It is even possible to write CUDA kernels, e.g., in case the user wants to use advanced CUDA features that are currently not yet supported by Quasar. Both kernel and device functions can be defined from within C/C++. In fact, the Quasar compiler can parse C/C++ modules and map functions that are named with the ENTRY-macro (such as ENTRY(some_function_name)) onto functions that can directly be called from Quasar. The first approach allows existing C/C++ functions to be used from within Quasar, given that the function signatures are defined in a way that the Quasar compiler can recognize. Running a C/C++ program as a host and accessing Quasar modules from C/C++: see Using the Quasar C++ host API. ![]()
0 Comments
Leave a Reply. |