Explicit Effects and Effect Constraints in ReML

on YouTube ↗21m

" summary

ReML is a Standard ML dialect that makes memory and other effects explicit while staying backwards compatible; it integrates region-polymorphic inference with the MLKit compiler, emits region annotations for allocation/deallocation, and enforces disjointness effect constraints so parallel subcomputations provably avoid interfering allocations or mutations. The paper gives a formal type-and-effect system with a soundness proof, an implementation linked to MLKit, evaluated artifacts and benchmarked parallel region-based examples that the artifact archive reproduces.

AI-written summary · about

" chapters (10)

1. 0:00ReML introduction
2. 2:09Region-based memory overview
3. 4:40ReML language design
4. 6:29Region inference and compiler
5. 9:44Region parameters and calls
6. 11:49Polymorphic functions and copying
7. 13:11Effect constraints
8. 14:39Parallelism and disjoint effects
9. 16:02Formalization and benchmarks
10. 17:06Q&A