ICS 142 Klefstad

Review

Outline Compiler Design Lexical Analysis Syntax Analysis Error Handling and Recovery Syntax-Directed Translation Type Checking Run-time Environments Intermediate Code Generation Code Generation Code Optimization Interpreters

Compiler Design

Separation of activities is good design
- know the purpose, input, output, and alternatives of each
Compiler activities
- Lexical Analysis
- Syntax Analysis
- Semantic Analysis
- Error Handling
- Intermediate Code Generation
- Code Optimization
- Code Generation
General goal of compiler science
- automatic generation of quality compiler from a language specification

Lexical Analysis

Primary purpose
- convert input characters into a sequence of tokens
Class of languages
- regular languages (only repetition, concatenation, disjunction)
Notations
- regular expressions
Formal model
- Finite-State Automaton

Issues
- must be fast (looks at every character in source program)
- Algorithms: RE --> NFA --> DFA --> Minimal DFA
Alternatives
- hand-coded state machine
- generated automatically from regular definition

Syntax Analysis

Primary purpose
- group tokens into constructs, to see if program has correct syntax
- builds a parse-tree showing the derivation from start symbol
Input
- a stream of tokens
Output
- a parse tree or a syntax tree
Class of languages
- context free languages (RL + nesting/balancing)
- can't handle for correct use of identifiers (EG types, subprograms)

Notations
- Context-Free Grammars
Formal model
- push-down automata - FSA with unbounded stack
Alternatives
- CKY, Earley's
- Operator-Precedence parsing
- LR Parsing ( handle-pruning (shift/reduce))
  - SLR, LALR
- LL Parsing
  - With back-tracking
  - Predictive Parsing (Must compute First and Follow)

Issues
- Algorithms: LR parser generation
- Writing a grammar:
  - associativity
  - precedence
  - left-factoring
  - removing right recursion
  - Ambiguity

Error Handling and Recovery

Purpose
- to detect and report program errors, then continue to check program
Alternatives
- panic mode
  - discard input tokens until one of an designated set is found
- phrase-level recovery
  - replace a prefix of the input with a correct string
- error productions
  - add productions to the grammar for common errors
- global correction
  - choose a minimal sequence of changes in the input to obtain a correct string

Syntax-Directed Translation

Primary purpose
- a notation for specifying translation along with syntax
Input
- a syntax-directed definition (CFG with semantic actions)
Two Varieties of Attributes
- inherited attributes
- synthesized attributes
Two Restrictions of SDD
- S-attribute definition - only synthesized attributes
- L-attribute definition - synthesized + inherited from left
- any L may be converts to S
- These attribute definitions can be evaluated while doing LR parsing

Symbol Tables

Purpose
- for managing inherited attributes
Functions
- insert, lookup, enter scope, exit scope
Representation
- For Algol-like languages: a stack of tables

Type Checking

Formal model
- Type Systems
Notations
- Type Expressions
Alternatives
- structural equivalence - types have same constructors
- name equivalence - types have same name
Efficient Representation for Type Systems
- a Directed-Acyclic-Graph (DAG)
- allows type checking with pointer equality comparison

Run-time Environments

Language issues affect storage organization and allocation
- recursion, lifetime of variables, parameter passing modes, non-local references, dynamic storage allocation and deallocation, multi-tasking

Typical organization of memory

 code segment,  initialized data segment,  uninitialized data segment,  stack,  heap

Issues
- activation record
- alignment, padding, packing

Intermediate Code Generation

Purpose
- allows machine independent optimizations and easier porting of compiler
Alternatives
- three address code
- syntax trees
- DAGs
- postfix

Code Generation

Purpose
- to generate machine specific instructions from intermediate code
Issues
- instruction selection
- register allocation
- instruction scheduling (for RISC)
Sethi-Ullman register numbering and generation from expressison trees
Code Generator Generators
Peep-Hole Optimization - to fix local inefficiencies

Code Optimization

Purpose
- apply techniques to yield space/time improvements
Requirement
- must not introduce errors
- must provide measurable improvement
Some Optimizations
- Subprograms
  - static allocation of activation records
  - eliminating recursion
  - in-line substitution of subprogram bodies
- Loops
  - loop unrolling
  - code motion (moving loop invariants out of loops)
  - reduction in strength
  - induction-variable elimination
- Blocks
  - assigning variables to registers
  - algebraic transformations
  - switching order of evaluation within a block (to minimize registers)
  - eliminating common subexpressions
  - constant and copy propagation
  - constant folding
  - eliminating dead code
Techniques
- DAGs for basic blocks
- Flow Graphs - to represent control flow
- Data-Flow analysis - optimizations based on Flow Graph
- Call graph - shows who can call whom and when
- focus optimizations on inner loops

Interpreters

Purpose
- an alternative means of executing programs
- provides quicker, more interactive program development
Components
- representation for interpreted programs - REP
- a scanner/parser - TEXT --> REP
- a pretty-printer - REP --> TEXT
- a symbol table manager - NAMES --> ATTRIBUTES
- a memory manager - storage for parameters, activations
- a compiler - REP --> CODE
- The Lisp interpreter