break points in CUDA code are skiped when debug with external program

I have CUDA visual studio solution. it outputs a DLL file. This DLL module is triggered by python code. I set all parameters in “Nsight User Settings” correctly. such as launch external program to C:\Anaconda\python.exe, and other related command line args, working directory…

When I start to run “start CUDA Debugging”, the break points in CUDA code are skipped at all.

The python code call CUDA DLL is:

self.libmodel = import(MyDLL)

My output DLL is in Release version.

Does anyone know how to resolve this issue?

hi civllman,
you should use the debug version for DLL, due to Release version don’t have debug symbols info.