I have a question,
Is the input data of such kernel: kernel void mul(int width, double2 A, double2 B, out double2 result<> always destroyed (on GPU memory) after every invocation of that kernel or only when destroy the Stream object?
In my particular case I need call that kernel, say 20000x times and only data of "B" changes, "A" is always the same and that is why I am asking. Copying 20000x data of "A" within host memory and GPU memory is a waste of time in this case.
thanks in advance for the reply.