COURSE DESCRIPTION:

 

Current developments in computer architecture covering  advanced concepts in sequential and parallel architectures.  Topics include memory hierarchy,  homogeneous and heterogeneous architectures,  shared-memory (SMP and DSM) and distributed-memory machines  (Beowulf cluster to high-end parallel machines),  dataflow and multithreaded architectures, ILP and VLIW,  pipelining,  and vector machines. Systolic arrays and application/language specific  architectures. Networks. Programming models for parallel machines.  Programming for performance on different architectures.

 

This course has the main purpose to give you a sound understanding of up-to-date computer architecture with the main focus on being able to exploit the architectural features optimally for performance programming or for systems programming. The course is not directly addressing the development of own architectures. On the other hand, it discusses architecture really at an advanced level with research orientation and may thus not be suitable for those who just want also to know a bit about computer architecture.

 

TEACHING MATERIAL:

 

• Dezsoe Sima, Terence Fountain, Peter Kacsuk: Advanced Computer Architectures - A Design-Oriented Approach, Addison Wesley, 1997
• Kai Hwang, Zhiwei Xu: Scalable Parallel Computing, McGraw-Hill, Boston etc., 1999

 

PREREQUISITES:

 

This course requires basic knowledge in standard computer architecture and strongly recommends a good understanding of basic operating system concepts. The course will only briefly discuss basic architectural concepts but focus on approaches in modern computer architecture to improve performance. And since most modern approaches involve some kind of parallelism, processes/threads, concurrency and communication play a major role. Without a basic knowledge in computer architecture a participation in this course appears meaningless. If you lack basic knowledge in the software topics mentioned (as typically taught in Operating Systems) you may try to fill the gap on your own.  Do this in the first 3 weeks of the semester – or you will always stay behind. The course will not be slowed down to accommodate weak prerequisites.

 

TOPICS COVERED

 

1. What is computer architecture? – Introduction
2. Why parallelism in architectures?
3. Von Neuman architecture, CISC and RISC, instruction scheduling, instruction pipelining
4. Concurrency, dependencies, levels of concurrency, multi-processor architectures and taxonomies
5. Memory, caches, I/O, memory hierarchy, asynchrony
6. Superscalar and VLIW processors
7. Qantification: performance metrics (runtime, speedup, scalability etc.), benchmarks
8. Vector processing
9. Programming models: MPI, multithreading and others
10. Shared-memory multiprocessors: SMP and DSM, consistency, VSM
11. Networks for HPC, network interfaces
12. Distributed memory - Dedicated MPP machines, HPC and Beowulf clusters
13. Dataflow and multithreading machines
14. Other architectures: language architectures, neural computers, systolic arrays
15. Distributed and grid computing, heterogeneous systems

 

COURSE REQUIREMENTS:

 

1.        Exams: One exam (midterm) will test conceptual understanding (not memorization) of the material presented in the class.

2.        Project: You will have to do practical work on a project

3.        Presentation/Report: Each student will be required to give a presentation and prepare a report on research papers. The presentation has to use Powerpoint or typewritten transparencies. This part of the requirement is supposed to train you ability in understanding papers, critically judging them and understanding alternatives. Additionally it gives you the possibility to train presentation and report writing. The report has to be  2500 to 4000 Words. Both a copy of this report and your presentation transparencies have to be handed in.

4.        Reports: You will have to write short reports on the student presentations.

5.        Small Assignments: There will be two small assignments to train some basic skills.

6.        Active Participation: You are expected to attend the class (the material is not easy and the class notes are not intended to be a replacement for coming to class). Furthermore, you are encouraged to actively participate by questions, answering questions, making comments (especially to student presentations).

 

COMMENTS ON ASSIGNMENTS:

 

Presentation:     A list of topics will be provided. You may not select a topic for which you fully or partially already got credit or will get credit this semester in a different course. Exceptionally, own suggestions for topics may be accepted if approved. The topics provided are mainly case studies for the concepts presented in the lectures. The presentation time will be announced later.

Small assignments: For training some basic skills. Will not be marked in detail.

Reports:             Mainly to make sure you will also listen to the student presentations plus it trains understanding

Project:               The project will have to be demonstrated.

Active participation: Supposed to get you involved and out of the role of passive consumption.

 

 

GRADE COMPOSITION:

 

 

 

IMPORTANT GENERAL NOTICES:

 

·         All of the material presented in the presentation or report must be in your own words such that no section of the report has been copied from any other student in this class, web, journal or conference proceeding. The presentation and report should be based on your understanding of the information and should train finding your own interpretation and formulation – even on the risk that the English is imperfect.

·         Any information taken from a book, journal, magazine or Internet site has to be properly cited according to the general scientific rules (in case of using information provided on the Internet, give the URL for the site from where you got the information).

·         The project and the presentation are to be done in a group of two. However, the individual contributions need to be clearly stated. In the paper presentation and project presentation each of you has to give an equal share of it. For the paper preparation and presentation, it is meaningful each of you should focus on one paper (but, of course, you should care about the other one, too). Though in most cases the mark will be the same (as it is hard to figure out who has exactly contributed how much and as good team work involves lots of mutual influences and some convergence in thought) you may get different marks if there are clear indications that the quality of the contributions is different (no “free ride”). Both partners have to sign a statement of what was the focus in their contribution. You are strongly encouraged to look for an equally strong partner as you depend to some degree on the partner. The partners for the project and the presentation may be different.

 

Copying from others or failure to provide proper references is plagiarism and will be a matter of disciplinary action. Similarly, putting an improperly higher workload on your group partner is considered a violation of work ethics and can lead to a serious reduction in marks.

 

___________________

 

“If you would hit the mark, you must aim a little above it; every arrow that flies feels the attraction of the earth.”

H.W.Y. Longfellow