Compare Serge's multi-core run:

[CODE]Using generic reduction FMA3 FFT length 1440K, Pass1=320, Pass2=4608, 18 threads, a = 11
11^((N-1)/5119)-1 is coprime to N!
11^((N-1)/7)-1 is coprime to N!
11^((N-1)/2)-1 is coprime to N!
143332^786432-143332^393216+1 is prime! (4055114 decimal digits) Time : 105281.520 sec.
[/CODE]

to UTM's single core:

[CODE]Phi(3,-143332^393216) is prime! (606135.7958s+1.1064s)
[Elapsed time: 7.02 days][/CODE]

It looks about right. The scaling was flat as soon as you leave the boundaries of a physical chip. I could have just as well run only 9 threads (it was a 9-core Xeon), with the effective scaling ~6x.

On a side, I have run a dozen simple-form smaller known primes for testing "no false negatives"/"no lost functionality"; all passed.

[QUOTE=paulunderwood;452029]:w00t: These/it could be over 5 million digits![/QUOTE]
...not "could". These/it actually [I]are[/I]. :rolleyes:

[QUOTE=Batalov;453409]...not "could". These/it actually [I]are[/I]. :rolleyes:[/QUOTE]

You've actually found one or more and are in the process of proving it/them? :unsure:

Yes. Still proving with the vanilla N-1 method though... and that takes quite some time even with parallel LLR.

Time for parallel-PFGW testing! :rolleyes:

[QUOTE=Batalov;453415]Yes. Still proving with the vanilla N-1 method though... and that takes quite some time even with parallel LLR.

Time for parallel-PFGW testing! :rolleyes:[/QUOTE]

Is there end of your prime findings? :bow: :bow: :bow:

[QUOTE]Starting N-1 prime test of 123447^1048576-123447^524288+1
Using generic reduction FMA3 FFT length 1920K, Pass1=320, Pass2=6K, 16 threads, a = 7
123447^1048576-123447^524288+1 may be prime, trying to compute gcd's
7^((N-1)/41149)-1 is coprime to N!
7^((N-1)/3)-1 is coprime to N!
123447^1048576-123447^524288+1 is prime! (5338805 decimal digits) Time : 187808.030 sec.[/QUOTE]

[URL="http://primes.utm.edu/primes/page.php?id=123041"]5,338,805 decimal digits[/URL]! :banana: :banana: :banana: :banana: :george:

[QUOTE=paulunderwood;453529][URL="http://primes.utm.edu/primes/page.php?id=123041"]5,338,805 decimal digits[/URL]! :banana: :banana: :banana: :banana: :george:[/QUOTE]

In just a little more than 2 days... Congratulations Paul! :bow:

[QUOTE=ET_;453533]In just a little more than 2 days... Congratulations Paul! :bow:[/QUOTE]

:no: Luigi, it is Serge and Ryan et al you should be congratulating :smile:

[QUOTE=paulunderwood;453537]:no: Luigi, it is Serge and Ryan et al you should be congratulating :smile:[/QUOTE]

Oh, I thought it was just another proven PRP... next time I will check the link before posting, sorry Serge and Ryan :redface:

[QUOTE=Batalov;453415]
Time for parallel-PFGW testing! :rolleyes:[/QUOTE]

Any joy there, Serge?

That was a hypothetical...

Someone else reported that it does work, didn't they?

[QUOTE=Batalov;453574]That was a hypothetical...

Someone else reported that it does work, didn't they?[/QUOTE]

Yeah, me. See: [url]http://www.mersenneforum.org/showthread.php?p=452599#post452599[/url]

One line with [c]sed[/c] to alter it.

I would run it myself, on your new prime, but there is little point -- besides it would take more than ~3 days on my 4770k :wink:

[QUOTE=pepi37;453423]Is there end of your prime findings? :bow: :bow: :bow:[/QUOTE]

:bump:

What is the status of 49152? Is it nearing completion? 26 primes and counting...

Yes, near finishing. (It was run on a few EC2 nodes and now I only have a clean up run closing small unfinished gaps; the largest being ~562k-600k.)

The slope of the sequence of [I]b[/I] values is reasonably close to the ad hoc expected [FONT=Times New Roman][SIZE=2]γ [/SIZE][/FONT][SIZE=2]• (2m) [FONT=&quot]≈ [/FONT]56721[/SIZE]

[QUOTE=Batalov;401873]I've systematically scanned the Phi(3,-b^2^15) P.I.E.S. series and here is the chart for the numerologists to ponder.

This is the distribution of the 30 b values for which Phi(3,-b^2^15) is prime, and it is fairly clumpy in parts: there are seven primes for b<46,000; ...[/QUOTE]
There are now 48 of the Phi(3,-b^2^15) P.I.E.S. and the slope is remarkably close to γ • (2m) ≈ 37814

While filling in missing values (a(11)-a(13)) in [url]https://oeis.org/A298206[/url] found a rather "late" bloomer prime (165394^49152-165394^24576+1, a 256501-digit prime).
a(13) = [B]165394[/B] stands out quite far from the "expected range to find it" (~30,000).

Just an illustration for statistical inferences about "where the next Mersenne is" or the like.
It is where it is, no matter where it was "expected". :rolleyes:

P.S. Left the scripts running for a few minutes and the next one showed up almost immediately: a'(13) = [B]165836[/B].
165394^49152-165394^24576+1 is a 256501-digit prime.
165836^49152-165836^24576+1 is a 256558-digit prime.
No smaller b<165394 with b^49152-b^24576+1 prime.

That's even funnier: this is directly related to silliness of those who stop searches for certain "k" values in Riesel search project - just because "there was a prime found just now. Now there will not be another one for miles." No, silly, the probability is exactly the same for each next candidate!