Computer Science 252.
Graduate Computer Architecture.
(4 units)

Fall 1996

Announcements

14-Dec

Exam #2 is graded and available outside of Dave's door, as well as solutions. Please take this opportunity to pick up any other remaining items (homeworks) of yours that are present.

Catalog Description

Three hours of lecture and one hour of discussion per week. Prerequisites: CS 152. Graduate survey of contemporary computer organizations covering: early systems, CPU design, instruction sets, control, processors, busses, ALU, memory, pipelined computers, multiprocessors, and case studies. Term paper or project required.

Expanded Description

This course focuses on the techniques of quantitative analysis and evaluation of modern computing systems, such as the selection of appropriate benchmarks to reveal and compare the performance of alternative design choices in system design. The emphasis is on the major component subsystems of high performance computers: pipelining, instruction level parallelism, memory hierarchies, input/output, and network-oriented interconnections. Students will undertake a major computing system analysis and design project of their own choosing.

Course Grading

10% Class Participation
30% Homeworks (work in pairs)
30% Examinations (2 Midterms)
30% Research Project (work in pairs)

Instructors, Fall 1996

Lecturer: Professor David Patterson
Teaching Assistant: Richard Fromm

Location

Lecture: Wed, Fri 12:30 - 2 PM, 203 McLaughlin Hall
Discussion: No regular discussion meetings until further notice.
Instead, office hours will be available during discussion times, in 479 Soda.
The scheduled rooms for discussion are still listed below for reference (in case we have discussion again), but they are generally no longer applicable.
Tues 2 - 3 PM, 247 Cory Hall; Thurs 3 - 4 PM, 3111 Etcheverry Hall

Roster

Click here to see who's who.

Communication

The class mailing list is cs252@cs.berkeley.edu
The class newsgroup is ucb.class.cs252

Textbook

J. L. Hennessy and D. A. Patterson, Computer Architecture: A Quantitative Approach, 2nd Edition, Morgan Kaufmann Publishing Co., Menlo Park, CA.

The errata sheet is finally available!!!

Note that the 2nd edition is significantly different than the 1st edition, and it is not recommended that you attempt to use the 1st edition as a textbook for this course.

Course Projects

XSPECulation: Benchmarks for the Future , David Andre and Nancy Chang
Comparison of Reed Solomon Codec Implementations , Heather Bowers and Hui Zhang
The Performance of Real Applications and Operating Systems on Simple IRAM Architectures , Wangeci Bowman, Neal Cardwel, and Cynthia Romer
FLECKmarks: Measuring Floating Point Performance using a Full IEEE Compliant Arithmetic BenchmarK , Joseph Darcy and David Gay
IDEA as a Benchmark for Reconfigurable Computing , Eylon Caspi and Nicholas Weaver
Compressed DRAM , Khang Dao
Java Cache Behavior: Cache performance of Java Interpreters and Just In Time compilers , Jeff Foster and Charles Zapata
an eXperimental SPEClike CPU benchmark , Joseph Gebis and Tajh Taylor
Analysis of an SRAM Circuit in a DRAM Process , Ben Gribstad and Jason Golbus
A Case Study of Compression on a Gate Array + Risc Processor Architecture , David Gibson and William Tsu
Evaluation and Comparison of Existing Cache Designs Implemented as an IRAM , Christoforos Kozyrakis and Helen Wang
On the Fly Cache Compression and Decompression , Eric Kusse and Adrian Isles
Design and Evaluation of a Voting Data Prefetch Engine , Gurmeet Singh Manku and Mukul Prasad
Cache Latency and Out of Order Execution , David Martin and Jordan Smith
Towards a Scalable Z-Buffer Benchmark and FBRAM Performance Measurements , David Parks and Yizhou Yu
Programmable Prefetch Unit , Roberto Passerone and Suet Fei Li
Microprocessor Performance Degradation in a DRAM Process , Stephen Tang and Kevin Wang
HYPE: A Hybrid Data Prefetch Engine , Carter Wendelken and Hanna Pasula

Course Schedule

Lecture notes will be available in both postscript and pdf formats. Postscript files are good for printing, but pdf files are usually better for viewing on the screen. Note that sometimes ghostview has trouble properly displaying postscript files, although they often print okay even when that occurs. Every effort will be made to get the notes on the web prior to the lecture. Note, however, that the notes may be updated slightly following the lecture. Click here for instructions regarding how to view pdf files.

