Engineering a Compiler, 2012, Cooper, Torczon

1 - Overview

• Compiler: a computer program that translates other computer programs to prepare them for execution
• Front end to deal with the source language, back end to deal with the target language
• Intermediate form to translate between the two
• Intermediate form is often optimized
• Some compilers produce source code as an output, for example, c
• Known as source-to-source compiler
• Interpreter is different from a compiler, but many of the same concepts apply

1.2

• There are two fundamental principles
• The compiler must preserve the meaning of the program being compiled
• The compiler must improve the input program in some discernible way
• Front end converts source to Intermediate representation (IR)
• Can (in theory) take IR and convert to multiple different back end targets
• Or add in more phases between front end and backend, such as optimization
• Optimization problems usually don't have a closed form solution
• Will produce "optimized" code but not "optimal" code

1.3

• Source form is "syntax"
• Source meaning is "semantics"
• Front end checks the syntax and semantics
• Syntax is defined by a grammar
• Two passes in the front end, for the scanner and the parser
• Scanner converts stream of characters into words
• The parser then checks that the words follow the rules of the grammar
• Most optimization consists of analysis and transformation
• The backend will decide what goes into registers and what goes into memory
• Rewriting to machine operations is "instruction selection"
• Will make another pass to translate "virtual registers" into actual registers on the target hardware, and update any required code as needed
• Cost of instructions is considered during instruction scheduling