Prime95 30.8 (big P-1 changes, see post #551)

[storm5510]

Originally Posted by pepi37

Or better to say / kindly ask please write one line what I should put in worktodo.txt to get benefit from using Pfactor. You have all info you need to know. Thanks in advance

Below is a Pfactor sample:

Pfactor=1,2,13977461,-1,69,1

The blue is your exponent. The green is how far it has been trial factored. Prime95 may say that the exponent is composite or does not need to be tested with P-1. If this happens then you may have to go back to your Pminus1 form. If Prime95 will run the way you have it now, just let it run and do not worry about the memory usage. Good luck.

[pepi37]

Originally Posted by storm5510

Below is a Pfactor sample:

Pfactor=1,2,13977461,-1,69,1

The blue is your exponent. The green is how far it has been trial factored. Prime95 may say that the exponent is composite or does not need to be tested with P-1. If this happens then you may have to go back to your Pminus1 form. If Prime95 will run the way you have it now, just let it run and do not worry about the memory usage. Good luck.

Like I was write Prime95 tell me that

64*920^300082+1 does not need P-1 factoring.

So we are at the start... Does anybody know how to setup Prime95 for stuff I need?

[kruoli]

Originally Posted by pepi37

and how I can say that Prime95 use all memory for every test.

If Prime95 thinks it can be most efficient by not using all available RAM, there is most likely a reason for this. In you case, B1 is quite small so the possibilities for stage 2 with different parameters are limited. In general, it is not advisable to enforce higher memory usage when Prime95 correctly detects available memory. The only exception would be a bug in the functions that chose stage 2 parameters.

For really small work (small exponents, small B1), it can even be most efficient to run stage 2 with 1 % of available memory.

If memory usage with the same number and the same B1 increases, D will increase, which will in turn increase the time to build the polynomials, and there is always a point at which the time to build the polynomials will take longer than the whole stage 2 would take with lesser settings.

[kriesel]

It appears to me prime95 computes its completion time estimates for each worker independently, without consideration of constraints other workers may impose.

In prime95 v30.8b17, MaxHighMemWorkers=2, example attached, the newly added assignment (worker 3) to P-1 factor a 137M Mersenne projects completion in less than a day. However, worker 2 and worker 4 are both engaged in continuing stage 2 P-1 on larger exponents projected to require 1.5-1.7 days further to complete. So stage 2 on the 137M wouldn't begin for at least 1.5 days, and so can't complete in <~1.9 days, unless one of worker 2 or 4 are stopped early long enough to allow worker 3 to acquire sooner, large memory for its stage 2 on 137M, by user intervention.

Presumably this effect will also occur for P+1 and ECM stage 2.

[pepi37]

Originally Posted by kruoli

If Prime95 thinks it can be most efficient by not using all available RAM, there is most likely a reason for this. In you case, B1 is quite small so the possibilities for stage 2 with different parameters are limited. In general, it is not advisable to enforce higher memory usage when Prime95 correctly detects available memory. The only exception would be a bug in the functions that chose stage 2 parameters.

For really small work (small exponents, small B1), it can even be most efficient to run stage 2 with 1 % of available memory.

If memory usage with the same number and the same B1 increases, D will increase, which will in turn increase the time to build the polynomials, and there is always a point at which the time to build the polynomials will take longer than the whole stage 2 would take with lesser settings.

I finally solve mystery: for those candidate B1 was to small. Make upgrade in memory ( now it has 32GB) and rise B1 from 100K to 1M.
Now Prime95 use all memory ( given 29GB) every time. Hope will find some nice factors :)