Efficient Parallel Functional Programming with Effects

Explore a groundbreaking approach to efficient parallel functional programming with effects in this 21-minute video presentation from PLDI 2023. Delve into techniques that allow unrestricted mutation in functional programs while maintaining performance and safety. Learn how researchers from Carnegie Mellon University tackle the challenge of entanglement in hierarchical memory architectures, distinguishing between disentangled and entangled objects to optimize memory management. Discover a formalized semantics for entanglement at the object level and understand the time and space cost metrics associated with this approach. Examine the implementation of these techniques in the MPL compiler for Parallel ML, supporting all language features including unrestricted effects. Analyze experimental results that demonstrate competitive performance with languages like C++, Go, Java, and OCaml, showcasing how functional programming's safety benefits can be combined with high performance in parallel computing environments.

[PLDI'23] Efficient Parallel Functional Programming with Effects