Welcome to Principles of Computer

1
Welcome to Principles of Computer Information
Technology!
Course Overview - Goal This course will
familiarize students with fundamental
concepts related to information technology with
an emphasis on computer architecture
performance. Required background Basic
familiarity with modern digital computer. Some
math required but no programming
necessary. Course homepage http//mccormick.nort
hwestern.edu/awolff/
2
What 421 is about

• 411 deals with how data is communicated.
• 412 deals with what happens at the end points
  how data is stored, displayed, and processed.

421
411
3
Grading

• Absolute Scale
• 91.50 100.00 A
• 90.00 91.49 A-
• 88.50 89.99 B
• 81.50 88.49 B
• 80.00 81.49 B-
• 78.50 79.99 C
• 71.50 78.49 C
• 70.00 71.49 C-

Grade Components 40 Midterm exam 10 HW 25
Final exam 25 Project Homework is to be done
in groups. One HW turned in per group. Discuss
with group members before coming to me. - The
midterm will be held on the 6th week of class.
4
MSIT 421 Schedule
5
Other Information
Textbooks Computer Organization Architecture,
William Stallings, 7th Edition, 2006. How
Computers Work, Ron White, Ninth Edition,
2008. Other Books Computer Architecture A
Quantitative Approach, John L. Hennesy, David A.
Patterson, 4th Edition, 2006. Introduction to
Information Technology, Third edition, Turban,
Rainer, Potter, 2005. Instructor Alan Wolff,
Ph.D. Phone number 847- 491-4443 (W),
847-502-1680 (Cell) Email a-wolff_at_northwestern.ed
uRoom Number L293 of TechEmail is the best way
to get in touch with me and get a quick
response!
6
Class 1 Intro / Computer Performance

• Intro to Computer Systems
• A. IT Matters
• B. Performance Basics
• C. Hierarchy, Structure and Function
• II. Technological Developments
• A. Generations
• B. Moores Law
• C. Enhancing Performance
• IV. Performance Measures
• A. MIPS, MFLOPS
• B. Price/ Performance
• C. Speedup Amdahls Law

7
Does IT Matter?

• Perspectives
• The Digital Age Storms the Corner Office
• Business Week, September 6, 2001.
• Info tech is so crucial to business operations
  today -- and so expensive -- that CEOs have no
  choice but to understand it.
• "IT Doesn't Matter," Harvard Business Review,
  Vol. 81, No. 5, May 2003.
• This provocative Harvard Business Review excerpt
  suggests that IT no longer conveys competitive
  advantage, so invest your capital elsewhere.
• Worrying about what might go wrong may not be
  as glamorous a job as speculating about the
  future, but it is a more essential job right
  now.
• The End of Corporate Computing, Spring 2005 issue
  of the MIT Sloan Management Review
• As with the factory-owned generators that
  dominated electricity production a century ago,
  today's private IT plants will be supplanted by
  large-scale, centralized utilities.

8
IT Creates Sustains Wealth
Americas Richest People Forbes, 9/08

• IT gives people what they want. Thus there is a
  lot
• of money in this business.

Seven of the 15 richest people in America are
directly linked to IT.
9
Performance Basics
Commercial Airplane Performance
Which Airplane has the best performance?
Computing Performance
Which data center has the best performance?
10
Computer Performance

• Response time and throughput are common ways to
  measure computer performance
• For the time being we will use response time as a
  measure of performance

PerformanceX 1/(Execution Time)X
11
Performance Comparison

• What does X is n faster than Y mean?
• Example
• If machine A runs a program in 10 seconds, and
  machine B
• runs the same program in 15 seconds, which of the
  following
• statements is true?
• A is 50 faster than B
• A is 33 faster than B

Solution PA 1/10 0.10 PB 1/15
0.06667 PA/PB 0.10/0.06667 1.50 n 50
12
What is a Computer?

• Webster a programmable usually electronic
  device that can store, retrieve, and process
  data.
• Date 1646
• A digital computer is a machine that can solve
  problems by carrying out instructions given to
  it.
• A sequence of instructions program
• The electronic circuits of each computer can
  recognize and directly execute a limited set of
  simple instructions into which all its programs
  must be converted before they can be executed.
• The instructions are rarely more complicated
  than
• Add two numbers
• Check a number to see if it is zero
• Copy a piece of data from one part of the
  computers memory to another

13
Machine Language

• A computers primitive instructions form a
  language called machine language.
• What is convenient for computers and what is
  convenient for people are not the same.
• Two languages instructions for computer form
  language L0 instructions that are more
  convenient for people L1.
• Two approaches for programs written in L1 to be
  executed by L0
• Translation
• Interpretation

14
Translation
program gcd(input, output) var i, j
integer begin read(i, j) while i j
do if i j then i i j else j j
i writeln(i) end.

