CS 61C: Great Ideas in Computer Architecture Course Introduction, Number Representation

Instructor: Justin Hsia

6/24/2013 Summer 2013 -- Lecture #1 1

Introducing Your Instructor

Im not a professor call me Justin Upbringing: Born and raised in the Bay Education: I bleed Blue and Gold (Go Bears!)
B.S. ME, B.S. EECS, M.S. EECS Ph.D. EECS (expected 2014)

Research: Synthetic Biology Teaching: EE128, EE40, CS61C (Su11, Su12, Sp13) Interests:
6/24/2013 Summer 2013 -- Lecture #1 2

Agenda
Course Overview
Six Great Ideas in Computer Architecture

Administrivia Number Representation

Number Bases Signed Representations Overflow Sign Extension
6/24/2013 Summer 2013 -- Lecture #1 3

What is CS 61C about?

It is about the hardware-software interface
What does the programmer need to know to achieve the highest possible performance?

Use low-level programming languages (closer to underlying hardware)

Allows us to talk about key hardware features in higher-level terms Allows programmers to harness underlying hardware parallelism for high performance
6/24/2013 Summer 2013 -- Lecture #1 4

Machine Structures
Application (ex: browser) Compiler Operating System (Mac OSX) I/O system

CS 61C
Instruction Set Architecture

Software Hardware

Assembler Processor

Memory

Datapath & Control Digital Design Circuit Design Transistors

6/24/2013

Summer 2013 -- Lecture #1

New-School: Parallelism
Software Parallel Requests
Assigned to computer e.g. search Garcia Assigned to core e.g. lookup, ads

Hardware
Warehouse Scale Computer Smart Phone

Parallel Threads Parallelism &

Achieve High Performance
Core Computer Memory Core Instruction Unit(s) Input/Output Functional Unit(s)
A0+B0 A1+B1 A2+B2 A3+B3

Leverage

Parallel Instructions
> 1 instruction @ one time e.g. 5 pipelined instructions

Core

Parallel Data
> 1 data item @ one time e.g. add of 4 pairs of words

Hardware descriptions
All gates functioning in parallel at same time
6/24/2013 Summer 2013 -- Lecture #1

Cache Memory

Logic Gates
6

Post-PC Era: Late 2000s - ???

Personal Mobile Devices (PMD):

Enabling Tech: Wireless networking, smartphones Big Players: Apple, Nokia, Cost: $500, Target: Consumers on the go Using: Objective C, Android OS
6/24/2013 Summer 2013 -- Lecture #1 7

iPhone Innards
I/O

Processor

I/O

Memory You will learn about multiple processors, the memory hierarchy, and I/O in 61C
I/O
6/24/2013 Summer 2013 -- Lecture #1 8

Post-PC Era: Late 2000s - ???

Cloud Computing:

Enabling Tech: Local Area Networks, broadband Internet Big Players: Amazon, Google, Target: Transient users or users who cannot afford highend equipment
6/24/2013 Summer 2013 -- Lecture #1 9

Post-PC Era: Late 2000s - ???

Datacenters and Warehouse Scale Computers (WSC):

Enabling Tech: Local Area Networks, cheap servers Cost: $200M clusters + maintenance costs Target: Internet services and PMDs Usages: MapReduce, Ruby on Rails
6/24/2013 Summer 2013 -- Lecture #1 10

Six Great Ideas in Computer Architecture

1. Layers of Representation/Interpretation 2. Technology Trends 3. Principle of Locality/Memory Hierarchy 4. Parallelism 5. Performance Measurement & Improvement 6. Dependability via Redundancy
6/24/2013 Summer 2013 -- Lecture #1 11

Great Idea #1: Levels of Representation/Interpretation

Higher-Level Language Program (e.g. C) Compiler Assembly Language Program (e.g. MIPS) Assembler Machine Language Program (MIPS) Machine Interpretation temp = v[k]; v[k] = v[k+1]; v[k+1] = temp;
lw lw sw sw $t0, 0($2) $t1, 4($2) $t1, 0($2) $t0, 4($2)
1001 1111 0110 1000

Anything can be represented as a number!

