Has anyone discovered a way (ATI?) to run GPU accelerated stream/opencl code on linux _without_ having to run a console X session?
nVIDIA CUDA/opencl is able to do this by using a script that creates the correct devices /dev/*.
This is greatly facilitate large scale deployments (HPC) of GPU accelerated codes. A large cluster of nodes will likely not have running displays attached. Worse yet, each node will be running a stripped down os/kernel so overcoming this limitation would be of tremendous value.
Finally, is there any way to install the driver & libs & xorg bits on linux in a non-standard location? For example, in /usr/local/ati?
This would make it targeted development easier since multiple SDK copies could be maintained simultaneously.