Now I wonder if you can extract a statement from nVidia about the possibility of moving PhysX over from CUDA (C for CUDA) to OpenCL? I guess the chance of them throwing ATI, Intel, and other GPGPU makers a bone are pretty low.

Manju Hegde is seriously deceived. OpenCL is the first general purpose GPU programming language that is at all attractive to an application software developer because (a) it is a clean design that does not refer to any byzantine graphics concepts; (b) it provides a simple interface to an extremely wide range of hardware, including CPUs, Cell, and GPUs from different vendors. CUDA doesn't do that and as a result its market is limited to people who are writing throwaway applications, like students. And anyone who is writing a throwaway application will prefer to program in a "higher-level style".

Fortran?

What no COBOL support?