output:  about 86% of peak performace (128*1.2 GB/s = 153.6 GB/s) in best case 

[gshi@blade1 eib_bandwidth]$ numactl -l --cpunodebind=0 ./eib_bandwidth -p 0 
We are testing pattern: 
0<->2
1<->3
2<->0
3<->1
time spent 0.111772, EIB bandwidth=117.267294 GB/s

The program has successfully executed.
[gshi@blade1 eib_bandwidth]$ numactl -l --cpunodebind=0 ./eib_bandwidth -p 1
We are testing pattern: 
0<->4
1<->5
2<->6
3<->7
time spent 0.099559, EIB bandwidth=131.652588 GB/s

The program has successfully executed.
[gshi@blade1 eib_bandwidth]$ numactl -l --cpunodebind=0 ./eib_bandwidth -p 2
We are testing pattern: 
0<->1
1<->0
2<->3
3<->2
time spent 0.099164, EIB bandwidth=132.177000 GB/s

The program has successfully executed.
[gshi@blade1 eib_bandwidth]$ numactl -l --cpunodebind=0 ./eib_bandwidth -p 3
We are testing pattern: 
0<->3
1<->2
2<->1
3<->0
time spent 0.100136, EIB bandwidth=130.893984 GB/s

The program has successfully executed.
[gshi@blade1 eib_bandwidth]$ numactl -l --cpunodebind=0 ./eib_bandwidth -p 4
We are testing pattern: 
0<->7
1<->6
2<->5
3<->4
time spent 0.220670, EIB bandwidth=59.397290 GB/s

The program has successfully executed.
[gshi@blade1 eib_bandwidth]$ numactl -l --cpunodebind=0 ./eib_bandwidth -p 5
We are testing pattern: 
0<->5
1<->4
2<->7
3<->6
time spent 0.181861, EIB bandwidth=72.072627 GB/s

The program has successfully executed.
[gshi@blade1 eib_bandwidth]$ numactl -l --cpunodebind=0 ./eib_bandwidth -p 6
We are testing pattern: 
0<->6
1<->7
2<->4
3<->5
time spent 0.097371, EIB bandwidth=134.610921 GB/s

The program has successfully executed.