I'm not sure if I really agree - the level of abstraction used for each is extremely similar. There's not really any "Graphics Pipeline Specifics" pollution in

I'm not sure if I really agree - the level of abstraction used for each is extremely similar. There's not really any "Graphics Pipeline Specifics" pollution in OpenCL or CUDA. You can pretty much translate something like https://github.com/jcupitt/opencl-experiments/blob/master/OpenCL_Hello_World_Example/hello.c https://github.com/jcupitt/opencl-experiments/blob/master/Op... with string replace function names.