Factoring secondpass ATT:MikeH

Printable View

10-04-2005, 04:25 PM
vjs

Factoring secondpass ATT:MikeH

Well Mike,

Looks like you beat me to it...

4847*2^2801151+1 has a factor: 7438415959930633

Just to let people know I think Mike Has started factoring from...

2900000 to probably 2920000 by now....

Not sure how far he is going to go....

___________________

It would be nice to know how far secondpass is going to be run but just so that people know n-values 3M<n<4M have not been P-1'ed as far as I know.

For reference...

21181*2^2801252+1 completed P-1, B1=20000, B2=280000, Wc1: 4878FF80
[Mon Oct 03 22:31:01 2005]

Took about 15 minutes on my 2.4Ghz Barton with 2G of memory.

Should we start a co-ordination thread for these secondpass - n's???

Also these factors will score around 20K points...
10-04-2005, 11:33 PM
Hades_au

If you think it's worthwhile, I'm happy to allocate a couple of boxes to this.

I'll take 3050000 3070000 and I'll see how it goes :)
10-05-2005, 01:38 AM
Nuri

It's a good idea to cover 2900000-4000000 range...

If I was on to the 999 thing, I'd have definitely participated.

We're talking about roughly 26500 tests to be done, right?

It'd take roughly a month for a computing power of 8-10 PCs.

I think, as this is a limited task and there will be a limited number of participants, I guess this thread itself would be sufficient to keep track. Just an idea.
10-05-2005, 06:40 AM
Greenbank

This is why I thought about a p-1 factoring website with assigned ranges/k,n pairs/etc. With records for B1/B2 used (and Sigma for ECM/P+1).

I'm on a training course next week in the US so I've got plenty of time to myself in the evenings. I'll see what I can knock up.

At first I wouldn't be able to host a copy of the residue file although once I sort out my DSL at home I might be able to host something for this. Even without this there would be a record of who worked on a specific range so someone could PM them here to ask for the residue files if they still have them.

I'd also include a way for people to upload log files, completed ini files, etc if people have them lying around so we can get an idea of what has been done.

Obviously we won't have all of the information but it's a start.
10-05-2005, 10:39 AM
vjs

Well I'm no expert when it comes to P-1...

So IMHO, I think people should only do them if they feel like it. It's probably a good opportunity for a first timer to grab a small range and see how thinks work. I personally believe we are better off spending our time pushing ahead the firstpass prp with P-1 since each factor found there will eliminate two tests.

As for bounds I wouldn't use very large bounds no more than B1=30K B2=300k for n<4M, I was personally considering B1=20k, B2=160k. Tests should take about 25 minutes on an average machine.

Just remember that these tests are only worth about 1.1 tests at most probably more like 1.05. So for those prime95 default people sieve=49.8 test value=1.1 would be a max honest input.

Oh and just to clairify...

Greenbank, we have never before P-1'ed for n's 1M<n<4M, check the archieve thread... perhaps I don't get the point.
10-06-2005, 07:15 PM
Hades_au

I'll also take 3030000 3050000.

I've finished 3050000 3060000 with B1/B2 at 30000/300000 with no factors.

It only took 2 days for that on a 2.4 P4, so I'll let the rest run over the weekend and have another look on Monday.
10-07-2005, 11:19 AM
MikeH

Quote:

Just to let people know I think Mike Has started factoring from...

2900000 to probably 2920000 by now....

Not sure how far he is going to go....

Actually I was just doing this to check the suitability of using a couple of PCs to do P-1 stage 1, with stage 2 being done elsewhere.

Experiment complete (and successfulish :) ). I'm now back on to big n values.

I don't have the right PC with we right now, but I think I did 2900000 - 2918000, with B1/B2 of 30000/270000 for some and 95000/1235000 for others.

That was before the enforced second pass strategy. Now things have changed, there is probably good mileage to be had out of P-1 testing anything up to 4M.

BTW, sieve/factor scoring updates have been a big sporadic because the next.txt file is empty! Thus I've been manually bodging things to make it work. Since it looks like we're staying this way for more than a few days, I've now put a better bodge in place.

..and sorry that the reservations are very out of date, and that my main web page is also out of date. At some point I will get some time, just not right now :cage:
10-07-2005, 11:49 AM
engracio

Sorry for being such a noob but reading thru the factoring threads, different people are getting different completion times on their factoring. I've asked around already and it seems like my completions times seems to be okay. Can some one explain the difference. I have a dual Athlon running 2000 mhz or XP2400, 512 mb memory, one cpu is sieving the other is factoring. Sieving speed is good. On factoring with 10000 range B1=65000, B2=747500, 425mb memory allocated I'm getting about 6000sec to complete. Am I doing something wrong, have I set up P95 incorrectly? Thanks.

e :)
10-09-2005, 09:15 PM
Hades_au

From doing P1 on the secondpass ranges I've grabbed, 5 factors in 582 tests over the weekend, so it's probably worth doing. :)

Quote:

Originally posted by engracio
Am I doing something wrong, have I set up P95 incorrectly? Thanks.

e :)

Is that 6000sec to complete both Stage 1 and Stage 2?

If it's for both, that sounds about right. My p4 2.8 completes a P1 test in about 5300sec using B1=65000 and B2=780000 with 920mb ram available. I know my Dual Xeon is slower as the cpu's still share memory, memory buses etc. Dual cpu's take a big hit using Prime95 as it heavily uses memory.
10-09-2005, 10:37 PM
engracio

Hades_au,

Thanks, after waiting a few days its looking like factoring and sieving are almost similar in speed. So as suggested I am doing factoring on 1 cpu and sieving on the other. My Xeons are straight up prp'ng. Should cover all of SOB.

