Note especially pages 52 and 53 of Chapter 3. Consider the following simple version of this in an obvious Fortran like pseudocode
N=100000; Sum=0.;
DO I=0,N-1
Sum = Sum + A(I);
ENDDO
PRINT Sum;
Assume we run this in parallel on a computer with 100 nodes such that the J'th Node holds location 1000*J to 1000*J+999 for J = 0 to 99
Sketch the MPI code that should be used to implement this parallel code basing in on early examples of slides 20 to 29 of MPI Lectures