Fall 2005 Programming Language Seminar: GCC Internals

Department of Computer Science

Synopsis:
The GNU Compiler Collection (GCC) is a widely popular optimizing compiler for a number of architectures that supports C, C++, Java, Objective C, FORTRAN, and Ada.

This seminar will be an examination of how GCC works under the hood. At each meeting, we will walk through parts of the GCC source tree and try to make sense of them. In particular, we will focus on changes to the Java front-end and library required to make them a suitable real-time Java platform. We'll also investigate how optimization passes are coded, how the Boehm garbage collector operates, and the requisite GCC basics like the build system, the tree-based intermediate representations, structure of front-ends and back-ends, etc. Other topics of interest to the seminar attendees will also be explored.

Many researchers are intimidated by GCC's size, initial learning curve, and the fact that its structure isn't that of a clean textbook compiler. However, GCC is a good fit for many research implementations in the area of compilation, languages, and runtime systems, and its full support of language features and numerous back-ends make it attractive for many projects. The purpose of this seminar is to demystify GCC so that it can be used as a basis for research and experimentation.

We will be using the most recent version of the GCC compiler suite, which is (at the time of writing) 4.0.1. We'll probably also be peeking at 4.1 development work.

Important Information:

Seminar Calendar:
The Fall 2005 programming language seminar meets Mondays at 1:00pm in Lopata 400.

Links:

• GCC home page
• Wikipedia entry for GCC, containing historical and overview information etc.
• The GCC wiki, containing lots of good technical information about GCC, plans for the future, etc.
• Steven Muchnick's book on compiler design and implementation, with a lot of great stuff about optimization and other advanced topics. Buy it. Read it. Love it. (Though if you get the same printing I have, I recommend you check out the errata.)
• Speaking of textbooks, you may also want to check out Engineering a Compiler by Cooper and Torczon if you need to brush up on your parsing theory and other compiler topics. I've heard great things about it, but I don't have it myself. (Errata.)
• Static single assignment (SSA) paper: Cytron et al. Efficiently computing static single assignment form and the control dependence graph.
• McCAT paper (using SIMPLE, on which GIMPLE is based): Hendren et al. Designing the McCAT Compiler Based on a Family of Structured Intermediate Representations.
• Paper on GCC IR: Merrill. GENERIC and GIMPLE: A New Tree Representation for Entire Functions.
• Paper on retargettable YC compiler and its use of RTL, machine descriptions, and instruction matching: Davidson and Fraser. Code selection through object code optimization.
• Bowen Alpern et al. Efficient Implementation of Java Interfaces: Invokeinterface Considered Harmless
• (Old) paper by Hans-J. Boehm describing various aspects of the Boehm collector: Hans-Juergen Boehm. Space-efficient conservative garbage collection. Even older paper: Hans-Juergen Boehm and Mark Weiser. Garbage collection in an uncooperative environment.
• ELF: Executable and Linkable Format
• The GCC Tree Browser
• Pentium 4 reference
• PowerPC reference
• ACM Digital Library -- access to full papers available from the Washington University network

Looking for something else?

• Programming language seminar information prior to Spring 2002
• Current and prior programming language seminar information
• Courses in which I'm involved