Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore the theoretical limits of garbage collection algorithms through a rigorous mathematical proof demonstrating that no "holy grail" garbage collector can exist. Learn how this 14-minute conference presentation from OOPSLA 2025 by Matthew Sotoudeh from Stanford University provides the first formal proof that every garbage collection algorithm implementing a realistic mutator-observer interface must exhibit pathological behavior under certain conditions. Discover how the research establishes mathematical certainty that all GC algorithms face an unavoidable trade-off: they must either introduce long pause times or reject allocations despite available memory when programs create heap cycles and operate near memory limits. Understand the formal reductions between garbage collection and dynamic graph connectivity problems in complexity theory, creating a bridge that allows algorithms and lower bounds from either field to transfer to the other. Examine the implications for language designers who must choose between accepting pathological scenarios with heuristic mitigation strategies like generational collection, or restricting programs through type systems that prevent cycles or require memory overprovisioning. Gain insight into techniques adapted from graph data structures that can build garbage collectors with improved worst-case guarantees for memory-constrained scenarios, though with practical overhead limitations that prevent recommendation for typical use cases.
