Consider the following multithreaded algorithm for performing pairwise addition on n-element arrays A[1 . . n] and B[1. . n], storing the sums in C[1. . n].

SUM-ARRAYS (A, B, C)

a. Rewrite the parallel loop in SUM-ARRAYS using nested parallelism (spawn and sync) in the manner of MAT-VEC-MAIN-LOOP. Analyze the parallelism of your implementation.

Consider the following alternative implementation of the parallel loop, which contains a value grain-size to be specified:

SUM-ARRAYS€² (A, B, C)

ADD-SUBARRAY (A, B, C, I, j)

b. Suppose that we set grain-size = 1. What is the parallelism of this implementation?

c. Give a formula for the span of SUM-ARRAYS€² in terms of n and grain-size. Derive the best value for grain-size to maximize parallelism.

Transcribed Image Text:

1 parallel for i = 1 to A.length C[i] = A[i] + B[i] 1 n = A.length 2 grain-size = ? = [n/grain-size] 4 for k = 0 to r – 1 spawn ADD-SUBARRAY(A, B, C,k · grain-size + 1, // to be determined 3 r min((k + 1) · grain-size, n)) 6 sync