BOINC Sieving

[Matt]

The PrimeGrid project is now running a boinc project which includes the PSP project, which also includes our sieving as far as I am aware. This means we effectively have a BOINC client for SOB (sieving) which is pretty awesome if you ask me.

You can see the ranges reserved by PrimeGrid here (http://www.sierpinskisieve.com/userinfo.php?user=203)

[hhh]

That's right. And you an in fact go to www.primegrid.com , get the boinc client, and get crunching. The sieving part is joint sieve for Seventeenorbust, and PrimeSierpinskiProject, so of great overall efficiency.

In something like two weeks, they found more than 1300 factors already, meaning that we eliminate about as many candidates by sieving as by LLR now.

H.

[ltd]

Looking into the 2007 stats ( http://www.henleyclan.co.uk/sobsieve/2007/scores.htm )
you can see that primegrid has already reached place 14 in the yearly stats.
And that after only 2 weeks.

[engracio]

Looking into the 2007 stats ( http://www.henleyclan.co.uk/sobsieve/2007/scores.htm )
you can see that primegrid has already reached place 14 in the yearly stats.
And that after only 2 weeks.

That is not counting the factors which have not reached the "active windows" yet. Which is worth more in points. Once it caught up with the "active window", watch it rocket to the moon.:thumbs:


e:)

[Joe O]

PrimeGrid now supports Linux sieving as well as Windows 32 bit and Windows 64 bit.

PSP Sieve
Platform Current version Installed
Microsoft Windows (98 or later) running on an Intel x86-compatible CPU 1.02 13 Oct 2007 12:21:18 UTC
Microsoft Windows running on an AMD x86_64 or Intel EM64T CPU 1.02 13 Oct 2007 10:49:17 UTC
Linux running on an Intel x86-compatible CPU 1.02 31 Oct 2007 10:14:27 UTC

[MikeH]

Looking into the 2007 stats ( http://www.henleyclan.co.uk/sobsieve/2007/scores.htm )
you can see that primegrid has already reached place 14 in the yearly stats.
And that after only 2 weeks.
Make that 5th, and looks like by tomorrow they'll have submitted the largest number of sieved factors for the year.
Sieving has gone mainstream :rock:

[MikeH]

Ah....and now we can all see that PrimeGrid is 5th. Moved to new computer - upload wasn't working. :bonk:

[Joe O]

Ah....and now we can all see that PrimeGrid is 5th. Moved to new computer - upload wasn't working. :bonk:

http://www.henleyclan.co.uk/sobsieve/tmp/results_duplicates_excluded_marked.zip still has not been uploaded.

[Brucifer]

Make that 5th, and looks like by tomorrow they'll have submitted the largest number of sieved factors for the year.
Sieving has gone mainstream :rock:

Should keep in mind here that all the boinc users fall under the primegrid account. So all the "manual" participants would fall under another "account" to keep things in perspective. But yes, primegrid will take over as there will be more and more participants than the non-boinc crowd. :)

[MikeH]

http://www.henleyclan.co.uk/sobsieve...ded_marked.zip still has not been uploaded.

Joe, hopefully that should finally be sorted. And it should get updated a little more often now. Sorry that's taken so long to fix.:spank:

[Matt]

PrimeGrid is now 4th and still moving, place bets now on when it will take 1st!

[MikeH]

PrimeGrid is now 4th and still moving, place bets now on when it will take 1st!
Not this year. Should take 2nd for the year, and then I guess will lock out the 2008 stats.

In the all time stats http://www.henleyclan.co.uk/sobsieve/alltime/scores.htm a lowly 56th and a long way to go.

[Joe O]

So the current standings are:

1 engracio
2 glennpat
3 PrimeGrid

Will this be the standings at years end? or will Primegrid move up to 2nd? or even 1st?

[Joe O]

At 11:20 this morning (November 12, 2007) over 100T had been sieved with 4036 factors found. This is just about 30 days after starting.

[Joe O]

145T sieved at 11/23/2007 10:20 am EST, 5672 factors found, just 41 days after starting.

[MikeH]

...and the standings are now


1 engracio
2+ PrimeGrid
3- glennpat

[b2riesel]

It's good to see you guys utilizing the boinc platform to sieve. I thought our success over at Riesel Sieve would catch your attention sometime. We are closing in on 5P and almost 4P of that is boinc sieve. Since we still have 66 k values left...it would have taken forever to sieve to this depth without boinc. Currently we are down to just 14T/day as we are moving more focus to LLR.