0000 1010 1100 0101 1100 0101 1010 0000 0110 1000 1111 1001 1010 0000 0101 1100

1111 1001 1000 0110

0101 1100 0000 1010

1000 0110 1001 1111

e.g. words, colors, data, logic, Hardware Architecture Description (e.g. block diagrams) code, even other numbers
Architecture Implementation Logic Circuit Description (Circuit Schematic Diagrams)
6/24/2013 Summer 2013 -- Lecture #1 12

Great Idea #2: Technology Trends

# of transistors on an integrated circuit (IC)

Predicts: Transistor count per chip doubles every 2 years

Gordon Moore Intel Cofounder B.S. Cal 1950 Year:

6/24/2013

Summer 2013 -- Lecture #1

13

Great Idea #3: Principle of Locality/ Memory Hierarchy

Increasing distance from application/user Trade-off in speed and cost vs. capacity!

6/24/2013

Summer 2013 -- Lecture #1

14

Storage Latency Analogy: How Far Away is the Data?

10 9

Andromeda Tape /Optical Robot

Pluto

2,000 Years
Jim Gray Turing Award B.S. Cal 1966 2 Years Ph.D. Cal 1969!

10

6 Disk

100

Memory

Sacramento
This Campus This Room My Head

1.5 hr 10 min 1 min

10 On Board Cache 2 On Chip Cache 1 Registers (ns)

Great Idea #4: Parallelism

6/24/2013

Summer 2013 -- Lecture #1

16

Great Idea #5: Performance Measurement and Improvement

Allows direct comparisons of architectures and quantification of improvements Most common measures are time to finish (latency) and rate of execution (throughput)
Latency includes both setup and execution time

Match application and hardware to exploit:

Locality, parallelism, special hardware features
6/24/2013 Summer 2013 -- Lecture #1 17

Great Idea #6: Dependability via Redundancy

Redundancy so that a failing piece doesnt make the whole system fail
1+1=2
2 of 3 agree

1+1=2
6/24/2013

1+1=2
Summer 2013 -- Lecture #1

1+1=1

FAIL!

18

Great Idea #6: Dependability via Redundancy

Applies to everything from datacenters to storage to memory
Redundant datacenters so that can lose 1 datacenter but Internet service stays online Redundant disks so that can lose 1 disk but not lose data (Redundant Arrays of Independent Disks/RAID) Redundant memory bits of so that can lose 1 bit but no data (Error Correcting Code/ECC Memory)

Increasing transistor density reduces the cost of redundancy

6/24/2013 Summer 2013 -- Lecture #1 19

Agenda
Course Overview
Six Great Ideas in Computer Architecture

Administrivia Number Representation

Number Bases Overflow Signed Representations Sign Extension
6/24/2013 Summer 2013 -- Lecture #1 20

Course Information
This information can also be found on the course syllabus Website: http://inst.eecs.berkeley.edu/~cs61c/su13 Instructor: Justin Hsia Teaching Assistants: Albert Magyar, Jeffrey Dong, Justin Fu, Kevin Yeun, Sagar Karandikar, Shaun Benjamin Textbooks: average 15 pages of reading/week
Patterson & Hennessey, Computer Organization and Design, Revised 4th Edition (not 3rd Edition, not Asian 4th Edition) Kernighan & Ritchie, The C Programming Language, 2nd Edition Barroso & Holzle, The Datacenter as a Computer, 1st Edition (free!)

Piazza (http://piazza.com/class#summer2013/cs61c) is the class forum

Every announcement, discussion, clarification happens there
6/24/2013 Summer 2013 -- Lecture #1 21

Course Assignments and Grading

Homework (15%) 5 total, weighted by difficulty Labs (6%) 12 total, weighted equally
Done in partners, good learning environment

Projects (24%) 3 total, weighted equally

1. MIPS Instruction Set Simulator 2. Performance Tuning of a Parallel Application 3. Computer Processor Pipelined Design

Midterm (24%): Friday, July 19 @ 9am Final (26%): Friday, August 16 @ 9am Participation and Altruism (5%)
6/24/2013 Summer 2013 -- Lecture #1 22

Projects
Software Parallel Requests
Assigned to computer e.g. search Garcia Assigned to core e.g. lookup, ads

