Logo EPFL
LAMP
Ecole Polytechnique Fédérale de Lausanne
Advanced Compiler Techniques (Compilation Avancée)
English only
 NEWS
10-06-2005
The exam will take place in room BC 344.
10-06-2005
New slides lecture 11-12, exercises.
08-04-2005
Schedule change. Lectures 11 and 12 moved to 10.06.05.
08-04-2005
Updated slides for lectures 1--2, 4--8.
07-04-2005
Added preliminary slides for Lecture 5 & 6.
29-03-2005
Added Java 1.4 version of eins compiler.
18-03-2005
Added newsgroup epfl.ic.cours.act.
Lectures: Friday 10:15-12:00 and 13:15-15:00, room INM010
Lab: Friday 10:15-12:00, room BC 07/08, and 13:15-15:00, room INF1
Newsgroup: news://epfl.ic.cours.act
  Goals

The student will learn about techniques used to implement high level languages, and compilation techniques used to obtain high performance on modern computer architectures. He or she will also get the opportunity to study one of these techniques in depth and gain experience with implementation issues through a project in the context of an actual compiler.

  • Optimization techniques
    • Data-flow analysis, program optimization, and code generation across basic blocks, procedures, and complete programs.
    • Interprocedural and intraprocedural analysis, intermediate representations, register allocation, and instruction scheduling.
    • Dependence analysis and loop transformations.
  • Implementation of high level languages
    • Implementation of higher order functions, coroutines, and processes.
    • Uniprocessor garbage collector techniques.
    • Virtual machines and the efficient implementation of their interpreters.

The course will include two practical projects.

Lecturer: Dr. Erik Stenman, stairsnone, phone+46-702705615
Assistant: Iulian Dragos, stairsINR321, phone36864
Course schedule
Date 10:15-12:00 and 13:15-15:00
room INM010 (Lecture)
10:15-12:00 and 13:15-15:00
room INF1, BC 07/08 (Project)
11.03.05 Lecture 1 & 2   --  
18.03.05 Lecture 3 & 4 -- Project 1 spec --  
25.03.05 Easter Break  
01.04.05 Easter break  
08.04.05 Lecture 5 & 6  
15.04.05 --   Project 1
22.04.05 Lecture 7 & 8  
29.04.05 --   Project 2
06.05.05 Lecture 9 & 10 --  
13.05.05 --   Project 2
20.05.05 --   Project 2
27.05.05 --   Project 2
03.06.05 --   Project 2
10.06.05 Lecture 11 & 12 --  
17.06.05 Wrap-up @ 10:15 --  

Exams: 13.06.05 -- 16.06.05 in Room BC 344

Exam schedule (paper chosen in brackets)
(Mon) 13.06.2005 (Tu) 14.06.2005 (Wed) 15.06.2005 (Thu) 16.06.2005
09:00-09:20 --- Jerome Caffaro [8] Pascal Jermini [8] Marc Torruella [8]
09:20-09:40 --- Philippe Moret [8] Martin Rubli [8]
09:40-10:00 --- Lukas Rytz [4] Claudia Rappo [2]
10:00-10:20 Ali Al-Shabibi [2] Laurent Haan [8] Florian Hof [3] Laurent Grangier [8]
10:20-10:40 Cedric Jeanneret [8] Jean-Philippe Pellet [4] Jérôme Rousselot [8]
14:00-14:20 Eyal Kaspi [3] Thierry Monney [7] Marc Moser [4]
14:20-14:40 David Wenger [2] Sébastien Cuendet [4] Ouafi Khaled [8]
14:40-15:00 Ariane Pasquier [4] Nils Raning [3] Pascal Perez [8]
15:00-15:20 Jérôme Pasquier [8] Matthieu Dietrich [8]
  1. Introduction to Optimizing Compilers.
    Slides as PDFs 6/page, 1/page.
    Read: M: Ch 1, A: Ch 1, E: Ch 8.1,8.2,8.4
  2. Terminology, foundations of dataflow analysis (partial orders and lattices) & introduction to abstract interpretation.
    Slides as PDFs 6/page, 1/page.
    Read: M: Ch 8.2, 17.1, A: Ch 7.1-7.2, 8.1-8.2, E: Ch 9.1-9.2.

  3. Local (one basic block) optimizations: CSE, constant propagation, copy propagation, dead code elimination, (algebraic simplification, strength reduction). Using dataflow analysis (abstract interpretation) for global optimizations (one CFG): reaching definitions, available expressions, and liveness analysis.
    Slides as PDFs 6/page, 1/page.
    Read: M: Ch 17.2-17.5, A: Ch 7.3-7.5, 8.3-8.5. E: Ch 8.3, Ch. 8.5, Ch 8.6.
  4. Static Single Assignment Form (SSA) & Dominators & SSA-based Dead Code Elimination
    Slides as PDFs 6/page, 1/page.
    Read: M: Ch 18.0-18.1, 19.0-19.3, A: Ch 8.11. E: Ch 9.3. Ch. 10.3.3, Paper 1 & 2.

  5. Sparse Conditional Constant Propagation.
    Slides as PDF 6/p or 1/p.
    Read: M: Ch. 19.3, 19.5, A: Ch 12.6. E: Ch 10.3, 10.4.1., Paper 3.
  6. (Global) Register Allocation
    Slides as PDF 6/p or 1/p.
    Read: M: Ch 11 A: Ch 16. E: Ch 13.

