This sample demonstrates implementation of matrix multiplication in Ada. The goal is to translate the corresponding example from NVIDIA cuda-samples repository (https://github.com/NVIDIA/cuda-samples/tree/master/Samples/0_Introduction/matrixMul) as close as possible. The original C code is preserved in comments to illustrate the mapping from one language to the other.

As a notable difference, the translation doesn't implement the -device argument. The code chooses the default GPU.

The testing algorithm fills the input matrices with a constant value. It also contains a simple iterative computation of matrix multiplication, which can can be used for validating the computation when a si,le constant input is replaced with random values and/or for performance comparisons.

Usage

$ cd <sample_dir> $ make

Run with the default settings that multiply 320x320 matrix with 640x320 matrix

$ ./main

The matrix dimensions can be changes with corresponding parameters:

$ ./main -wA=640 -hA=640 -wB=640 -hB=640

Switches -? and --help print usage information:

$ ./main -? [Matrix Multiply Using CUDA] - Starting... Usage [? | --help] [-wA=WidthA] [-hA=HeightA] (Width x Height of Matrix A) [-wB=WidthB] [-hB=HeightB] (Width x Height of Matrix B) Note: Outer matrix dimensions of A & B matrices must be equal.