CS 4240

1
CS 4240

• Nate Clark

(many slides from Bodik and August)
2
About Me

• Prof. Nate Clark
• But just call me Nate
• I swear a lot
• 2nd year here at Ga. Tech. (be gentle)
• Hardware guy who likes compilers
• Dynamic program optimization
• Before this Grad student at UMich
• IBM, ARM, HP
• Before before Undergrad at UMich

3
About Class

• http//www.cc.gatech.edu/ntclark/4240f08/
• Give a man fire and hes warm for a day set a
  man on fire and hes warm for the rest of his
  life
• Background

4
Planned topics

• Lexing/parsing
• Symbol tables
• Error detection/recovery
• Semantic analysis
• IR generation
• OO languages
• Some optimizations

5
Classy Class Class

• Book
• Compilers Principles, Techniques, and Tools,
  Aho, Lam, Sethi, and Ullman

6
Grades

• Project X
• 2 Exams Y
• Cheating
• Winnowing Local Algorithms for Document
  Fingerprinting by Schleimer, Wilkerson, and Aiken

7
What is a Compiler?
8
The Stack

• Almost 100 application code goes through
  compiler
• 99.5 Linux OS Kernel

9

• the structure of a compiler

10
Three execution environments

• Interpreters
• Scheme, lisp, perl, python
• the more popular of these later got compilers
• Compilers
• C/C to machine code
• Java to Java bytecode
• Virtual machines
• interpreter, often aided by a just-in-time (JIT)
  compiler
• Java bytecode runs on a VM

11
The Structure of a Compiler

• Scanning (Lexical Analysis)
• Parsing (Syntactic Analysis)
• Type checking (Semantic Analysis)
• Optimization
• Code Generation
• The first 3, at least, can be understood by
  analogy to how humans comprehend English.

12
Lexical Analysis

• Lexical analyzer divides program text into
  words or tokens
• if x y then z 1 else z 2
• Units
• if, x, , y, then, z, , 1, , else, z, , 2,

13
Parsing

• Once words are understood, the next step is to
  understand sentence structure
• Parsing Diagramming Sentences
• The diagram is a tree

14
Diagramming a Sentence
15
Parsing Programs

• Parsing program expressions is the same
• Consider
• If x y then z 1 else z 2
• Diagrammed

16
Semantic Analysis in English

• Example
• Jack said Jerry left his assignment at home.
• What does his refer to? Jack or Jerry?
• Even worse
• Jack said Jack left his assignment at home?
• How many Jacks are there?
• Which one left the assignment?

17
Semantic Analysis I

• Programming languages define strict rules to
  avoid such ambiguities
• This Java code prints 4 the inner definition
  is used

• int Jack 3
• int Jack 4
• System.out. print(Jack)

18
Semantic Analysis II

• Compilers also perform checks to find bugs
• Example
• Jack left her homework at home.
• A type mismatch between her and Jack
• we know they are different people (presumably
  Jack is male)

19
Code Generation

• A translation into another language
• Analogous to human translation
• Compilers for C, C
• produce assembly code (typically)
• Code generators
• produce C or Java

20

• New Languages will Keep Coming

21
A survey how many languages have you used?

• Lets list them here

22
Be prepared to program in new languages

• Languages undergo constant change
• FORTRAN 1953
• ALGOL 60 1960
• C 1973
• C 1985
• Java 1995
• Evolution steps 12 years per big language
• are we overdue for the next big one?
• ... or is the future already here?
• are we in a major change in what programs
  express, and how?

23
Develop your own language

• Are you kidding? No. Guess who developed
• PHP
• Ruby
• JavaScript
• perl
• Done by smart hackers like you
• say these were done in a garage
• not in ivory tower

24

• Trends in Programming Languages

25
Trends in programming languages

• the programming language and its compiler
• programmers primary tools
• you must know them inside out
• languages has been constantly evolving ...
• what are the forces driving the change?
• ... and will keep doing so
• to predict the future, lets examine the history

26
ENIAC (1946, University of Philadelphia)
ENIAC program for external ballistic equations
27
Programming the ENIAC
28
ENIAC (1946, U of Philadelphia)

• programming done by
• rewiring the interconnections
• to set up desired formulas, etc
• Problem whats the tedious part?
• programming rewiring
• slow, error-prone
• Lesson
• store the program in memory!
• birth of von Neuman paradigm

