Bit-Stealing Made Legal: Compilation for Custom Memory Representations of Algebraic Data Types

Explore a 32-minute conference talk from ICFP 2023 that delves into a novel approach for specifying data layouts of rich data types in low-level programming. Learn about the dual view concept, combining high-level source types with memory types for better control over memory representation. Discover compilation algorithms for constructors, destructors, and pattern matching, along with their correctness proofs. Gain insights into the implementation of these algorithms in the ribbit tool, which compiles to LLVM IR. Examine the potential benefits of this approach for optimizing Algebraic Data Types (ADTs) in low-level programming languages, addressing limitations in current mainstream languages like Rust.