complex support would be particularly helpful in the next coming releases due to it's applications in simulations (even at a basic level).
I too would like complex support for single and double. It's reserved right now so I at least assume it's under consideration. Same goes for quad right?
Would be useful indeed!
Yes, please provide complex number support!
It's absolutely necessary for any kind of serious signal processing, and is one of the major disadvantages that opencl has in regard to CUDA.
I currently use a set of functions written for pyopencl, available at:
But this is cumbersome, and makes it hard to spot errors.