Just starting.

[maniacken]

Hi i am new to factoring but have done some sieving.

I was wondering which client to install (SBFactor v1.2 or SBFactor v1.25.5 (for Windows) ).

this was from 1.25.5 config

LogToConsole = 1
type = P-1
test = range
form = Sierpinski
ram = 256
priority = 5
cpus = 1
cpu = 0
# using optimal bounds / only possible for P-1
bounds = optimal
fdepth = 47
fvalue = 1.5
#
# or using manual bounds
# bounds = manual
B1 = 100
B2 = 2000

I know i need to reserve a range so i reserved 8500000 8505000

also what is fdepth and fvalue mean, and what B1 and B2 should i use?

Thx for the help Kenny.

[maniacken]

so i think i downloaded all that i need.
in a folder i put
sbfactor.exe
sbfactor.cfg
sob.dat
results.txt

i think i changed the config file correctly, but when i try to run sbfactor.exe it pops up a dos box and then disappears right away. I get a file sbfactor.log and all the information i get is

SBFactor v1.25.5
P-1 and ECM factoring for number of the form k*2^n+1 and k*2^n-1.
Adapted from GIMPS v23.5 by George Woltman and Louis Helm
Configuration file support by Henrik Christian Grove
Unknown CPU Type detected.


do i need to create a work file with range and b1 b2 numbers. If so what format do i do this.

[Keroberts1]

once again this is the problem wiht the factoring clients. They are not easy to run and there are no instructions on how to figure it out. Personally i would like to use my P4 for factoring but don't want to deal wiht all of the extra work. So I'll just leave it on prime hunting and let my AMDs sieve.

[maniacken]

well i got it working . had to remember how to do the dos commands to get into the folder i created.

i ended up using

sbfactor.exe 8500000 8505000 sbfactor.cfg

in the config file i used

bounds = optimal
fdepth = 47
fvalue = 1.5

do i need to change these? if so i can restart the range.

[vjs]

well I think you would have to restart but the optimal bounds now are

bounds = optimal
fdepth = 49
fvalue = 1.5

I'd actually set the bounds manually and use b1 and b2

Generally I try to use a larger B1 than suggested and keep b2 the same or higher if you have alot of memory.

[maniacken]

it suggested b1 = 30000 and b2 = 255000 should i go higher?
i am allocating 256Mb.

so with your suggestion should i go 50000 and 300000?

[Joe O]

Originally posted by maniacken
it suggested b1 = 30000 and b2 = 255000 should i go higher?
i am allocating 256Mb.

so with your suggestion should i go 50000 and 300000?

Yes, or even 60000 and 300000. Note, Only increased B1.
What are the elapsed times at the different settings?

[maniacken]

at 30000 and 255000 stage 1 takes about 75 minutes, and after stage 2 it says total factoring time is 129 minutes.

[Joe O]

Various places state that B2 should be chosen relative to B1 so that Stage2 takes from 50% to 100% of the time Stage1 takes. So your B1 B2 relationship is close to optimal.
Given your timings, if it were my machine, I would be conservative in making changes. I would leave B2 at 255000 or slightly increase it to 260000 if the total running time remains under 150 minutes. Then, keeping B2 constant, I might try to increase B1 by small amounts, keeping the Stage1 time under 90 minutes and the total time under 180 minutes. The idea is to strike a balance between running time and factoring depth. If there were more factorers, then perhaps a longer running time would be acceptable, but there are so few factorers and so many k n pairs to test.

[maniacken]

Thanks for the help Joe.

[hhh]

Hello, I would like to start, too, but need a little help. I started trying with sbfactor.exe, but other people always close the window, even when I rename it to dontcloseplease.exe.

On hcgrove's excellent page on factoring (
http://www.sslug.dk/~grove/sbfactor/ ), the links to George Woltman's versions (sobpm1.zip) are broken, but with google, one can find a file with this name somewhere else, which contains prime95.exe, 2656KB large, changed 2004/09/17.

