Comparison of some parallel matrix multiplication algorithms
Jurij Franc Tasič, Matej Zajc, Andrej Košir · 2002
Systolic array for the matrix multiplication is presented by a mesh of the locally connected special processors. In the paper the optimisation of the mapping procedure is presented due to minimisation of the execution time. As a result of this procedure a new mesh structure with shifted data input string is presented. Computation time in the proposed solution is faster than in the original organisation of the array for (N-1) cycles.