Explore a 25-minute conference talk from PLDI 2023 that addresses the challenge of priority inversions in concurrent programming. Delve into a novel approach combining static and dynamic techniques to prevent priority inversions when using mutexes and condition variables. Learn about a type system ensuring safe usage of condition variables while dynamic techniques adjust thread priorities at runtime. Examine the soundness proof based on cost models for parallel programs and discover practical implementations in Rust and C++. Gain insights into the application of this system through case studies, including a port of the Memcached object server. Presented by researchers from Illinois Institute of Technology, Washington University in St. Louis, and Carnegie Mellon University, this talk offers valuable knowledge for those interested in concurrent programming, priority management, and synchronization mechanisms.
Overview
Syllabus
[PLDI'23] Responsive Parallelism with Synchronization
Taught by
ACM SIGPLAN