  7. PART 1: Partial Redundancy Elimination.
    Read: A: Ch 13.0-13.2. E: Ch 10.3.5.
    PART 2: Loop Optimizations.
    Read: M: Ch. 18 A: Ch 13.2. .
    PART 3: Lazy Code Motion.
    Read: A: Ch 13.3. E: Ch 10.3.2., Papers 4,5 & 6.
    Slides as PDF 6/p or 1/p.
  8. Code Scheduling
    Slides as PDF 6/p or 1/p.
    Read: M: Ch 20 A: Ch 17. E: Ch 12.

  9. Summary of part 1 of the course: "Optimizations techniques".
    Introduction to part 2: "Implementation of high level languages"
    Implementation of Objects and FPL (higher order functions, laziness). Slides as PDF 6/p or 1/p.
    Read: M: Ch 14, Ch 15 E: Ch 7.10.
  10. Implementation of Concurrency

  11. Slides as PDF 6/p or 1/p.

  12. (Automatic) Memory Management
    Slides as PDF 6/p or 1/p.
    Read: M: Ch 13, (Ch 21) A: (Ch 20.) E: Ch 6.7.
  13. Virtual Machines, Interpretation Techniques, and Just-In-Time Compilers.
    Slides as PDF 6/p or 1/p.

  14. Bits and pieces, such as implementation of exceptions, linkers and loaders.
    Q&A.
    Summary.
    Slides as PDF 6/p or 1/p.
You can download the solution to the exercise session of April 22 (pdf). And Exercise3.pdf, and Exercise4.pdf.

There will be two projects during the course. The performance on these projects will directly influence the grade. Students are allowed and encouraged to work in groups of two.

Like for the first compilation course, we will use Sygeco to manage the course. You are kindly asked to register yourself and your group on this system before the deadline for the first project.

The first project will be to implement a simple register allocator using Sethi-Ullman register allocation for trees. (See Chapter 11.5 in "Modern Compiler Implementation").

The second project will be to implement any number of optimizations in an existing compiler framework. To pass the performance on the benchmark-suite has to be improved by a certain percentage.

Suggested optimizations: see the project detail page.

  Exam

The exam will take place in room BC 344.

There will be an oral exam during the last week of the course. This exam will cover the lecture notes (the slides) and the corresponding information that can be found in all three books. (I.e., the intersection of the information in the books. The lazy student can try to find the minimal set to read, while the highly motivated student can read all three books to get three different views on the material.) It will also cover one of the articles mentioned below (the handouts), which you will need to understand.

You will have to register for a 20 minute slot for the individual exam, by sending an email to Iulian or Erik.

In order to encourage you all to come to the last lecture there will be one written question during that lecture that will also count towards your final grade.

Course book

Reference books (can also be used instead of the official course book)

The slides

The slides from all the lectures are available as one big (ca 5 MB) pdf file.

Handouts

  1. Ron Cytron, Jeanne Ferrante, Barry K. Rosen, Mark N. Wegman, and F. Kenneth Zadeck.
    Efficiently Computing Static Single Assignment Form and the Control Dependence Graph.
    ACM Transactions on Programming Languages and Systems (TOPLAS), 13(4):451-490, October 1991.
  2. Preston Briggs, Keith D. Cooper, Timothy J. Harvey, and L. Taylor Simpson.
    Practical improvements to the construction and destruction of static single assignment form.
    Software: Practice & Experience, 28(8):859-881, July 1998.
  3. Mark N. Wegman and F. Kenneth Zadeck.
    Constant Propagation with Conditional Branches.
    ACM Transactions on Programming Languages and Systems (TOPLAS), 13(2):181-210, April 1991.
  4. Jens Knoop, Oliver Rüthing, and Bernhard Steffen.
    Lazy Code Motion.
    Proceedings of the SIGPLAN Conference on Programming Language Design and Implementation (PLDI), pages 212-223, July 1992.
  5. Karl-Heinz Drechsler and Manfred P. Stadel.
    A Variation of Knoop, Rüthing, and Steffen's Lazy Code Motion.
    SIGPLAN Notices 28(5), pages 29-38, 1993.
  6. Robert Kennedy, Sun Chan, Shin-Ming Liu, Raymond Lo, Peng Tu, and Fred Chow.
    Partial redundancy elimination in SSA form.
    ACM Transactions on Programming Languages and Systems (TOPLAS), 21(3):627-676, May 1999.
  7. Paul Wilson.
    Uniprocessor Garbage Collection Techniques.
    (Read only Sections 1, 2, 3.1-3.3, 4.1-4.3, 4.5-4.7, 8, and 9.)
  8. Anton Ertl and David Gregg.
    Branch prediction and VM interpreters.
  9. Xavier Leroy.
    The ZINC experiment: an economical implementation of the ML language.

The Deadly Sins from P. J. Brown's Writing Interactive Compilers and Interpreters, Wiley 1979.

  • The first deadly sin is to code before you think.
  • The second deadly sin is to assume the user has all the knowledge the compiler writer has.
  • The third deadly sin is not to write proper documentation.
  • The fourth deadly sin is to ignore language standards.
  • The fifth deadly sin is to treat error diagnosis as an afterthought.
  • The sixth deadly sin is to equate the unlikely with the impossible.
  • The seventh deadly sin is to make the encoding of the compiler dependent on its data formats.
  • The eighth deadly sin is to use numbers for objects that are not numbers.
  • The ninth deadly sin is to pretend you are catering for everyone at the same time.
  • The tenth deadly sin is to have no strategy for processing break-ins.
  • The eleventh deadly sin is to rate the beauty of mathematics above the usability of your compiler.
  • The twelfth deadly sin is to let any error go undetected.
  • The thirteenth deadly sin is to leave users to find the errors in your compiler.