PDA

View Full Version : How to get rolling on PSP Sieve



NeoGen
12-24-2005, 10:56 AM
(1)First, as all projects need an account to keep track of your individual progress, you have to create an account, but the catch in this project is that everything is ran from the forums, so first of all you should go to the Marsenne forums and register yourself:
http://www.mersenneforum.org/register.php?
(The marsenne forums host subsections for plenty on Maths projects, including PSP)

Now that you got your registration done, you better bookmark the PSP forums, as you will have to go there regularly:
http://www.mersenneforum.org/forumdisplay.php?f=48

(2)On the PSP forums, you can join a team by leaving a post on their sticky thread called "Teams : Join a Team". You just go there and post "Please join me on <team name>" and the admins will take in account that all your results will benefit that team.
Tip: Always join AMD Users ;)

Back on the PSP forums you'll see a sticky thread called "Sieve:- Pick A Range". Click on it to make the reservation of a range to sieve. (Same as a workunit to crunch but you get to select the size you want)

(3)To reserve a range you only have to post on that thread asking the range you want. Here's a typical example of a request

13200-13250 reserved by vaughan
And the project admins thank you if you reserve the lower ranges possible. Here's an example list of reserved ranges:

13150-13200 reserved by NeoGen
13200-13250 reserved by vaughan
13250-13300 reserved by magnav0x
Unless other ranges were already reserved in posts below, the best range you could select would start at 13300. So for example to reserve a range of 100G you would just have to post on that thread something like:
"13300-13400 reserved by <your_nickname>"

So now you claimed your workunit, what to do with it? To crunch your workunit you'll need two files.

(4)The first file you need is the application - Proth Sieve (download here (http://www.teamprimerib.com/vjs/proth_sieve_042_all.zip))
The zip file contains compiled versions for each OS/CPU type. Use the one that most suits you. The best one is always the one with SSE2 support, if your cpu only supports SSE, then use the CMOV application... and if you have an older machine, then your last (and slowest) solution is the regular one with no optimizations

The second file is required for the sieving - sob50m.dat (download here (http://www.ldausch.de/test/sob50m.zip))
Just download and extract it into the proth sieve folder.

So now you've got your registration done, your workunit claimed, it's time to get to work!

(5)Just click twice on proth sieve to start it, a command line window should appear asking what is the range to sieve. Here you insert the range you claimed at the forums.
Using the example numbers from above, when it asks you "Sieve from" you insert "13300" and click Enter. And then it'll ask you "Sieve to" and you insert "13400" and click Enter again.
(You will insert your own reserved range's values. Not these example ones)

And that is that. Now you only have to wait until it finishes.
If you need to shutdown the computer, you can close the command line application. When you restart, proth sieve will continue from where it left off.

(6)Once finished, do as instructed in the range reservations thread:

Result submission:
Please send fact.txt and if you have factrange.txt to: psp"at"ldausch.de
And don't forget to post on the range reservations thread once more to announce that you finished the range. Just post something like "13300-13400 finished by <your_nickname>"
(Of course, post with your range numbers, not these example ones)

Clear the folder of all files other than the proth sieve application and sob.dat, reserve another range and repeat the process again.

Ototero
12-24-2005, 12:19 PM
Nice instructions Neo, I'm sieving now as well.

vaughan
12-24-2005, 12:48 PM
Good work NeoGen. I think I followed what you said :!:

I've got an xp2800+ running PSP Sieve now too.

NeoGen
12-24-2005, 11:09 PM
Thanks guys! :)

It's pretty easy to make simple sieving. For starters and people that just want to try it out it's more than enough.
One of these days I'll add the extra steps for getting into combined sieving with SOB. I'm trying it just now. One thing I can tell already is that it's at least 20% slower than normal sieving.

Brucifer
12-24-2005, 11:50 PM
Okay, you fished another one in. I've got 13400-13800 reserved.

So since I'm a newbie on this project, what is the range above -- single pass, non-combined? What does all that mean? Show me the stats! Gotta see the official psp stats page

Is this effort tied in with SOB? Do they share data? How much work is left?