MapReduce Smart Phone

Hardware
Warehouse Scale Computer

Parallel Threads Parallelism &

Achieve High Performance
Core Computer Memory Core Instruction Unit(s) Input/Output Functional Unit(s)
A0+B0 A1+B1 A2+B2 A3+B3

Leverage

Parallel Instructions
>1 instruction @ one time e.g. 5 pipelined instructions

Core

Parallel Data
> 1 data item @ one time e.g. add of 4 pairs of words

Hardware descriptions
All gates functioning in parallel at same time
6/24/2013 Summer 2013 -- Lecture #1

Project 2: Matrix Multiply Logic Gates

Cache Memory Project 3: CPU Design

23

Participation and Altruism

Participation
Asking great questions in discussion and lecture and making it more interactive Attending office hours, completing all assignments Peer instruction voting (more on this next)

Altruism
Helping others in lab, discussion, and on Piazza

This is a subjective score (internal)

Reward for helping, not a penalty for not helping The point is to encourage class-wide learning!
6/24/2013 Summer 2013 -- Lecture #1 24

Peer Instruction
Increase real-time learning in lecture, test understanding of concepts vs. details
mazur-www.harvard.edu/education/pi.phtml

Multiple choice question at end of a segment

1 minute to decide yourself 2 minutes in pairs to reach consensus Learn by teaching!

Vote using i>clicker remote

Can buy from bookstore or online Will start using tomorrow Counts towards P&A grade
6/24/2013 Summer 2013 -- Lecture #1 25

Question: Which statement is FALSE about Great Ideas in Computer Architecture?

(A) To offer a dependable system, you must use

components that almost never fail (B) Memory hierarchy goal is to look as fast as most expensive memory, as big as cheapest (C) Moores Law means computers will continue to get put twice as many transistors/chip every 2 years without fail
(D)

26

Late Policy Slip Days

Assignments due at 11:59:59pm (time-stamped)

You have 3 slip day tokens

Token used for every day your project or homework is late (even by a second) Keep in mind how much each assignment is worth

After tokens, its 33% deducted per day

No credit if more than 2 days late Cannot retroactively reassign slip days
6/24/2013 Summer 2013 -- Lecture #1 27

Policy on Assignments and Independent Work

With the exception of Project 2, all homework and projects are to be YOUR work and your work ALONE. You are encouraged to discuss your assignments with other students (ideas), but we expect that what you hand in is yours. It is NOT acceptable to copy solutions from other students. It is NOT acceptable to copy (or start your) solutions from the Web.
6/24/2013 Summer 2013 -- Lecture #1 28

Policy on Assignments and Independent Work

We have tools and methods, developed over many years, for detecting this. You WILL be caught, and the penalties WILL be severe. Both the cheater and the enabler receive 100% for the assignment. Letter to your university record documenting the incidence of cheating.
Possible automatic F in the course

People are caught every semester of 61C

6/24/2013 Summer 2013 -- Lecture #1 29

Comments on Summer Version

Summer is EXTREMELY hectic!
Double the standard pace Less time to synthesize material Falling behind just a little can be fatal

No MapReduce project Starts deceptively slowly (first two weeks)

If the course begins to overwhelm you, dont wait, contact me or your TA immediately!
6/24/2013 Summer 2013 -- Lecture #1 30

Pedagogic Comments
Deep learning does not happen in lecture
Learn by doing: labs, discussions, and assignments

Engaging the material outside of class and lab is critical to your success
Study groups, testing out your own questions Talking with others helps solidify your own understanding

You learn best from your mistakes

Dont be afraid to be wrong; only you lose if you remain silent
6/24/2013 Summer 2013 -- Lecture #1 31

Architecture of a Lecture
Full

Attention

Administrivia Tech break + stretch break + GTKYS

Summary + Bonus

20 25

50 53

78 80

Time (minutes)
6/24/2013 Summer 2013 -- Lecture #1 32

Last Things
Discussions, labs, and OHs start immediately
Yes, that means today! Switching sections: if you find another 61C student willing to swap lab, talk to your TAs Attend whichever discussion you want, as long as there is enough physical room

