View previous topic :: View next topic |
Author |
Message |
Yankton Prince


Joined: 27 Sep 2008 Posts: 1702 Location: California
|
Posted: Sat Jan 31, 2009 11:40 am Post subject: Sieving for PSP and SoB at the same time |
|
|
These are 2 of our worst vault rankings and I think part of the reason is the instructions are horrible.
There are a few things you need to keep in mind. Running a batch for these projects takes quite a while and you have to be patient to get your credit until after your work is submitted. On older machines a decent batch can take a month! I'm experimenting running this on a newer quad right now and it appears the run times are down to a couple of days on it. You also have to manually submit your work and modify some text files to run this. It's not difficult.
First you should download the software. http://www.geocities.com/g_w_reynolds/sr2sieve/
There is a note at the bottom that version 1.7.15 is the latest stable verstion so that is the one you should run. But if they change it between now and when you start this take note of it in case it has changed.
There are four files of the current verstion available. Two for windows and two for linux. For each of these there is a 32-bit and a 64-bit client. Download the one most appropriate to you.
The linux versions are tarballs. You can extract them using the archive manager or using the command line. Place them wherever you are most comfortable (/opt or your home directory or whatever you like)
The windows versions are zip files. You can extract them to a directory that suits your preferences. I made a directory in my Program Files directory or you can just place them in a directory under c:
Once you have it extracted you need the main DAT file in that directory as well. This can be downloaded at http://www.psp-project.de/test/sievecomb.zip
This zip file is used for all versions. Extract it and place it in the directory where you placed your software.
Now before we get ahead of ourselves you need to make a couple of accounts.
First go to http://www.mersenneforum.org and register yourself. Once you are registered go to http://www.mersenneforum.org/showthread.php?t=2692 and make a post asking to be added to the Knights who Say NI!. One of the mods will add you.
Next go to http://www.seventeenorbust.com/ and click the Sign Up link on the left. Once you have an account log into the site and select preferences on the left. On the page that comes up click Team Affiliation. On that page there is a drop down list of teams to join. I'm sure you all know which one to pick! If you are feeling looney you can also set up your public profile.
Now that you have software and accounts you need a range of data to work on. Go to http://www.sierpinskisieve.com/ and bookmark it. You will be back. Sign up for an account here as well. This is so you can have your name attached to the range that you select so no one else wanders in and does it and takes credit. Once you are signed in click the overview link and it will take you to the main reservation page. Click on the reserve a range link.
Now, it has a calculation system down at the bottom for helping you pick a range size based on how fast your computer is and how long you want it to take for one range.
For a taste, if you have a slower computer you may want to start with a 50G range and on faster computers a 200G range.
Enter your choice in the range size and select reserve range. Keep this page open so that you can see your numbers.
Now the client has switches to enter the ranges in the command line, but I never got them to work properly, fortunately there is a simple way to run these. With the range handy go to the directory where you have your software installed and make a new file called sr2work.txt
You will make one line in this file. The start of your range,the end of your range.
For example, my most recent reservation was Quote: | 12703300-12703500 G reserved by Yankton |
So my sr2work.txt file will have the following line:
Now I'm sure this can be automated as a service but I like to watch the progress (especially as when it finishes the range it just exits). so open a terminal window (Windows and linux all work the same for this software) and go to the directory where your installed this.
For windows type:
For Linux type:
The bottom line will allow you to monitor progress, expected finish time, how many factors are found, etc.
One thing, this will run on a single core only, so if you have a multi-core machine and you want to run this on more than one core, make a copy of the directory and rename it to something else (copy it after you have the dat file installed - less work), reserve a range for it, make your sr2work.txt file for the additional core and let it loose!
Now, when the program ends you will have one of two things happen. It found factors or it didn't. If it didn't, you log back onto the reservation system and view your ranges. Edit the range that found nothing, select the finished box, make a comment of no factors found and save it. Unfortunately you get no points for no factors.
On the other hand, you make good points if you do find factors but you need to submit them in two places as some of what you find will be PSP factors and others will be SoB factors. Submit them all to both, they will figure it out.
Go to the directory with your software and you will find a new text file that has the word 'factors' and the end of your range in it. Open it and copy the contents.
To Submit to PSP go to http://psp-project.de/sieveimport.php
Paste the contents of the factors file in the box and select your name from the drop down, select SoB factors submitted by user and submit.
Then to submit to SoB go to http://www.seventeenorbust.com/sieve/
You will need to log in and then return to that page. While logged in paste your factors into the provided box and press Submit.
You can then go back to the reservation system and gather a new range to work on!
If you have any questions please feel free to submit them here.
Happy Sieving! [/code][/quote] _________________ Some days are worse than others. |
|
Back to top |
|
 |
ohiomike Prince