Is this the right one? Or should I get another version of prime95.exe?

Unfortunately, it comes without readme or help.
This program, as it becomes an icon when minimize, would of course have more chances of surviving.
But as I understood, it gets its instructions from a worktodo.ini, which I have to create. On hcgrove's page, there is a PERL program to create it, but I never did PERL, so neither I can use it nor I understand it. There seem to be no PERL compiler neither.

Is it possible to create the worktodo.ini manually? What to put in? A range? The numbers to factor?

That's it, I think, for the moment. If I have more questions, I will post later, don't worry.

By the way, you complain to be few factorers. One reason is, sure, that it's not easy. Another one is that it's almost impossible to get started for a beginner with the information in the forum.
I promise that　when everything is working, I write a manual (which you will have to correct, of course), unless you tell me not to do so.

Thank you in advance, H.

[prime95]

The latest executable is at ftp://mersenne.org/gimps/p95v246.zip

Others can help you create the worktodo.ini file.

[Joe O]

The format of the worktodo.ini entry is

Pfactor=k,2,n,1,47,1.5

where k is one of the 10 k left and n is one of the remaining values for that k in the range you reserved.

Now how to get the k n pairs.
You could use one of the sbfactor programs, inputing the range and capturing the screen output and manually changing the format.
You could just use the Perl program. By the way Perl is interpreted, not compiled. Please post your operating system of choice, and I will tell you where/how to get the interpreter for it.
HCGrove, or someone, could write a very quick program in C or C++. By the way the Perl syntax is very close to these languages. I don't have a C or C++ compiler or I would do it. (My computer crashed, taking with it my Visual Studio 6 Enterprise Edition. I still have the 3 CDs, but have misplaced the key.)

Oh, yes the 47 is the sieve depth, and 1.5 is the number of tests that will be saved.
If you don't like the B1 B2 values chosen by the program, you can play with them.
Decreasing the 47 to 43 and increasing the 1.5 to 3 should give you B1=35000 and B2=455000 if you have 400MB assigned. B1=35000 and B2=300000 would be good if you have 256MB to spare for the program. Remember, If you assign too much memory to the program, there will not be enough for the operating system and any other programs you want to run. Leave them at least 128MB if not more, unless this is a dedicated machine, then you can leave only 64MB. And don't forget about the video memory. If it is shared you have to allow for that as well.

[hhh]

First, thank you for the help.
I think, there is no program for the creation of worktodo.ini needed, the replace function does it very well. (For the beginning at least).

I am using three Celeron 2 GHz, assignating 100 MB at daytime and 150 MB during the night (it's not much, but factoring is the only thing the machines can be used for).

I played a little bit with the numbers, first adjusting the tine estimation to 3 hours approximately, but I got huge B1 and B2 (B2=1000000)
So I pushed down the last value (Probability of finding a factor?) until
Pfactor=21181,2,8990012,1,49,1.6
getting
B1=45000, B2=427500, est. prob. 0.97%
When I put 1.5 instead of 1.5, it says that no factoring is needed and abords. Normal?

As I read somewhere that the optimal are B1=30000 and B2=260000 (ca.), I should get down the factoring depths perhaps.
Like this, Stage 1 will take 100 minutes. (I'am not yet at stage 2)

What do you think?
Sincerely, H.

[Nuri]

I think 49 and 1.6 are both reasonable.

If I recall correctly, one can use the second digit as well. If you want to trim a bit more, try using figures like 1.58, 1.55, ect.

[garo]

Also, someone posted a script to convert exponents from the sbfactor to prime95 format. It should be in the P-1 client discussion thread. 49, 1.6 seem to be reasonable numbers. Go for it.

[hhh]

Yes I do. It is a pleasure to view the computers working.
One last question (I hope so at least)
Do I need to report the factors as soon as theiy appear or is it enough to report them before PRP reaches the level?
Thank you very much, again, Yours H.

