cse240a: Graduate Computer Architecture Steven Swanson Hung-Wei - - PowerPoint PPT Presentation

▶
cse240a graduate computer architecture
SMART_READER_LITE
LIVE PREVIEW

cse240a: Graduate Computer Architecture Steven Swanson Hung-Wei - - PowerPoint PPT Presentation

Jul 07, 2023 103 likes •725 views

cse240a: Graduate Computer Architecture Steven Swanson Hung-Wei Tseng 1 Todays Agenda What is architecture? Why is it important? At the highest level, where is architecture today? Where is it going? Whats in this class?

slide-1
SLIDE 1

cse240a: Graduate Computer Architecture

Steven Swanson Hung-Wei Tseng

1
slide-2
SLIDE 2

Today’s Agenda

  • What is architecture?
  • Why is it important?
  • At the highest level, where is architecture today?

Where is it going?

  • What’s in this class?
2
slide-3
SLIDE 3

What is architecture?

  • How do you build a machine that computes?
  • Quickly, safely, cheaply, efficiently, in technology X, for

application Y, etc.

Civilization advances by extending the number of important

  • perations which we can perform

without thinking about them.

  • - Alfred North Whitehead
slide-4
SLIDE 4

Orientation

The internet

slide-5
SLIDE 5

Orientation

The internet

slide-6
SLIDE 6

Orientation

System Bus (PCI) IO Power Memory Power Memory Memory Memory Architecture begins about here.

slide-7
SLIDE 7

Orientation

System Bus (PCI) IO Power Memory Power Memory Memory Memory Architecture begins about here.

slide-8
SLIDE 8
slide-9
SLIDE 9
slide-10
SLIDE 10

You are here

slide-11
SLIDE 11

You are here

cse240a

slide-12
SLIDE 12

The processors go here…

slide-13
SLIDE 13

The processors go here…

slide-14
SLIDE 14

Abstractions of the Physical World…

Physics/Materials Devices Micro-architecture Architectures Processors

slide-15
SLIDE 15

Abstractions of the Physical World…

Physics/Materials Devices Micro-architecture Architectures Processors

This Course cse241a/ ECE dept Physics/ Chemistry/ Material science

slide-16
SLIDE 16

…for the Rest of the System

Architectures

JVM

Processor Abstraction Compilers Languages Software Engineers/ Applications

slide-17
SLIDE 17

…for the Rest of the System

Architectures

JVM

Processor Abstraction Compilers Languages Software Engineers/ Applications

slide-18
SLIDE 18

Why study architecture?

11
  • As CEs or CSs you should understand how computers

work

  • Processors are the basis for everything in CS (except theory)
  • They are where the rubber meets the road.
  • Performance is important
  • Faster machines make applications cheaper
  • Understanding hardware is essential to understanding how

systems behave

  • It’s cool!
  • Microprocessors are among the most sophisticated devices

manufactured by people

  • How they work (and even that they work) as reliably and as

quickly as they do is amazing.

  • Architecture is undergoing a revolution
  • The future is uncertain
  • Opportunities for innovation abound.
slide-19
SLIDE 19

Performance and You!

  • Live Demo
12
slide-20
SLIDE 20

Processor are Cool!

  • Chips are made of silicon
  • Aka “sand”
  • The most abundant element in the

earth’s crust.

  • Extremely pure (<1 part per billion)
  • This is the purest stuff people make
slide-21
SLIDE 21

Building Chips

slide-22
SLIDE 22

Building Chips

  • Photolithography

Silicon Wafer

slide-23
SLIDE 23

Building Chips

  • Photolithography

Silicon Wafer Silicon Wafer SiO2

Grow silicon dioxide

slide-24
SLIDE 24

Building Chips

  • Photolithography

Silicon Wafer Silicon Wafer SiO2

Grow silicon dioxide

Silicon Wafer SiO2 Resist

Apply photo resist

slide-25
SLIDE 25

Building Chips

  • Photolithography

Silicon Wafer Silicon Wafer SiO2

Grow silicon dioxide

Silicon Wafer SiO2 Resist

Apply photo resist

Silicon Wafer SiO2 Resist Mask Mask

Expose to UV

slide-26
SLIDE 26

Building Chips

  • Photolithography

Silicon Wafer Silicon Wafer SiO2

Grow silicon dioxide

Silicon Wafer SiO2 Resist

Apply photo resist

Silicon Wafer SiO2 Resist Mask Mask

Expose to UV

Silicon Wafer SiO2

Patterned resist

slide-27
SLIDE 27

Building Chips

  • Photolithography

Silicon Wafer Silicon Wafer SiO2

Grow silicon dioxide

Silicon Wafer SiO2 Resist

Apply photo resist

Silicon Wafer SiO2 Resist Mask Mask

Expose to UV

Silicon Wafer SiO2

Patterned resist

Silicon Wafer

Etch SiO2

slide-28
SLIDE 28

Building Chips

  • Photolithography

Silicon Wafer Silicon Wafer SiO2

Grow silicon dioxide

Silicon Wafer SiO2 Resist

Apply photo resist

Silicon Wafer SiO2 Resist Mask Mask

Expose to UV

Silicon Wafer SiO2

Patterned resist

Silicon Wafer

Etch SiO2

Silicon Wafer Met

Deposit metal

slide-29
SLIDE 29

Building Chips

  • Photolithography

Silicon Wafer Silicon Wafer SiO2

Grow silicon dioxide

Silicon Wafer SiO2 Resist

Apply photo resist

Silicon Wafer SiO2 Resist Mask Mask

Expose to UV

Silicon Wafer SiO2

Patterned resist

Silicon Wafer

Etch SiO2

Silicon Wafer Met

Deposit metal

Silicon Wafer Met

Etch SiO2 (Or not)

