This is a small library that adapts GPU allocated memory into streams. This helps abstract away the implementation of GPU bound I/O. It implements the following:
- Interfaces for GPU memory readers and writers (in)
- Basic GPU Buffer to inject the readers and writers into (in: opencl, todo: cuda)
- "Factory" for specifying an OpenCL or CUDA memory stream (todo)
Mostly, because I can. Also, who knows when you want to read/write formatted locale oriented information to a GPU with a decent level of abstraction so you don't have to think about it.
TODO: get them written up and running
TODO: learn it, make it, build it
- Some kind of OpenCL SDK (AMD, Intel, NVIDIA) (Note: I built it using OpenCL 1.2 with the NVIDIA SDK)
- Link to the libs and dlls
- Create a main() to execute the code!
- Martin Hollstein