CS/ECE 3330Computer Architecture

Kim HazelwoodFall 2009

CS/ECE 3330 – Fall 2009

Computer Architecture IS MY RESEARCH AREA

My Background

PhD from Harvard in 2004

Post-Doc at Intel Massachusetts 2004-2005

Started at UVa in 2005

Consultant for Intel 2005-presentTelecommute - 1 day/week

On site – once per quarter

Other Industry ExperienceHP Cupertino CA

HP Labs Cambridge MA

IBM Research NY

About Your Instructor

2

CS/ECE 3330 – Fall 20093

What will we learn in this course?

The “interface” between hardware and software

– Good design requires an understanding of both

Microprocessor design from a high level– Architecture … not implementation

Software from a low level– Assembly, bit manipulation

Design considerations, metrics, and evaluation

How it all fits together

HW

SW

CS/ECE 3330 – Fall 20094

Course Materials

The Book: 4th Edition Patterson & Hennessy– 3rd edition won’t suffice– CD contains SW we’ll be using

Everything is on Collab– The schedule– Problem sets– Lab materials

Lectures– Slides on the course website– Attending class is in your best interest

CS/ECE 3330 – Fall 2009

Starts in a couple weeks• Ignore locations• Ignore enrollment limits

Five labs• Work in teams of 2• Will have 2 weeks for most labs• Can do some of it at home ahead of time

Tools•PCSpim – MIPS ISA simulator•SMOK – Machine organization simulator

• ALUs, register files, logic gates, …

The Lab

5

CS/ECE 3330 – Fall 20096

Required Work

Three Exams (55%)– Two tests in class (30%)

– Tentatively: 9/28 and 11/2– Cumulative exam (25%)

– Sat 12/12 @ 2PM

Problem Sets (20%)– About 7 assignments

Labs (25%)– 5 labs

CS/ECE 3330 – Fall 2009

What do you look for when buying a computer?

Class Survey

7

CS/ECE 3330 – Fall 2009

Classes of Computers

Desktop computers• General purpose, variety of software• Subject to cost/performance tradeoff

Server computers• Network based• High capacity, performance, reliability• Range from small servers to building sized

Embedded computers• Hidden as components of systems• Stringent power/performance/cost constraints

CS/ECE 3330 – Fall 2009

Intel’s 1.6 GHz Atom

CS/ECE 3330 – Fall 2009

How can you make your programs run faster?

Class Survey

10

CS/ECE 3330 – Fall 2009

Understanding Performance

Algorithm• Determines number of operations executed

Programming language, compiler, architecture• Determine number of machine instructions executed

per operation

Processor and memory system• Determine how fast instructions are executed

I/O system (including OS)• Determines how fast I/O operations are executed

CS/ECE 3330 – Fall 2009

Moore’s Law

But watch out for Gates' Law…

How Have We Been Doing?

12

CS/ECE 3330 – Fall 200913

Hicham El Guerrouj1999 3:43.13

Roger Bannister

1954 3:59.4

7% Improvement in 45 years

By Comparison: The Mile Run

CS/ECE 3330 – Fall 2009

A Successful Endeavor

14

1/1/

1913

1/1/

1917

1/1/

1921

1/1/

1925

1/1/

1929

1/1/

1933

1/1/

1937

1/1/

1941

1/1/

1945

1/1/

1949

1/1/

1953

1/1/

1957

1/1/

1961

1/1/

1965

1/1/

1969

1/1/

1973

1/1/

1977

1/1/

1981

1/1/

1985

1/1/

1989

1/1/

1993

1/1/

1997

1/1/

2001

1/1/

2005

1/1/

2009

0

10

20

30

40

50

60

70

80

World Record Hot Dog Eating (12 Minutes)

Hot

Dog

s

CS/ECE 3330 – Fall 2009

Application software• Written in high-level language

System software• Compiler: translates HLL code to

machine code• Operating System: service code

– Handling input/output– Managing memory and storage– Scheduling tasks & sharing

resources

Hardware• Processor, memory, I/O controllers

Below Your Program

CS/ECE 3330 – Fall 2009

Levels of Program Code

High-level language• Level of abstraction closer to

problem domain• Provides for productivity and

portability Assembly language• Textual representation of

instructionsHardware representation• Binary digits (bits)• Encoded instructions and

data

CS/ECE 3330 – Fall 2009

Opening the Box

CS/ECE 3330 – Fall 2009

Inside the Processor (CPU)

Datapath: performs operations on data

Control: sequences datapath, memory, ...

Cache memory• Small fast SRAM memory for immediate access to

data

CS/ECE 3330 – Fall 2009

Inside the Processor

AMD Barcelona: 4 processor cores

CS/ECE 3330 – Fall 2009

12 inch wafer of AMD Opteron X2 chips

Fabbing the Chips

20

CS/ECE 3330 – Fall 2009

How long do you think it takes to design and build a modern processor?

Processor Design Cycle

21

CS/ECE 3330 – Fall 2009

“The Multicore Challenge”

Power, temperature, and reliability

What’s New and Cool in the Field?

22

CS/ECE 3330 – Fall 2009

Heterogeneous systems• Special purpose processors on chip, e.g.

GPGPUs

Many big companies have an architecture team• Microsoft• Google• D.E. Shaw

What’s New and Cool in the Field?

23

CS/ECE 3330 – Fall 2009

Performance analysis

First problem set is on collab

Due one week from today at 2PM

Next Time …

24