[QUOTE=xilman;87332]I'm getting useful results now. Several more p11 (which aren't useful) and a p13 and p19 which are. They are both of the same number too!
[/QUOTE]
I retract that claim. :blush:

Turns out there was a typo in the input file. I had something like (a^b-1)/(2*23*3*11), omitting a crucial '*' between digits in '23'.

Still chugging along but nothing found which is bigger than p11.


Paul

[QUOTE=xilman;87321]Desite some difficulties (including, perhaps, discovering a bug in ecm-gmp) I've made a start on the numbers with under 100K digits.

The first pass, now complete, was to identify all tiny factors. Two larger ones were found but, at p10 and p11, aren't quite large enough.

Paul[/QUOTE]

Are you sure it's a bug in GMP-ECM? I have run into an issue as well, but I have tracked down to the millerrabin test of GMP. I need to rebuild GMP with -g so that I can debug it to see where it is failing. Let me know if you are seeing the same problem.

[QUOTE=rogue;87348]Are you sure it's a bug in GMP-ECM? I have run into an issue as well, but I have tracked down to the millerrabin test of GMP. I need to rebuild GMP with -g so that I can debug it to see where it is failing. Let me know if you are seeing the same problem.[/QUOTE]

I just reported this as a bug to the GMP bug list. I could reproduce the error on both PowerPC G5 and Pentium 4. AFAICT there is an infinite loop in mpz_powm, which is called by the Miller-Rabin test that is part of mpz_probab_prime_p.

Hi,

All this discussion motivated me to fire up my hacked-up trial division code again, and this appeared.

4186707324070231 | 821^2091403-1

Greg

[QUOTE=rogue;87364]I just reported this as a bug to the GMP bug list. I could reproduce the error on both PowerPC G5 and Pentium 4. AFAICT there is an infinite loop in mpz_powm, which is called by the Miller-Rabin test that is part of mpz_probab_prime_p.[/QUOTE]

:blush:
I was kindly pointed to the fact that I didn't do enough research on the problem. Check out [url]http://swox.com/list-archives/gmp-bugs/2005-March/000189.html[/url]

Here's another...

27337802079423499 | 881^11192861-1

Greg

We've got seven new cases that can be factored. The first is from an error on my part. I had previously shown that the order of 7672759 in 971 was 1759, so Pace needed factors of 971^1759-1. The correct order is 3 * 727 * 1759, so several additional small exponents are suitable.

The other cases come because Pace has asked me to include smaller values of "p" from the Keller and Richstein list of Vanishing Fermat Quotients. Most of these are already factored using Brent's factors, but some cases are incomplete. These are

(197, 653)
(331,359)
(401 347)
(691, 509)
(761, 907)
(863, 467)

Found by gmp-ecm:
p=232553733131897 divides 401^347-1
p is prime

Another factor:
p=5731313378419 divides 863^467-1

[QUOTE=R. Gerbicz;87410]Found by gmp-ecm:
p=232553733131897 divides 401^347-1
p is prime[/QUOTE]

please look at [url]http://oddperfect.org/FermatQuotients.html[/url] :
you have to factor 401^(2*173)-1 instead of 401^347-1

found by P-1:
107832403857149897 | 971^727-1

[QUOTE=Pascal Ochem;87414]please look at [url]http://oddperfect.org/FermatQuotients.html[/url] :
you have to factor 401^(2*173)-1 instead of 401^347-1
[/QUOTE]

Yes, Pascal is correct. I listed the pairs that define the Vanishing Fermat Quotient, not the parmeters for the algebraic factors that need factors. I apologize for not having made that clearer when I showed the additional cases.

William