gpuOWL for Wagstaff

[kriesel]

Originally Posted by axn

Here is cudaWagstaff v0.03 (adapted from cudalucas)
This is a cut down version with lot of the non-critical functionalities missing. For comparable FFT it is about 5% slower than CUDALucas (on my 1660Ti - YMMV).

Assignment syntax is
PRP=<exponent> or PRP=<exponent>,<fft>

Implements GEC PRP.
GEC failure / rollback logic is untested, but should work.
Supports cufftbench functionality, but no threadbench.

I have only built/tested in Linux; may or may not build/run under windows.
Please give it a try. Let me know if you have any questions / issues / feedback.

EDIT:- Attachment updated with minor change

Wow, outstanding, axn.
Care to post an executable? (Most of us don't have working CUDA build environments or CUDA app programming or build experience.)
Did you preserve the nonzero offset present in CUDALucas?
Maybe CUDAWagstaff should have its own thread at some point.

The speed with which you adapted it to both Wagstaff and PRP suggests you could also produce CUDAPRP (Mersenne numbers tested by PRP on CUDA) without huge effort.
CUDAPRP with pseudorandom offset would be a nice gpu complement to Gpuowl.
And possibly even run on older CC2.x or 3.x level cards that can't run Gpuowl, take a long time to do an assignment and so may produce more errors that CUDALucas won't catch.
Your time to spend how you like, of course. (I could make some runs for you in exchange.)