Question about Prime95 24.12 Problem???

VJS · 2005-06-23, 17:03

I would like to try a very large P-1 factoring attempt. However I've been having difficulties determining the largest B1 possible.

I've tried several different worktodo.ini

Pminus1=24737,2,991,1,29000000000,1

However the max B1 attainable is roughtly B1=1.2G anything larger gives me an unpredictable B1 value generally much less than the input.

Sorry if this has been addressed in the past.

ewmayer · 2005-06-23, 17:55

Any time you start looking at p-1 bounds that large, you should first think about whether you'd be better off running ECM with appropriate bounds for whatever minimum factor size range seems likely.

VJS · 2005-06-23, 18:41

Thanks, we are currently looking at the 50 digit level for this number. Testing the required number of curves for 50 digits will take quite some time.

It only seems reasonable to exhaust or at least spend at least a week or two doing a P-1 with the newer/faster/better Prime95.

Looking at the record's for P-1 and P+1, a B1 bound of 10^10 don't seem too unreasonable. It can be done with ecm6.0 but prime95 is quite a bit faster.

I just think it would be a shame to miss a large P-1 or P+1 factor through lack of trying while investing effort into ECM. You only have to do one P-1 at those bounds anyways then bounds less than those are also completed.

Note sure if the same is true for P+1.

Prime95 · 2005-06-23, 19:41

Maximum B1 value is 4 billion.

VJS · 2005-06-23, 20:00

Thank you very much, and especially thank you for the speed up with 24.12.

ewmayer · 2005-06-23, 22:06

Quote:

Originally Posted by Prime95

Maximum B1 value is 4 billion.

4 billion, or 2^32-1?

VJS · 2005-06-24, 16:38

I'll let you know when it gets there.

2^32-1 - 4G =

If so Looks like I get an extra 294967294

Is it possible to convert/use the P-1 stage 1 result in ecm6.0, like stage1 ecm? Is there a sticky or some brief instructions somewhere.

Same goes for P+1 any "experience with posts" for the noob?

akruppa · 2005-06-24, 17:57

Personally, I don't think P+1 is really all that useful for numbers in general. There's a 50% chance you'll merely do P-1 again and the smoothness of the group order is not as good as for ECM. If you want to factor Lucas or Fibonacci numbers, P+1 with -x0 23/11 is great and should be done instead of P-1.

Alex

VJS · 2005-06-24, 18:23

Thanks for the comments,

I was just curious about other factoring techniques for

24737*2^991+1

10% of the ECM curves have been done for the 50-digit level already, and the concenses is to continue with ecm if we want to find a factor.

BTW P-1 B1=2^32-1 complete starting stage2, no joy thus far.

2005-06-23, 17:03	#1
VJS Dec 2004 100101011₂ Posts	Question about Prime95 24.12 Problem??? I would like to try a very large P-1 factoring attempt. However I've been having difficulties determining the largest B1 possible. I've tried several different worktodo.ini Pminus1=24737,2,991,1,29000000000,1 However the max B1 attainable is roughtly B1=1.2G anything larger gives me an unpredictable B1 value generally much less than the input. Sorry if this has been addressed in the past.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
prime95 problem	Kurtbusch	PrimeNet	1	2014-09-09 03:22
Problem with prime95	bertie	Software	1	2012-12-16 17:56
Prime95 problem	boldenc	Software	2	2009-08-09 14:33
problem with Prime95	ixfd64	Software	1	2005-09-24 08:59
Prime95 Client problem	GSV3MiaC	Software	21	2004-01-14 14:02

2005-06-23, 17:55	#2
ewmayer ∂²ω=0 Sep 2002 República de California 5×2,351 Posts	Any time you start looking at p-1 bounds that large, you should first think about whether you'd be better off running ECM with appropriate bounds for whatever minimum factor size range seems likely.

2005-06-23, 18:41	#3
VJS Dec 2004 299₁₀ Posts	Thanks, we are currently looking at the 50 digit level for this number. Testing the required number of curves for 50 digits will take quite some time. It only seems reasonable to exhaust or at least spend at least a week or two doing a P-1 with the newer/faster/better Prime95. Looking at the record's for P-1 and P+1, a B1 bound of 10^10 don't seem too unreasonable. It can be done with ecm6.0 but prime95 is quite a bit faster. I just think it would be a shame to miss a large P-1 or P+1 factor through lack of trying while investing effort into ECM. You only have to do one P-1 at those bounds anyways then bounds less than those are also completed. Note sure if the same is true for P+1.

2005-06-23, 19:41	#4
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 41·199 Posts	Maximum B1 value is 4 billion.

2005-06-23, 20:00	#5
VJS Dec 2004 13×23 Posts	Thank you very much, and especially thank you for the speed up with 24.12.

2005-06-24, 16:38	#7
VJS Dec 2004 13·23 Posts	I'll let you know when it gets there. 2^32-1 - 4G = If so Looks like I get an extra 294967294 Is it possible to convert/use the P-1 stage 1 result in ecm6.0, like stage1 ecm? Is there a sticky or some brief instructions somewhere. Same goes for P+1 any "experience with posts" for the noob?

2005-06-24, 17:57	#8
akruppa "Nancy" Aug 2002 Alexandria 2,467 Posts	Personally, I don't think P+1 is really all that useful for numbers in general. There's a 50% chance you'll merely do P-1 again and the smoothness of the group order is not as good as for ECM. If you want to factor Lucas or Fibonacci numbers, P+1 with -x0 23/11 is great and should be done instead of P-1. Alex

2005-06-24, 18:23	#9
VJS Dec 2004 13×23 Posts	Thanks for the comments, I was just curious about other factoring techniques for 24737*2^991+1 10% of the ECM curves have been done for the 50-digit level already, and the concenses is to continue with ecm if we want to find a factor. BTW P-1 B1=2^32-1 complete starting stage2, no joy thus far.