Par Lab Seminar Series: Madan Musuvathi, Microsoft Research
Lecture | October 12 | 11 a.m.-12:30 p.m. | Soda Hall, Wozniak Lounge, 438
Madan Musuvathi, Microsoft Research
Par Lab Seminar for Tuesday, Oct 12 at 11am-12:30pm in the Woz, 438 Soda
Speaker: Madan Musuvathi, Microsoft Research
Title: DRFx: Simplifying the Semantics of Concurrent Programs Through Efficient Data-Race Detection
Abstract:
Modern programming languages, such as Java and C++, provide weak or no semantics to programs with data races. This compromises the safety and debuggability for large programs, which are likely to have data races. In this talk, I will present DRFx, a language runtime that provides fail-stop semantics to racy programs by using a low-overhead data-race detector. Under DRFx, programs either execute with sequentially consistent semantics or will be terminated by a “memory-model” exception. DRFx will never terminate a data-race-free program. I will discuss the challenges and tradeoffs in building such a language runtime and provide (correctness and performance) arguments as to why such runtimes require support from *both* the compiler and the hardware.
This is joint work with Dan Marino (UCLA), Abhay Singh (UMich), Todd Millstein (UCLA), and Satish Narayanasamy (UMich).
Bio:
Madan Musuvathi is a researcher at Microsoft Research. His research interests are primarily in the analysis of complex large-scale systems and includes systems, networking, program analysis, model checking, verification, and theorem proving. He spends a lot of time at Microsoft building analysis tools to improve the productivity of software developers and testers. He received his Ph.D. from Stanford University in 2004.
roxana@eecs.berkeley.edu, 510-643-1455