Quote:
Originally Posted by Andi47
It seems that your results of proving the compositeness of F25, F26 (and now F27) did not make it to Wilfried Keller's Page. Can you please send him an email?

I had an email from Wilfrid Keller just a few days ago in which he discussed this exact topic. He complained that the status of the Fermat cofactors is somewhat murky, although the smaller ones have undoubtedly been tested independently enough times that their status as composites is not in doubt. But he says that even the composite cofactor of F22 does not meet his standard of two matching tests using different hardware and different software.
A simple prp test done on two different machines using different software should verify this status as composite. Doesn't Ernst's MLucas code also contain routines for doing calculations modulo Fermat numbers?
On the other hand, historically, the following test has often been done, and has the advantage that if the full result of the Pepin test is saved, and another factor is discovered in the future, the new cofactor can be tested easily without repeating another long Pepin test. The test is as follows:
1) Compute R
_{1} as 3 raised to the 2
^{2[SUP]n}[/SUP] power modulo F
_{n}=2
^{2[SUP]n}[/SUP]+1 (the Pepin residue.)
2) Compute R
_{2} as 3 raised to the power of P1 mod F
_{n} where P is the product of all known prime factors of F
_{n}.
3) Reduce both of these residues mod C, where C is the remaining cofactor of F
_{n}. If they are not equal, C is composite.
4) Take the GCD of the difference of these two residues R
_{1}R
_{2} with C. If the GCD is equal to 1, C cannot be a prime power. (If it is not equal to 1, we have discovered a new factor of C.)
Note that computing R
_{1} is costly for large Fermat numbers, but for small factors P, R
_{2} is easily computed. Therefore, it would be quite quick, given R
_{1}, to test a new cofactor should a new small factor be discovered in the future.