mersenneforum.org > Math Shouldn't people who want the $100,000 prime-prize be going to Fatphil for help?  User Name Remember Me? Password  Register FAQ Search Today's Posts Mark Forums Read  2005-11-03, 21:30 #1 jasong "Jason Goatcher" Mar 2005 350710 Posts Shouldn't people who want the$100,000 prime-prize be going to Fatphil for help? Maybe I'm just being a retard monkey for asking this, but since Fatphil's methods seem to yield more primes for less work, shouldn't people be attempting to either recruit him or imitate some of the stuff he does? Not having a clue how Phi(x,y) works, or what it means, I'm willing to be told it doesn't apply, but I'm a little curious.
 2005-11-03, 21:45 #2 rogue     "Mark" Apr 2003 Between here and the 2×7×467 Posts The FFT's Phil is using don't work for numbers of that size and they probably won't anytime soon.
2005-11-04, 10:09   #3
fatphil

May 2003

3·7·11 Posts

Quote:
 Originally Posted by jasong Maybe I'm just being a retard monkey for asking this, but since Fatphil's methods seem to yield more primes for less work, shouldn't people be attempting to either recruit him or imitate some of the stuff he does? Not having a clue how Phi(x,y) works, or what it means, I'm willing to be told it doesn't apply, but I'm a little curious.
They almost certainly don't yield more primes for less work than the most efficient searches out there. George's FFTs are faster, and the limiting factor in this game is the FFTs.

I, with my Phi(3*2^t,b) am merely imitating Yves Gallot, and his GFN (Phi(2^n,b)) project. By virtue of the fact that my calculations take place in the Eisenstein integers rather than the Gaussian integers (or transformed floating point approximations thereto), my code _has_ to be slower than Yves', as I _have_ to do more work than him (mapping between the 2 spaces).

That's not to say that I defer entirely to Yves in this field. I've recently achieved something that, after several months of discussion and trial and error (all error) a few years back, Yves concluded was impossible. Fortunately I'm more stubborn than him.

 2005-11-04, 21:56 #4 jasong     "Jason Goatcher" Mar 2005 66638 Posts I only understood part of what Fatphil said, but I got the answer I was seeking. Having a math education that was cut way short, I'm hoping my college can help me get the understanding I want in, maybe, 5 years. I love math, but my government check I get(because of mental illness) is very small.
2005-11-06, 01:13   #5
fatphil

May 2003

3478 Posts

Quote:
 Originally Posted by jasong I only understood part of what Fatphil said, but I got the answer I was seeking. Having a math education that was cut way short, I'm hoping my college can help me get the understanding I want in, maybe, 5 years. I love math, but my government check I get(because of mental illness) is very small.
Gaussian integers are just x+iy where x and y are integers.
Eisenstein integers are x+wy where x and y are integers, and w is the complex cubic root of 1. See the project logo for a definition of w.

2005-11-06, 01:23   #6
fatphil

May 2003

111001112 Posts

Quote:
 Originally Posted by rogue The FFT's Phil is using don't work for numbers of that size and they probably won't anytime soon.
I use completely different FFTs -- DJB's and YG's. DJB's have hit the brick wall, but in doing so have found hundreds of top-5000 primes. YG's are better utilised in his own projects for the reasons I previously mention.

 Similar Threads Thread Thread Starter Forum Replies Last Post Dubslow Hardware 8 2016-04-15 09:14 Explorer09 Software 1 2014-08-18 19:44 ixfd64 Lounge 10 2012-09-22 15:55 ixfd64 Lounge 22 2010-03-24 02:44 Unregistered Information & Answers 73 2007-08-11 11:38

All times are UTC. The time now is 10:40.

Thu Jan 27 10:40:28 UTC 2022 up 188 days, 5:09, 1 user, load averages: 1.41, 1.27, 1.27