1. Advanced Compilers CMPSCI 710 Spring 2003 Lecture 1
Emery Berger
University of Massachusetts, Amherst

2. Administrivia Class format: Course website: Grades: Materials:
I present material & work examples
You take notes, read book & papers
Course website:
Grades:
40% - 4 homework assignments (due dates to be announced)
60% - project
Materials:
Required text: Muchnick, Advanced Compiler Design Implementation (ACDI)
Supplementary material: selected papers

3. Project Stuff 1 to 2 person teams Implement optimization/analysis in:
Jikes RVM (IBM’s research Java compiler)
Broadway (UTexas “metacompiler”)
other (subject to approval)
Due dates:
02/11/03: One-page project description.
02/25/03: 2-4 page project design.
03/25/03: Project implementation review.
04/29/03: Implementation due.
05/06-13/03: In-class presentations.
05/13/03: Project report.

4. Why Compilers Matter

5. What are Compilers, Anyway?
Compiler: translates program in one language to executable program in other language
Typically lowers abstraction level
E.g., Java/C++ to assembler
Optimizing compiler:
Misnomer!
Optimal compilation intractable
Improves program performance

6. Performance Anxiety But does performance really matter?
Computers are really fast
Moore’s law (roughly): hardware performance doubles every 18 months
Real bottlenecks lie elsewhere:
Disk
Network
Human! (think interactive apps)
At most 120 cps typing
Waste time “thinking”

7. Compilers Don’t Help Much
Do compilers improve performance anyway?
Proebsting’s law (Todd Proebsting, Microsoft Research):
Difference between optimizing and non-optimizing compiler ~ 4x
Assume compiler technology represents 36 years of progress (actually more)
Compilers double program performance every 18 years!
Not quite Moore’s Law…

8. Argumentum Ad Absurdum
Compilers don’t help much (Proebsting’s Law)
Don’t use optimizer
Computers are getting faster (Moore’s Law)
Buy a new one – helps the economy too!
Or: wait until one’s fast enough, then run program
Hmmm…

9. A Big BUT Why use high-level languages anyway?
Easier to write & maintain
Safer (think Java)
More convenient (think libraries, GC…)
But: people will not accept massive performance hit for these gains
Compile with optimization!
Still use C and C++!!
Hand-optimize their code!!!
Even write assembler code (gasp)!!!!
Apparently performance does matter…

10. Why Compilers Matter
Key part of compiler’s job: make the costs of abstraction reasonable
Remove performance penalty for:
Using objects
Safety checks (e.g., array-bounds)
Writing clean code (e.g., recursion)
Use program analysis to transform code: primary topic of this course

11. Analysis & The Holy Grail
Other great uses for program analysis:
Static error-checking
Detect information leaks
Avoid security holes
Informing runtime system (e.g., GC)
Even better optimizations!
E.g., locality-improving transformations (“memory wall”)
Holy Grail for compiler research:
Programmer: writes simple but O(2^n) algorithm
Compiler: changes it to O(n log n) (or O(1)!)

12. Levels of Analysis (in order of increasing detail & complexity)
Local (single-block) [1960’s]
Straight-line code
Simple to analyze; limited impact
Intraprocedural [1970’s – today]
Whole procedure
Dataflow & dependence analysis
Interprocedural [late 1970’s – today]
Whole-program analysis
Tricky:
Very time and space intensive
Hard for some PL’s (e.g., Java)

13. Optimization = Analysis + Transformation
Key analyses:
Control-flow
if-statements, branches, loops, procedure calls
Data-flow
definitions and uses of variables
Representations:
Control-flow graph
Control-dependence graph
Def/use, use/def chains
SSA (Static Single Assignment)

14. Next Time Read ACDI Chapters 1, 2, 7
Next meeting might be in CS building