I have an AMD CPU Kaveri A10 7850K (CPU 4C + GPU 8C) with 6GB RAM installed and a discrete GPU alonside R7 265 2GB.
I've installed Ubuntu 14.04 LTS on the system and then the HSA kernel and runtime .
I've got the sample working on HSA and from what I understand in order to use OpenCL with the HSA I have to compile the CL code into BRIG/HSA .
Looking at the schematics of the software stack describing HSA I don't understand where the GPU driver actually fits in . I mean I can use the GPU in a HSA program without the amdgpu kernel module ? Or is it already installed ? How do OpenCL commands translate into this picture  ?
I installed AMD APP SDK 3.0 beta, compiled the provided samples and ofc tried to run them. I get no valid platform/device found.
I then turned to install the fglrx driver from stable or updates - that crashed the X server and I had to remove them. I tried to run OpenCL while they were installed but again received errors no platform/device found.
So to summarize the questions are:
1. Where does the GPU driver fit in the picture  from both HSA and OpenCL point of view ?
2. Does the amdkfd control both CPU and GPU ? Who for instance controls the device init and clocks ? Who builds the GPU command queue and sends commands to the GPU ?
3. Can't I use OpenCL and HSA at the same time ? If so what is the proper/working configuration to do this ?
4. Is OpenCL 2.0 supported at the GPU level and if yes in what software configuration (driver, app sdk etc...) ?
5. Is there a place that summarizes all of the configurations aside from wiki/forum/release notes ? There is a lot of information and configurations and you can easily mess things up.
6. Aside from Kaveri A10 7850 have the other CPUs been tested ? For instance on A10 7850 I get "The maximum queue size is 131072." What will I get on A6 7400K ?