HW0 due this Tuesday, June 25th

Find a small digital image of yourself

HW1 due this Sunday, June 30th

6/24/2013 Summer 2013 -- Lecture #1 33

Get To Know Your Staff

Category: Cal

6/24/2013

Summer 2013 -- Lecture #1

34

Agenda
Course Overview
Six Great Ideas in Computer Architecture

Administrivia Number Representation

Number Bases Signed Representations Overflow Sign Extension
6/24/2013 Summer 2013 -- Lecture #1 35

Number Representation
Numbers are an abstract concept!
Recognize that these are arbitrary symbols

Inside a computer, everything stored as a sequence of 0s and 1s (bits)

Even this is an abstraction!

How do we represent numbers in this format?

Lets start with integers

6/24/2013

Summer 2013 -- Lecture #1

36

Number Bases
Key terminology: digit (d) and base (B) Value of i-th digit is dBi where i starts at 0 and increases from right to left
n digit number dn-1dn-2 ... d1d0 value = dn-1Bn-1 + dn-2Bn-2 + ... + d1B1 + d0B0

In base B, each digit is one of B possible symbols Base is indicated using either a prefix or a subscript
6/24/2013 Summer 2013 -- Lecture #1 37

Commonly Used Number Bases

Decimal (base 10)
Symbols: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 Notation: 9472ten = 9472

Binary (base 2)
Symbols: 0, 1 Notation: 101011two = 0b101011

Hexadecimal (base 16)

Symbols: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, A, B, C, D, E, F Notation: 2A5Dhex = 0x2A5D
6/24/2013 Summer 2013 -- Lecture #1 38

Number Base Examples

Examples:
9472ten = 9000 + 400 + 70 +2 9x1000 + 4x100 + 7x10 + 2x1 9x103 + 4x102 + 7x101 + 2x100
9472ten = 2x163 + 5x162 + 0x161 + 0x160 9472ten = 2500hex
3210

0xA15 = 0b 1010 0001 0101

6/24/2013 Summer 2013 -- Lecture #1 39

Bits Can Represent Anything

n digits in base B can represent at most Bn things!
Each of the n digits is one of B possible symbols Have more things? Add more digits!

Example: Logical values (1 bit) 0 is False, 1 is True Example: Characters

26 letters require 5 bits (25 = 32 > 26)

Example: Students in this class (8 bits)

For convenience, can group into nibbles (4 bits) and bytes (8 bits)
6/24/2013 Summer 2013 -- Lecture #1 40

So What Number Is It Anyway?

Here we are using binary bit patterns to represent numbers
Strictly speaking they are called numerals and have no meaning until you interpret them Is CAB a word (taxi) or a number (3243ten)? Is 0x999999 a number or a color (RGB)?

Keep in mind that the same bit pattern will mean different things depending on how you choose to interpret it
6/24/2013 Summer 2013 -- Lecture #1 41

Numbers in a Computer (1/2)

The number line:

Alt text: The Wikipedia page "List of Numbers" opens with "This list is incomplete; you can help by expanding it. (Source: http://xkcd.com/899)

What do we choose to represent?

6/24/2013 Summer 2013 -- Lecture #1 42

Numbers in a Computer (2/2)

Numbers really have digits, but hardware can only store a finite number of them (fixed)
Usually ignore leading zeros Leftmost is most significant bit (MSB) Rightmost is least significant bit (LSB)

Circle of binary numerals:

(3-bit example)

6/24/2013

Summer 2013 -- Lecture #1

43

Unsigned Integers
Represent only non-negative (unsigned) integers:

0000two = 0ten 0001two = 1ten ... 0110two = 6ten 0111two = 7ten 1000two = 8ten 1001two = 9ten ... 1110two = 14ten 1111two = 15ten
6/24/2013

Zero? 00two = 0ten

Most neg number? 00two = 0ten Most pos number? 11two = (2n-1)ten Increment?
Summer 2013 -- Lecture #1 44

Signed Integers
n bits can represent 2n different things
Ideally, want the range evenly split between positive and negative

How do we want to encode zero? Can we encode them in such a way that we can use the same hardware regardless of whether the numbers are signed or unsigned?
e.g. incrementing pos/neg numbers
6/24/2013 Summer 2013 -- Lecture #1 45

