Welcome to 6.004!

Computation Structures

Spring 2021
 6.004 Course Staff

Instructors

Song Han Silvina Hanono Wachman Mengjia Yan

[email protected] [email protected] [email protected]

Teaching Assistants

Isabelle Miles Kendall Zack Soumya Sarah Catherine

Chong Dai Garner Pitcher Ram Spector Zeng

February 16, 2021 MIT 6.004 Spring 2021 L01-2

 Computing Devices Then…

ENIAC, 1943 30 tons, 200KW, ~1000 ops/sec

February 16, 2021 MIT 6.004 Spring 2021 L01-3

 Computing Devices Now

Typical 2020 laptop

1kg, 10W, 10 billion ops/s

February 16, 2021 MIT 6.004 Spring 2021 L01-4

 An Introduction to the Digital World

Computer programs
Virtual machines
Computer systems
Operating systems, virtual memory, I/O
Instruction set + memory
Computer architecture
Processors, caches, pipelining
Digital circuits
Digital design
Combinational and sequential circuits
Bits, Logic gates
Devices
Materials
Atoms
February 16, 2021 MIT 6.004 Spring 2021 L01-5
 The Power of Engineering
Abstractions
Good abstractions let us
reason about behavior while
shielding us from the details Virtual machines
of the implementation

Corollary: Implementation
technologies can evolve Instruction set + memory
while preserving the
engineering investment at
other levels Digital circuits

Leads to hierarchical design:

• Limited complexity at each
Bits, Logic gates
level ⇒ shorten design
time, easier to verify
• Reusable building blocks

February 16, 2021 MIT 6.004 Spring 2021 L01-6

 Course Outline
§ Module 1: Assembly language
§ From high-level programming languages
to the language of the computer

§ Module 2: Digital design

§ Combinational and sequential circuits

§ Module 3: Computer architecture

§ Simple and pipelined processors
§ Caches and the memory hierarchy

§ Module 4: Computer systems

§ Operating system and virtual memory
§ Parallelism and synchronization

February 16, 2021 MIT 6.004 Spring 2021 L01-7

 Our Focus: Programmable
General-Purpose Processors
§ Microprocessors are the basic building block of
computer systems
§ Understanding them is crucial even if you do not plan to
work as a hardware designer
§ Microprocessors are the most sophisticated digital
systems that exist today
§ Understanding them will help you design all kinds of
hardware

By the end of the term you will have

designed a simple processor from scratch!

February 16, 2021 MIT 6.004 Spring 2021 L01-8

 We Rely on Modern Design Tools
§ We will use RISC-V, a simple
and modern instruction set

Minispec
source
§ We will design hardware
using Minispec, a new Minispec compiler
hardware description
Bluespec yosys
language built for 6.004 compiler synthesis
§ Based on Bluespec,
but heavily simplified
Compiled
Simulation circuit
(gates)

February 16, 2021 MIT 6.004 Spring 2021 L01-9

 Course Mechanics
§ 2 lectures/week: slides, videos, and reference
materials on website
§ 2 recitations/week: work through tutorial problems
using skills and concepts from previous lecture

§ 7 mandatory lab exercises

§ Online submission + virtual check-off meetings
§ Due throughout the term (7 free late days meant to give
you flexibility, cover short illnesses, etc.; see website)
§ One optional open-ended design project
§ Due at the end of the term
§ 3 quizzes: Mar 18, Apr 22, May 13 (7:30-9:30pm
ET)
§ If you have a conflict, contact us to schedule a makeup
February 16, 2021 MIT 6.004 Spring 2021 L01-10
 Lectures
§ Lectures are synchronous, through Zoom
§ Video recordings will be available on the website

§ Two ways of asking questions:

§ Unmute yourself and ask – for questions specific to lecture
§ Zoom chat – for relevant but less direct questions

§ If you can, please enable video J

§ Helps you stay engaged, helps us get to know you
and get nonverbal feedback like in an in-person lecture
§ Your video won’t appear on recordings, and recordings
won’t be publicly available

