Computer Science ETDs

Author

David Mohr

Publication Date

5-1-2009

Abstract

It is inherently difficult for static analyses to make precise decisions about dynamic features of modern object-oriented languages. This makes it more difficult to apply optimizations aggressively. This thesis introduces the D.U.P.O. framework to facilitate the use of dynamic analyses to enable performance optimizations. Since dynamic analyses cannot guarantee complete code coverage, a two part strategy is employed: unit tests are used as a de facto specification, and the programmer provides final verification. The interaction can be kept at a minimum by using the rich information provided by a dynamic analysis. Object inlining is discussed as an example instance of the framework.

Language

English

Keywords

java, optimization, dynamic anaylsis, offline, unit tests, programmer interaction

Document Type

Thesis

Degree Name

Computer Science

Level of Degree

Masters

Department Name

Department of Computer Science

First Committee Member (Chair)

Diwan, Amer

Second Committee Member

Bridges, Patrick G.

Project Sponsors

National Science Foundation

Share

COinS