Ride the boinc sieve wave as long as you can. As long as those factors keep falling out...I'd ride that pony until it died.

Good luck and see you at the prime race finish line.

Lee Stephens
www.rieselsieve.com

[Matt]

PrimeGrid is #1 for this year in sieving and has reached position #6 over all time - not bad!

[vjs]

Sieve really going strong with bionic and the manual ranges.


PSP just eliminated a k so that should speed things up a touch as well. On wards and upwards. It's possible to beat riesel in sieve since our dat is smaller we would just need more people.

[balachmar]

Should all sieving now be done with the PSP Boinc client?
Because I see very little activity here.
And one of the reasons for me to start sieving is to lift the place of the Dutch Power Cows in sieving. But since I have to send in the results to both SoB sieving and PSP, will I have to register at both sites, and be getting points on both projects or how will that work?
Thanks for the information.

[Matt]

Manual sieving is still continuing.

You can think of combined sieving as operating in two sieve projects at once, it makes more efficient use of your processing power than sieving in the different projects seperately.

[hhh]

If you have a 24/7 internet connection, I guess BOINC is your choice. You will get BOINC credit.

If you have more of a standalone-PC, you should prefer manual sieving. If you are concerned about credit, you need in this case to get a username for the semiautomatic reservation system, for PSP (post a message in the mersenneforum-->PSP-->join-a-team thread), and at SoB (forgot how that works). You will submit factors at PSP, at SoB, and get separate credit.

Did that help? H.

[balachmar]

If you have a 24/7 internet connection, I guess BOINC is your choice. You will get BOINC credit.

If you have more of a standalone-PC, you should prefer manual sieving. If you are concerned about credit, you need in this case to get a username for the semiautomatic reservation system, for PSP (post a message in the mersenneforum-->PSP-->join-a-team thread), and at SoB (forgot how that works). You will submit factors at PSP, at SoB, and get separate credit.

Did that help? H.
Yes that did help.
So the BOINc client only gives BOINC credit. And doing it manually gives credit to both projects.
Right?
And since I am doing this to help DPC getting up in the DC vault, I'll go with the manual sieving then.
Thanks for the info!

[hhh]

Well, actually, there might still be confusion, as there are two kinds of credits at SoB. A compendium of all the different kinds of credit:

http://www.seventeenorbust.com/stats/teams/ ; here the DuchPowerCows are indeed doing fairly well; these are the stats for PRP-Tests, i.e. running the normal SoB-Client that you get here:
http://www.seventeenorbust.com/download

The SoB-sieving-stats look like this: http://www.henleyclan.co.uk/sobsieve/2007/scores.htm ; here I don't see any notion of team whatsoever. This one is the credit I referred to in my above post.

The PSP-sieving credit can be seen here: http://www.psp-project.de/stats.html ; the Power Cows have not yet shown particular interest for neither of the two sieving- and PRP-parts.

And concerning Primegrid, they score a mere rank 62 here: http://www.primegrid.com/top_teams.php ; being passed by even by third world countries like Canada and ... Belgium.



BUT: Sieving is the most efficient to bring the whole two projects further, and the gentlemen's choice, if you ask me.
Do whatever pleases you most. But one request: if you allready reserved a manual sieve range, please finish it, as not doing so causes manual work, I guess.

Cheers, H.

[balachmar]

Well, actually, there might still be confusion, as there are two kinds of credits at SoB. A compendium of all the different kinds of credit:

http://www.seventeenorbust.com/stats/teams/ ; here the DuchPowerCows are indeed doing fairly well; these are the stats for PRP-Tests, i.e. running the normal SoB-Client that you get here:
http://www.seventeenorbust.com/download

You call a number one position doing fairly well? :rotfl:
But this was not what I want to do, because this is what I am doing at the moment, because the Dutch Power Cows are on a stampede there.
I am looking ahead to find something to do afterwards.



The SoB-sieving-stats look like this: http://www.henleyclan.co.uk/sobsieve/2007/scores.htm ; here I don't see any notion of team whatsoever. This one is the credit I referred to in my above post.

Actually in the teams I do see:
9 Dutch_Power_Cows 943315.34 ( 0.30) 0 ( 1 ) 0 ( 1 ) 1 0 0 0 189/289
However we are not in this:
http://www.henleyclan.co.uk/sobsieve/2008/scores_t.htm



The PSP-sieving credit can be seen here: http://www.psp-project.de/stats.html ; the Power Cows have not yet shown particular interest for neither of the two sieving- and PRP-parts.

But with manual sieving I will be sending in results for both the SoB sieving and the PSP-sieving? right?



