Overview
This course will cover architectural issues for parallel computing, including parallel program design for performance, performance metrics, shared memory and message passing issues, interconnection network design, and latency tolerance techniques. It assumes that students are already familiar with the material covered in CSE 431 (and to a certain extent in CSE 530) including processor architectures, pipelining, caches and memory hierarchies, and I/O system architectures. The text for this course is "Parallel Computer Architecture: A Hardware/Software Approach" by D.E. Culler, J. P. Singh and A. Gupta (published by Morgan Kaufmann). Projects and exams will include all material covered in the lectures and active class participation is essential.

Projects
In addition to learning the concepts in class, you will be working with a commercial simulator (Simics) - using it for architectural studies as well as adding/changing modules for architectural enhancements. There will be 2 programming projects for the course and a detailed description of each project will be handed out separately. The project reports are due in class on the designated day (no extensions will be given). All projects should be implemented on SUN workstations running Solaris using Simics. Appropriate documentation/manuals for will be made available. You will be working in teams (atmost 2 per team) for each of the projects. You are free to choose your partner for each of the projects (could be different for each project). If either of you choose to drop out of your team for any reason at any time, each of you will be individually responsible for implementing and demonstrating the entire project and writing the project report by the designated deadline. In any case, both of you will need to understand and be able to explain the entire design and implementation even though you may have worked on only one part of the project.