Instructor
Prof. Antonia Zhai
E-mail: zhai at cs dot umn dot edu
Phone: 612-626-1285
Office: EE/CSci 6-205
Office Hour: Fri. 1:00pm-3:00pm
Time and location
4:00AM-5:15AM, M/W, EE/CSci 3-111 and UNITE.
Online Information:
- Course Webpage: http://www-users.itlabs.umn.edu/classes/Spring-2008/csci8205/
- Mailing List: csci8205@itlabs.umn.edu, visit here to subscribe
- Course Forum: https://wwws.cs.umn.edu/it-forum/list.php?397
- Lecture Notes: Password protected. User Id and password will be announced in class.
- Grades: available on webct.
Prerequisite
EE 4363 / CS 4203 or instructor's permission
Description
With advances of technology, parallel machines are becoming ubiquitous from
desktops to supercomputers, and even handhold devices. Thus, as researchers,
we are facing the pleasant challenge of exploiting the potential processing
power of these machines.
The goal of CS8205/EE8367 (Parallel Machine Organization) is to provide an
in-depth study of the design, engineering, and evaluation of modern parallel
computers.
This course is for graduate students who are interested in advanced topics
on multi-cores and multiprocessor systems.
In this course, we will cover both the software and the hardware aspects of
parallel architectures. We begin with an overview of the field by
focusing on the convergence of many diverse architectural approaches.
We will then discuss fundamental design issues, including: naming shared data,
synchronizing threads, and trading-off the latency and bandwidth associated
with communication. The following topics will be covered in this class:
Parallel Programming
Shared Memory Multiprocessors
Cache Coherence Protocol
Snoop-Based Multiprocessor
Scalable Distributed Memory
Interconnection Network Design
Latency Tolerance
Synchronization
Readings
Textbook:
David Culler, Jaswinder Singh and Anoop Gupta, "Parallel Computer Architecture: A Hardware/Software Approach", Mogran Kaufmann Publishers, 1999. (Errata and Corrections)
Various research papers: Available online from the course webpage.
Course Work and Grading
- Project:
One major component of this course is the project. The project must involve
some experimental components.
We encourage you to come up with your own topic for your project, although we
will post suggested projects to the class web page. Refer to the project page
for more information.
- Midterm:
There will be one 75-minute midterm. If you miss the midterm, you will obtain
a score of zero, unless you make prior arrangement with the instructor.
- Reading summaries:
Several classes are set aside entirely to discuss papers on active areas of
research and research papers will be assigned to these classes.
All students are expected to read the papers before class and select
two papers to write summaries for.
|
| Item | Fraction of Grade | |
| Reading Summaries | 10% |
| Project | 60% |
| Midterm | 30% |
|
Misc.
- Homework will be posted on the class homepage. You are encouraged to
work through these problems to prepare for the midterm. However, these
assignments will NOT be graded and do not need to be turned in.
- Late reports/summaries will receive a reduction of 15% of the
maximum possible score for each day they are late and no
submission is allowed after three days, except for
documented illnesses.
- You will have access to the IT workstation labs to use in your
final course project. See http://www.itlabs.umn.edu/ for information
about accessing
these workstations.
- The contents of this document and everything on the course webpage are subject
to change, and you are responsible for monitoring such changes.
- Incompletes are to be given only when a student who has successfully completed
all the summaries and the midterm and is unable to finish the final
project for reasons beyond their control such as a medical emergency.
- Cheating will not be tolerated.
|
