Comparison of Multi-Threading and Message-Passing

MPI Comm size(MPI COMM WORLD, nproc) for (i=0 i nproc i++)  

MPI Allgatherv (c local, local n, MPI DOUBLE, c, n on proc, offset on proc, MPI DOUBLE, MPI COMM WORLD)  

A routine to compute the matrix vector product c = Ab using message-passing. The input data b is replicated, and A is distributed with local rows in A local. The output data c is replicated among all of the processes. The number of processes is designated nproc, and local n is the number of elements of the c vector that wiU be computed locally. 

---