This appendix introduces CUDA FORTRAN, the FORTRAN language programming interface for CUDA GPU devices. It demonstrates the advantage of FORTRAN multidimensional arrays. It also shows how the generic interfaces and compiler directives can be used to improve performance with minimal programmer effort. It also shows how to achieve asynchronous data transfers in CUDA FORTRAN programs.
CUDA FORTRAN; multidimensional array; generic interfaces; directives; asynchronous data transfer