(a) Suppose that you wish to use a group of threads to form the sum, in...

 

Transcribed Image Text:

(a) Suppose that you wish to use a group of threads to form the sum, in a shared variable result, of an array of numbers. This can be done by giving each thread responsibility for a portion of the array, such that it forms, in its local memory space, the sum of its portion of the array. Outline (using pseudo-code), with explanation, an algorithm that can be executed in parallel by each task in an MPI program. The array should be distributed, and global sum should be computed, using suitable collective operations. Include in your explanation a diagram showing the respective tasks and address spaces, in particular, the placement of the array, and the variables representing local and global sums. Odd Even Sort is based on serial Bubble sort. Consider a hybrid soring algorithm that combines serial merge sort, Parallel bubble sort of blocks and local serial merge of blocks. Analyze this algorithm for the following: A) Parallel Time in terms of N and P B) Fastest Parallel Time in terms of N and P in relation to N C) Speed-Up in terms of P, N D) Fastest Speed-Up in terms of N E) Cost (a) Suppose that you wish to use a group of threads to form the sum, in a shared variable result, of an array of numbers. This can be done by giving each thread responsibility for a portion of the array, such that it forms, in its local memory space, the sum of its portion of the array. Outline (using pseudo-code), with explanation, an algorithm that can be executed in parallel by each task in an MPI program. The array should be distributed, and global sum should be computed, using suitable collective operations. Include in your explanation a diagram showing the respective tasks and address spaces, in particular, the placement of the array, and the variables representing local and global sums. Odd Even Sort is based on serial Bubble sort. Consider a hybrid soring algorithm that combines serial merge sort, Parallel bubble sort of blocks and local serial merge of blocks. Analyze this algorithm for the following: A) Parallel Time in terms of N and P B) Fastest Parallel Time in terms of N and P in relation to N C) Speed-Up in terms of P, N D) Fastest Speed-Up in terms of N E) Cost