slide-30
SLIDE 30

Building Blocks: Transistors

slide-31
SLIDE 31

Building Blocks: Wires

slide-32
SLIDE 32

State of the art CPU

  • 1-2 Billion xtrs
  • 45nm features
  • 3-4Ghz
  • Several 100 designers
  • >5 years
  • $3Billion fab
  • 70 GFLOPS
18
slide-33
SLIDE 33

Current state of architecture

slide-34
SLIDE 34

Since 1940

slide-35
SLIDE 35

Since 1940

  • Plug boards -> Java
  • Hand assembling -> GCC
  • No OS -> Windows Vista
slide-36
SLIDE 36

Since 1940

  • Plug boards -> Java
  • Hand assembling -> GCC
  • No OS -> Windows Vista

Flexible performance is a liquid asset

  • 50,000 x speedup
  • >1,000,000,000 x density

(Moore’s Law)

slide-37
SLIDE 37

Moore’s Law: Raw transistors

slide-38
SLIDE 38

Computer Performance

22
slide-39
SLIDE 39

Computer Performance

22 1 10 100 1000 10000 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Relative Performance Year specINT95 specINT2000 specINT2006
slide-40
SLIDE 40

Computer Performance

22 1 10 100 1000 10000 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Relative Performance Year specINT95 specINT2000 specINT2006 1 10 100 1000 10000 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Relative Performance Year specINT95 specINT2000 specINT2006 47% per year
slide-41
SLIDE 41

Computer Performance

22 1 10 100 1000 10000 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Relative Performance Year specINT95 specINT2000 specINT2006 1 10 100 1000 10000 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Relative Performance Year specINT95 specINT2000 specINT2006 47% per year 1 10 100 1000 10000 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Relative Performance Year specINT95 specINT2000 specINT2006 47% per year 39% per year
slide-42
SLIDE 42

Computer Performance

22 1 10 100 1000 10000 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Relative Performance Year specINT95 specINT2000 specINT2006 1 10 100 1000 10000 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Relative Performance Year specINT95 specINT2000 specINT2006 47% per year 1 10 100 1000 10000 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Relative Performance Year specINT95 specINT2000 specINT2006 47% per year 39% per year 1 10 100 1000 10000 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Relative Performance Year specINT95 specINT2000 specINT2006 47% per year 39% per year 25% per year
slide-43
SLIDE 43

The clock speed addiction

23
  • Clock speed is the biggest contributor to power
  • Chip manufactures (Intel, esp.) pushed clock speeds very

hard in the 90s and early 2000s.

  • Doubling the clock speed increases power by 2-8x
  • Clock speed scaling is essentially finished.
1000 2000 3000 4000 5000 1996 1998 2000 2002 2004 2006 2008 2010 Clock speed (Mhz) Year specINT2000 specINT2006
slide-44
SLIDE 44

Power

24

Watts/cm

2 1 10 100 1000 1.5µ 1.5µ 1µ 1µ 0.7µ 0.7µ 0.5µ 0.5µ 0.35µ 0.35µ 0.25µ 0.25µ 0.18µ 0.18µ 0.13µ 0.13µ 0.1µ 0.1µ 0.07µ 0.07µ
slide-45
SLIDE 45

Power

24

Watts/cm

2 1 10 100 1000 1.5µ 1.5µ 1µ 1µ 0.7µ 0.7µ 0.5µ 0.5µ 0.35µ 0.35µ 0.25µ 0.25µ 0.18µ 0.18µ 0.13µ 0.13µ 0.1µ 0.1µ 0.07µ 0.07µ
slide-46
SLIDE 46

Power

24

Watts/cm

2 1 10 100 1000 1.5µ 1.5µ 1µ 1µ 0.7µ 0.7µ 0.5µ 0.5µ 0.35µ 0.35µ 0.25µ 0.25µ 0.18µ 0.18µ 0.13µ 0.13µ 0.1µ 0.1µ 0.07µ 0.07µ
slide-47
SLIDE 47

Power

24

Watts/cm

2 1 10 100 1000 1.5µ 1.5µ 1µ 1µ 0.7µ 0.7µ 0.5µ 0.5µ 0.35µ 0.35µ 0.25µ 0.25µ 0.18µ 0.18µ 0.13µ 0.13µ 0.1µ 0.1µ 0.07µ 0.07µ
slide-48
SLIDE 48