[Mystwalker]

The latter is suficient - but sending the factors in as soon as possible is a good way not to forget them.

[hhh]

By the way, what happens if I find a factor with prime95? I mean, what is the output?
Fact.exe as with sbfactor?
It seems to me I found nothing so far, but I wouldn't like to miss it by my own stupidity.
H.

[Nuri]

sample output at fact.txt file*

4847*2^8995767+1 completed P-1, B1=30000, B2=255000, Wa1: E8C102B0
[Thu Dec 09 16:33:53 2004]
P-1 found a factor in stage #2, B1=30000, B2=255000.
4847*2^8995863+1 has a factor: 240667294106777359
[Thu Dec 09 17:53:26 2004]
4847*2^8996007+1 completed P-1, B1=30000, B2=255000, Wa1: E8C802B4
[Thu Dec 09 19:13:07 2004]

* check for your prime.ini file. it will appear at fact.txt if prime.ini has results.txt=fact.txt in it. Otherwise, I guess it would be in results.txt.

[hhh]

You are right, it goes to results.txt.
I am lucky to have asked! So I have anly to redo one range (as I replaced the results.txt (the large one)).
Well, 2 days for nothing is not so hard.

H.

[Joe O]

That's a good point. We should probably encourage everyone factoring using Prime95 to put

results.txt=fact.txt

in their prime.ini file. It is too easy to download results.txt and overlay the file.

[hhh]

Yuhee, I found something:
P-1 found a factor in stage #1, B1=40000.
4847*2^8387871+1 has a factor: 68788181892978001

But I don't know how to put it into the form.
It is http://www.seventeenorbust.com/sieve ,isn't it?

I am sorry to bother you with all these questions, but on the other side, I think it's not that much work to write a line on the submission-page how to submit. It is mot on MikeH's all sieve site (or I am blind).

(Actually, I tried several quite stupid, I admit, ways to put it in, and got

Factor table setup returned 1
Test table setup returned 1

0 of 1 verified in 0.01 secs.
0 of the results were new results and saved to the database.

So, I think it it is my fault as the input was strange, but as I don't know how to write it... )
Please help me again.

[Mystwalker]

It should be

68788181892978001 | 4847*2^8387871+1

[vjs]

That's pretty smooth if I'm not mistaken

68788 181892 978000 = 2 ^ 4 x 3 ^ 2 x 5 ^ 3 x 23 x 1559 x 5059 x 21067

All you have to do is enter that line

68788181892978001 | 4847*2^8387871+1

into the box on this page

http://www.seventeenorbust.com/sieve

make sure your logged in.

Congrats

[MikeH]

I know there's a great bit of pearl script around (can't find a link right now), but for anyone that doesn't want to have to install pearl, here is a regular windows .exe that does basically the same thing.

http://www.aooq73.dsl.pipex.com/utils/makewtd.zip

I've included the source, so anyone that wants to improve it (not difficult), or use it on another platform is free to do so.

[hhh]

Ironbits, your answer:

You will get a lot of these:

[Thu Dec 09 16:33:53 2004]
4847*2^8995767+1 completed P-1, B1=30000, B2=255000, Wa1: E8C102B0
[Thu Dec 09 17:53:26 2004]
4847*2^8996007+1 completed P-1, B1=30000, B2=255000, Wa1: E8C802B4

These are residues of the calculation and show you only that no factor has been found.

You are looking for those:
[Thu Dec 09 19:13:07 2004]
P-1 found a factor in stage #2, B1=30000, B2=255000.
4847*2^8995863+1 has a factor: 240667294106777359

This is what to submit:
240667294106777359 | 4847*2^8995863+1

Hope to have been of help... , Yours H.

[IronBits]

Thanks for that!
That was when I was using prime95

I'm using sbfactor dated 1-26-2004
sbfactor 9475000 9480000 49 2.5 512

I have 8 files with no extension and no logfile in the directory
examples
48479475071
48479475143
48479475167
1mg to 2mg in size, all binary garbage when I look at them in notepad

I have a fact.txt with one entry
49560960642980711 | 24737*2^9475327+1
Would that be a factor I would submit?

