Concurrent Mutation must go
summary
Argument: concurrent mutation causes the costly read-reclaim race so the simple claim is remove concurrent mutation for temporal safety. Evidence: survey of practical techniques (GC, reference counting, hazard pointers, epoch reclamation) shows tradeoffs where reads become expensive or reclamation delayed. Construction: a language design (Project Rona) partitions heap into regions-as-capabilities so regions can use tracing, RC or arenas independently and a matching concurrency model removes memory shifting under readers, simplifying reclamation and enabling predictable memory strategies.
AI-written summary · about
chapters (9)
- 0:00Temporal safety challenges
- 2:08Read reclaim race issues
- 6:02Solutions for read reclaim race
- 8:06Ownership types in Rust
- 10:14External uniqueness and partitioning
- 11:41Concurrent mutation must go
- 12:05Q&A: Rust and separation logic
- 15:16Q&A: Capabilities and memory tagging
- 17:25Q&A: Concurrency and distrust
related talks
55m“A lock-free concurrency toolkit for deferred reclamation and optimistic speculation"
58m“The Landscape and Exciting New Future of Safe Reclamation for High Performance”
27mThe Mutability Matrix of Pain
23mAn Operational Approach to Library Abstraction under Relaxed Memory Concurrency
30mA Tale of Shared Memory Concurrency in Programming Languages
17mConcurrent Immediate Reference Counting
38mImmutability is everything!
51mMemory Model: Get your shared data under control