AUTOMAP at OOPSLA 2024

Our paper AUTOMAP: Inferring Rank-Polymorphic Function Applications with Integer Linear Programming, with Robert Schenck, Nikolaj Hey Hinnerskov, Troels Henriksen, and Magnus Madsen, was accepted for presentation at OOPSLA 2024 in Pasadena, with the talk scheduled for October 23.

Array languages often let an operation on a single value work on a whole array automatically. AUTOMAP brings that convenience to a statically typed language: it works out where to insert mapping and replication operations, while retaining compile-time type checking. The system uses integer linear programming to find a solution requiring as few inserted operations as possible. The paper develops the theory and describes an implementation in Futhark.

Read the paper (PDF) or see the published article and OOPSLA programme.