P-1 factoring anyone?

[henryzz]

Originally Posted by cheesehead

Module gwnum.h in the Windows source has the definition

#define MAX_PRIME_SSE2 596000000L /* SSE2 bit limit */

595999993 is probably the largest prime < 596000000

LL, ECM and P-1 all use FFTs, and 596000000 is the maximum exponent the FFTs can now handle AFAIK.

Perhaps the higher exponents were processed with a Mac version of the software rather than a Windows version.

AFAIK the non-SSE2 code supports higher exponents than SSE2 code. The sacrifice was thought worthwhile at the time.

[Mr. P-1]

Originally Posted by xilman

To compensate, rate the top 1000 by the size of P-1 after division by all powers of the Mersenne exponent.

Why all powers? Why not one power?

Note that this proposal scales down composite P-1 factors also.

Composite factors should be resolved into their constituent primes.

[cheesehead]

Originally Posted by henryzz

AFAIK the non-SSE2 code supports higher exponents than SSE2 code.

I haven't found the definition of MAX_PRIME yet.

[xilman]

Originally Posted by Mr. P-1

Why all powers? Why not one power?



Composite factors should be resolved into their constituent primes.

Purely to give a way of dealing with composites. If they are factored, multiple powers are not the issue.

Paul

[axn]

Originally Posted by xilman

Purely to give a way of dealing with composites. If they are factored, multiple powers are not the issue.

Paul

Slighly confused. Are you suggesting that if C = P1*P2 be a composite factor of a mersenne M(p), then C-1 will be divisible by p^2?

Obviously (P1-1)*(P2-1) will be divisible by p^2, but, in general, C-1 won't be.

[xilman]

Originally Posted by axn

Slighly confused. Are you suggesting that if C = P1*P2 be a composite factor of a mersenne M(p), then C-1 will be divisible by p^2?

Obviously (P1-1)*(P2-1) will be divisible by p^2, but, in general, C-1 won't be.

You're right and I was confused (read: wrong).

The correct approach, IMO, is to report only prime factors and to scale them by dividing by p.

Division by p rewards hard work, because higher B1 & B2 gives a higher chance of finding a factor, and good luck.

Paul

[drh]

Originally Posted by petrw1

And if I put my entire (albeit meagre) team of P-1 capable PCs (3 Quads and 6 Duals and 1 P4=25 cores) I estimate I could complete only about 10 or 12 per day or 5 - 6%.

Mind you, all 4 cores of a Quad on P-1 could be quite taxing.

I now have 5 cores running P-1's and over the last 30 days I'm averaging 2.6 per day in the 53M range all in the 4.6 - 4.7 GHz-Day range.

[science_man_88]

Originally Posted by drh

I now have 5 cores running P-1's and over the last 30 days I'm averaging 2.6 per day in the 53M range all in the 4.6 - 4.7 GHz-Day range.

I remember being told by someone that p-1 would only take me 7 days but I stopped it ~1 day in because it had only done like 3% (100/3 =33.3 days not 7)

[petrw1]

Originally Posted by science_man_88

I remember being told by someone that p-1 would only take me 7 days but I stopped it ~1 day in because it had only done like 3% (100/3 =33.3 days not 7)

Depends completely on what CPU you are running on and how much RAM you allocate.

I have a i5-750 that will complete a P-1 in the 50M range in 31 hours and a PIV 2.4 Ghz that takes over 100 hours.

[petrw1]

Originally Posted by drh

I now have 5 cores running P-1's and over the last 30 days I'm averaging 2.6 per day in the 53M range all in the 4.6 - 4.7 GHz-Day range.

Is that 5 cores on 1 PC? Or 1 core on each of 5 PC's or a combination of such?

If you have more than 1 core on a PC doing P-1 do you notice any conflict for CPu resources; for example, do the run times drop noticeably if there are more than 1 cores doing stage 2?

[drh]

Originally Posted by petrw1

Is that 5 cores on 1 PC? Or 1 core on each of 5 PC's or a combination of such?

If you have more than 1 core on a PC doing P-1 do you notice any conflict for CPu resources; for example, do the run times drop noticeably if there are more than 1 cores doing stage 2?

Here is what I'm doing right now -

1 Intel Core i5 M 540 @ 2.53GHz 4Gig RAM -- (1 Core P-1 & 1 Core mafktc)
1 Intel Core i7 Q 720 @ 1.60GHz 8Gig RAM -- (2 Cores P-1 & 2 Cores LL)
1 Intel Core2 Quad Q8400 @ 2.66GHz 8Gig RAM -- (2 Cores P-1 & 2 Cores LL)

The memory on the systems running (2 P-1's) automatically adjusts when they are both in stage 2, but the throughput remains the same, as well as the throughput of the LL's. I'm contemplating trying (3 P-1's) at a time, and if I do, I'll let you know those results as well.