And concerning Primegrid, they score a mere rank 62 here: http://www.primegrid.com/top_teams.php ; being passed by even by third world countries like Canada and ... Belgium.

Well, that might be another thing for the future then :)


BUT: Sieving is the most efficient to bring the whole two projects further, and the gentlemen's choice, if you ask me.
Do whatever pleases you most. But one request: if you allready reserved a manual sieve range, please finish it, as not doing so causes manual work, I guess.

Cheers, H.
I haven't reserver anything yet, I am planning the switch to manual sieving in a week or so. After the Stampede has finished and the tests of the normal SoB are completed.

I hope that I now understand things correctly.

[hhh]

To get the most credit out of manual sieving, yes, you need to submit your factors at

http://www.seventeenorbust.com/sieve (don't forget to log in)
and
http://www.psp-project.de/sieveimport.php (check "Sob factors reported by user"; you need to drop a message in the "Join a team"-thread so that he can affect you to your team)
H.

[Matt]

Actually in the teams I do see:
9 Dutch_Power_Cows 943315.34 ( 0.30) 0 ( 1 ) 0 ( 1 ) 1 0 0 0 189/289
However we are not in this:
http://www.henleyclan.co.uk/sobsieve/2008/scores_t.htm


This is probably because the team has not submitted any factors yet this year?

If you look at http://www.henleyclan.co.uk/sobsieve/alltime/scores_t.htm you will see the all time stats for the team.

[balachmar]

I know that DPC wasn't in the rankings because they submitted no results. Well, know... I thought so.
That is why I am going to do it as a member of DPC. Just to get them active in a somewhat smaller project.
And who knows, maybe others DPC'ers find their way here as well.

[hhh]

I know that DPC wasn't in the rankings because they submitted no results. Well, know... I thought so.
That is why I am going to do it as a member of DPC. Just to get them active in a somewhat smaller project.
And who knows, maybe others DPC'ers find their way here as well.

If you want to do some good to SoB, and the fame of the DPC, you direct as many people as possible to Primegrid, crunching PSP/SoB joint work.

The reasons:

- sieving helps Sob even more than doing PRP tests;
- becoming number 1 should not be so difficult over there.
- everything is automatic, and easy, hence suited for a big crowd of people
- it is easier to manage new BOINC member for the admins than 50 new manual sieving members
- I think the manual sieving stats are somehow obsolete after the enormous success of the BOINC sieving revolution. (See Primgrid is already the number 2 all time user?)

H.

[balachmar]

just one more question, where do I register my username and team for the psp project?

[Matt]

For manual sieving you should post on the PSP forum on this thread ( http://www.mersenneforum.org/showthread.php?t=8162 ) to get your username added to the factor submission form. You can then post on this thread ( http://www.mersenneforum.org/showthread.php?t=2692 ) to join a team.

[balachmar]

OK, me again.
I just completed sieving of a range:
The reserved range was:
12303000G - 12303100G

I ran this command (on linux):
./sr2sieve -s -p 12303000000000000 -P 12303100000000000 --intel

And the result on screen was:
sr2sieve 1.7.9 -- A sieve for multiple sequences k*b^n+/-1 and b^n+/-k.
Read 1963642 terms for 16 sequences from dat format file `SoB.dat'.
Split 16 base 2 sequences into 160 base 2^360 subsequences.
Expecting to find factors for about 0.43 terms in this range.
sr2sieve 1.7.9 started: 991 <= n <= 49999997, 12303000000000000 <= p <= 12303100000000000
p=12303099997319357, 671093 p/sec, 0 factors, 100.0% done, ETA 15 May 04:07
sr2sieve 1.7.9 stopped: at p=12303100000000000 because range is complete.

What do I need to put into the form to send in the results?

OK, from what I have read so far, there is no factor found, so I have nothing to submit.
Because if it does find a factor, there would have been a factors.txt and this content I need to submit, right?
But How do I let you guys know that there is no factor to be found in this range?

[Joe O]

When you edit your range to mark it finished, put 0 in the Factors Found.

[balachmar]

Do I need to let PSP sieve know as well?
Or do you only get points for factors you have found?

[hhh]

Do I need to let PSP sieve know as well?
Or do you only get points for factors you have found?

No. That's right. At PSP, instantly; and for SoB factors only, when First(orSecond-)Pass passes them (makes use of them (to eliminate a candidate). (You get a little bit of credit, but the most is to come later).

(Remember someone could find a prime for that k, and then we wouldn't continue at all do LLR on your number).

