[QUOTE=Charles]Pari certifies this as prime (4.9 seconds) with Pocklington-Lehmer:
[/QUOTE]

N-1 certified it in less than 1/10 of a second.

PARI can compete with Proth.exe in the long run.

Certification w/PARI took 1/2 that time for me.

[QUOTE=3.14159;228180]N-1 certified it in less than 1/10 of a second.[/QUOTE]

Yes, Pari isn't good for [i]certifying[/i] primality of small numbers; I usually use Primo (on Windows) or François Morain's ECPP (on Linux) for that. Of course when you have a special-form number Brillhart-Lehmer-Selfridge is going to be faster!

[QUOTE=3.14159;228180]Certification w/PARI took 1/2 that time for me.[/QUOTE]

I ran the test on an old Windows machine instead of my fast Linux box.

[QUOTE=Charles]Yes, Pari isn't good for certifying primality of small numbers; I usually use Primo (on Windows) or François Morain's ECPP (on Linux) for that. Of course when you have a special-form number Brillhart-Lehmer-Selfridge is going to be faster!
[/QUOTE]

Right..

[QUOTE=Charles]I ran the test on an old Windows machine instead of my fast Linux box.
[/QUOTE]

I tested on my Windows machine.

How fast is your "Fast Linux Box"?

[QUOTE=3.14159;228208]How fast is your "Fast Linux Box"?[/QUOTE]

1. Not that fast.
2. Much faster than the Windows machine.

[QUOTE=Charles]1. Not that fast.
2. Much faster than the Windows machine.[/QUOTE]

Bummer.

Also: Anyone willing to help out with writing a simple script for PFGW, for numbers k * b[sup]n[/sup] + 1?

I made a copy of Karsten's script, and this is where I wish to change everything accordingly, so I can make a working script for k * b[sup]n[/sup] + 1, where b and n vary, a la Proth.exe. I'll use the N-1 test (if possible) to see if it outperforms Proth at its own game. It probably will.

Or perhaps just trial-divides to prove the primality of the tiny primes? (1-10 digits)

I ran the +1 side b=1 to 750 in 8 hours and b=1 to 1000 in 27 hours running PFGW using the simplest PFGW ABC2 script on one modern core of an I7. I simply set trial factoring to 100% with the -f100 switch. I then proved all of the PRPs with a second run using the -t switch in < 30 mins.

Everyone made this much too difficult with sieving analysis and different programs. It was extremely simple with PFGW. No sieving needed; only trial factoring that is already built into PFGW. The tests are not big enough to justify spending the personal time to mess with sieving.

A full quad could run this to b=2500 in a few days with minimal sieving.

Attached are all primes for b<=1000. All have been proven prime. Note that these were tested with PFGW 3.3.4. Since the GWNUM libraries have a known problem with them, I'm now in the process of retesting the entire thing with PFGW 3.3.6. I have reached base 800 with version 3.3.6 and have found no problems with the 3.3.4 list. I should be complete with the doublecheck early Tuesday.

Karsten, you might use the attachment to check and extend your page.

Has anyone run this for k*b^b-1 ? I thought I read where someone was going to do that. Karsten, have you created a page for the Riesel side?

Edit: I found it interesting to note that the median prime was for base 22. It just goes to show how many n=1 and n=2 primes there are and how few there are at the higher bases and exponents.


Gary

Dougal has done the Riesel side (see post #95) but only for n=7 to 1000.

I've not yet created a page for the Riesel side but that is no problem.

If you've checked all results with PFGW 3.3.6 I can update/complete the +1-page and also include the -1 page!

[QUOTE=kar_bon;229665]Dougal has done the Riesel side (see post #95) but only for n=7 to 1000.

I've not yet created a page for the Riesel side but that is no problem.

If you've checked all results with PFGW 3.3.6 I can update/complete the +1-page and also include the -1 page![/QUOTE]

-1 is currently up to 2000,il post results tomorrow.i might take it further,ill decide tomorrow.