Instructor: Dr. Abdelkader Baggag
Email: baggag@clumeq.mcgill.ca
Phone: (514) 398-3293 Fax: (514) 398-2203
Course Schedule: May 12-13, 2004
Course Location: Carleton University 
Registration: www.cfd2004.org/registration.html
Course Fee: $250CAD/$193US before March 22 and $300CAD $231US after March 22. 

Prerequisites: No prior experience with MPI or parallel programming is required.

Course Instructor:

BIOGRAPHY

Dr. Abdelkader Baggag completed his doctoral degree in Computer Science with a Minor in Scientific Computation from the university of Minnesota / Purdue university, receiving renown and praise from his faculty advisors and peers as well as those in the research community with whom he worked at NASA Langley Research Center in Hampton, Virginia.

Dr. Abdelkader Baggag is an expert in High-Performance Computing: He has tremendous Masters, Ph.D. and Research-work experience in large-scale applications on parallel computers. He has worked with the best in Montreal, Minnesota, Purdue and NASA Langley Research Center at the Institute for Computer Applications in Science and Engineering (ICASE).

He intimately knows CFD, and has studied under the best in Minnesota and Purdue: Iterative Solvers, Discontinuous Galerkin, Finite Volumes, Finite Elements.

He intimately knows parallel computing and has done it successfully on a variety of scientific applications, from aero-acoustics to particulate flows.

He has prepared and taught university level courses at the undergraduate and graduate level at Hampton university in Virginia.

At McGill university, he is associated with the department of Mechanical Engineering as a Professor, teaches "High-Performance Computing" courses, and does research in "Parallel Robust Preconditioners for Large-Scale problems" in collaboration with the McGill's CFD laboratory. He also provides "Parallel Numerical Algorithms" for the researchers at large at McGill university, since he is also "Senior HPC Analyst" at the CLUMEQ supercomputer center, which is in the top500 fastest machines in the world.

EXPERTISE

Large-Scale Applications on Parallel Computers 
Computational Fluid Dynamics (CFD) 
Iterative Solvers 
Discontinuous Galerkin 
Finite Volumes 
Finite Elements 
Aero-Acoustics 
Parallel Numerical Simulation of Particulate Flows 
Robust Preconditioners 

Parallel Computing Explained

This module is an introduction to parallel computing. It provides a resource that is useful to both beginners and more experienced users.

Outline:

Parallel Computing Overview 
SGI Origin3800 Overview 
Cluster Computing Overview 
Porting Issues 
Timing and Profiling 
Scalar and Cache Tuning 
How to Parallelize a Code 
Parallel Performance Analysis 
Parallel Code Tuning 

This module is an introduction to parallel programming through the Message Passing Interface (MPI), a standard library of subroutines, or function calls that can be used to implement a message-passing program.

Outline:

Message-passing fundamentals 
Getting started with MPI 
MPI program structure 
Point-to-Point communication 
Communication modes 
Derived datatypes 
Collective communications 
Communicators 
Virtual topologies 

This module covers "intermediate" level topics in MPI, and is both useful and relevant to expand MPI knowledge acquired in the previous module.

Outline:

Message Probing 
Derived Datatypes 
User-defined Reduction Operations 
Inter-communicators 
Parallel Input/Output 
Graph Topologies 

This module describes a hybrid-type programming using both OpenMP on computational nodes, and MPI across nodes.

Outline:

Distributed-Shared Memory machines 
OpenMP Programming Review 
Laplace Solver using OpenMP 
Laplace Solver using MPI 
Combining OpenMP and MPI Programming 
Laplace Solver using OpenMP and MPI 
Multi-level Parallelism and Performance 
Common Multi-Level Parallelism Errors

Domain Decomposition Techniques 
Parallel Implementation of Discontinuous Galerkin Method for Aeroacoustics (Explicit) 
Parallel Iterative Solvers and Preconditioners (Implicit) 
Parallel mathematical libraries