View Single Post
2005-08-08, 06:16   #3
trilliwig

Oct 2004
tropical Massachusetts

10001012 Posts

To Washuu's first question, the simple answer is "no". Or "only by accident". At least with "pol51m0b", it chooses non-monic linear polynomials which I do believe will eliminate any reasonable possibility of finding simple polynomials for numbers of this form. I can't really give a better answer than that; the polynomial selection programs in GGNFS were never intended to check for SNFS numbers.

Second, I don't even think such a program would solve a real problem.
Quote:
 Originally Posted by Washuu let's assume we have a general number N, that can be represented as SNFS k*x^p+c (example: 3049*1067^37-12321) But, somehow, user forgot to check this.
What conceivable circumstance could exist where the user would "forget"? If a user wants to factor a number, he or she should be intimately familiar with where it comes from! It's saner all around to write a program that parses "3049*1067^37-12321" as input and does the right thing automatically, than expecting the user to give the raw decimal expansion and expecting the program to figure out if a good SNFS polynomial exists for it.

Quote:
 Originally Posted by akruppa this will only detect numbers with positive c. For negative c, look for a lot of b-1 digits in the middle. Neither will work if known factors have been divided out of N already.
That will be a fatal deficiency. The proper input value of N requires that all known factors be divided out. If this program is to have any hope of guessing if SNFS is possible on a particular N, it would need the algebraic form in addition to N, and the most convenient method of input should already tell the program everything it needs to know.

--
Sam