Sieve Discussions

[BlisteringSheep]

When marking ranges done through the automated system, it asks for the number of factors found. Should this be the number of SoB-only factors or the number of combined SoB+PSP factors?
If it should be the SoB-only factors, then my range 1224300-1224340 needs to be changed to indicate only 3 instead of 5 (the comment on the range is correct).

[vjs]

Sheep,

That's an interesting question.

The feild for factors found is not critical in anyway, it was used to judge how many factors one could expect to find or for personal boasting. I don't nessarily trust those numbers and we have better methods for determining factor density.

[jasong]

I'm sorry if I caused a problem, but I reserved 630000-630500 over at Mersenne Forum and forgot to come here. Unfortunately, I can't get to the SOB sieve reservation site for some reason, so I'm not sure if that's been reserved yet.

I'll try again later.

[Joe O]

I'm sorry if I caused a problem, but I reserved 630000-630500 over at Mersenne Forum and forgot to come here. Unfortunately, I can't get to the SOB sieve reservation site for some reason, so I'm not sure if that's been reserved yet.

I'll try again later.

The only place you need to reserve that range is on the Mersenne Forum, where you did. Just sieve it, submit it and post it complete on the Mersenne Forum.

[jasong]

Correct me if I'm wrong, but the range 630000-630500 shouldn't have any any factors found for k's below 78557, right?(I think it's 78557) Well, I'm about a quarter of the way through the range, and I've found two.

[Joe O]

Correct me if I'm wrong, but the range 630000-630500 shouldn't have any any factors found for k's below 78557, right?(I think it's 78557) Well, I'm about a quarter of the way through the range, and I've found two.

You are wrong, sorry. That's the reason we are using the combined dat. There are factors to be found for k's below 78557, not many, but some.
Please submit them here (after logging in, of course) http://www.seventeenorbust.com/sieve/

[paleseptember]

Not sure if I should be starting a new thread for this...

I want to get involved in the Sieving project (mainly as my work computer is firewalled, and I can't use the regular SoB client here).

So, after much confusing jumping between explanation pages, I *think* I have got this set up correctly.

Have d/led the Sieve_setup and combined.dat zips linked in the first message in this thread. Have extracted to C:\SieveA (not worrying about SieveB as this is an old P4 2GHz machine). Headed over to the Automatic Reservation System, created an account (paleseptember, same as SoB), reserved a 1000G range (as recommended by the README that came with the Sieve_setup zip). Have setup Sobistrator to look at this range, which is now chugging away with jjsievesse2 sieve.

My question is - is there anything else I have to do? Do I need to post my reserved range anywhere, or does the Automated Reservation system take care of that?

Annnnnd, once it's finished a range, I need to export some output data to the SoB and PSP sites (which file is that?), and zip up fact, factexcl, and factrange and email off to factrange-at-yahoo.com, yes?

[Joe O]

My question is - is there anything else I have to do? Do I need to post my reserved range anywhere, or does the Automated Reservation system take care of that?

Annnnnd, once it's finished a range, I need to export some output data to the SoB and PSP sites (which file is that?), and zip up fact, factexcl, and factrange and email off to factrange-at-yahoo.com, yes?

You do not need to post your reserved range anywhere else, the Automated reservation system is the only place.

The fact.txt file is the only one that you submit at the SoB and PSP sites.

Since you are running on an Intel Piv, you might want to try out SR2sieve to see if it is faster. See attached.

[paleseptember]

Thanks Joe!

Am trying to work out how to set up sr2sieve (on a different computer. Am not exactly brilliant with computers >.< ) I've extracted the sr2sieve zip to C:\sieveA, setup sobistrator, and run sr2sieve -s from the command line. It crunches the sample range included with sobistrator, but then wants the sr2work.txt file

The log file lists:
"
09/05/07 12:26:16 ERROR: Failed to open input file `sr2work.txt'.
"

I followed the readme, and created a sr2work text file with 1000,1001. The error is now
"
09/05/07 12:28:18 ERROR: Invalid or unrecognised file `sr2data.txt'.
"

It's about this point where I have no clue...

(PS. If an admin or mod wants to split this discussion off into a separate thread so it doesn't clutter up the Info on Getting Start..?)

