The P-1 factoring CUDA program
 

6 GB of non-shared memory mean decent P-1 power, right?

to **** with it. i'll try something, but it will be unclever, slow, and useless at first.

Plans for a P-1 factoring CUDA program
 

[QUOTE=firejuggler;330225]to **** with it. i'll try something, but it will be unclever, slow, and useless at first.[/QUOTE]

Hey, deliver a prototype P-1 GPU program and I'll give you my first born child.

(Please note: I'm intentionally childless.)

[QUOTE=firejuggler;330225]to **** with it. i'll try something, but it will be unclever, slow, and useless at first.[/QUOTE]

Are you planning to use CuFFT or a GPU-enabled GMP?

I have the second, but I'm afraid I can reach a decent efficiency using montgomery multiplication...

What I need is how to apply the FFT multiplication routines to the code.

Yes, I'm dreaming about coding something myself...

Firejuggler, if you like you may share your ideas with us.
We could open a dedicated thread on how to design a GPU-enabled P-1 program (and invite people with experience...).

What do you think? :boxer::big grin::sos::et_:

Luigi

please note : I have a lot of free time, and a less than average programming capacity...
the important part being lot of free time.

[QUOTE=firejuggler;330229]please note : I have a lot of free time, and a less than average programming capacity...[/QUOTE]

Fine! I have some programming capacities, and already worked out some CUDA and math programs, I can read and check blueprints, code pseudo-algorithms and read specifications, but my time to code is limited (no CUDA 2.0 at work).

If you have freetime, I'd like to have links of different, working, easy to read P-1 programs...

Luigi

Using CUDALucas' FFT/mul code would be a good place to start. I believe there is also other work being done on this particular front of the GPU P-1 issue.

[QUOTE=Dubslow;330233]Using CUDALucas' FFT/mul code would be a good place to start. I believe there is also other work being done on this particular front of the GPU P-1 issue.[/QUOTE]

Care to expand?

Belief is one thing. Knowledge is quite another....

[QUOTE=chalsall;330235]Care to expand?[/quote]Nope.
[QUOTE=chalsall;330235]
Belief is one thing. Knowledge is quite another....[/QUOTE]
I did just double check that I was remembering correctly, and I was. One more knowledgeable than I in such large-number arithmetic on GPUs has "been thinking about p-1 on gpus".

I don't know if that's going anywhere though.

Plans for a P-1 factoring CUDA program
 

I've been thinking about a CUDA program for P-1 factoring for quite a bit, and think that many other Mersennaries had.

First of all, note that I have only a limited knowledge about the math involved, but I'm willing to expand this limitation studying under the guide of more informed people, and eventually start coding something with the ir help.

I'd like to gather ideas about how such a program should be designed. Some questions will be trivial, some other maybe deeper, but all of them will be enclosed in this thread.

Some naif subjects to talk about:

- Parallelization of tasks
- Limitations due to the memory factor of the GPU (how far may we go having 0.5, 1, 2,3 or 6GB of memory?
- Limitations of the GPU shared memory.
- Description of steps 1 and 2 (from MersenneWiki I got a grasp of it, but a talk would explain more).
- use of streams to pass chunks of bytes to analyze.
- How to apply CuFFT library to the algorithm.
- Is a parallel Montgomery multiplication algorithm out of question for such algorithm?

I hope it may help both people in need for a CUDA P-1 program, programmers, mathematicians.

Luigi

[QUOTE=Dubslow;330237]Nope.

I did just double check that I was remembering correctly, and I was. One more knowledgeable than I in such large-number arithmetic on GPUs has "been thinking about p-1 on gpus".

I don't know if that's going anywhere though.[/QUOTE]

Would you mind posting [URL="http://www.mersenneforum.org/showthread.php?p=330241#post330241"]here[/URL]?

Luigi