A question about porting CUDA to OpenCL

Hi all,

I’m currently trying to port an CUDA application to OpenCL, but I have difficulty if I’m porting correctly while porting.
The problem is I cannot check the intermediate progress of porting.
The program has about 25 kernels, so I need to check quite often during porting.
I cannot guarantee that I convert 100% correctly.
Do you have any idea to check intermediate progress while porting?
I appreciate all your suggestions.

Write unit tests and regression tests.