Compilers That (Intentionally) Change the Result Your Program Produces
Martin C. Rinard
Massachusetts Institute of Technology
Thrusday, November 10, 2011
11:00 a.m., 1000 SEO Building
To view the video recording of this seminar please contact: email@example.com
We present a new program analysis framework that uses probabilistic and statistical reasoning to justify the application of transformations that may change, within probabilistic and statistical?bounds, the result that the program produces. ?We call such transformations accuracy-aware transformations because they are designed to manipulate the accuracy of the computation. ?This broad scope gives the transformations great freedom to influence desirable program properties such as performance and reliability. We show how to apply this approach to justify several program transformations. We also present experimental results that illustrate the potential benefits of this approach.
Martin Rinard is a Professor in the MIT Department of Electrical Engineering and Computer Science and a member of the MIT Computer Science and Artificial Intelligence Laboratory. His research interests include parallel and distributed computing, programming languages, program analysis, program verification, software engineering, and computer systems. Much of his current research focuses on techniques that enable software systems to survive otherwise fatal errors or security attacks. Professor Rinard is an ACM Fellow and holds many awards including the Most Influential Paper in 20 Years Award in the area of Concurrent Constraint Programming (awarded by The Association for Logic Programming in 2004). For more information see: http://people.csail.mit.edu/rinard/