What's supported in Nvidia's OpenCL?

Can someone point me to a list of what’s supported/not supported in Nvidia’s OpenCL implementation? In particular, I’m looking for CL_DEVICE_TYPE_CPU support and support for task parallelism.


Our implementation doesn’t support running OpenCL kernels on the CPU, although I would expect the CPU vendors will have their own implementations soon. Note that optimizing OpenCL code for the GPU and CPU will be very different.

Current GPUs don’t really support task-level parallelism. But there’s a lot you can do with data-parallel!