P-1 factoring anyone?

[Jwb52z]

Thank you. I'll try to remember from now on so I won't do that again.

[bcp19]

Found a nice 102 bit factor: http://mersenne-aries.sili.net/expon...etails=8491921, was just missed by the old P-1

[kladner]

Originally Posted by Jwb52z

Thank you. I'll try to remember from now on so I won't do that again.

Speaking only for myself, such reports are interesting wherever they turn up.

[bcp19]

Found another B-S factor http://mersenne-aries.sili.net/expon...etails=8499577

[MrHappy]

When searching for big P-1 factors do I necessarily need to use high B1/B2 bounds? I can get big factors when k is a product of a bunch of small primes, right?
Does the probability of finding a big factor increase by using higher bounds?

[James Heinrich]

Originally Posted by MrHappy

Does the probability of finding a big factor increase by using higher bounds?

The probability of finding a factor of any kind increases with higher bounds. You are correct that it is possible to find large factors with small bounds (M31001941 and M11865241 are some more extreme examples of this). Your probability of finding any factor should be somewhere around 5% for a P-1 test, but whether you find a 130-bit monster or a 73.01-bit one that just missed TF is out of your control.

I would advise against choosing artificially small bounds (or manually choosing bounds at all unless you're very certain what you're doing) in hopes of finding large factors with small bounds. If such a small factor exists, P-1 will likely find it in stage1 anyways, saving you half a P-1 test.

[flashjh]

Is it true that any factor found in stage 1 would always be found in stage 2 also?

[Dubslow]

Originally Posted by flashjh

Is it true that any factor found in stage 1 would always be found in stage 2 also?

Yes, though the only way this would happen is if you skip the Stage 1 GCD.

[flashjh]

Originally Posted by Dubslow

Yes, though the only way this would happen is if you skip the Stage 1 GCD.

I'm just looking at how much time all my systems spend in S1 vs S2. I know S2 is slower, but I think the time wasted in S1 when no factor is found would generally be made up by just running S2 and finding the factors anyway.

I know it has been discussed before, but what would be required to get future versions of P95 to just do S2 on all P-1, as an option? In a project like GPU72 (and many others, I'm sure), P-1 is always* done with enough memory, so skipping S1 would save me time. Just a thought...

* Not true, but it sounds great

[Dubslow]

Originally Posted by flashjh

I'm just looking at how much time all my systems spend in S1 vs S2. I know S2 is slower, but I think the time wasted in S1 when no factor is found would generally be made up by just running S2 and finding the factors anyway.

I know it has been discussed before, but what would be required to get future versions of P95 to just do S2 on all P-1, as an option? In a project like GPU72 (and many others, I'm sure), P-1 is always* done with enough memory, so skipping S1 would save me time. Just a thought...

* Not true, but it sounds great

Oh, I thought you meant doing the S1 work without checking for a factor. I don't think you can do S2 without S1, which is why changing S1 bounds means you toss any S2 work you had. (Edit: More proof is that you need the full S1 save file to run S2.)

[flashjh]

Originally Posted by Dubslow

Oh, I thought you meant doing the S1 work without checking for a factor. I don't think you can do S2 without S1, which is why changing S1 bounds means you toss any S2 work you had. (Edit: More proof is that you need the full S1 save file to run S2.)

Right, but is that a necessary thing or a legacy of systems running P-1 with 8Mb years ago? So S1 generates necessary data for S2 that can not be computed before hand? Based on what I have read so far, I could only find what you said, that without the save file, P95 won't run S2.