CUDA Fortran for Scientists and Engineers

Appendix D

Source Code

CUDA Fortran source code that was deemed too long to include in its entirety in earlier chapters is listed in this appendix. Each section in this appendix contains all the relevant code, both host code and device code, for the particular application.

D.1 Texture memory

The following is the CUDA Fortran code used in Section 3.2.3 to discuss how textures can be advantageous in accessing neighboring data on a 2D mesh using four- and eight-point stencils:

D.2 Matrix transpose

The following is the complete matrix transpose CUDA Fortran ...