[Gpundarray] as_{contiguous,fortran}array and opencl

[Frédéric Bastien]

I have finished the gpu code modif and the python code to call
pyopencl instead of pycuda. I have one problem left and I don't find
documentation about it.

The problem is when calling the gpu function generated by pyopencl.
How can I pass him a gpu pointor made by compyte? It accept a Buffer
object or MemoryObject only I think. I can't create a MemoryObject
from python. I only found a way to create a Buffer that allocate
memory. but the memory is already allocated by compyte, how can I
reuse it? I tried passing the pointer as a python long int as done
with pycuda, but this didn't worked. Any tips?

I'm looking into modifing wrap_cl.hpp, but that is not the type of
solution I like. If I need to do that, I will accept python long as
pointer. Comments/Objections on that?

Fred