Week Lect. Date Day Lecture Notes Due Chptr(s)
1 1 28-Aug Wed Review: Performance, Cost, DLX Instruction Set ps, pdf 1, 2
2 30-Aug Fri Review: Pipeline, Cache ps, pdf 2, 3
2 4-Sep Wed Prerequisite Quiz
3 6-Sep Fri ILP, loop unrolling, scoreboard ps, pdf 4
3 4 11-Sep Wed ILP, Tomasulo , Dynamic Branch Prediction ps, pdf 4
5 13-Sep Fri ILP, trace scheduling, Speculation ps, pdf 4
4 16-Sep Mon HW #1 due by 5:00 PM HW #1
6 18-Sep Wed Vector processors ps, pdf Apndx. B
7 20-Sep Fri Memory Hierachy: Caches ps, pdf 5
5 23-Sep Mon Pick project partner and choice, submit by e-mail Proj. choice
8 25-Sep Wed Array Processing and Reconfigurable Computing
(André DeHon)		 ps
9 27-Sep Fri Caches and Main Memory/Main Memory ps, pdf 5
6 30-Sep Mon HW #2 due by 5:00 PM HW #2
10 2-Oct Wed IRAM/How to do (Architecture) Research in the Electronic Age ps, pdf Meeting signup
3-Oct Thurs Project Survey due by 5:00 PM Proj. Survey
11 4-Oct Fri Memory Hierarchy Exmaple and Introduction to I/O ps, pdf Proj. Meetings 5/6
7 9-Oct Wed Quiz 1 (5:45PM - 8:45PM, 306 Soda)
Optional review during lecture
12 11-Oct Fri I/O 1: Metrics, Queuing theory, Busses ps, pdf 6
8 13 16-Oct Wed I/O 2: Busses and RAID ps, pdf 6
14 18-Oct Fri I/O 3: Tertiary Storage and Network Intro ps, pdf 6
9 21-Oct Mon HW #3 due by 5:00 PM HW #3
15 23-Oct Wed Networks 2: Interface, Switches, Routing, Examples ps, pdf Proj. Survey #2 7
16 25-Oct Fri Networks 3: Internetworking, Examples ps, pdf 7
10 30-Oct Wed Visitor (no slides) : 7
17 1-Nov Fri Multiprocessor 1 Introduction: ps, pdf 8
11 4-Nov Mon HW #4 due by 5:00 PM HW #4
18 6-Nov Wed Multiprocessor 2 Snooping vs. Directory: ps, pdf Proj. Checkpoint 8
20 8-Nov Fri Networks Of Workstations
(Remzi Arpaci) 		ps,
12 13-Nov Wed Multiprocessor 3: ps, pdf 8
21 15-Nov Fri 20 minute Project Review
(no lecture)
13 20-Nov Wed Quiz 2 (5PM - 8PM, 306 Soda)
Optional review during lecture
22 22-Nov Fri How to Have a Bad Academic Career ps, pdf
14 23 27-Nov Wed Video: Sun Architecture Futures, Gordon Bell on Computer History
29-Nov Fri Vacation
15 4-Dec Wed Poster Session (11AM - 3PM)
6th floor Soda Corridor
(no lecture)
25 6-Dec Fri Final Lecture (12:30-1:30)/
End of class celebration at LaVal's (1:30-3:00)		 ps, pdf
16 9-Dec Mon Project Papers/URLs due by 4:30 PM

Handouts

Homework 1
Project Suggestions
Homework 2
Resources at your disposal for project work
Homework 3

Related Course Pages

This course is part of CalVIEW, Video Instruction for the Engineering World, which supports UC Berkeley's activity as a member school in the National Technological University, NTU, which is a consortium of 47 universities and colleges.

Fall 1996

Spring 1996

Fall 1995

Fall 1994

Other Useful Links

This page is located at http://http.cs.berkeley.edu/~patterson/252/
Last Modified, 14-Dec-1996