[Joe O]

Three of the command line arguments that you will find of interest are:

 -s --sierpinski       Sieve sequences k*2^n+1 in dat format file 'SoB.dat'
 -r --riesel           Sieve sequences k*2^n-1 in dat format file 'riesel.dat'
 -i --input FILE       Read sieve from FILE instead of `sr2data.txt'.
                         Instead of `SoB.dat',`riesel.dat' when used with -s,-r.

You are now looking for the dat file. If it is named sob.dat then all you need is the -s on the command line. If it has another name, then you need -s -i 'another name' on the command line.

These posts will be moved to the sieve discussion thread when we are done.

[paleseptember]

Okay, I reserved a new range (just 100G), made a new sr2work.txt file with "1254600,1254700" and ran "sr2sieve-intel -s".

Seems to work (needed the sr2work.txt file though).

It's hammering away at about 580kp/s compared to about 400kp/s I was getting using JJsieveSSE2. 45% speed boost - nice! (pity this is a library computer (am working off my thumb drive), and is vastly superior to my work computer and home computer :P)

[engracio]

I created a batch file and saved it on the same directory so that it is easier to just double click on it instead of opening a command prompt and typing the command. -v gives you more output. I saved it as sr2sieveintel.bat or sr2sieveamd.bat for amd boxes.

sr2sieve-intel -s -v

I'm not sure how user friendly the proggie is, so I normally alt-crt-del when closing instead of just X-ing. It saves the last point on the checkpoint.txt

Okay, I reserved a new range (just 100G), made a new sr2work.txt file with "1254600,1254700" and ran "sr2sieve-intel -s".

Seems to work (needed the sr2work.txt file though).

It's hammering away at about 580kp/s compared to about 400kp/s I was getting using JJsieveSSE2. 45% speed boost - nice! (pity this is a library computer (am working off my thumb drive), and is vastly superior to my work computer and home computer :P)

[paleseptember]

Just comparing the outputs for sr2sieve and jjsieve -- and I noticed that running

sr2sieve-intel -s

only produces an equivalent for fact.txt, but not factexcel.txt and factrange.txt

Is this a -switch for these? Is it important? Just trying to do my bit


---
Edit: Reading the README for sr2sieve a little closer, looks like -D shows the duplicates in the cmd window, but doesn't record them to a txt file. In about 29G, I've picked up 16 duplicates, but only 1 unique factor :P Interesting

[vjs]

fact.txt is all that is important right now.

Thanks for helping out

[geoff]

Just comparing the outputs for sr2sieve and jjsieve -- and I noticed that running

sr2sieve-intel -s

only produces an equivalent for fact.txt, but not factexcel.txt and factrange.txt

Is this a -switch for these? Is it important? Just trying to do my bit


---
Edit: Reading the README for sr2sieve a little closer, looks like -D shows the duplicates in the cmd window, but doesn't record them to a txt file. In about 29G, I've picked up 16 duplicates, but only 1 unique factor :P Interesting

There is no way for sr2sieve to produce factrange.txt because the algorithm it uses doesn't ever find out-of-range factors.

It would be possible to produce an equivalent to factexcl.txt, but it would probably not be useful because sr2sieve wouldn't necessarily find the same duplicate factors as jjsieve.

[vjs]

Hey Guys!

I just started my Q6600...

With sr2 I'm getting 667 kps per core with the 18k dat!!!

Wow!!!! That is totally insane if real.

I'd also like to say that this is at the default 2.4G, I was going to increase it to 3.0G since they appear to o/c to at least 3.2 on air.

833 per core just sounds wrong...

O.K. quick update...

905-915 kps at 3.3 GHz....

[Joe O]

The most recent versions of SR2sieve are available at:

http://www.geocities.com/g_w_reynolds/sr2sieve/

Any references to the url http://www.geocities.com/g_w_reynolds/ should be replaced with the url http://sites.google.com/site/geoffre...olds/programs/.

[engracio]

The most recent versions of SR2sieve are available at:

http://www.geocities.com/g_w_reynolds/sr2sieve/
Any references to the url http://www.geocities.com/g_w_reynolds/ should be replaced with the url http://sites.google.com/site/geoffre...olds/programs/.

