SoCal: A Language for Memory-Layout Factorization of Recursive Datatypes
Published in Preprint, under submission, 2026
SoCal is a formal language for expressing array-of-structs to structure-of-arrays factorization over packed algebraic datatypes. The structure-of-arrays transformation enables selective traversal, with benchmark speedups of 1.1x to 12x, and the type system establishes memory correctness (type safety) for the resulting programs.
Recommended citation: Vidush Singhal, Mikah Kainen, Artem Pelenitsyn, Michael H. Borkowski, Mike Vollmer, Milind Kulkarni. (2026). "SoCal: A Language for Memory-Layout Factorization of Recursive Datatypes." Preprint, under submission.
Download Paper