Hi Serban,
I'm having a weird problem with
the double-precision CUDA version of Boris on Linux: when I use 'cuda
1' to turn CUDA mode on, the value of every vector in every mesh seems
to disappear. Values which aren't stored in a mesh, such as time,
iteration, dM/dt and Ha remain as they were, but anything that is, like
average magnetisation or temperature, becomes zero. Turning CUDA mode
off recovers the data, unless the mesh is changed while CUDA mode is on,
in which case the values become corrupted - magnetisation becomes
~-1e32 A/m in each direction while temperature disappears.
I tried to include a screenshot of what I see in the GUI when I save the simulation with CUDA on and then open it on Windows, but I wasn't able to post this with that included. I'll try again in a later reply if that'd be helpful.
Thanks in advance,
Jack