GPU Computing: Spring 2008
Project 1: Due Mar 3, Monday 2008 at 5pm
Requirements
In this assignment you will implement basic algebra operations include:
1. Matrix addition\reduction: C=A+B
2. Matrix multiplication D=A*B,
3. Matrix-Vector multiplication y=Ax, here y and x are vectors
Advice
The input matrices and vectors can be generated automatically in your code by using
random number generator and scale the generated number to appropriate range (such as 0
to 1)
Please make sure about the compatibility of matrices in the operations. For example,
A+B is valid only if A and B have the same dimensions. For multiplications, please refer
to general linear algebra text for theoretical details.
Make you code as general as possible for definition of matrices. You can try different
dimensions of matrices for performance test.
Requirement
In your submitted version:
N by N matrix should be used with the dimension N = 500;
Each element of your matrix\vector should range between 0 and 1 in floats;
Please submit a performance chart on both GPU and CPU with increasing number of N,
and your short description and analysis.
Credits
Implementing all the requirements will give you full credits at 100 points.
Extra! Extra!
Compute the inversion of a matrix is still a difficult and active research topic for GPU
computing. If you can implement this on GPU, with an exceeding speed than CPU
version, you will get 30 extra points.
Submission
Please follow the submission rules of the course (see course website). Send submission to
[email protected] Please DO NOT include executive files.