The program activity demonstrates how C/C++ or even CUDA codes can be used in Python scripts.

1. The _pointers.cpp file is loaded with the corresponding header file named _pointers.h. The header file is very much important here! Declaration as extern "C" is needed in order for the function to be read by the Python API. The compilation using nvcc is as follows:

C:> nvcc _pointers.cpp -shared -o <outputfile.dll>

2. Run main.py to load function from the shared library. The function names can be checked by running dumpbin for Windows or nm for Unix.