Show simple item record

dc.contributor.advisorGriswold, Ralph E.en_US
dc.contributor.authorWalker, Kenneth William.
dc.creatorWalker, Kenneth William.en_US
dc.date.accessioned2011-10-31T17:42:47Zen
dc.date.available2011-10-31T17:42:47Zen
dc.date.issued1991en_US
dc.identifier.urihttp://hdl.handle.net/10150/185599en
dc.description.abstractThere are many optimizations that can be applied while translating Icon programs. These optimizations and the analyses needed to apply them are of interest for two reasons. First, Icon's unique combination of characteristics requires developing new techniques for implementing them. Second, these optimizations are used in variety of languages and Icon can be used as a medium for extending the state of the art. Many of these optimizations require detailed control of the generated code. Previous production implementations of the Icon programming language have been interpreters. The virtual machine code of an interpreter is seldom flexible enough to accommodate these optimizations and modifying the virtual machine to add the flexibility destroys the simplicity that justified using an interpreter in the first place. These optimizations can only reasonably be implemented in a compiler. In order to explore these optimizations for Icon programs, a compiler was developed. This dissertation describes the compiler and the optimizations it employs. It also describes a run-time system designed to support the analyses and optimizations. Icon variables are untyped. The compiler contains a type inferencing system that determines what values variables and expression may take on during program execution. This system is effective in the presence of values with pointer semantics and of assignments to components of data structures. The compiler stores intermediate results in temporary variables rather than on a stack. A simple and efficient algorithm was developed for determining the lifetimes of intermediate results in the presence of goal-directed evaluation. This allows an efficient allocation of temporary variables to intermediate results. The compiler uses information from type inferencing and liveness analysis to simplify generated code. Performance measurements on a variety of Icon programs show these optimizations to be effective.
dc.language.isoenen_US
dc.publisherThe University of Arizona.en_US
dc.rightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.en_US
dc.subjectCompilers (Computer programs)en_US
dc.titleThe implementation of an optimizing compiler for Icon.en_US
dc.typetexten_US
dc.typeDissertation-Reproduction (electronic)en_US
dc.identifier.oclc704393866en_US
thesis.degree.grantorUniversity of Arizonaen_US
thesis.degree.leveldoctoralen_US
dc.contributor.committeememberDowney, Peter J.en_US
dc.contributor.committeememberDebray, Saumyaen_US
dc.identifier.proquest9200048en_US
thesis.degree.disciplineComputer Sciencesen_US
thesis.degree.disciplineGraduate Collegeen_US
thesis.degree.namePh.D.en_US
refterms.dateFOA2018-08-23T04:46:39Z
html.description.abstractThere are many optimizations that can be applied while translating Icon programs. These optimizations and the analyses needed to apply them are of interest for two reasons. First, Icon's unique combination of characteristics requires developing new techniques for implementing them. Second, these optimizations are used in variety of languages and Icon can be used as a medium for extending the state of the art. Many of these optimizations require detailed control of the generated code. Previous production implementations of the Icon programming language have been interpreters. The virtual machine code of an interpreter is seldom flexible enough to accommodate these optimizations and modifying the virtual machine to add the flexibility destroys the simplicity that justified using an interpreter in the first place. These optimizations can only reasonably be implemented in a compiler. In order to explore these optimizations for Icon programs, a compiler was developed. This dissertation describes the compiler and the optimizations it employs. It also describes a run-time system designed to support the analyses and optimizations. Icon variables are untyped. The compiler contains a type inferencing system that determines what values variables and expression may take on during program execution. This system is effective in the presence of values with pointer semantics and of assignments to components of data structures. The compiler stores intermediate results in temporary variables rather than on a stack. A simple and efficient algorithm was developed for determining the lifetimes of intermediate results in the presence of goal-directed evaluation. This allows an efficient allocation of temporary variables to intermediate results. The compiler uses information from type inferencing and liveness analysis to simplify generated code. Performance measurements on a variety of Icon programs show these optimizations to be effective.


Files in this item

Thumbnail
Name:
azu_td_9200048_sip1_w.pdf
Size:
5.963Mb
Format:
PDF

This item appears in the following Collection(s)

Show simple item record