Hello everyone. I need to build an OpenCL dedicated computation server. I have a motherboard with 2 PCIE16s, AMD Radeon 4850 and a Radeon 5450. I connected both and installed a fresh Linux, CCC drivers and AMD-APP-SDK. I proceeded to compile the computation application and it (the compiled application) didn't detect those GPUs. I tried to "lspci | grep VGA" and they are detected by the kernel. What could be wrong?
Edit: The computation application in question depends on AMD-APP-SDK.
Did you configure the CCC driver, e.g., "aticonfig --initial --adapters=all"? Other than that, don't forget that for the GPU to be recognized by an OpenCL application you must have started an X server on that GPU. I find that requirement particularly frustrating, but it's what we're stuck with at the moment.
After installing and setting up the drivers (as described by settle) you might want to run the clinfo program the comes with the amd app sdk to see whether the opencl runtime can detect your devices. You can also use 'lspci -k' to see whether the amd driver (rather than nouveau) is managing the devices.
Also note that not all OpenCL programs will run on the 4850. That generation of gpus has several hardware limitations (e.g. the texture units behave differently than required by OpenCL) that make it incompatible with the full OpenCL specification.
As settle mentioned, run "aticonfig --initial --adapters=all" then post your xorg.conf for us to look at. It will also help if you let us know your Linux distribution and method used to install Catalyst (AMD installer or package within your distro).
Depending on how things are set up I have had to run "export DISPLAY=:0." to get multiple devices to show up.