Which libraries I need to include when deploying?

Same question, different title - hope to get some answers here.


I’m new to CUDA programming.
I have a project with .cu file, this project have additional dependencies specified: cuda.lib, cublas.lib, cudart.lib, cufft.lib.
I create .dll file from this project, which I later add to my Win32 program.
On my development machine (Win 7 x64) I have NVIDIA GPU Computing kit installed which has this directory:
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v4.1\bin, in there there are cudart32_41_28.dll etc…

The executable works fine on my computer.

Now, when I move the executable to a client machine (Win 7 x64), which doesn’t have this toolkit installed (and it is not suppose I guess) the program stops working. When I debug it shows that it is missing these *32_41_28.dlls.

For now I’m adding these dlls to the installer, which solves the problem, but I have a feeling that this is not a good solution - if NVIDIA drivers are updated on the client machine in the future it can affect how my program works.

Can you please comment?