mersenneforum.org  

Go Back   mersenneforum.org > Prime Search Projects > Five or Bust - The Dual Sierpinski Problem

Reply
 
Thread Tools
Old 2008-10-10, 22:35   #1
philmoore
 
philmoore's Avatar
 
"Phil"
Sep 2002
Tracktown, U.S.A.

100010111002 Posts
Default Sieving discussion thread

Please use the other thread only for coordination of sieving reservations. All discussion of sieving can go in this thread.
philmoore is offline   Reply With Quote
Old 2008-10-15, 00:40   #2
geoff
 
geoff's Avatar
 
Mar 2003
New Zealand

13×89 Posts
Default

Between sr2sieve versions 1.7.x and 1.8.x there was a change in the way sr2sieve handles the dual form (b^n+/-k) sequences.

The main difference is that 1.8.x can't sieve standard and dual forms together, but that doesn't affect this project.

However with the current version 1.8.2 you will need to add the -d or --dual switch to the command line to sieve dual form sequences. With earlier versions this switch is not needed (or recognised). I will probably make this switch optional in future versions when using ABCD format sieve files.
geoff is offline   Reply With Quote
Old 2008-10-16, 04:05   #3
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

792 Posts
Default

Quick question: I noticed that the reservations in this thread started at 0T and go up from there in 1T blocks. Yet in the PRP thread, it is stated that the candidates there have already been sieved to 17.59T. Does this mean that in fact there is a complete overlap between the n<2.1M candidates that were already sieved, and this sieving effort? If that is the case, then why not simply remove anything for n<2.1M from this sieve file, and simply use the existing sieve file for doublechecks of n<2.1M?

Last fiddled with by mdettweiler on 2008-10-16 at 04:06
mdettweiler is offline   Reply With Quote
Old 2008-10-16, 06:44   #4
hhh
 
hhh's Avatar
 
Jun 2005

22·3·31 Posts
Default

In other projects, like PSP, removing candidates like this (n-wise) does not affect the speed. The same might be the case here, so we sieve that low stuff again for free.

geoff, the sr2sieve(1.8.2) processes were running at normal priority on my Win XP 32. Can you change that default, please?

H.
hhh is offline   Reply With Quote
Old 2008-10-16, 11:03   #5
philmoore
 
philmoore's Avatar
 
"Phil"
Sep 2002
Tracktown, U.S.A.

22×32×31 Posts
Default

I recall that the sieving rate is supposed to be inversely proportional to the square-root of the n-range, as well as the square-root of the number of sequences. So the rate would be about 2% faster if we dropped n<2.1M. On the other hand, I did many of the earlier PRP tests with the development version of pfgw (version 24 FFTs), and we did encounter some problems with FFT cutoffs which George has fixed in version 25. I especially had trouble with the 2^n+2131 sequence, and had to redo many tests with the stable version of pfgw (version 23 FFTs), so at some point, we probably do want to do some double-checking. I am guessing that it would pay off to resieve to a higher limit, maybe to 100T, before dropping n<2.1M from the range.
philmoore is offline   Reply With Quote
Old 2008-10-16, 11:52   #6
Lennart
 
Lennart's Avatar
 
"Lennart"
Jun 2007

21408 Posts
Default

Quote:
Originally Posted by hhh View Post
In other projects, like PSP, removing candidates like this (n-wise) does not affect the speed. The same might be the case here, so we sieve that low stuff again for free.

geoff, the sr2sieve(1.8.2) processes were running at normal priority on my Win XP 32. Can you change that default, please?

H.

Yes you can. (from 1.8.1)

"New process priority behaviour is incompatiple with previous versions:
Default is not to change process priority (previous default was idle).
-zz sets lowest priority (nice 20)
-z sets low priority (nice 10)
-Z sets high priority (nice -10)
-ZZ sets highest priority (nice -20) "

Lennart

Last fiddled with by Lennart on 2008-10-16 at 11:54 Reason: Version
Lennart is offline   Reply With Quote
Old 2008-10-16, 14:36   #7
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

792 Posts
Default

Quote:
Originally Posted by philmoore View Post
I recall that the sieving rate is supposed to be inversely proportional to the square-root of the n-range, as well as the square-root of the number of sequences. So the rate would be about 2% faster if we dropped n<2.1M. On the other hand, I did many of the earlier PRP tests with the development version of pfgw (version 24 FFTs), and we did encounter some problems with FFT cutoffs which George has fixed in version 25. I especially had trouble with the 2^n+2131 sequence, and had to redo many tests with the stable version of pfgw (version 23 FFTs), so at some point, we probably do want to do some double-checking. I am guessing that it would pay off to resieve to a higher limit, maybe to 100T, before dropping n<2.1M from the range.
Ah, I see--I get it now. In that case, I agree that it probably would be wisest to leave the lower ranges in for now, for doublechecking purposes as you suggested.
mdettweiler is offline   Reply With Quote
Old 2008-10-17, 03:39   #8
geoff
 
geoff's Avatar
 
Mar 2003
New Zealand

13×89 Posts
Default

The time to complete a 1T range should drop a little as the factor size increases (because less primes in each range), but here are estimated times to complete the 3-4T range with sr2sieve 1.8.2 on my machines:

2.66GHz Core 2 Duo:
(64 bit, 1 core): 3 days 16 hr,
(32 bit, 1 core): 6 days 13 hr.

