I am trying to step through host code that calls several kernels during its course. I was wondering if I can set an arbitrary breakpoint in the host code and manually cudaDeviceSynchronize from cuda-gdb command line. Is this possible?