will have finished 330-340 by then.
fivemack is offline   Reply With Quote
Old 2012-12-07, 21:15   #94
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3·29·83 Posts
Default

Quote:
Originally Posted by fivemack View Post
360-385 finished at 1102 this morning; it didn't give enough relations for a matrix. So I'm starting 340-360 now. It'll probably take almost until Christmas (I'm doing a few C159 GNFS jobs for aliquot sequences in parallel); I hope dubslow will have finished 330-340 by then.
I'm doing the 6M range in 11 days and 5 hours, with 5 days 8 hours left; to do the extra 4M then would be another ~7.5 on top of that, and 7+5+7.5 ~ 20, or five days before Christmas. Sounds like a deal.
Dubslow is offline   Reply With Quote
Old 2012-12-23, 22:39   #95
fivemack
(loop (#_fork))
 
fivemack's Avatar
 
Feb 2006
Cambridge, England

23·797 Posts
Default Won't be done by Christmas

With 330-360 merged in, I do get a matrix.

However, it's a Godzillasaurus of the matrix world, with 45.2 million cycles and an expected run-time of 1100 wall-time hours on 24 CPUs.

So I'll sieve 80-95, which should take about 250 wall-time hours on 24 CPUs (but I'll use 48) and see how much smaller that makes the matrix. Could I interest anyone else in doing 60-80 ?

Last fiddled with by fivemack on 2012-12-23 at 22:40
fivemack is offline   Reply With Quote
Old 2012-12-23, 22:49   #96
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3·29·83 Posts
Default

Quote:
Originally Posted by fivemack View Post
Could I interest anyone else in doing 60-80 ?
I'll byte. I'll take the top half (70-80).
Dubslow is offline   Reply With Quote
Old 2013-01-19, 10:17   #97
fivemack
(loop (#_fork))
 
fivemack's Avatar
 
Feb 2006
Cambridge, England

23×797 Posts
Default et sic incipit

Thanks to Dubslow for providing 14621712 relations.

I sieved 60-70 and 80-95, Dubslow sieved 70-80. The result was about 51.6 million relations, of which 28.7 million survived uniqification.

Code:
driver@tractor:/scratch/M929.l$ grep cycles msieve.log
Sun Dec 23 22:23:57 2012  found 45398613 cycles, need 45251317
Sun Dec 23 22:24:40 2012  weight of 45251317 cycles is about 3167856145 (70.01/cycle)
Sat Jan 19 09:43:21 2013  found 38739067 cycles, need 38550605
Sat Jan 19 09:43:54 2013  weight of 38550605 cycles is about 2698848446 (70.01/cycle)
So, by an estimated 15000 CPU-hours of computation (I'm measuring a cost of 1.18 CPU-seconds per relation) spread over 635 hours of wallclock time, I have reduced the expected wallclock time of the next phase by (using my curve-fit 1.554e-10 * N^2.143 for runs on 24 cores) 327 hours.

I'm not sure that really counts as success. It would have advanced the time to completion only very marginally if I'd run on all 48 cores and not performed five quite large linear algebra jobs for NFS@home in the interim.

I've started the linear algebra and anticipate results before the end of February.
fivemack is offline   Reply With Quote
Old 2013-03-01, 18:33   #98
fivemack
(loop (#_fork))
 
fivemack's Avatar
 
Feb 2006
Cambridge, England

23×797 Posts
Default

2^929-1 == 0 (mod 1213545181384297818479668246537957920979209321925502086596377 )
fivemack is offline   Reply With Quote
Old 2013-03-01, 18:36   #99
unconnected
 
unconnected's Avatar
 
May 2009
Russia, Moscow

2×5×251 Posts
Default

ECM miss?
unconnected is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Samsung: smaller SRAM cells with EUV, better than Intel VictordeHolland Hardware 2 2018-02-16 15:13
Slow progress! Xyzzy Miscellaneous Math 62 2017-03-02 00:35
Some questions about M929 thomasn Factoring 5 2011-05-29 01:13
Prime 95 kills CPUs? BlackOmega Information & Answers 6 2010-03-01 07:32
Windows XP logoff sometimes kills Prime95 markhl Software 8 2003-07-07 16:00

All times are UTC. The time now is 01:48.

Sun Dec 6 01:48:24 UTC 2020 up 2 days, 21:59, 0 users, load averages: 3.10, 2.90, 2.71

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.