The AMD OpenCL Programming Guide page A-15 says all APUs support double precision. But this is not correct afaik?
http://developer.amd.com/sdks/AMDAPPSDK/assets/AMD_Accelerated_Parallel_Processing_OpenCL_Programmin...
Now you might say the CPU side supports it, but then same page shows APUs do not support device_fission which is supported by CPU side.
Is this a mistake in the guide? Who should we inform?
Thanks,
Evren
Evren, this post is enough. I've notified the documentation person about this issue and asked him to update the documentation.