Generalizing matrix multiplication for efficient computations on modern computers

Abstract. Recent advances in computing allow taking new look at ma-trix multiplication, where the key ideas are: decreasing interest in re-cursion, development of processors with thousands (potentially millions) of processing units, and influences from the Algebraic Path Problems. In this context, we propose a generalized matrix-matrix multiply-add (MMA) operation and illustrate its usability. Furthermore, we elaborate the interrelation between this generalization and the BLAS standard