Joined: 20 May 2007 Posts: 858 Location: Sometimes
|
Posted: Sat Jan 31, 2009 2:54 pm Post subject: |
|
|
Do we have to stand on our left foot with our eyes covered while we do that,
or will simply slaughtering a goat on the keyboard suffice? _________________

Resident Linux fan and credit ho >My Shrubbers<
Proud member of the "Fry a CPU for breakfast club"
Last edited by ohiomike on Sat Jan 31, 2009 3:12 pm; edited 1 time in total |
|
Back to top |
|
 |
Yankton Prince


Joined: 27 Sep 2008 Posts: 1702 Location: California
|
Posted: Sat Jan 31, 2009 3:11 pm Post subject: |
|
|
Honestly, after you go through it once, it gets VERY easy. It's just annoying the first time through. Did I leave something that needs more clarifying? _________________ Some days are worse than others. |
|
Back to top |
|
 |
PhastPhred Prince


Joined: 22 Mar 2006 Posts: 6017 Location: Northwest AR (USA)
|
Posted: Sat Jan 31, 2009 3:41 pm Post subject: |
|
|
ohiomike wrote: |
Do we have to stand on our left foot with our eyes covered while we do that,
or will simply slaughtering a goat on the keyboard suffice? |
You ALMOST got it! Slaughter the goat WITH the keyboard... (It works better than a Herring)  _________________
 |
|
Back to top |
|
 |
Mildew KWSN ArchBishop


Joined: 13 Oct 2002 Posts: 2617 Location: Sweden
|
Posted: Sun Feb 01, 2009 4:14 am Post subject: |
|
|
You don't have to create a work file. You can just give the range at the command-line when you start the program.
Code: | ./sr2sieve -s -p 12703300e9 -P 12703500e9 |
Last time I ran this project was on a Q6600, and then I reserved a 4000G range and ran 1000G on each core. Took about two weeks I think, but the numbers were a lot smaller then, and each 1G step takes longer the higher up you get, so it would probably take longer than that now.
Since a large range can take a long time to complete, there may come a time when you want to restart your computer. I would recommend taking a backup of the sieve folder before starting it again. Most (hopefully all) times you will have no problems at all, but if for some reason you happen to screw up, you may lose a lot of work. (I forgot to add the "e9" after the numbers once when restarting, which resulted in a very small range being completed in no time, and then all my earlier progress was lost) |
|
Back to top |
|
 |
Yankton Prince


Joined: 27 Sep 2008 Posts: 1702 Location: California
|
Posted: Sun Feb 01, 2009 7:37 am Post subject: |
|
|
That might be a point for the workfile then, if you restart you just run sr2sieve -s when it comes back up and it restarts at it's checkpoint. I tried using the -p -P when I first attempted to use the program and couldn't get it to work right. No e9 meant it wouldn't run (outside the range of the dat or some such) yet none of the help files explained that. It looks like 200G on a Q6600 (1 core) is going to take 4 days, so 1000G should take about 3 weeks in the current ranges. Such an annoying program to use, but the points are worth it. _________________ Some days are worse than others. |
|
Back to top |
|
 |
Idan UN-Smitten


Joined: 07 Dec 2005 Posts: 2993 Location: Tel-Aviv, Israel
|
Posted: Sun Feb 01, 2009 3:05 pm Post subject: |
|
|
I followed the instructions, and after countless registrations and sign ups, I started crunching a modest 100G range for this weird project on one cpu, once it will finish and I'll have a taste of it, I will know what it takes to continue...
Good luck!  _________________ Anyone for Crunch?
 |
|
Back to top |
|
 |
Yankton Prince


Joined: 27 Sep 2008 Posts: 1702 Location: California
|
Posted: Sun Feb 01, 2009 3:28 pm Post subject: |
|
|
Most excellent. The hardest part really is signing up for everything, once it's going it's rather simple. Good Luck! _________________ Some days are worse than others. |
|
Back to top |
|
 |
ohiomike Prince


Joined: 20 May 2007 Posts: 858 Location: Sometimes
|
Posted: Sun Feb 01, 2009 3:44 pm Post subject: |
|
|
OK- I also have one 100G range running on a VM on my MacPro. We shall see!
Setting it up is just somewhat easier than giving yourself a root canal with a hammer and chisel! _________________

Resident Linux fan and credit ho >My Shrubbers<
Proud member of the "Fry a CPU for breakfast club" |
|
Back to top |
|
 |
Idan UN-Smitten


Joined: 07 Dec 2005 Posts: 2993 Location: Tel-Aviv, Israel
|
Posted: Sun Feb 01, 2009 3:51 pm Post subject: |
|
|
When I will be able to see my name at http://psp-project.de/sieveimport.php?
And where do you see the credits for this projects?  _________________ Anyone for Crunch?
 |
|
Back to top |
|
 |
Yankton Prince