Sign and Magnitude

MSB gives sign, rest treated as unsigned (magnitude):

0000two = +0ten 0001two = +1ten ... 0110two = +6ten 0111two = +7ten 1000two = -0ten 1001two = -1ten ... 1110two = -6ten 1111two = -7ten
6/24/2013

Zero? 00two and 100two = 0ten

Most neg number? 11two = -(2(n-1)-1)ten Most pos number? 011two = (2(n-1)-1)ten Increment?
Summer 2013 -- Lecture #1 46

Biased Notation
Like unsigned, but shifted so zero is in the middle:

0000two = -7ten 0001two = -6ten ... 0110two = -1ten 0111two = 0ten 1000two = +1ten 1001two = +2ten ... 1110two = +7ten 1111two = +8ten
6/24/2013

Zero? 011two = 0ten

Most neg number? 00two = -(2(n-1)-1)ten Most pos number? 11two = 2(n-1)ten Increment?
Summer 2013 -- Lecture #1 47

Ones Complement
New negation procedure complement the bits:

0000two = +0ten 0001two = +1ten ... 0110two = +6ten 0111two = +7ten 1000two = -7ten 1001two = -6ten ... 1110two = -1ten 1111two = -0ten
6/24/2013

Zero? 00two and 11two = 0ten

Most neg number? 100two = -(2(n-1)-1)ten Most pos number? 011two = (2(n-1)-1)ten Increment?
Summer 2013 -- Lecture #1 48

Twos Complement
Like Ones Complement, but shift negative #s by 1:

0000two = 0ten 0001two = +1ten ... 0110two = +6ten 0111two = +7ten 1000two = -8ten 1001two = -7ten ... 1110two = -2ten 1111two = -1ten
6/24/2013

Zero? 00two = 0ten

Most neg number? 100two = -2(n-1)ten Most pos number? 011two = (2(n-1)-1)ten Increment?
Summer 2013 -- Lecture #1 49

Twos Complement Summary

Used by all modern hardware

Roughly evenly split between positive and negative

One more negative # because positive side has 0

Can still use MSB as sign bit To negate: Flip the bits and add one
Example: +7 = 0b 0000 0111, -7 = 0b 1111 1001
6/24/2013 Summer 2013 -- Lecture #1 50

Twos Complement Review

Suppose we had 5 bits. What integers can be represented in twos complement? (A) -31 to +31 need 6 bits (B) -15_to +15 ones complement (C) -0_to +31 unsigned (D) -16_to +15 twos complement (E) -32 to +31 need 6 bits
6/24/2013 Summer 2013 -- Lecture #1 51

Overflow
Overflow is when the result of an arithmetic operation cant be represented by the hardware bits
i.e. the result is mathematically incorrect

Examples:
Unsigned: Twos: 0b11 + 110 = 0b00 = 0? 0b011 + 110 = 0b100 = -2(n-1)ten?

0b 0000

0001

0010 unsigned

1110

1111

6/24/2013

Summer 2013 -- Lecture #1

52

Sign Extension
Want to represent the same number using more bits than before
Easy for positive #s (add leading 0s), more complicated for negative #s Sign and magnitude: add 0s after the sign bit Ones complement: copy MSB Twos complement: copy MSB

Example:
Sign and magnitude: 0b 11 = 0b 1001 Ones/Twos complement: 0b 11 = 0b 1111
6/24/2013 Summer 2013 -- Lecture #1 53

Summary (1/2)
CS61C: Learn 6 Great Ideas in Computer Architecture to enable high performance programming via parallelism
1. 2. 3. 4. 5. 6.
6/24/2013

Layers of Representation/Interpretation Technology Trends Principle of Locality/Memory Hierarchy Parallelism Performance Measurement and Improvement Dependability via Redundancy
Summer 2013 -- Lecture #1 54

Summary (2/2)
Number Representation: How to represent positive and negative integers using binary
Unsigned: Interpret numeral in base 2 Signed: Twos Complement Biased: Subtract bias Sign extension must preserve signed number

6/24/2013

Summer 2013 -- Lecture #1

55