Question

Text: Complete implementing the following barrier algorithms in a single MPI program. Use output messages printed by the master process to indicate the beginning and ending of each barrier algorithm. Implement the tree barrier algorithm. Note that the barrier algorithm has separate arrival and departure phases. The output messages printed by the master process must clearly mark each important time point. 2. Implement the butterfly barrier described in Section 6.1.4. Note that in the butterfly barrier, the arrival phase and departure phase are merged into one phase. 3. Investigate the time taken by each of your barrier algorithms by using the timing functions, e.g.: t1 = MPI_Wtime(); butterfly_barrier(); // or tree_barrier() - call to your tree and butterfly barrier algorithms. t2 = MPI_Wtime(); printf("Elapsed time = %d seconds ", t2-t1); Please note that timing is only taken by the master process. The master process should print out the timing information of each barrier algorithm on display. You can assume the number of processes is a power of 2, e.g., 2, 4, 8, 16, etc. The only MPI routines you can use in your codes are MPI_send, MPI_recv, MPI_sendrecv, or their asynchronous variants. Please submit in one source code C after the last step and screenshot the output.

          Text: Complete implementing the following barrier algorithms in a single MPI program. Use output messages printed by the master process to indicate the beginning and ending of each barrier algorithm. Implement the tree barrier algorithm. Note that the barrier algorithm has separate arrival and departure phases. The output messages printed by the master process must clearly mark each important time point. 

2. Implement the butterfly barrier described in Section 6.1.4. Note that in the butterfly barrier, the arrival phase and departure phase are merged into one phase. 

3. Investigate the time taken by each of your barrier algorithms by using the timing functions, e.g.: 
t1 = MPI_Wtime();
butterfly_barrier(); // or tree_barrier() - call to your tree and butterfly barrier algorithms.
t2 = MPI_Wtime();
printf("Elapsed time = %d seconds
", t2-t1);

Please note that timing is only taken by the master process. The master process should print out the timing information of each barrier algorithm on display. You can assume the number of processes is a power of 2, e.g., 2, 4, 8, 16, etc. The only MPI routines you can use in your codes are MPI_send, MPI_recv, MPI_sendrecv, or their asynchronous variants. Please submit in one source code C after the last step and screenshot the output.

Added by James P.

Question

Please give Ace some feedback