This project is read-only.

Does cudafy support unified memory or not?

Jan 29, 2015 at 4:17 PM
CUDA SDK 6.0 provide unified memory. developer do not need move variables between CPU and GPU explicitly. these job are done by cuda run time.

I just want to know if currently version cudafy support this feature or not.
Jan 29, 2015 at 6:02 PM
Well, first of all, managedCuda has nothing to do with cudafy except that both projects aim to bring Cuda to the C# world. But already the “how” differs quite a lot...

I’m also not sure what you mean by unified memory. Do you mean managed memory that gets copied automatically from host to device and vice versa by the runtime? If so, managedCuda does support managed allocations. But again, you’re in the managedCuda forum here and I’m not involved in cudafy, so I can’t tell what cudafy supports and what it doesn’t support.