National Compiler Infrastructure Tutorial

June 19th 1:30 p.m. to 5:30 p.m.

The Defense Advanced Research Projects Agency and the National Science Foundation's Division of Advanced Computational Infrastructure and Research (ACIR) is funding the development of a National Compiler Infrastructure (NCI) to support and accelerate the pace of research in compilers and experimental computer science. This tutorial will describe, in some technical detail, SUIF and Zephyr, the key components of the emerging infrastructure and how to use the infrastructure. This tutorial is free and open to any interested PLDI '98 attendees. The material that will be presented includes:

• Several detailed views of the SUIF compiler system, each aimed at a particular class of user:
  • using the system to compile applications,
  • building your own optimization/analysis pass, and
  • extending the system in some simple way.
• A less detailed presentation of the capabilities of SUIF and the axes of extensibility beyond what is covered in detail.
• An overview of Zephyr and the specification languages used in the Zephyr compiler construction system..
  • Abstract Syntax Description Language (ASDL) - a language for specifying intermediate representations.
  • Computer System Description Language (CSDL) - a language for specifying the machine
• Experience using ASDL to connect different compiler passes; Dave Hanson of Microsoft Research will describe his experience using ASDL to connect lcc's front and back ends.
• A description of how to connect vpo to an existing front or middle end of a compiler. In particular, the three interfaces needed to use the vpo optimizer will be described:
  • RTL creation,
  • Main code generation, and
  • Assembly-language support.

The tutorial will also provide an opportunity for interested researchers to provide input to insure that the infrastructure meets the current and future needs of the PLDI research community.

For more information about the tutorial, contact:

• Jack Davidson
  Department of Computer Science
  University of Virginia
  
  or

• Monica Lam
  Department of Computer Science
  Stanford University