Hope that makes sense... Yours H.

[balachmar]

No. That's right. At PSP, instantly; and for SoB factors only, when First(orSecond-)Pass passes them (makes use of them (to eliminate a candidate). (You get a little bit of credit, but the most is to come later).

(Remember someone could find a prime for that k, and then we wouldn't continue at all do LLR on your number).

Hope that makes sense... Yours H.
But do I have to send something in to the psp website?
I am doing manual sieving. And I didn't find a factor. So I don't know what I would be supposed to send in.

[Matt]

The manual sieving website is the only place where you need to reserve a range and mark it is complete. In the event that you find factors you should submit them to both PSP and SoB seperately.

[Death]

Hi guys. I need a team in psp sieve. I'm currntly in 35 place, but have no team ((

My team called Ukraine (as usual).

How can I have one?

[hhh]

Hi guys. I need a team in psp sieve. I'm currntly in 35 place, but have no team ((

My team called Ukraine (as usual).

How can I have one?

Death, you are alive!

I posted here

http://www.mersenneforum.org/showthread.php?p=133752#post133752

that should take care of it. H.

[balachmar]

I fear I am not getting in the SoB Sieve stats.
With sending in the result at SoB I get:
Factor table setup returned 1
Test table setup returned 1

0 of 1 verified in 0.02 secs.
0 of the results were new results and saved to the database.

and from PSP I get:
User: balachmar **SOB Flag: PSP_Only
Input contained 1 total lines.
Input contained 1 valid lines.
Input contained 1 valid lines from PSP k values.
Input contained 0 valid lines from SOB k values.
Input contained 0 invalid lines.
Input contained 0 lines with unknown K.
Input contained 0 lines with already known factor.
Input contained 0 wrong factors.

Am I sieving in the wrong range or what?

[glennpat]

I fear I am not getting in the SoB Sieve stats.
With sending in the result at SoB I get:
Factor table setup returned 1
Test table setup returned 1

0 of 1 verified in 0.02 secs.
0 of the results were new results and saved to the database.

and from PSP I get:
User: balachmar **SOB Flag: PSP_Only
Input contained 1 total lines.
Input contained 1 valid lines.
Input contained 1 valid lines from PSP k values.
Input contained 0 valid lines from SOB k values.
Input contained 0 invalid lines.
Input contained 0 lines with unknown K.
Input contained 0 lines with already known factor.
Input contained 0 wrong factors.

Am I sieving in the wrong range or what?

From the PSP side it says there was one factor and it was a PSP factor. When a PSP factor is submitted to SOB it will not be saved into the SOB database. It also will not say it is a PSP factor so it looks like it is ok. You will only get points in PSP for PSP factors and only get points in SOB for SOB factors.

[balachmar]

ok, so SoB factors are a different kind of factors than psp?
I didn't know that. But I am doing the right range, but sometimes a factor is a PSP factor and other times it is a SoB factor. right? Has it something to do with the formula being used?

[Sloth151]

The DAT file has 16 different K that it is looking at. 10 are from PSP, the other 6 from SOB. Every factor you find will be from one of the K. The one found happened to be from PSP. You will find a random mix of factors for different K in your range.

When you find factors you want to submit to both sites. As from http://www.free-dc.org/forum/showthread.php?s=&threadid=3501

You must use the combined dat for a reservation gotten through Matt's system.
Then you must submit the factors at two different places:
1)At SB's submit page
2)At PSP's submit page

NOTE:
If you only submit at the PSP site, your factors are not lost, but you won't get credit at the SB site.
If you only submit at the SB site, and do not send your factors to factrange at yahoo, your PSP factors will be lost.


S.

[balachmar]

OK, now lately I have sent in 2 results to SoB sieve. And I see 0 points on the stats. And it says they are excluded. What went wrong?

[Matt]

Maybe they were PSP factors, did you submit them to here: http://psp-project.de/sieveimport.php

[balachmar]

I submit them to both sites.
Bot they were SoB factors. Else they don't turn up in these stats:
http://www.henleyclan.co.uk/sobsieve/alltime/ti/0034.htm
I don't know what went wrong...
One of the factors was:
12309442058453173|55459*2^21693514+1
(I resubmitted it, just to check if that was the one, I hope this doesn't mess up anything)

[balachmar]

Can nobody help me out on why those factors were excluded?

