Overview of Compilation Programming Language Translators Prepared by

Overview of Compilation Programming Language Translators Prepared by Manuel E. Bermúdez, Ph. D. Associate Professor University of Florida

Overview of Translation • Definition: A translator is an algorithm that converts source programs into equivalent target programs. Source Translator Target • Definition: A compiler is a translator whose target language is at a “lower” level than its source language.

Overview of Translation (cont’d) • When is one language’s level “lower” than another’s? • Definition: An interpreter is an algorithm that simulates the execution of programs written in a given source language. input Source Interpreter output

Overview of Translation (cont’d) • Definition: An implementation of a programming language consists of a translator (or compiler) for that language, and an interpreter for the corresponding target language. input Source Compiler Target Interpreter output

Translation • A source program may be translated an arbitrary number of times before the target program is generated. Source Translator 1 Translator 2. . . Translator. N Target

Translation (cont’d) • Each of these translations is called a phase, not to be confused with a pass, i. e. , a disk dump. Q: How should a compiler be divided into phases? A: So that each phase can be easily described by some formal model of computation, and so the phase can be carried out efficiently.

Translation (cont’d) Q: How is a compiler usually divided? A: Two major phases, with many possibilities for subdivision. • Phase 1: Analysis (determine correctness) • Phase 2: Synthesis (produce target code) • Another criterion: • Phase 1: Syntax (form). • Phase 2: Semantics (meaning).

Typical Compiler Breakdown • Scanning (Lexical analysis). • Goal: Group sequences of characters that occur on the source, into logical atomic units called tokens. • Examples of tokens: Identifiers, keywords, integers, strings, punctuation marks, “white spaces”, end-of-line characters, comments, etc. , … Source Scanner (Lexical analysis) Sequence of Tokens

Example (see diagram) • Probably must deal with end-of-line and end-of-file characters. • A preliminary classification of tokens is made. For example, both ‘program’ and ‘Ex’ are classified as Identifier. • Someone must give unambiguous rules forming tokens.

Screening • Goals: • Remove unwanted tokens. • Classify keywords. • Merge/simplify tokens. Sequence of Tokens Screener Sequence of Tokens

Example (see diagrams) • Keywords recognized. • White spaces (and comments) discarded. • The screener acts as an interface between the scanner and the next phase, the parser.

Parsing (Syntax Analysis) • Goals • To group together the tokens, into the correct syntactic structures, if possible. • To determine whether the tokens appear in patterns that syntactically correct.

Parsing (Syntax Analysis, cont’d) • Syntactic structures: • Expressions • Statements • Procedures • Functions • Modules • Methodology: • Use “re-write” rules (a. k. a. BNF).

String-To-Tree Transduction (see diagrams) • Goal: To build a “syntax tree” from the sequence of rewrite rules. The tree will be the functional representation of the source. • Method: Build tree “bottom-up, ” as the rewrite rules are emitted. Use a stack of trees.

Contextual Constraint Analysis • Goal: To analyze static semantics, e. g. , • Are all variables declared before they are used? • Is there assignment compatibility? • e. g. , a: =3 • Is there operator type compatibility? • e. g. , a+3 • Do actual and formal parameter types match? • Enforcement of scope rules.

Contextual Constraint Analysis (cont’d) • Method: Traverse the tree recursively, deducing type information at the bottom, and passing it up. – Make use of a DECLARATION TABLE, to record information about names. – “Decorate” tree with reference information.

Example Chronologically, 1. Enter x into the DCLN table, with its type. 2. Check type compatibility for x = 5. 3. X 2 not declared! 4. Verify type of ’>’ is boolean. 5. Check type compatibility for ‘+’. 6. Check type compatibility between x and int (‘+’).

Code Generation • Goal: Convert syntax tree to target code. Target code could be: • Machine language. • Assembly language. • Quadruples for a fictional machine: • label • opcode • operands (1 or 2)

Code Generation (cont’d) • Example: • “pc” on UNIX generates assembly code • “pi” on UNIX generates code for the “p” machine, which is interpreted by… an interpreter. • pc: slow compilation, fast running code. • pi: fast compilation, slow running code. • Method: Traverse the tree again.

Code (for a stack machine): See diagrams) L 1 LOAD STORE LOAD BGT COND LOAD BADD STORE GOTO L 2. . . L 3 5 X X 10 L 1 X L 3 L 2

Code Optimization • Goals: • Reduce the size of the target program. • Decrease the running time of the target. • Note: “Optimization” is a misnomer. Code improvement would be better. • Two types of optimization: • Peephole optimization (local). • Global optimization (improve loops, etc. ).

Code Optimization (cont’d) • Example (from previous slide): LOAD 5 STORE X LOAD X can be replaced with LOAD 5 STND X Store non-destructively, i. e. , store in X, but do not destroy value on top of stack.

Summary Source Scanner Tokens Screener Tokens Parser Table Routines Error Routines Tree Constrainer Tree Code Generator Code (for an abstract machine) Input Interpreter Output

Overview of Compilation Programming Language Translators Prepared by Manuel E. Bermúdez, Ph. D. Associate Professor University of Florida