Conflicted use of DISPLAY variable in SSH

Discussion created by bananafish on Nov 19, 2010
Latest reply on Nov 20, 2010 by nou
OpenCL's use of DISPLAY for GPU conflicts with application's use of DISPLAY for GUI

Is it possible to create an `ssh -XY` connection to a headless machine running an OpenCL application using that headless machine's GPU, while tunnelling that application's GUI up the SSH connection?

The two use the same DISPLAY variable, and so far I have not found a way to tunnel X11 from the application to my laptop while at the same time using the headless GPU.

Could it be necessary to have OpenCL use a seperate environment variable for deciding which GPU display(s) to use (i.e. CL_DISPLAY) so that the GUI's DISPLAY can be set to another value?