I found a newer sbfactor.exe dated 4-27-2004 which uses a .cfg file
Here's what I have in it
LogToConsole = 1
type = P-1
test = range
form = Sierpinski
ram = 512
priority = 3
cpus = 1
cpu = 0
bounds = optimal
fdepth = 49
fvalue = 2.5
B1 = 9475000
B2 = 9480000
curves = 2

I think I should be using the newer one because I am using an FX55, but it wouldn't start up for some reason.
Maybe results.txt was missing, I don't know, but, if I should be using the newest one, what do I have to do to upgrade it, and should I use the same command line as above to continue where the old one left off so I don't lose any work?
Getting closer! Thanks!

I'll be upgrading this box soon to an AMD Athlon 64 X2 4400+ 2.2GHz w/ 1MB L2 cache-per-core, and Windows Pro XP64, so I hope I get this all figured out and stable in the next month

[Joe O]

Yes, fact.txt is the file to submit.

As far as the newer factor program, yes I find it a little faster.
And it does need the same files as the older program, in addition to its cfg file.

[Joe O]

Ironbits,
I've been thinking about it and if you are running windows you probably should be running Prime95 V24.11 I've found it to be faster on an Athlon 64 than any of the SBfactor programs.
The relevant worktodo.ini entry is
Pfactor=k,2,n,1,43,3
e.g.
Pfactor=4847,2,2005551,1,43,3

There are Perl scripts and MikeH'sprogram available to generate the entries.

Don't forget to put
results.txt=something.txt
in prime.ini
or Prime95 will append to your results.txt file

If you are running *nix, try mprime.

There are many opinions on the 43,3 which is sievebitdepth,PRPsaved

Adjust them to get the B1 B2 you want.

[hhh]

Originally posted by Joe O

There are many opinions on the 43,3 which is sievebitdepth,PRPsaved

Adjust them to get the B1 B2 you want.

I thought like this until I got 'snubbed' by prime95.
In fact, the values represent the depth until which has been already sieved, and the number of PRPtests which will be saved by the factor.
On the base of these (and the memory allocated, and ???), prime95 calculates the optimal bounds for factoring. To tell the truth, I don't know exactly in which way optimal, but I guess for througput.

Sieving is somewhere between 2^49 and 2^50, so, 49 is the appropriate choice. A factor will save 2 PRPtests, so it's 2.0.
Putting the right numbers will be the most efficient on the long run. If you have other priorities, like the ambition to get lage factors, you can increase the bounds. But as for the project, Pfactor=k,2,n,1,49,2.0 should be the right choice.

I factored with 1.6 instead of 2.0, a while ago, too, in order to have more tests done. So... you see.

H.

[prime95]

Originally posted by hhh
I factored with 1.6 instead of 2.0, a while ago, too, in order to have more tests done. So... you see. [/B]

1.6 is a fairly intelligent choice. The argument goes that if a prime is found for that k, then a second PRP test will not be saved (i.e. you should have used 1.0).
So 1.6 really says a factor will save 1 PRP test plus a 60% chance it will save two PRP tests.

A lot of P-1 users choose values lower than 49 and above 2.0 to force prime95 to run P-1. The truth is that with the great job you guys have done seiving, P-1 is just on the border between no throughput gain and barely worthwhile.

[Nuri]

Exactly.

In fact, the only reason I'm currently running P-1 on my P4 is that I get cool points for the factors I find.

On second note, I personally believe 49.4,1.6 is the optimum selection for the moment. This is really an on the limit choice though, based on memory allocated, one might see the client skip tests altogether.

In practice though, I generally run at 49.4,1.8 or (49.4,1.7), whicever keeps B1 just above 40,000 and B2 just above 400,000.

[Nuri]

Just for info for those interested.

98% sieve point has just passed 2^49.5