[glennpat]

I submit them to both sites.
Bot they were SoB factors. Else they don't turn up in these stats:
http://www.henleyclan.co.uk/sobsieve/alltime/ti/0034.htm
I don't know what went wrong...
One of the factors was:
12309442058453173|55459*2^21693514+1
(I resubmitted it, just to check if that was the one, I hope this doesn't mess up anything)

On SoB sieve anything over 2^20000000+1 doesn't get recorded in a persons stats and you don't get points for them. I believe it does say when submitted that they are saved to the data base.

If it was a PSP factor then you do get points for it and the bigger the number the more points you get.

I didn't have much luck with these new ranges in the 12300000 range and have switched over to P-1 factoring to find factors described at:

http://www.free-dc.org/forum/forumdisplay.php?f=52

[Max Dettweiler]

On SoB sieve anything over 2^20000000+1 doesn't get recorded in a persons stats and you don't get points for them. I believe it does say when submitted that they are saved to the data base.

If it was a PSP factor then you do get points for it and the bigger the number the more points you get.

I didn't have much luck with these new ranges in the 12300000 range and have switched over to P-1 factoring to find factors described at:

http://www.free-dc.org/forum/forumdisplay.php?f=52
Hmm....wouldn't factors for the whole <50M range be credited, since the dat file goes up that far? (Or has the system just not been updated from when the dat file used to go to 20M?)

[balachmar]

Or else the software should stop checking for SoB factors above the threshold where you don't get any points. (Else I am wasting time) If they however want these factors then credit should be given.
I obviously prefer to get credit for the work I send in.
Could maybe hhh explain a bit, what is up with the credits?

[hhh]

If you call me like this, I will pop in, even if my memory about the subject is bad.
Here we go. (All the following info is modulo "IIRC"):Pokes:

PSP gives credit for factors as a function of n and p the factorsize, end of discussion. Note that even factors get credit that might reveal to be useless, because a prime is found before they save an LLR test.

At SoB, the story is more difficult. Here, factors are like a bank deposit that you can't withdraw immediately. The thing is that a factor only scores when it has actually saved an LLR test.

Say, you submit a factor for n=16M. You get no (or only marginal, I don't know) score for the moment. When in one or two years, firstpass LLR reaches 16M, your factor will save a LLR test, and you get your score. Now, imagine, a prime is found for the k of your factor. So doublecheck will never be done, and your factor does not score again. If no prime is found, however, it will, a couple of years later.

Why do the factors above 20 M not appear at all? Because a couple of years back, we switched from sieving until 20M to until 50M, and the script and the scoring system and everything has not been updated yet. However, factors above 20M will save a test only in 10 years or so, so until then, this might become fixed.

As you know, factors can be found by sieving, and P-1. While sieving just tries all p for any n, and hence gets factors for randomly scattered n, with P-1, we search any factor p for n that we can fix. So, we search for factors for n just above the firsrpass level, in order to save these 2 LLR tests before it is too late. To make this more interesting, there is something called "active window". Factors in this active window (n between firstpass and firstpass+300000 or so) score twice as much. You might be lucky and find such a factor by sieving, but you are sure about that for P-1. Additionally, P-1 saves LLR tests soon, and hence you get youre score soon.

PLEASE don't be angry with this system, and SUBMIT EVERYTHING YOU FIND, even if you have to wait a little eternity until you see the effects. If you want sieveing credits immediately, I suggest BOINC, which gives you even credits if you don't find a factor.

Lengthy post, but it's no easy subject.

Glad to help, H.

[balachmar]

Thanks for the reply this clarifies it.
So in time the script will be changed and I might get credit for the "excluded" factors. But if I want to get credit soon, I need to do P-1.
I'll check P-1 out then :)

[vjs]

balachmar,

Humm.. if memory serves me correctly one will not recieve credit for excluded factors.

Excluded factors are a little weird and why the client reports them I'm not sure.

The reason why excluded factors don't count is pretty simple really.
An excluded factor is not needed since another (generally) smaller factor has already be found for the same k/n pair. So basically even though its a real factor it wasn't needed because we already factored out the k/n pair.

If you look at the k/n pair that your factor sieved out you will actually see that that k/n pair does not exist in the database.

Why these factors are generated I don't know... I think it has something to do with the fact that one factor cannot factor more than 1 valid k/n pair, so if a k/n pair is found that your p matches there is no need to factor further with that p and the client moves on. I guess this makes the client a little faster. ( On why exactly, i'm not sure (recalling from memory) but I know they won't credit.)

I probably have a couple thousand of these...

[Death]

Well, excluded factors like double check in general SoB. I prefer to have two different factors for k*n pair, just to be sure. And AFAIK, some SoB primes come from second pass )))