29
UDSAC (1947, Cambridge University)

• the first real computer
• large-scale, fully functional, stored-program
  electronic digital computer (by Maurice Wilkes)
• this was a von Neuman computer
• problem Wilkes realized
• a good part of the remainder of my life was
  going to be spent in finding errors in ...
  programs
• solution procedures (1951)
• reusable software was born
• procedure the first (implemented) language
  construct

30
Assembly the language (UNIVAC 1, 1950)

• Idea translate mnemonic code (assembly) by hand
• by hand, they did not have a compiler yet
• write programs with mnemonic codes (add, sub),
  with symbolic labels,
• then assign addresses by hand
• Example of symbolic assembler
• clear-and-add a
• add b
• store c
• translate it by hand to something like this
  (understood by CPU)
• B100 A200 C300

31
Assembler the compiler (Manchester, 1952)

• it was assembler nearly as we know it, called
  AutoCode
• a loop example, in MIPS, a modern-day assembly
  code
• loop addi t3, t0, -8
• addi t4, t0, -4
• lw t1, theArray(t3) Gets the last
• lw t2, theArray(t4) two elements
• add t5, t1, t2 Adds them together...
• sw t5, theArray(t0) ...and stores result
• addi t0, t0, 4 Moves to next "element
• of theArray
• blt t0, 160, loop If not past the end of
• jr ra theArray, repeat

32
Assembly programming caught on, but

• Problem Software costs exceeded hardware costs!
• John Backus Speedcoding
• An interpreter for a high-level language
• Ran 10-20 times slower than hand-written assembly
• way too slow

33
FORTRAN I (1954-57)

• The first HLL compiler
• Produced code almost as good as hand-written
• Huge impact on computer science
• Modern compilers preserve its outlines
• FORTRAN (the language) still in use today
• By 1958, gt50 of all software is in FORTRAN
• Cut development time dramatically
• 2 weeks ? 2 hrs
• thats more than 100-fold

34
FORTRAN I (IBM, John Backus, 1954)

• Example nested loops in FORTRAN
• a big improvement over assembler,
• but annoying artifacts of assembly remain
• labels and rather explicit jumps (CONTINUE)
• lexical columns the statement must start in
  column 7
• The MIPS loop in FORTRAN
• DO 10 I 2, 40
• AI AI-1 AI-2
• 10 CONTINUE

35
Side note designing a good language is hard

• A good language protects against bugs, but
  lessons take a while.
• An example that cause a failure of a NASA
  planetary probe
• buggy line
• DO 15 I 1.100
• what was intended (a dot had replaced the
  comma)
• DO 15 I 1,100
• because Fortran ignores spaces, compiler read
  this as
• DO15I 1.100
• which is an assignment into a variable DO15I,
  not a loop.
• This mistake is harder (if at all possible) to
  make with the modern lexical rules (white space
  not ignored) and loop syntax
• for (i1 i lt 100 i)

36
Goto considered harmful

• L1 statement
• if expression goto L1
• statement
• Dijkstra says gotos are harmful
• use structured programming
• lose some performance, gain a lot of readability
• how do you rewrite the above code into structured
  form?

37
Object-oriented programming (1970s)

• The need to express that more than one object
  supports draw()
• draw(2DElement p)
• switch (p.type)
• SQUARE // draw a square
• break
• CIRCLE // draw a circle
• break
• Problem
• unrelated code (drawing of SQUARE and CIRCLE)
  mixed in same procedure
• Solution
• Object-oriented programming with inheritance

38
Object-oriented programming

• In Java, the same code has the desired
  separation
• class Circle extends 2DElement
• void draw() ltdraw circlegt
• // similar for Square
• the dispatch is now much simpler
• p.draw()

39
Review of historic development

• wired interconnects ? stored program (von
  Neuman machines)
• lots of common bugs in repeated code ? procedures
• machine code ? symbolic assembly (compiled by
  hand)
• tedious compilation ? assembler (the assembly
  compiler)
• assembly ? FORTRAN I
• gotos ? structured programming
• hardcoded OO programming ? inheritance, virtual
  calls
• Do you see a trend?

40
Where will languages go from here?

• The trend is towards higher-level abstractions
• express the algorithm concisely!
• which means hiding often repeated code fragments
• new language constructs hide more of these
  low-level details.
• Also, detect more bugs when the program is
  compiled
• with stricter type checking

41
Why software developers need PL

• New languages will keep coming. But why do
  they?
• Understand them, choose the right one.
• Write code that writes code. When to write a
  codegen?
• Be the wizard, not the typist.
• Develop your own language. How do I do that?
• Are you kidding? No.
• Learn about compilers and interpreters.
• Programmers main tools.
• One answer to these questions (to) hide the
  plumbing.