Joined: 27 Sep 2008 Posts: 1702 Location: California
|
Posted: Sun Feb 01, 2009 4:05 pm Post subject: |
|
|
I'm not sure how long it took before my name showed up on that list, but it was before I submitted anything - give it a day.
As for shrubs shrubbed, you can find the stats for PSP at http://www.psp-project.de/stats.html
For the SoB Sieve, you can check the stats at http://www.henleyclan.co.uk/sobsieve/alltime/ui/19999.htm
This page doesn't get updated as often as it could, they may lake a certain amount of automation. But as Sir OhioMike pulls the data for the stats page it would probably be easier to just check there  _________________ Some days are worse than others. |
|
Back to top |
|
 |
ohiomike Prince


Joined: 20 May 2007 Posts: 858 Location: Sometimes
|
|
Back to top |
|
 |
Yankton Prince


Joined: 27 Sep 2008 Posts: 1702 Location: California
|
Posted: Sun Feb 01, 2009 5:16 pm Post subject: |
|
|
They are a pain. The sieving is from both projects but you can use one program to sieve for both (which is good). PSP also has the prp which can be done in primegrid on boinc but we need actual project credits for the vault. SoB also has a prime search but happily that is completely seperate.
Just remember...Fire Hot! _________________ Some days are worse than others. |
|
Back to top |
|
 |
Idan UN-Smitten


Joined: 07 Dec 2005 Posts: 2993 Location: Tel-Aviv, Israel
|
Posted: Sun Feb 01, 2009 5:22 pm Post subject: |
|
|
What does this line mean:
Quote: | Expecting to find factors for about 0.39 terms in this range. |
Are those the odds for a factor to be found in this range?
furthermore, what do I do when I don't find a factor? I should submit something to both project so they'll know there are no factors for this range (even if I won't get credit)... _________________ Anyone for Crunch?
 |
|
Back to top |
|
 |
Yankton Prince


Joined: 27 Sep 2008 Posts: 1702 Location: California
|
Posted: Sun Feb 01, 2009 5:58 pm Post subject: |
|
|
It's the odds that you'll find something. You can't rely on that. Basically it thinks you have a 39% chance of finding one factor. You might find more, or you might find none.
Quote: | Now, when the program ends you will have one of two things happen. It found factors or it didn't. If it didn't, you log back onto the reservation system and view your ranges. Edit the range that found nothing, select the finished box, make a comment of no factors found and save it. Unfortunately you get no points for no factors. |
_________________ Some days are worse than others. |
|
Back to top |
|
 |
Idan UN-Smitten


Joined: 07 Dec 2005 Posts: 2993 Location: Tel-Aviv, Israel
|
Posted: Sun Feb 01, 2009 6:00 pm Post subject: |
|
|
Will do, Thanks Yankton, didn't notice that line, and I've read it like 5 times today just to understand how to get some work for this project  _________________ Anyone for Crunch?
 |
|
Back to top |
|
 |
Mildew KWSN ArchBishop


Joined: 13 Oct 2002 Posts: 2617 Location: Sweden
|
Posted: Mon Feb 02, 2009 1:40 am Post subject: |
|
|
The really fun part is that you get credit per found factor.
In other words, you have a 39% chance to be awarded credit for all your hard work  |
|
Back to top |
|
 |
Idan UN-Smitten


Joined: 07 Dec 2005 Posts: 2993 Location: Tel-Aviv, Israel
|
Posted: Mon Feb 02, 2009 3:26 am Post subject: |
|
|
39% to be credited... Those odds are sure looney
So, let me get this straight, if I take a batch of 1000G and find 2 factors, and find about the same factors within a 100G range (2 factors with the same importance) so I'll get awarded with the same credit for both batches?  _________________ Anyone for Crunch?
 |
|
Back to top |
|
 |
Mildew KWSN ArchBishop


Joined: 13 Oct 2002 Posts: 2617 Location: Sweden
|
Posted: Mon Feb 02, 2009 4:06 am Post subject: |
|
|
Idan wrote: | 39% to be credited... Those odds are sure looney
So, let me get this straight, if I take a batch of 1000G and find 2 factors, and find about the same factors within a 100G range (2 factors with the same importance) so I'll get awarded with the same credit for both batches?  |
Correct.
However, if you run enough work, the probabilities will even out so that you get lucky about as many times as you get unlucky.
This is where finding primes differ from finding girls.
 |
|
Back to top |
|
 |
ohiomike Prince


Joined: 20 May 2007 Posts: 858 Location: Sometimes
|
Posted: Tue Feb 03, 2009 7:50 am Post subject: |
|
|
OK, Returned my first results (2 factors), we will see if I reported them right.
(What a pain).  _________________

Resident Linux fan and credit ho >My Shrubbers<
Proud member of the "Fry a CPU for breakfast club" |
|
Back to top |
|
 |
|