2.9GHz Pentium 4 HT:
(32 bit, 1 thread): 12 days 8 hr,
(32 bit, 2 threads): 9 days 8 hr.
geoff is offline   Reply With Quote
Old 2008-10-23, 05:23   #9
paleseptember
 
paleseptember's Avatar
 
Jun 2008
Wollongong, .au

3×61 Posts
Default

Hopefully someone can help me. I'm ... challenged when it comes to these sort of things.

I want to help :) I have a Q6600 ready for some sieving action, I've downloaded sr2sieve, I've got the .abcd file, I can see that the next available range is 5-6T.

What do I do next? Should I create four copies of the folder and run a cmd window for each 250G slice? (And what should I type? I've sieved before for PSP/SOB, but used a work.txt file to input the numbers).

Thanks in advance. (It's mainly dealing with the quad-core that I'm asking about)
paleseptember is offline   Reply With Quote
Old 2008-10-23, 05:36   #10
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

792 Posts
Default

Quote:
Originally Posted by paleseptember View Post
Hopefully someone can help me. I'm ... challenged when it comes to these sort of things.

I want to help :) I have a Q6600 ready for some sieving action, I've downloaded sr2sieve, I've got the .abcd file, I can see that the next available range is 5-6T.

What do I do next? Should I create four copies of the folder and run a cmd window for each 250G slice? (And what should I type? I've sieved before for PSP/SOB, but used a work.txt file to input the numbers).

Thanks in advance. (It's mainly dealing with the quad-core that I'm asking about)
Yes, create four copies of the sr2sieve folder, and simply use an sr2work.txt file as always--using 5T-6T as an example, you'd simply enter "5000-5250", "5250-5500", "5500-5750", and "5750-6000" in the four respective sr2work.txt files.

Assuming you've dedicated those particular sr2sieve folders to Five or Bust, you may wish to rename the ABCD file to "sr2data.txt" for simplification of necessary command line flags.

Hope this helps!

Max

P.S.: One more thing to watch: if you're using sr2sieve v1.8.x then you'll need to run sr2sieve with the -d flag on the command line. That is, instead of just double-clicking the sr2sieve.exe icon, you'll want to create a file called "run_sr2sieve.bat" (or something along those lines) which contains one line, "sr2sieve -d", and run that instead. (Or, if your preferred method is to run sr2sieve directly from a command window, simply append the -d to the command in a similar fashion.) Also, if your machine doesn't have any particular issues with not clocking up the CPU properly for low-priority apps, you may want to add the -z option, too, so that sr2sieve runs in low-priority mode and stays out of the way of foreground programs without any issues.

Of course, if you're using an earlier version of sr2sieve (i.e. prior to 1.8.0, I think? or was it 1.7.something?), then the issues of both the -d and -z options become a moot point.
mdettweiler is offline   Reply With Quote
Old 2008-10-23, 05:48   #11
geoff
 
geoff's Avatar
 
Mar 2003
New Zealand

13·89 Posts
Default

Quote:
Originally Posted by paleseptember View Post
I want to help :) I have a Q6600 ready for some sieving action, I've downloaded sr2sieve, I've got the .abcd file, I can see that the next available range is 5-6T.

What do I do next? Should I create four copies of the folder and run a cmd window for each 250G slice? (And what should I type? I've sieved before for PSP/SOB, but used a work.txt file to input the numbers).

Thanks in advance. (It's mainly dealing with the quad-core that I'm asking about)
Once you have reserved the 5-6T range, In Linux you can run:

sr2sieve -v -d -i dualSierp.abcd -p 5e12 -P 6e12 -t4

which will use all 4 cores and write all factors to factors.txt

In Windows you will need to run four separate processes, but you can run them all from the same directory by using the -u switch:

sr2sieve -v -d -i dualSierp.abcd -p 5000e9 -P 5250e9 -u1
sr2sieve -v -d -i dualSierp.abcd -p 5250e9 -P 5500e9 -u2
sr2sieve -v -d -i dualSierp.abcd -p 5500e9 -P 5750e9 -u3
sr2sieve -v -d -i dualSierp.abcd -p 5750e9 -P 6000e9 -u4

this will write factors to factors-1.txt, factors-2.txt, factors-3.txt, factors-4.txt. Or else you can run one process in 4 different directories without the -u switch.

If you stop sr2sieve you can resume from where you left off by running the same command again, i.e. with the same range given by -p and -P and with the same argument to the -u switch, and in the same directory.

You can use a work file called sr2work.txt instead of using command-line arguments if you wish. If you use the -u1 switch then name the work file sr2work-1.txt, etc.

The work files should contain the range as a multiple of 10^9. e.g. if you gave the range as "-p 5000e9 -P 5250e9" in the example above then the work file would contain "5000-5250" (without quotes).
geoff is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
S9 and general sieving discussion Lennart Conjectures 'R Us 31 2014-09-14 15:14
Sieving discussion thread jasong Twin Prime Search 311 2010-10-22 18:41
Combined sieving discussion ltd Prime Sierpinski Project 76 2008-07-25 11:44
Sieving Discussion ltd Prime Sierpinski Project 26 2005-11-01 07:45
Sieving Discussion R.D. Silverman Factoring 7 2005-09-30 12:57

All times are UTC. The time now is 18:22.

Fri Nov 27 18:22:04 UTC 2020 up 78 days, 15:33, 3 users, load averages: 1.48, 1.54, 1.50

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.