From Capabilities to Regions: Enabling Efficient Compilation of Lexical Effect ...
summary
A typed, semantics- and typability-preserving lift-inference translation converts a source language with second-class capabilities into a region calculus that exposes evidence for how many handlers to skip, enabling a CPS pipeline via shift0 and System F. A full compiler pipeline is implemented and evaluated on community benchmarks, reporting 1.6–23× speedups over alternative compilers on many cases while noting a few slowdowns due to runtime differences and exceptional-case optimizer strengths. Artifacts are available.
AI-written summary · about
chapters (9)
- 0:00Effect handlers compilation
- 1:26Effect polymorphism in Effect
- 4:16Lexical scoping via capabilities
- 6:02Translate lexical handlers to shift zero
- 7:24Goal of iterated CPS
- 9:56Evidence measures distance
- 13:49Benchmark results and gains
- 15:12Conclusion and future work
- 15:34Q&A
related talks
21mLift Inference for Lexical Effect Handlers with Second-Class Functions
21mLift Inference for Lexical Effect Handlers with Second-Class Functions
12mCompiling Effect Handlers in Capability-Passing Style
30mEffects,capabilities,and boxes: from scope-based reasoning to type-based reasoning and back
18mParallelism in a Region Inference Context
18mLexical Effect Handlers, Directly
15mEfficient Compilation of Algebraic Effect Handlers
29mLogical Relations for Effect Capabilities