ECE 453
Advanced Computer Architecture II:
Multiprocessor Design

Winter 2007
Mondays and Wednesdays 3:30pm-5:00pm
Tech L158


Course Description

This course examines fundamental issues and design trade-offs in multiprocessor architecture. It covers both quantitative and qualitative analysis of parallel computer systems and serves as a entry point to further research in high performance computing.

Course material covers both traditional topics in parallel architectures such as parallel programming models, symmetric multiprocessors, coherency policies, distributed shared memories, and scalable multiprocessors, as well as important emerging topics like chip multiprocessors and on-chip interconnection networks. The project component is open-ended and students are encouraged to draw on their own research interests and prior background for inspiration.


News

  • Exam 1 will be in class Wednesday, Feb 7

Course Information

For detailed course information, including grading policies, and course structure, please see the syllabus.

Required Text

David E. Culler and Jaswinder Pal Singh with Anoop Gupta, Parallel Computer Architecture: A Hardware/Software Approach, Morgan Kaufmann Publishers, 1999

Instructor

Professor Russ Joseph
Office: Tech L467
Phone: 1-3061
Office Hours: Friday 2:00pm-4:00pm

Assignments


Projects


Tentative Schedule

The schedule below is a best guess given the likely sequence of material covered in this course. Slight deviations are inevitable.

Date Topic
1/3 Introduction to Parallel Architectures
1/5 Communication Architectures
1/8 Programming Models
1/10 Memory Consistency
1/15 Basic Bus Based Coherence
1/17 More Bus Based Coherence
1/22 Evaluating Coherence Strategies
1/24 Implementing Snoopy Caches
1/29 Discussion: CMP Cache Strategies
1/31 Exam 1
2/5 Scaleable Systems
2/7 Directory Based Coherence
2/12 Basics of Interconnection Networks
2/14 Advanced Interconnection Networks
2/19 Transactional Memory
2/21 Paper Discussion
2/26 Paper Discussion
2/28 Paper Discussion
3/5 Exam 2
3/7 Project Presentations

Paper Discussion Topics

  • Coherence Ordering for Ring-based Chip Multiprocessors: Michael R. Marty, University of Wisconsin - Madison Mark D. Hill, University of Wisconsin - Madison [pdf]
  • In-Network Cache Coherence: Noel Eisley - Princeton University, Li-Shiuan Peh - Princeton University, Li Shang - Queens University [pdf]
  • An Analysis of Efficient Multi-Core Global Power Management Policies: Maximizing Performance for a Given Power Budget: Canturk Isci - Princeton University, Alper Buyuktosunoglu - IBM, Pradip Bose - IBM, Margaret Martonosi - Princeton University, Chen-Yong Cher - IBM [pdf]
  • Utility-Based Cache Partitioning: A Low-Overhead, High-Performance, Runtime Mechanism to Partition Shared Caches: Moinuddin K. Qureshi - University of Texas - Austin, Yale N. Patt - University of Texas - Austin [pdf]
  • Molecular Caches: A Caching Structure for Dynamic Creation of Application-Specific Heterogeneous Cache Regions: Keshavan Varadarajan - Indian Institute of Science, Bangalore, S. K. Nandy - Indian Institute of Science, Bangalore, Vishal Sharda - Indian Institute of Science, Bangalore, Amrutur Bharadwaj - Indian Institute of Science, Bangalore, Ravi Iyer - Intel Corporation, Srihari Makineni - Intel Corporation, Donald Newell - Intel Corporation [pdf]
  • ASR: Adaptive Selective Replication for CMP Caches: Bradford M. Beckmann - University of Wisconsin - Madison, Michael R. Marty - University of Wisconsin- Madison, David A. Wood - University of Wisconsin- Madison [pdf]
  • Managing Distributed, Shared L2 Caches through OS-Level Page Allocation: Sangyeun Cho - University of Pittsburgh, Lei Jin - University of Pittsburgh [pdf]
  • Transactional Memory (Read these for in class discussion on 2/28):
    • Transactional Memory: Architectural Support for Lock-Free Data Structures: Maurice Herlihy and J. Eliot B. Moss [pdf]
    • Lowering the Overhead of Nonblocking Software Transactional Memory: Virendra J. Marathe, Michael F. Spear, Christopher Heriot, Athul Acharya, David Eisenstat, William N. Scherer III, and Michael L. Scott [pdf]