Good idea, updated my X2 4600+ from srsieve 1.5.14 to 1.5.18 and it jumped from 520kps to 535kps. Every little speed helps.

e

Edit: Geocities closed down

[Keroberts1]

trying ot work with the new sr2sieve what do i need for support files? I setup sr2work but it says i need sr2data. is that a new dat file i need to download?

[Joe O]

trying ot work with the new sr2sieve what do i need for support files? I setup sr2work but it says i need sr2data. is that a new dat file i need to download?

Please go up 10 posts and read post #10.

[SlicerAce]

Hello,

I factored the range:

1269200G to 1269210G

using sr2sieve and the combined SoB.dat file found at:

http://www.psp-project.de/test/sievecomb.zip

and found 0 facts. I am using sr2sieve, so it does not output a fact.txt, so I do not know what I am precisely supposed to submit to the PSP and SoB submit forms to let both projects know that no factors were found in that range. Thanks!

[Joe O]

Hello,

I factored the range:

1269200G to 1269210G

using sr2sieve and the combined SoB.dat file found at:

http://www.psp-project.de/test/sievecomb.zip

and found 0 facts. I am using sr2sieve, so it does not output a fact.txt, so I do not know what I am precisely supposed to submit to the PSP and SoB submit forms to let both projects know that no factors were found in that range. Thanks!

If you had found factors, you would have submitted your factor file to the two submit forms. Since you found no factors, just mark your range finished in the automated reservation system as you have done:

1269200, 1269210, dchris, , Pending, 2007-10-14

[glennpat]

I just reserved a range and got 12300000-12303000 G. The number is a digit bigger than what I have seen before. I assume it is correct.

[Matt]

Sounds about right to me, we've allocated all the ranges up to 1,500,000 now and there are large chunks after that which are reserved for PrimeGrid. Manual sieving is now taking the next available chunk from 12,300,000 to 12,400,000.

[hhh]

With all the due respect, but that .........makes no sense (to me, at least). It will take Primegrid a long time to catch that up, and in the meantime, manual sieving will be a lot less effective, where it should be equally or more effective, IMHO. Or is it on purpose to bring the people to PrimeGrid? I have got a not so stable Internet connection; I could work with PG, but it's too much work. So I sieve manually, which is work as well. OK. But not under these conditions, good Lord!
Why not unreserve a small range around 5P?

Puzzled, H.

[Joe O]

It's where I computed BOINC sieve will reach by the end of the year.

[Matt]

I understand what you're saying, I think the reason we have ended up this way is because the ranges are assigned to PrimeGrid quite some way in advance in order to ensure they never run out of work. This means that they have reserved all of the ranges up to around 10M.

[hhh]

Well, I figure there are some reasons, but it's not very motivating. It will be once PG has passed the range, which is a somehow stupid thought of mine, I am aware of that. Anyway, never mind: The world's end is sceduled for later, even with this change. H.

[glennpat]

For the 12300000 area sr2sieve is telling me to expect 4.31 factors in a 1000G range. For the 1498000 area it was 37.5 factors in a 1000G range. It does run the 12300000 area 20% faster, but I hope the 4.31 factors in 1000Gs is not correct. I've only run a few hours in the new range so don't have any real data.

[KWSN-SpongeBobSquarePants]

Ni!

Hello, new to this but first one went well new one is

12303100-12303600
12303600-12304100

I get the client shutdown and this in the log.txt file...

PRS52 v0.102b for processor type SSE2
Command line:
End of Command Line.
File Names Used
Inputs : sob.dat
Status : SoBStatus.dat
Factor : fact.txt
Ranges : nextrange.txt
Progress: progress.txt
Debug : debug.txt
Log : log.txt
Stats : stat.txt
Work : workarea.tmp
Found SoBStatus.dat with unfinished work.
Continuing from last save point.
Setting priority to Process 0 Thread 3
Running in Sierpinski mode
Using CMOV optimizations
Using SSE optimizations
Using SSE2 optimizations
Verbose level: 0
Constants and limits:
Performance Counter Frequency 14318180
Clocks per sec = 1000
Max # k = 75
Max p = 4503599627370495
Hash bits = 14
Hash mask = 16383
Hash factor limit = 268435456
Reporting Interval = 10000000
Dat file opened.
Grand Total for ALL k 1963642 n values.
Expected 16 k values.
Found 16 k values.
Compute the regular grange.
Done with dat.
Setup took 3.611909 seconds.
This program only supports p up to 4503599627370495
You have specified pmax=12303600000000000 Truncating
Contradiction: pmax=4503599627370495 <= pmin=12303100000000000


What am I doing wrong? My number seems to be too big.