February 16, 2021 MIT 6.004 Spring 2021 L01-11

 Recitations
§ 12 recitation sections on Wed & Fri
§ If you have a conflict with your assigned section,
or prefer the 9pm section, let us know via Piazza

§ Recitations
§ Solve tutorial problems with instructor or in small groups
§ You must watch the preceding lecture and attempt
a few short post-lecture problems before each recitation

§ Recitation participation is worth 5% of your grade

§ Everyone has 4 excused absences

§ You may opt out of recitation participation by Mar

19
§ If you opt out, that 5% of grade will come from quizzes
§ We believe recitations are the right choice for most of you
February 16, 2021 MIT 6.004 Spring 2021 L01-12
 Grading
§ 80 points from labs,
20 points from design project,
90 points from quizzes,
10 points from recitation participation

§ Fixed grade cutoffs:

§ A: Points >= 165
§ B: Points >= 145
§ C: Points >= 125
§ F: Points < 125 or not all labs complete

February 16, 2021 MIT 6.004 Spring 2021 L01-13

 Online Resources
§ The course website has up-to-date information,
handouts, and references to supplemental reading:
http://6004.mit.edu

§ We use Piazza extensively

§ Fastest way to get your questions answered
§ All course announcements are made on Piazza

§ We will hold regular virtual office hours to help you

with labs, infrastructure, and any other questions
§ Schedule and lab queue are available on the website

February 16, 2021 MIT 6.004 Spring 2021 L01-14

 Binary Number Encoding
and Arithmetic

February 16, 2021 MIT 6.004 Spring 2021 L01-15

 Digital Information Encoding
§ Digital systems represent and process information
using discrete symbols or digits

Voltage
1
§ These are typically binary digits
(bits): 0 and 1 0
Time

§ We can implement operations like +, >, AND, etc.

on binary numbers in hardware very efficiently

February 16, 2021 MIT 6.004 Spring 2021 L01-16

 Encoding Positive Integers
It is straightforward to encode positive integers as
a sequence of bits. Each bit is assigned a weight.
Ordered from right to left, these weights are
increasing powers of 2. The value of an N-bit
number is given by the following formula:
%&' 211210 29 28 27 26 25 24 23 22 21 20
𝑣 = # 2" 𝑏" 0 1 1 1 1 1 0 1 0 0 0 0
"#$

What value does 011111010000 encode?

V = 0*211 + 1*210 + 1*29 + …
= 1024 + 512 + 256 +128 + 64 + 16 = 2000

Smallest number? 0 Largest number? 2N-1

February 16, 2021 MIT 6.004 Spring 2021 L01-17
 Hexadecimal Notation
Long strings of bits are tedious and error-prone to transcribe,
so we often use a higher-radix notation, choosing the radix so
that it’s simple to recover the original bit string.

A popular choice is to transcribe numbers in base-16, called

hexadecimal. Each group of 4 adjacent bits is represented as a
single hexadecimal digit.

Hexadecimal - base 16 211210 29 28 27 26 25 24 23 22 21 20

0000 - 0 1000 - 8
0001 - 1 1001 - 9 0 1 1 1 1 1 0 1 0 0 0 0
0010 - 2 1010 - A
0011 - 3 1011 - B
0100 - 4 1100 - C
0101 - 5 1101 - D 7 D 0
0110 - 6 1110 - E
0111 - 7 1111 - F
0b011111010000 = 0x7D0

