Title: Advances in Determinacy Race Detection for Task-Parallel Code
Task parallelism is designed to simplify the job of writing parallel code that can utilize the multicore hardware efficiently. With task parallelism, the programmer expresses the logical parallelism of the computation using high-level parallel control constructs, and lets the the underlying runtime system automates the necessary scheduling and synchronizations. Even with task parallelism, writing correct and efficient parallel code can still be challenging.
One of the challenges is to deal with determinacy races, which occur when logically parallel parts of the computation access the same memory location and at least one of the accesses is a write. Determinacy races are generally bugs in the program since they lead to non-deterministic program behavior --- different schedules of the program can lead to different results. Moreover, they are difficult to detect and debug, since a race may manifest itself in one run but not in another.
In this talk, I will discuss recent advances in detecting determinacy races for task-parallel code. Most prior work on detecting determinacy races has focused on fork-join parallelism, which generates series-parallel dags, a class of computations that is well structured. We will discuss race detection algorithms that can race detect programs with more general structures.
Bio: I-Ting Angelina Lee is an assistant professor in the Computer Science and Engineering department in Washington University in St. Louis. Prior to that, she worked with the Supertech research group in Massachusetts Institute of Technology lead by Professor Charles Leiserson for her graduate study and subsequently as a postdoctoral associate.
Dr. Lee's research focuses on advancing software technologies for parallel computing. She is interested in many aspects of parallel computing, including designing programming models and linguistic constructs to simplify parallel programming, developing runtime and operating system support to execute multithreaded programs efficiently, and building software tools to aid debugging and performance engineering of multithreaded code.