Brucifer
12-25-2005, 12:26 AM
When we go to email the results in, can we compress them and send them together?

NeoGen
12-25-2005, 12:31 AM
Yep. :)

I actually did that the first time I submitted back results. I zipped the files. The results were received and accepted.

Anonymous
12-25-2005, 01:02 AM
72CJ5,

PLease reserve ranges on our forum (www.mersenneforum.org). As you see some one else had already reserved 13400-13800. If you check.

Citrix

NeoGen
12-25-2005, 02:22 AM
Okay, you fished another one in. I've got 13400-13800 reserved.

So since I'm a newbie on this project, what is the range above -- single pass, non-combined? What does all that mean? Show me the stats! Gotta see the official psp stats page

Is this effort tied in with SOB? Do they share data? How much work is left?

Ok, one question at a time... :)
The range above is indeed non-combined, and second-pass (I think)
Non-combined means that the results are only used by PSP, as opposed to combined where the results fit for both PSP and SOB
The stats are here: http://www.ldausch.de/stats.html
PSP and SOB share data yes, especially about the combined ranges. And lots of work is left, don't worry because there's plenty to go around for a few years still. :)


Citrix... you are right, 72CJ5 (here) is Brucifer (there). There's no conflict about the ranges.
And hey, nice to see you around. I sure didn't expect you to pick up this thread so fast. If I got something wrong up there in the tutorial, please point it out and I'll correct it. :)

And to everyone else that is not aware... Citrix is the PSP founder! :cool:

Brucifer
12-25-2005, 04:28 AM
So how is the second pass stuff scored?

Anonymous
12-25-2005, 08:59 AM
So how is the second pass stuff scored?
There is no difference in points if you do a second pass range or a first pass range in sieving.
The thing is that second pass sieving is needed to find all factors for our expandet test range. We ran upto n=20Mil in the past and now we run upto n=50Mil to have compatible ranges with SOB.

Lars

PSP DB Maintainer AKA as ltd

Anonymous
12-25-2005, 09:33 AM
When we go to email the results in, can we compress them and send them together?

I can handle all the "normal" compressions. (zip,7zip,rar,lzh....)

Lars

carlos
12-25-2005, 10:38 AM
I really would like to defend Free-DC position but it's getting difficult over here to stabilize the client.

Carlos

vaughan
12-25-2005, 10:50 AM
... but it's getting difficult over here to stabilize the client.

Carlos

What do you mean "stabilize"? Can you elaborate please?

Do you use Sobistrator? I have been unable to reach the site that used to host it.

carlos
12-25-2005, 10:56 AM
From time to time the client crashes and I don't know why. The funny thing is that it only happens at night. There's another client, more stable, but slower on PSP.

Carlos

AMDave
12-26-2005, 12:48 AM
...
Do you use Sobistrator? I have been unable to reach the site that used to host it.

Vaughan,

you can pick up the SoBistrator software here from the author:
http://81.233.34.242/sob.php

vaughan
12-26-2005, 01:57 AM
Thanks AMDave - got it downloaded.

AMDave
12-26-2005, 04:08 AM
I'd like to welcome citrix and ltd from the PSP project to the AMD Users forum.

Thanks for dropping in fellows.

Ototero
12-26-2005, 10:07 AM
Mine crashed last night also. I restarted it OK thos morning. A few more wasted cycles.

NeoGen
12-26-2005, 10:19 AM
Also to Carlos it crashes sometimes and he says it's only during the night.
If not hardware failure, then it's a really odd bug... only striking at night? lol

I don't understand how that happens. and mine has been running flawlessly for days and nights in a row.

carlos
12-26-2005, 10:48 AM
So I'm not the only one.

Ototero,

Try using the client that can be found in this thread (http://free-dc.org/forum/showthread.php?t=10069). It's slower but it's stable and gives more factors. Proth has a few bugs. Can't tell you much about it. To run the client in idle mode you have to create a batch file with the following content:


start /low jjsieveSSE2.exe -d -w

NeoGen,



I don't understand how that happens. and mine has been running flawlessly for days and nights in a row.

Just wait....for the vampire bug....

