Rounding

dc.contributor.advisorGrossman, Dan
dc.contributor.advisorTatlock, Zach
dc.contributor.authorZorn, Bill
dc.date.accessioned2022-01-26T23:23:22Z
dc.date.available2022-01-26T23:23:22Z
dc.date.issued2022-01-26
dc.date.submitted2021
dc.descriptionThesis (Ph.D.)--University of Washington, 2021
dc.description.abstractComputer number systems are one of the most fundamental interfaces between software and hardware, but despite recent interest they are rarely studied. We present a suite of tools and techniques to make it easier for both application-level software developers and hardware architects to study number systems and design new ones. Our key theoretical contribution is the use of rounding as an abstraction to describe the behavior of a wide variety of number systems in terms of real arithmetic. By leveraging this abstraction, we can build tools that simulate the behavior of many different number systems, efficiently track error through large computations, and automatically search for number system configurations that are optimized for a particular application.
dc.embargo.termsOpen Access
dc.format.mimetypeapplication/pdf
dc.identifier.otherZorn_washington_0250E_23717.pdf
dc.identifier.urihttp://hdl.handle.net/1773/48230
dc.language.isoen_US
dc.rightsCC BY
dc.subjectFloating-point
dc.subjectNumber systems
dc.subjectRounding
dc.subjectComputer science
dc.subject.otherComputer science and engineering
dc.titleRounding
dc.typeThesis

Files

Original bundle

Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
Zorn_washington_0250E_23717.pdf
Size:
1.61 MB
Format:
Adobe Portable Document Format