P-1 Missed factor
 

I have been doing additional P-1 in the 11,000,000 - 11,100,000 range. From the mersenne.ca site I took all exponents with P-1 below 3% chance of finding a factor and brought them up to around 5%.

Today my machine did M11023153 and [URL="http://www.mersenne.ca/exponent.php?exponentdetails=11023153"]found a factor[/URL]. The exponent had P-1 previously been done up to B1=60,000 and B2=1,020,000. My machine redid it with B1=250,000 and B2=5,000,000. The factor should have been found in the previous run. I renamed the output and backup files to make them invisible to mprime and redid the exponent with the same settings as the original run. It now found the factor too. Which means the previous machine missed it for some yet unknown reason.

Also, when prime95 or mprime reports a found factor to the server it seems to skip sending the B1 and B2 values to the server. Since it does send them when no factor is found it would be of some help if it would always send the B1 and B2 value, regardless of a factor having been found or not.

Don't ask me how or why, but there is a huge list of P-1 misses:
[URL]http://www.mersenne.ca/p1missed.php[/URL]

Hardware glitch. Old software glitch. Who knows?

Nice find, though.

For some reason this made me think of the Boxcar Children or Hardy Boys mystery books I read in grade school--

"The Case of the Missing Factors"

Hmm...not sure if would have held the same appeal back then...

Get Hercule Poirot on this.

NOW!

I like Philip Marlowe better...

[edit: except when Peter Ustinov plays Poirot, he is one of my favorites, ever!]

I would imagine that P-1 has the same potential problems when running a fft as LL testing. It could even be worse with P-1 as memory use is much higher. I am not surprised that some factors have been missed.

Hmm, I am thinking to delve into this and see if we can figure out which machines did the original P-1 and have proven to be unreliable in LL testing. Next step then could be co compile a list of suspect P-1 exponents and do a confirmation run on them. We do have a nice list of nearly a 1000 exponents with corrected P-1 runs on them to start with.