Power

24

Watts/cm

2 1 10 100 1000 1.5µ 1.5µ 1µ 1µ 0.7µ 0.7µ 0.5µ 0.5µ 0.35µ 0.35µ 0.25µ 0.25µ 0.18µ 0.18µ 0.13µ 0.13µ 0.1µ 0.1µ 0.07µ 0.07µ
slide-49
SLIDE 49

Power

24

Watts/cm

2 1 10 100 1000 1.5µ 1.5µ 1µ 1µ 0.7µ 0.7µ 0.5µ 0.5µ 0.35µ 0.35µ 0.25µ 0.25µ 0.18µ 0.18µ 0.13µ 0.13µ 0.1µ 0.1µ 0.07µ 0.07µ
slide-50
SLIDE 50

What’s Next: Brainiacs

  • Hold the clock rate steady.
  • Be smarter in silicon
  • More sophisticated processors
  • More clever algorithms
  • This continues to deliver about 25% per year.
  • But for how long?
25
slide-51
SLIDE 51

What’s Next: Parallelism

  • This is all the rage right now
  • You probably own a multi-processor, they used to

be pretty exotic.

  • They provide some performance, but it’s hard to

use.

  • There aren’t that many threads
  • Remember, flexible performance is a liquid asset
  • Remember or look forward to OS
26
slide-52
SLIDE 52 27

Intel P4 1 core Intel Core 2 Duo 2 cores AMD Barcelona 4 cores SPARC T1 8 cores Intel Prototype 80 cores Cell BE 8 + 1 cores Intel Nahalem 4 cores

slide-53
SLIDE 53

Course Staff

  • Instructor: Steven Swanson
  • Lectures Tues + Thurs
  • TA: Hung-Wei Tseng
  • See the course web page for

contact information and

  • ffice hours.
28
slide-54
SLIDE 54

Who am I?

  • BA/BS at University of Puget Sound
  • PhD at the University of Washington
  • Computer architecture
  • Ubiquitous computing
  • Thesis: “The WaveScalar Architecture”
  • At UCSD since 2006
  • Heterogeneous architectures
  • Non-volatile, solid-state memories
  • Multi-processor memory system optimizations
slide-55
SLIDE 55

Course Outline

1.

Performance

  • 2. Pipelining basics
  • 3. Caches and memory systems
  • 4. More pipelining
  • 5. Data and Control Hazards
  • 6. Exception handling
  • 7. Speculation/Branch prediction
  • 8. ISAs
  • 9. Instruction Level Parallelism/Out-of-Order Execution

10.Multi-threaded architectures Please watch the website for course updates, reading assignments and homework assignments!

slide-56
SLIDE 56

What you can expect to get out of this class

  • Understand fundamental concepts in computer architecture and how

they impact computer and application performance.

  • Evaluate architectural descriptions of even today’s most complex

processors.

  • Learn experimental techniques used to evaluate advanced

architecture design – you will be primed to work on architecture problems – or other areas that benefit from architecture knowledge

slide-57
SLIDE 57

Grading

  • Grading is on a 13 point scale -- F through A+
  • You will get a letter grade on each assignment
  • Your final grade is the weighted average of the

assignment grades.

  • An excel spreadsheet calculates your grades
  • We will post a sanitized version online once a week.
  • It will tell you exactly where you stand.
  • It specifies the curves used for each assignment etc.
  • OpenOffice doesn’t run it properly.
32
slide-58
SLIDE 58

Course Work and Grading

  • Computer Architecture: A Quantitative Approach 4th Ed
  • by Hennessy and Patterson
  • Course Work
  • Homework assignments / statistically graded
  • No late assignments accepted, no re-grading of homework
  • Work in groups of one or two, submit one write-up
  • 1 Simulation projects
  • Optional class presentations
  • Grading
  • 10% homework
  • 20% Project/competition
  • 40% 2 midterms/in class presentation
  • 30% final (cumulative)
  • Tests
  • Closed books and no notes
33
slide-59
SLIDE 59

Presentations

  • More detail on Thursday
  • You can sign up in pairs to present 1-3

research papers (depending on length) in class.

– The grade will replace the lowest of your midterm grades. – The presentations should 45-60 minutes. – There are 8 presentation slots. Sign up is first come first serve. You cannot back out.

34
slide-60
SLIDE 60

Academic Honesty

  • Don’t cheat.
  • Cheating on a test will get you an F in the class and no
  • ption to drop, and a visit with your college dean.
  • Cheating on homeworks means you don’t have to turn

them in any more, but you don’t get points either. You will also take at least 25% penalty on the exam grades.

  • Copying solutions of the internet or a solutions

manual is cheating.

  • Review the UCSD student handbook
  • When in doubt, ask.
35
slide-61
SLIDE 61

Wait List

  • This class is over-booked
  • If you want in and are currently on the waitlist
  • Come everyday, and sign in.
  • I’ll start letting people in next week.
  • If you have a critical need to take this class
  • Send an email to Hung Wei and I explaining why it is

essential that you take the class this quarter.

  • And show up everyday.
36