CIS 570 Papers

We will read the following papers in CIS570 this semester. This list is (obviously) not complete. Note that these papers can only be accessed from machines in the penn.edu domain.

[Arnold05] Matthew Arnold, Stephen J. Fink, David Grove, Michael Hind, and Peter F. Sweevey. A Survey of Adaptive Optimization in Virtual Machines. Proceedings of THE IEEE, Vol. 93, No. 2, February 2005. (PDF)

Discussion Questions for Arnold05

These questions are to be answered before class. A paragraph (or so) for each question is sufficient. Email your text to eclewis+arnold05@cis.upenn.edu by noon on Thursday 30 November. Naturally, these questions are just a starting point; we will delve deeper during class. Come prepared with (i) an understanding of the paper, (ii) your assessment of its contributions and deficiencies, and (iii) your own discussion questions.

  • What is the difference between Just-In-Time and "deferred" compilation? Under what circumstances would you prefer the former? Latter?
  • What are the main challenges for practical adaptive optimization? Briefly elaborate on each, including your thoughts on whether/how they can be overcome.
  • Which proposed area of further research (Section VIII) do you find most compelling? Why? Or do you think there is another more important research direction? Why?
[Wegman91] Mark N. Wegman and F. Kenneth Zadeck. Constant Propagation with Conditional Branches. ACM Transactions on Programming Languages and Systems, Vol. 13, No. 2, April 1991. (pdf)

Discussion Questions for Wegman91

These questions are to be answered before class. A paragraph (or so) for each question is sufficient. Email your text to eclewis+wegman91@cis.upenn.edu by noon on Tuesday 10 October. Naturally, these questions are just a starting point; we will delve deeper during class. Come prepared with (i) an understanding of the paper, (ii) your assessment of its contributions and deficiencies, and (iii) your own discussion questions.

  • What is the difference (in terms of effect, not implementation) between simple constants analysis and conditional constants analysis?
  • How is sparse constant propagation improved via the use of SSA (versus du chains)?
  • Why does SCC require two work lists while SSC only requires one?
[Sites93] Richard L. Sites, Anton Chernoff, Matthew B. Kirk, Maurice P. Marks, and Scott G. Robinson. Binary Translation. CACM, February, 1993. (pdf)

Discussion Questions for Sites93

These questions are to be answered before class. A paragraph (or so) for each question is sufficient. Email your text to eclewis+sites93@cis.upenn.edu by noon on Tuesday 19 September. Naturally, these questions are just a starting point; we will delve deeper during class. Come prepared with (i) an understanding of the paper, (ii) your assessment of its contributions and deficiencies, and (iii) your own discussion questions.

  • What are the pros and cons of bounded translation systems? Open-ended translation systems?
  • Compare and contrast VEST/TIE and mx/mxr.
  • If you were an author on this paper, what would you have done to improve the evaluation? Why? (i.e., what would your evaluation demonstrate that the existing evaluation does not?)
E Christopher Lewis
Last modified: Tue Nov 28 09:01:34 EST 2006