Programming Massively Parallel Processors, 2nd Edition

Appendix A

Matrix Multiplication Host-Only Version Source Code

Appendix Outline

A.1 matrixmul.cu

A.2 matrixmul_gold.cpp

A.3 matrixmul.h

A.4 assist.h

A.5 Expected Output

This appendix shows a host-only source code that can be used as the base of your CUDA matrix multiplication code. We have already inserted timer calls in key places so that you can use the measurement to isolate the execution time of the function that actually performs the matrix multiplication. It also has the code that you can use to print out the matrix contents and verify the results.

A.1 matrixmul.cu

/∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗

 File Name   [matrixmul.cu]

 Synopsis    [This file defines the main function to do matrix-matrixmultiplication.] ...