• Goal Preserve original tone and meaning.
• Example Translation of classic literature from
  Greek or Latin to English.

15
Interpretation

• A compiler is a program that translates
  high-level source programs into target program

• Fuzzy difference
• A language is interpreted when the initial
  translation is simple
• A language is compiled when the translation
  process is complicated

16
Examples Translation Interpretation

• Low Level Languages
• emphasize detailed control, close to the machine
• programs are often slow to write but should run
  fast
• the translation (compilation) takes place before
  the program can be run
• programs are distributed as binaries
• used where performance and accuracy is important
• Examples C, C, Pascal, Fortran, Ada, Modula-2
• High Level Languages
• a.k.a. scripting languages
• emphasize broad detail and abstractions
• programs can be written quickly but ok to run
  slow
• the translation (interpretation) takes place
  while the program is run
• programs are distributed as source code
• used for one-shot programs, text processing,
  extensions internal langauges, macros. Programs
  are often called "scripts".
• Examples Python, Perl, Basic, Lisp, Scheme,
  Java, Tcl, R

17
Multilevel approach
18
Functional Perspective

• Generally, all computer functions can be
  classified into four categories
• Data processing arithmetic or logic operation
  (Add, True, etc).
• Data storage internal/external memory (RAM, HD,
  DVD, etc).
• Data movement between components or computers
  (ex RAM-CPU)
• Control Orchestrates the other functions listed
  above.

19
Structure - Top Level
Computer
Peripherals
Central Processing Unit
Main Memory
Computer
Systems Interconnection
Input Output
Communication lines
20
Structure - The CPU
CPU
Arithmetic and Login Unit
Computer
Registers
I/O
CPU
System Bus
Internal CPU Interconnection
Memory
Control Unit
21
Computer Model----Von Neumann Model

• Stored Program concept
• Heard of punch card ?
• Main memory storing programs and data
• Control unit interpreting instructions from
  memory and executing them.
• ALU operating on binary data
• Input and output equipment transmit information.
• All of todays computer designs are using Von
  Neumann model !

22
Computer History

• Generation 0 Mechanical
• Generation 1 Vacuum Tubes
• Generation 2 Transistors
• Generation 3 Integrated Circuits
• Generation 4 Very Large Scale Integration
• Generation 5 ??

23
Generation 0 Mechanical Implementation

• The mechanical implementation of computers used
  gears and wheels and other mechanical movements.
• There are inherent limitations and shortcomings
  of mechanical computation including
• Complex design and construction
• Wear, breakdown, and maintenance and mechanical
  parts
• Limits on operating speed

24
Early Mechanical Computers

• Early pioneers
• William Schickard (1623) mechanical (addition,
  subtraction, multiplication, division)
• Blaise Pascal (1642) mechanical (addition
  subtraction).
• Gottfreid Wilhelm von Leibnitz (1670s)
  mechanical (also multiplication division)
• Charles Babbage 1820s
• Difference engine mechanical, addition
  subtraction navigation tables
• Analytical machine mechanical (never functional,
  poorly documented)- first general purpose
  computer controlled by punched card programs
  (worlds first computer programmer Ada Augusta
  Lovelace)

25
Generation 1 Vacuum Tubes

• Necessity is the mother of invention. Plato
  in Republic
• Colossus 1943, GB
• First electronic digital computer
• Cracking the ENIGMA cyphers
• ENIAC 1946, US
• Electronic computer (18,000 vacuum tubes),
• 20 registers with 10 digit decimal numbers (not
  binary)
• Computation of trajectory tables for heavy
  artillery
• 1945, Legend of First Computer Bug at Harvard
  (Mark I)
• von Neumann Machine John von Neumann, Princeton
  1950. the IAS computer
• Basis of today architectures stored program
  concept
• Memory, ALU, control unit, input, output, binary
• 1953 first computer by IBM

26
IAS Memory/Instructions
Address Contents 08D 010AA210AB In
the program at address 08D, the two lines
are LOAD(0AA) STOR(0AB)
Note 01 in hex is 00000001 in binary, 21 in hex
is 00100001 in binary
27
Generation 2 Transistors

• The transistor was invented in 1948 at Bell Labs
• Made of silicon (sand)- mixed with certain
  chemical
• semiconductor (Nobel Prize in Physics, 1956)
• Less heat, lower cost, more reliable than vacuum
  tubes
• PDP-1 Digital Equipment Corporation, 1961
• First commercial transistorized computer
  (mini-computer)
• Visual display with 512 x 512 pixels
• PDP-8 DEC, 1965
• Single bus
• connected components
• 50,000 units sold
• Established DEC as a major player
• The first super-computers emerged
• Control Data Corporation (CDC)
• Seymour Cray CDC 6600, CDC 7600, Cray-1