BTW, do you run 4 instances of prp on your dual xeon boxes or run -o2 which is 3 instances on my xeon box HT enabled? Thanks.

e:)
10-10-2005, 07:10 PM
Hades_au

I'm abandoning 3040000 3050000, secondpass is running alot quicker than I thought.

I'll reserve 3500000 3550000 though.

Quote:

Originally posted by engracio
BTW, do you run 4 instances of prp on your dual xeon boxes or run -o2 which is 3 instances on my xeon box HT enabled? Thanks.

e:)

I actually tried running 2, 3 and 4 instances, just to see which had the highest total output and the result, for me, was that 2 instances (1 per cpu) had the highest. Most others with dual xeons say 3 or 4 is best, it probably depends on the motherboard and memory buses.
10-10-2005, 09:35 PM
engracio

Quote:

Originally posted by Hades_au

Most others with dual xeons say 3 or 4 is best, it probably depends on the motherboard and memory buses.

Thanks, tried 4 instances but kind of slow. 3 seems to be the best compromise for my mobo.

e:)
10-10-2005, 11:41 PM
vjs

Totall agreed with Hades_Au,

Probably not worth doing unless your interested in trying some quick factoring.

bounds of

B1=20K B2=180K would probably be a max if I were to try, we have just sieved fairly deep for these factors.
10-18-2005, 11:49 AM
garo

I am doing a quick pass over 3.7-3.75M. B1=8k, B2=80k. I am not testing k=4847 due to the prime rumours.
10-20-2005, 10:21 AM
garo

Looks like my gamble not to test 4847 paid off! However, second pass is moving really fast so I have ahd to abandon the range at 3720607. I'm now taking up 3.9-4M.
10-20-2005, 10:38 AM
vjs

Garo 4M is the lower limit of P-1, as I'm sure your aware. There are also a few smaller holes in the reservations starting around 4.2M.

Check the archieve thread if your interested.
10-20-2005, 10:46 AM
Greenbank

My 1.7GHz P4 will have:-

4349000 4360000 Greenbank

which is part of a previously passed by range or 232 numbers.
10-20-2005, 11:00 AM
garo

I am not quite sure what you mean by limit. If you mean the point at which P-1 started for first pass, I am well aware of that! I was one of the first to submit a P-1 factor for n=4.01M!! I have checked the archive already!

If you mean the point at which P-1 starts becoming useful, I suspect that the point lies higher than n=4M, at least for second pass as now most factors are worth only 1 test. That's why I am using relatively low bounds of 10k,80k.
10-20-2005, 11:46 AM
vjs

Garo,

Yes the post was more of an obvious one to everyone,

Quote:

the point at which P-1 started for first pass

Yes, sorry I didn't get much sleep last night.
10-20-2005, 11:50 AM
vjs

4203000 4210000 ? [passed by]
4211000 4250000 ? [passed by]
4255000 4297000 ? [passed by]
4301000 4348000 ? [passed by]
4360000 4530000 ? [passed by]
4548000 4570000 ? [passed by]
4578000 4620000 ? [passed by]
4623000 4645000 ? [passed by]
4651000 4700000 ? [passed by]
4702000 4749500 ? [passed by]
4768000 4775000 ? [passed by]
4782500 4820000 ? [passed by]
4834000 4850000 ? [passed by]
4852000 4875000 ? [passed by]
4882000 4921000 ? [passed by]
4935000 4980000 ? [passed by]
4983180 4985000 ? [passed by]
4985340 4990000 ? [passed by]
5000300 5006782 ? [passed by]
5008000 5100000 ? [passed by]
5110000 5118800 ? [passed by]
5126400 5150000 ? [passed by]
5150250 5200000 ? [passed by]
5206701 5220000 ? [passed by]
5222720 5240000 ? [passed by]
5246000 5300000 ? [passed by]
5294000 5300000 ? [passed by]
5338500 5340000 ? [passed by]
5347000 5480000 ? [passed by]
5524000 5550000 ? [passed by]
5553000 5595000 ? [passed by]
5763400 5765000 ? [passed by]
5810000 5840000 ? [passed by]
5900000 5945000 ? [passed by]
10-20-2005, 05:11 PM
garo

Thanks vjs, that's very useful. Let's see how far the project managers intend to run second pass though...
10-20-2005, 08:16 PM
vjs

Yes I wouldn't start P-1'ing out past 4.5M just yet... even if we do test out to 5M the secondpass n should start to slow down with larger tests fairly soon.
10-31-2005, 03:53 AM
Greenbank

4349000 4360000 Greenbank [complete] 1 factor
10-31-2005, 11:17 AM
vjs

I believe Garo commented in another thread that it is nolonger benifital to P-1 these numbers even with low bounds.

If I'm wrong please correct me. Considering one factor is pretty much equal to one test for these ranges probably better off just testing. Also considering that the main effort is currently running these secondpass tests we are probably better off running sieve at the moment.

IMHO, I'd like us to concentrate on sieve for a while, or P4's on P-1 factoring >10M.
10-31-2005, 11:42 AM
Greenbank

Yup, I just moved this box over to PRP. I just wanted to finish the range in case I was just lucky enough to get a few more factors.