BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20131121T204500Z
DTEND:20131121T210000Z
LOCATION:601/603
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: As scientific computation continues to scale, it is crucial to use floating-point arithmetic as efficiently as possible. Lower precision allows streaming architectures to perform more operations per second and can reduce memory bandwidth pressure. However, using a precision that is too low will result in inaccurate results. Thus, developers must balance speed and accuracy when choosing the floating-point precision of their subroutines and data structures.=0A=0AI have developed program analysis techniques to help developers learn about the runtime floating-point behavior of their programs, with the goal of helping them make decisions concerning the choice of precision in implementation. I have developed automated methods for cancellation detection, range tracking, mixed-precision configuration, mixed-precision search, and bit-level sensitivity analysis. This body of work improves the state of the field of program analysis and scientific computing by providing insights that were not previously available using automated analysis techniques.
SUMMARY:Automated Floating-point Program Analysis
PRIORITY:3
END:VEVENT
END:VCALENDAR