Lecture on Advance Parallel Computation

---------------------------------------------------------------------------------------------------------

Fifty per cent of the computing resources in the industry are nowadays parallel computers. As a matter of fact, efficient numerical simulation is becoming a key component in the design of new products and/or the production of energy. This lecture will review tools, algorithms and applications of parallel computing. Main emphasis will be on practical issues and we will be using alpha based parallel systems as well as so called Beowulf system.

We will review message passing programming, benchmarking and verification of existing codes as common tools. We will study the parallel version of fundamental algorithms to solve linear systems, fast Fourier transforms, and modern methods as domain decomposition and multigrids. Application will includes three basic families of problems that are heat transfer, wave propagation and potential flow. These examples will provide a basis to further development in grid computing.

Plan:

Introduction and Motivation

Tools Applications Methods

Message Passing Potential Flow Linear system solve

Benchmarking Heat Transfer Multilevel Method

Code verification Wave propagation Domain Decomposition