Operators both on host and device functions

C CUDA works just fine.
But I’m much more worried about the allocation of custom types:

type ( Vector ), allocatable, dimension( : ),device :: device_vector

That isn’t working either. Without being able to allocate stuff not being able to write into stuff doesn’t matter much.

I figured out the other problem (and I guess it’s interesting for you guys):

Completely unrelated to the simple variable assignment I had a huge array (100 cubed) to be used on the graphics card. It compiled just fine but I guess in the actual program it messed things up so that the first time I tried to access the GPU memory (which happened to be the harmless thing I described) it segfaulted. Making the array smaller solved the problem, all other “normal” variable assignments now work just fine.

The other problem is still unsolved: The only reason I had this huge array in the first place is that I can’t manage to allocate custom types dynamically, it only works for the primitive data types.

Any ideas on that?