Original Post
Is there a reasonably efficient way to send a progress update back from an OpenCL kernel? I have a x86 implementation of a very intensive math function right now, and I've set it up to update an integer that another thread can poll to update a UI progress bar. I figure I can't efficiently get the constant update like I have right now, but some sort of periodic (at least once a second) readback from OpenCL would go a long way. I'm mostly wondering if there's a an existing paradigm for doing this sort of thing (other than just breaking the task into smaller kernels and running those one at a time).