Also I was not logged in when I submitted my first factors to SOB, who do I write to get credit?

Thanks

KWSN-SpongeBob SquarePants

[paleseptember]

It looks like you're inputting your range incorrectly. Input your pmax in billions (or G, if you prefer)

So take off the last nine zeroes of your range in your input file.

[glennpat]

I tried your range on jjsieve and got the same message. From the error message it sounds like it doesn't support the large p value. I believe jjsieve is what your using. I use sr2sieve. Where to get it is described at:

http://www.free-dc.org/forum/showthread.php?t=3501

Any references to the url http://www.geocities.com/g_w_reynolds/ should be replaced with the url http://sites.google.com/site/geoffre...olds/programs/.

Geocities closed down.

[hhh]

For the 12300000 area sr2sieve is telling me to expect 4.31 factors in a 1000G range. For the 1498000 area it was 37.5 factors in a 1000G range. It does run the 12300000 area 20% faster, but I hope the 4.31 factors in 1000Gs is not correct. I've only run a few hours in the new range so don't have any real data.

Well, 10 times deeper sieving, 10 times less factors. That's what I complained about.

Yet, it is still more effective than LLR, on the long run; If at the end of your range, you add the LLR times of the candidates eliminated, you will see that it would have taken longer than the sieving of your range.
H.

[glennpat]

For the 12300000 area sr2sieve is telling me to expect 4.31 factors in a 1000G range. For the 1498000 area it was 37.5 factors in a 1000G range. It does run the 12300000 area 20% faster, but I hope the 4.31 factors in 1000Gs is not correct. I've only run a few hours in the new range so don't have any real data.

I have now completed 1000Gs and found 5 factors. They were all PSP factors and I got 12000 points each for them. Since there is going to be lot less SOB factors it going to be hard to get one that will be worth much points. You do get more points for factors in this range, but I don't think it will make up for the fewer factors.

Anyway, it was a lot of fun finding the factors I have found in the past and watching the scores increase. I've got another 2000Gs on my range to complete and hopefully I will find a good SOB factor.

[glennpat]

I have now completed 1000Gs and found 5 factors. They were all PSP factors and I got 12000 points each for them. Since there is going to be lot less SOB factors it going to be hard to get one that will be worth much points. You do get more points for factors in this range, but I don't think it will make up for the fewer factors.

Anyway, it was a lot of fun finding the factors I have found in the past and watching the scores increase. I've got another 2000Gs on my range to complete and hopefully I will find a good SOB factor.

I completed my 3000G range and found 1 SoB factor worth points. I got 539 points for it and it was of the type that will not be upgraded in value.

[[DPC]Frentik]

Can I already submit factors while I am still sieving? I currently sieving two 500G ranges and both have found a sob factor but I still need to sieve ~60% per range.

Can I submit the factors or should I finish the ranges first?

[hhh]

Can I already submit factors while I am still sieving? I currently sieving two 500G ranges and both have found a sob factor but I still need to sieve ~60% per range.

Can I submit the factors or should I finish the ranges first?

You can submit at any time. Make sure, though, that no factor gets lost. At least for me, the more I do, the more I make mistakes. But that's not necessarily the case with other people.... H.

[engracio]

Can I already submit factors while I am still sieving? I currently sieving two 500G ranges and both have found a sob factor but I still need to sieve ~60% per range.

Can I submit the factors or should I finish the ranges first?

Definitely, the sooner you submit the factor the sooner that factor is taken out of circulation and not tested at all. Therefore saved cpu cycles. I normally create a sub folder named as the current range. Cut and paste to that folder all of your confirmed submissions. The confirmation will be at the submission web page. At the end of the completed range, count how many factors you found and post it here as completed. Reserve another range.

[ST47]

Where do I sign up to get an account on http://psp-project.de/sieveimport.php to submit factors?

[Joe O]

Where do I sign up to get an account on http://psp-project.de/sieveimport.php to submit factors?

By posting in this thread: http://www.mersenneforum.org/showthread.php?t=8162

or emailing psp"at"ldausch.de