Recursive Algebraic Coloring Engine (RACE) published

Finally, our work on the Recursive Algebraic Coloring Engine (RACE) is officially published in the ACM Transactions on Parallel Computing:

C. L. Alappat, G. Hager, O. Schenk, J. Thies, A. Basermann, A. R. Bishop, H. Fehske, and G. Wellein: A Recursive Algebraic Coloring Technique for Hardware-Efficient Symmetric Sparse Matrix-Vector Multiplication. ACM Trans. Parallel Comput. 7(3), Article 19 (June 2020), 37 pages. Available with Open Access. DOI: 10.1145/3399732.

Our PhD student Christie Alappat had already scored some success with preliminary work on RACE in the ACM 2018 Student Research Competition, but this paper now details the whole algorithm and applies it to the important symmetric sparse MVM kernel. It solves some notorious performance problems of traditional coloring approaches in terms of data access and parallel efficiency; the resulting symmetric spMVM operation outperforms existing solutions, including Intel’s MKL, for a set of 31 test matrices.