Normalize, transpose, and distribute: An automatic approach for handling nonscalars

Abstract

SequenceL is a concise, high-level language with a simple semantics that provides for the automatic derivation of many iterative and parallel control structures. The semantics repeatedly applies a “Normalize-Transpose-Distribute” operation to functions and operators until base cases are discovered. Base cases include the grounding of variables and the application of built-in operators to operands of appropriate types. This article introduces the results of a 24-month effort to reduce the language to a very small set of primitives. Included are comparisons with other languages, the formal syntax and semantics, and the traces of several example problems run with a prototype interpreter developed in 2006.