Floating point precision
If I need more precision that float supports, what other options do I have in OpenCL? From what I see double precision may not be available to all platforms.
Re: Floating point precision
If double is not supported then you will have to divide up your calculations manually to retain more precision. (Effectively emulating double support.) Double support should be available on CPU devices since they all support CPUs, and my guess is that all GPUs will support it within 2 years.