6 Program Translation CS 100 The World of

6. Program Translation CS 100: The World of Computing John Dougherty Haverford College

Overview w The Problem w The “Source” – high-level code w The Target – low-level (machine) code w Types of translation w The translation algorithm/process w PIPPIN

The Problem w People communication in ambiguous, highlevel languages, using experience, context, and can ask for clarity interactively n e. g. , Thoreau threw through the tunnel. w Machines have no “sense” of context or experience, and need unambiguous instructions

The Source w High-level programming languages close to natural language (but not quite) n Alice, Javascript, C++, Java, C# w Known as Source code w Each instructions implies many lower-level instructions (as we’ll see …)

The Target w Low-level instructions that are clear and simple – typically fixed in size, with a command some reference to data n n Opcode Operand w Known as … n Machine code || binary code || executable

Program Translation w From High- to Low-Level w Recall “divide and conquer” in programming n n Input Process Output … then details of Process, then details of … w Typically many low-level operations per high-level instruction w From source code to machine/binary code w Two ways to translate …

Interpretation w Works with the source always w Translates and executes “on the fly” w Like a language translator at the UN w Easier to debug w Executes slower

Compilation w Works with executable w Translates the entire program from source to machine code once w Executes the machine code as many times as needed w Recompile often during development w Executes substantially faster w Most software is distributed (except open source) w Hides algorithm

Phases of Translation w Scanning – breaking text sequence into tokens (i. e. , meaningful chunks) n “while”, “=”, “For all together” w Parsing – organizing the tokens to discover the meaning of the program w Code Generation – writing the sequence of machine level operations n Opcodes, operands

Language Levels w High-level: one-to-many relation to machine language (e. g. , z = x + y is 4 PIPPIN ops) w Assembly language: one-to-one (roughly) relation to machine language (PIPPIN) w Low-level: machine, or binary, language of 0 s and 1 s

Arithmetic Instructions w To demonstrate this process, we’ll look at standard arithmetic expressions and statements in a high-level language w Expressions have a pattern, or (recursivelydefined) form Var = exp Where exp = value | exp + exp | exp – exp | … (demonstration of Rosetta)

PIPPIN instruction layout opcode operand Each box contains a byte

Sample PIPPIN Opcodes 00000100 LOD (load from X) 00000101 STO (store to X) 00001111 HLT (halt execution) 0000 ADD (acc = acc + X)

Example PIPPIN program ; PIPPIN code for Z = X + Y [1] [2] [3] [4] LOD X ADD Y STO Z HLT ; acc <= X ; acc <= acc + Y ; acc => Z ; halt ; other examples AE pp. 252 -4

Programming Paradigms w Imperative: procedures as abstractions, details of how to do a task (e. g. , FORTRAN, Pascal) w Functional: mathematical approach of inputprocess-return value – functions can be composed of other functions (including themselves), and can be evaluated (e. g. , LISP) w Declarative: describe the information, but not the way it is processed (e. g. , Prolog) w Object-Oriented: interacting objects (e. g. , Java, C++, C#, Smalltalk, Javascript, Alice)