On OS X 10.6, you will be using Apple’s own OpenCL implementation, not NVIDIA’s won’t you?. You might need to ask on the Apple developers forums about it.
I used OpenCL on my personal computer (that is a Debian Linux machine) with no problems at all, I could run programs using OpenCL “remotely” launching them from an ssh console.
But I found the same problem described in this post on a Debian Linux machine using the same configuration I used on my personal one. The only difference that I can see is that in this machine is used a “centralized users accounting” (and I am not the administrator of it); I think they are using LDAP.
I really can’t understand which could be the problem, I checked that all the environment variables are set correctly…
Also the example programs provided with the NVIDIA SDK are failing with this error when lunched from an ssh console:
OpenCL SW Info:
Error -1001 in clGetPlatformIDs Call !!!
!!! Error # -1000 () at line 42 , in file oclDeviceQuery.cpp !!!
Exiting...
This is an old problem, but as it is not solved yet, I believe it is a good idea to open it again. I solved it by giving the user who runs the software write access to /dev/nvidia0 (or whatever file there is) and /dev/nvidiactl. On my system, I could do this by adding the user to the “video” group but a chmod 777 of course also works (although probably only temporarily).