Carlos

Ototero
12-26-2005, 10:58 AM
cheers Carlos.

carlos
12-26-2005, 11:10 AM
cheers Carlos.

Thanks.

Are you using emule?

Carlos

Ototero
12-26-2005, 11:47 AM
yes, but not at this minute

ltd
12-26-2005, 01:21 PM
Is there any entry in the event viewer that gives a hint?
I find these crashes very strange as for example my longest runing proth_sieve client runs now for 63 days without stopping.

Lars

Ototero
12-26-2005, 01:36 PM
Yes Lars, here is a cut-and-paste.

Faulting application proth_sieve_sse2.exe, version 0.0.0.0, faulting module , version 0.0.0.0, fault address 0x00000000.



I have XP Pro SP2 running on AMD 64 3000+

Seems very little to go on.

ltd
12-26-2005, 01:50 PM
Yes Lars, here is a cut-and-paste.

Faulting application proth_sieve_sse2.exe, version 0.0.0.0, faulting module , version 0.0.0.0, fault address 0x00000000.



I have XP Pro SP2 running on AMD 64 3000+

Seems very little to go on.

This looks more like a programm error then a hardware problem to me.
"fault address 0x00000000" is typical for a invalid pointer operation which can happen if a local variable is not correctly initialised.

Have you tried the cmov variant of the application also or is it slower on the AMD 64?

Lars

Ototero
12-26-2005, 01:58 PM
Lars

So far this is the only version I've, Carlos, who posted earlier here, suggested another version.

ltd
12-26-2005, 02:11 PM
jjsieve is a new approach and is still in beta. I use it for stability testing at the moment but it is still slower then proth_sieve.
(Only marginal when used with a combined range but at least 10-15% slower on the second pass ranges)
jjsieve scales better the more "k" are in the test file.


If you want i can send you the proth_sieve_cmov executable by mail or you can find all available version here:

http://81.233.34.242/proth_sieve.php

Lars

Ototero
12-26-2005, 03:02 PM
Thanks Lars,

I'll consider other versions after this one crashes again. It's running nicely at around 400kp/s.

Ototero
12-28-2005, 04:09 PM
Strange, when I'm running PSP Sieve, my cpu temperature drops 3 degrees C :lol:

When it finishes dnet kicks in and it rises 3C again.

NeoGen
12-28-2005, 06:27 PM
I noticed that same 3 or 4 degrees difference between sieving and PRP'ing.

Seems like sieving is light work for the cpu :)

vaughan
12-28-2005, 11:44 PM
If you want some really light work for the cpu try eOn - it :sleeping: every few minutes.

I have also found that PSP runs better on a PC with a borderline hsf. My XP2800+ has a ThermalTake Volcano 7 "noise blaster version" and this rig powers itself off when it gets too hot. Running Dnet or BOINC and it lasts a few hours of run time. With PSP it will run 24/7; heck I even had eOn running on it in parallel and it kept going.

Lagu
12-29-2005, 04:08 PM
Hej

I have found MJ-12 and uFluid can run together because MJ-12 only use high usage when archiving but this demand a powerful computer with at least 2.200 MHz and 512 MB RAM

Lagu.
:)

Lagu
12-29-2005, 04:58 PM
Hej

I’m now running eOn thanks to Vaughan who told the agent only received a WU every 5 minutes. I have not known about that and it is good to run this together with other projects.

There have been talking about eOn earlier but I have been sitting in front of the monitor with my eyes closed. :D

I’m now running MJ-12, uFluid and eOn in a glorious mix.

Thanks
Lagu :)

Anonymous
12-29-2005, 08:16 PM
[quote=72CJ5]Citrix... you are right, 72CJ5 (here) is Brucifer (there). There's no conflict about the ranges.
And hey, nice to see you around. I sure didn't expect you to pick up this thread so fast. If I got something wrong up there in the tutorial, please point it out and I'll correct it. :)

And to everyone else that is not aware... Citrix is the PSP founder! :cool:

Thanks Neogen. Everything seems correct to me. Just to let you guys know that AMD can be used for PRP effort also, they do not have to be used for Sieve only. Unlike the P4's that should only be used for PRP.

AMD rocks! :D


Citrix