missed factor?

[tha]

I have the following question. If I look at this page requested from the server it appears to me as if trial factoring missed the exponent that was later found by P-1 factoring.
I was under the believe that that should not be possible. Any comments?

[kracker]

Originally Posted by tha

I have the following question. If I look at this page requested from the server it appears to me as if trial factoring missed the exponent that was later found by P-1 factoring.
I was under the believe that that should not be possible. Any comments?

P-1 factoring is not the same as trial factoring. Don't worry

[tha]

Originally Posted by kracker

P-1 factoring is not the same as trial factoring. Don't worry

As I understand it, trial factoring tries each possible factor up to a limit (2^amount of bits). P-1 factoring tries a small amount of easy to compute factors up to a higher range. Therefore P-1 should not be able to find a factor that is below the limit of the trialfactoring done. This factor is about 65 bits while trialfactoring went up to 70 bits already. Or am I missing something?

[kladner]

[deleted by user for excessive ignorance]

[Bdot]

Originally Posted by tha

I have the following question. If I look at this page requested from the server it appears to me as if trial factoring missed the exponent that was later found by P-1 factoring.
I was under the believe that that should not be possible. Any comments?

It seems weird indeed. If ECM found a 65.2 bit factor, then TF should have found it too. Hardware error, software error, cosmic radiation ... whatever. Lucky if the factor is found by other means.
Maybe Never Odd Or Even can retest it with the TF program of his choice ... if it really misses the factor reliably, then a programming bug can be uncovered. I'm running it through mfakto right now ...

[kladner]

Originally Posted by Bdot

It seems weird indeed. If ECM found a 65.2 bit factor, then TF should have found it too. Hardware error, software error, cosmic radiation ... whatever. Lucky if the factor is found by other means.

[Further confusion ensued..... .....will try to stop inanities cluttering the thread.]

[Bdot]

Originally Posted by kladner

Do the same rules apply to ECM and P-1? P-1 was the original topic, I thought.

P-1 was the original topic, but the factor has an F-ECM tag. But yes, a factor of a certain size should not be missed by TF when testing the range containing that size. Both P-1 and ECM do not test factors by their size but use other properties of potential factors to define the search space (and therefore can find very small factors as well as very big ones).

[kracker]

Originally Posted by kladner

[Further confusion ensued..... .....will try to stop inanities cluttering the thread.]

Ahh I'm with you... I need to learn to reed more throughly

I get what the OP's question. The factor should have been found when TF'ing...

[sdbardwick]

Adding more strangeness: James' page (also linked on the Primenet report) shows the exponent as TF'd to 64 bits.
I've got it TFing on Prime95v259; should have results in a couple of hours (IIRC, P95 tests in ascending 'bitness', so should catch the factor about 18% of the way through) or days...

[Bdot]

Originally Posted by sdbardwick

Adding more strangeness: James' page (also linked on the Primenet report) shows the exponent as TF'd to 64 bits.
I've got it TFing on Prime95v259; should have results in a couple of hours (IIRC, P95 tests in ascending 'bitness', so should catch the factor about 18% of the way through) or days...

It seems less likely that 174GHz-days of TF ( exp=1332151 bit_min=65 bit_max=70 (173.90 GHz-days)) have been done with prime95 ... my ETA for 2^65 to 2^66 is 1:30h ...

[sdbardwick]

Originally Posted by Bdot

It seems less likely that 174GHz-days of TF ( exp=1332151 bit_min=65 bit_max=70 (173.90 GHz-days)) have been done with prime95 ... my ETA for 2^65 to 2^66 is 1:30h ...

Indeed, but my concern for relatively low bit depth is the reliability of P95 tests done before GPUs became feasible.

Anyway, the factor was found, so now my concern is the reliability of NOOE's TF results (and Primenet's by association).