28
Generation 3 Integrated Circuits

• Silicon integrated circuit (chip) was invented in
  1958
• Dozens of transistors could be put on a single
  chip
• Computers became smaller, faster, cheaper
• System /360 series, IBM 1964.
• Both scientific and commercial applications
• Replaced two separate strands of system designs
  at IBM
• First machine that could emulate other computers
  by microprograms
• 32 bit computer whose memory was byte addressed
• PDP-11, DEC, end of 1960s
• - Highly successful, especially at universities

29
Generation 4 Very Large Scale Integration (VLSI)

• In the 1970s 1980s, VLSI emerged
• Tens of thousands of transistors could be put on
  a single chip
• Millions of transistors
• Computers became even faster and cheaper
• PC the personal computer
• Originally kits without software
• Xerox PARC graphical user interfaces, windows,
  mouse
• Steve Jobs, Steve Wozniak, Apple, 1977- first
  assembled computer
• IBM PC 1981 got into PC business,
• used Intel and a small company called Microsoft
• PC clones industry emerged
• Mid-1980s new processor designs
• - RISC architectures, super-scalar CPUs

30
Generation 5 ?

• What will come next?
• 3-dimensional circuit designs
• pack transistors in cubes or layers instead of
  chips
• Optical Computing
• Replace wires electronics by optics
• Molecular computing (DNA Computer)
• Use chemical/biological processes for computing
• Quantum computing
• Sub-atomic particle- quantum- two states
  simultaneously qubit, can do base 4 instead of
  base 2
• Special applications only, e.g. quantum
  cryptography

31
Technologies Performance
What if technology related to energy advanced at
the same rate??
32
Moores Law

• Increased density of components on chip
• Gordon Moore - cofounder of Intel
• Number of transistors on a chip will double every
  year
• Since 1970s development has slowed a little
• Number of transistors doubles every 18 months
• Cost of a chip has remained almost unchanged
• Higher packing density means shorter electrical
  paths, giving higher performance
• Smaller size gives increased flexibility

33
Moores Law
http//www.intel.com
34
Main Driver Device Sizes
35
Secondary Driver Wafer Size

• /cm2
• Wafer size improvements have offset the
  increasing costs of wafer processing

36
From Ingot to Chips
37
Fabricating Processors Wafers Chips
Dies per wafer p x (wafer diameter/2)2/(die
area) - p x (wafer diameter)/(v(2 x die area)
450mm wafer Image Source Intel Corporation
www.intel.com
38
Not just the Processor!

• Processor
• Logic capacity increases about 30 per year
• Performance 2x every 1.5 years
• Memory
• DRAM capacity about 60 per year
• Memory speed 1.5x every 10 years
• Cost per bit decreases about 25 per year
• Disk
• Capacity increases about 60 per year

39
Speeding it up even more

• Pipelining
• Increasing number of bits retrieved at a time
• Increase interconnection bandwidth (bus)
• On board cache
• Branch prediction
• Data flow analysis

40
Intel Microprocessor Performance
41
Increased Cache Capacity

• Typically two or three levels of cache between
  processor and main memory
• Chip density increased
• More cache memory on chip
• Faster cache access

Source Semico Research Corp. ASIC report, 2007
42
More Complex Execution Logic

• Enable parallel execution of instructions
• Pipeline works like assembly line
• Different stages of execution of different
  instructions at same time along pipeline
• Superscalar allows multiple pipelines within
  single processor
• Instructions that do not depend on one another
  can be executed in parallel

43
Diminishing Returns

• Internal organization of processors complex
• Can get a great deal of parallelism
• Difficult to take advantage of massive
  parallelism in software
• Benefits from cache reach a limit
• Interconnect (wires) speed limitation
• Increasing clock rate runs into power dissipation
  problem (heat)
• Some fundamental physical limits are being
  reached

44
New Approach Multiple Cores

• Multiple processors on single chip
• Large shared cache
• Within one processor, increase in performance
  proportional to square root of increase in
  complexity
• If software can use multiple processors, doubling
  number of processors almost doubles performance
• So, use two simpler processors on the chip rather
  than one more complex processor
• With two processors, larger caches are justified

45
CPU Performance Factors

• CPU Execution Time CPU clock cycles x Clock
  Cycle Time
• OR
• CPU Execution Time (CPU clock cycles)/(Clock
  rate)
• EX Computer A has a clock cycle rate of 4 GHz,
  so its clock cycle time is
• 1/(4 x 109) seconds 2.5 x 10-10 seconds 250
  picoseconds (ps). If it takes 40 billion CPU
  cycles to complete a program, what is the CPU
  execution time?
• CPU Execution Time (40 x 109 cycles) x (2.5 x
  10-10 seconds/cycle)
• 10 seconds
• OR
• CPU Execution Time (40 x 109 cycles) / (4 x
  109 cycles/second)
• 10 seconds

