Memory efficient parallel matrix multiplication operation for irregular problems