I've added the CUDA ppm solver and have it compiled.
I'm hoping someone will help to test this on some simple test problems now
and I'll work on replacing the old CUDA MHD solver with the new and improved version.
This runs fine (it flies) at the GPU cluster at Stanford.
Thanks to Sam for help with clean up ...
Since most of this doesn't affect other solvers, and it's a big improvement over the existing CUDA, I am +1 on including it. Are there any reasons to hold off?
Other than the last remaining conflict, this looks good to me! Thanks Tom!