1. CSE 503: Software Engineering Winter 2010 Lecturer: Michael Ernst

2. 503 Software Engineering Research Not: how to write good software – and get a good job at Amazon/Google/Microsoft Research methods and ideas in SE – (this may make you a more thoughtful developer) In alternating years: – broad overview – in-depth examination of one topic Examples: software evolution, program analysis

3. Type systems Checking Inference Non-standard type systems Example: view type system as a set of constraints to compute legal refactorings

4. Abstract interpretation A.k.a. dataflow analysis “Run” your program statically – Choose an abstract domain; e.g., { +, 0, - } – Assign semantics to operators – Start at beginning of program – Examine possible values of variables Used daily on aeronautics software

5. Model checking In simplest terms, exhaustive testing How to make this scale? – Choose abstractions that lose just the right amount of precision – Efficient encodings

6. Analysis back-ends Binary Decision Diagrams (BDDs) Reduction to SAT – 1979: “Problem X reduces to SAT, so it is hard.” – 2009: “Problem X reduces to SAT, so it is easy.” Datalog

7. Dynamic analysis Testing Type inference Model creation Counterexample-guided refinement Concolic execution

8. Applications Security Correctness Performance Rapid development System analysis Maintenance and evolution

9. Broader themes Precision vs. performance Power vs. transparency Static vs. dynamic Tuning analysis to the real problem

10. Format 50% of lectures: classic background 50% of lectures: current research Homework: – 1 in-class presentation – Group project to put the ideas into practice Makes you a better researcher, in any field

11. Who cares? Intellectually exciting and deep Spans both “hard” and “soft” areas of computing Connections to PL, security, systems, architecture, databases, … (many more!) Quals credit