solving large linear systems. We will focus on iterative linear solvers and its

parallelization, especially the algorithms for multi-GPU (graphics processing

unit). Students who take this coarse can learn various numerical linear solvers

and their implementations on massively parallel environments.

Textbook:

Iterative methods for sparse linear systems (2nd edition) by Yousef Saad 2003.

http://www-users.cs.umn.edu/~saad/IterMethBook_2ndEd.pdf

http://ec-securehost.com/SIAM/ot82.html

Topics:

1. Basic linear algebra review: LU, Cholesky, QR, Eigen-problems

2. Introduction to GPU programming

3. Stationary iterative linear solver

4. Krylov subspace methods

5. Lanczos based methods

6. Pre-conditioning techniques

7. Parallel preconditioner

8. Algebraic multigrid method

Grading:

1. 3 Programming assignments on GPU: 45%

2. Class note: 25%

3. Final project: 30%