- Test for correct communication mpirun -np 20 ./int_ring 1
0 rounds, rank 1 received from 0 the message 0
0 rounds, rank 2 received from 1 the message 1
0 rounds, rank 3 received from 2 the message 3
0 rounds, rank 4 received from 3 the message 6
0 rounds, rank 5 received from 4 the message 10
0 rounds, rank 6 received from 5 the message 15
0 rounds, rank 7 received from 6 the message 21
0 rounds, rank 8 received from 7 the message 28
0 rounds, rank 9 received from 8 the message 36
0 rounds, rank 11 received from 10 the message 55
0 rounds, rank 12 received from 11 the message 66
0 rounds, rank 13 received from 12 the message 78
0 rounds, rank 10 received from 9 the message 45
0 rounds, rank 14 received from 13 the message 91
0 rounds, rank 15 received from 14 the message 105
0 rounds, rank 16 received from 15 the message 120
0 rounds, rank 17 received from 16 the message 136
0 rounds, rank 18 received from 17 the message 153
0 rounds, rank 0 received from 19 the message 190
0 rounds, rank 19 received from 18 the message 171
- Test for latency about CIMS crunchy1.cims.nyu.edu with 20 nodes
mpirun -n 20 ./int_ring 1000
1000 rounds, rank 0, communication time 4.636791 sec.
Latency = 2.3 x 10^-4 sec
mpirun -n 10 ./int_ring 1000
1000 rounds, rank 0, communication time 0.014833 sec.
Latency = 1.5 x 10^-6 sec
- Test for large array data[500,000]
mpirun -n 20 ./int_ring 100
100 rounds, rank 0, communication time 59.738771 sec.
Bandwidth = 2 x 100 x 20 / 59.738771 = 66.95 MB/s
mpirun -n 10 ./int_ring 100
100 rounds, rank 0, communication time 14.280453 sec.
ndwidth = 2 x 100 x 10 / 59.738771 = 60.10 MB/s