February 16, 2021 MIT 6.004 Spring 2021 L01-18

 Binary Addition and Subtraction
§ Addition and subtraction in base 2 are performed
just like in base 10
Base 10 Base 2
carry
1 11 1 What does this mean?
14 1110 -23 + 0b110
+ 7 + 111 We need a way to
represent negative
21 1 0 10 1 numbers!
borrow
-1 -1-1-1 -1
14 1110 011
– 7 – 111 – 101
07 0 11 1 110
???
February 16, 2021 MIT 6.004 Spring 2021 L01-19
 Binary Modular Arithmetic
§ If we use a fixed number of bits, addition and other
operations may produce results outside the range that
the output can represent (up to 1 extra bit for addition)
§ This is known as an overflow
§ Common approach: Ignore the extra bit
§ Gives rise to modular arithmetic: With N-bit numbers, equivalent
to following all operations with mod 2N
§ Visually, numbers “wrap around”:
Example: (3 – 5) mod 23 ?
0
000 0
7 1
7 000 1
111 001
111 001
6 + 2
6 2
110 – 010
110 010
5 3
5 3
101 4 011
101 4 011
100
100
February 16, 2021 MIT 6.004 Spring 2021 L01-20
 Encoding Negative Integers
Attempt #1: Use a sign-magnitude representation for
decimal numbers, encoding the sign of the number
(using “+” and “-”) separately from its magnitude (using
decimal digits).
We could use the same approach for binary
representations:

1 1 1 1 1 1 0 1 0 0 0 0
“0” for “+”
“1” for “-”
=-2000

What issues might this encoding have?

Two representations for 0 (+0, -0)
Addition and subtraction use different algorithms
and are more complex than with unsigned numbers
February 16, 2021 MIT 6.004 Spring 2021 L01-21
 Deriving a Better Encoding

Can you simply relabel some of the digits to

represent negative numbers while retaining the
nice properties of modular arithmetic?
Yes!
0
-1 7 000 1
111 001

-2 6 2
110 010
-3 5 3
101 -4 4 011
100 This is called two’s
complement encoding
February 16, 2021 MIT 6.004 Spring 2021 L01-22
 Two’s Complement Encoding
In two’s complement encoding, the high-order bit of
the N-bit representation has negative weight:
N bits
!"'
𝑣 = −2!"#𝑏!"# + ' 2$ 𝑏$ -2N-1 2N-2 … … … 23 22 21 20
$%&

• Negative numbers have “1” in the high-order bit

• Most negative number? 10…0000 -2N-1
• Most positive number? 01…1111 +2N-1 - 1
• If all bits are 1? 11…1111 -1

February 16, 2021 MIT 6.004 Spring 2021 L01-23

 Two’s Complement and Arithmetic
§ To negate a number (i.e., compute –A given A), we
invert all the bits and add one
Why does this work?
-A + A = 0 = -1 + 1
-A = (-1 – A) + 1

1 ...1 1
- An-1...A1A0
An-1...A1A0

§ To compute A-B, we can simply use addition and

compute A+(-B)
§ Result: Same circuit can add and subtract!
February 16, 2021 MIT 6.004 Spring 2021 L01-24
 Two’s Complement Example
Compute 3 – 6 using 4-bit 2’s complement addition
§ 3: 0011 1
§ 6: 0110
0011
+ 1010
§ -6: 1010
1 10 1
Compute 3 – 2 using 3-bit 2’s complement addition
§ 3: 011 What does this 1 mean?
11
§ 2: 010 011 0
-1 000 1
§ -2: 110 + 110 111
-2
001
2
110 010
1 00 1 -3 3
101 -4 011
Keep only last 3 bits 100
Zero crossing
February 16, 2021 MIT 6.004 Spring 2021 L01-25
 Summary
§ Digital systems encode information using binary for
efficiency and reliability

§ We can encode unsigned integers using strings of bits;

long addition and subtraction are done as in decimal

§ Two’s complement allows encoding negative integers

while preserving the simplicity of unsigned arithmetic

February 16, 2021 MIT 6.004 Spring 2021 L01-26

 Post-Lecture Questions
6004.mit.edu > Material > Lectures > L1 questions

February 16, 2021 MIT 6.004 Spring 2021 L01-27

 Thank you!

Next lecture:
Introduction to assembly language
and RISC-V

February 16, 2021 MIT 6.004 Spring 2021 L01-28