46
Cycles Per Instruction (CPI)

• CPI is a term used to describe one aspect of a
  processor's performance the number of clock
  cycles that happen when an instruction is being
  executed. All programs consist of many
  instructions. Low CPI is good.
• CPU Clock Cycles (Instructions for Program) x
  CPI
• EX Computer A is 4 GHz with a CPI of 2.0 for a
  program, and Computer B is 2 GHz with a CPI of
  1.2. Which computer is faster and by how much?
• Answer Lets say the number of instructions is
  I for both computers.
• (CPU Clock Cycles)A I x 2.0
• (CPU Clock Cycles)B I x 1.2, then
• CPU TimeA (CPU clock cycles)A x (Clock cycle
  time)A
• (I x 2.0) cycles x 250 ps/cycle 500 ps
• CPU TimeB (I x 1.2) cycles x 500 ps/cycle
  600 ps
• Computer A is faster by 100 ps. The higher clock
  rate more than compensates for better CPI.

47
Example

• Consider a computer with an instruction set like
  this
• Which program will run faster if each is
  comprised of
• Answer
• Program 1 (1 instr/cycle)(2000 cycle) (2
  instr/cycle)(1000 cycle)
• (3 instr/cycle)(2000 cycle) 10,000 cycles
• Program 2 (1 instr/cycle)(4000 cycle) (2
  instr/cycle)(1000 cycle)
• (3 instr/cycle)(1000 cycle) 9,000 cycles
• -Program 2 is faster
• Notes
• MIPS means millions of instructions per second
  instructions / (106)
• MFLOPS means millions of floating point
  operations per second FLOPS / (106)

48
Measure of CPU Performance FLOPS

• FLOPS floating point operations per second.
• Floating point is a particular kind of
  non-integer mathematical instruction.
• Performance above measured in GFLOPS billions
  of floating point operations per second

Source http//www.top500.org/list/2008/06/100
49
Worlds Fastest Computer System
50
CPU Performance

• The real measure of performance is the time it
  takes for a task to run. So in many ways it
  depends on the application.
• In the industry, different benchmarks are used
  for performance. A good example is that of the
  Standards Performance Evaluation Corporation
  (SPEC). These benchmarks are real programs

51
Performance of Servers
Performance based on TPC-C Spec, September, 2008
Spec includes simulation of typical server
transactions in a database Payment,
Order-Status, Delivery, Stock-Level
http//www.tpc.org/
52
Price/Performance of Servers
Price/Performance based on TPC-C Spec/Cost, Sept,
2008
Spec includes simulation of typical server
transactions in a database Payment,
Order-Status, Delivery, Stock-Level
http//www.tpc.org/
53
Computer Market
4th quarter 2007
All 2007
54
Problems with MIPS, MFLOPS

• MIPS is a useless measure sometimes
• Different chips use different instruction sets
• Not all instructions are the same within an
  instruction set
• Performance also depends on register size, cache
  memory performance, memory bandwidth among
  other things!
• MFLOPS is a little less useless because it is
  more specific type of instruction but also
  depends on the above listed factors.
• These measures give a rough idea of computer
  performance

55
Designing for Performance

• Principle
• Make the common case fast.
• A computing process consists of different
  events, and
• uses different resources. To improve the whole
  computer
• system performance, the frequently happening
  events
• and the most often used resources should be speed
  up in
• the highest priority.
• Example
• Optimizing a rarely used instruction is not
  helpful to
• improve performance.

56
Amdahls Law

• The performance improvement to be gained from
  using some faster mode of execution is limited by
  the fraction of the time the faster mode can be
  used.
• Amdahls law defines the speedup that can be
  gained by using a particular feature.
• For example, to improve the computers
  performance, there are many options CPU speed,
  memory access time, hard disk rotate speed and
  access time, I/O access time, data bus width, and
  so on.
• How important is each one of them?
•  

57
Speedup Definition

• Execution time without
  using the enhancement
• Speedup------------------------------------------
  --------------------
• Execution time after using
  the enhancement

Speedup tells how much faster a task will run by
using the enhancement.
Suppose that we are considering an enhancement
that runs 10 times faster than the original
machine but is only usable 40 of the time. What
is the overall speedup gained by incorporating
the enhancement?
58
Types of Systems (1)

• Desktops and laptops the largest market
• Price-performance is most important
• Servers availability, reliability, scalability,
  throughput

59
Component Costs of PCs
60
Cost vs. Price

61
Types of Systems (2)

• Embedded computers fastest growing segment
• -includes palms, cell phones, video games,
  microwaves, printers, networking switches,
  automobiles, etc.
• - price, real-time performance, low memory, low
  power
• Summary