Matlab with CUDA v.5.0

This question has been addressed with previous versions of CUDA ( http://www.mathworks.com/support/bugreports/814923), but it's not working for v5.0 for some reason. I'm trying to compile ptx code with -arch sm_20 and i'm getting one of two errors when i run CUDAKernel:

1) When compiling only the kernel function:

No entries were found in the PTX file

2) When I've got an additional device function defined before the entry function:

The number of inputs to the PTX code (0) is NOT the same as the number of inputs in the C prototype (5).

Any clues?

Thanks,

-Matt