Q-Logic IB6054601-00 D Manuel D’Utilisation
A – Benchmark Programs
Benchmark 3: Messaging Rate Microbenchmarks
Benchmark 3: Messaging Rate Microbenchmarks
A-4
IB6054601-00 D
Q
benchmark (as shown in the example above). It has been enhanced with the
following additional functionality:
following additional functionality:
■
Messaging rate reported as well as bandwidth
■
N/2 dynamically calculated at end of run
■
Allows user to run multiple processes per node and see aggregate bandwidth
and messaging rates
and messaging rates
The benchmark has been updated with code to dynamically determine which
processes are on which host. This is an example showing the type of output you
will see when you run
processes are on which host. This is an example showing the type of output you
will see when you run
mpi_multibw
:
$ mpirun -np 8 ./mpi_multibw
This will run on four processes per node. Typical output might look like:
# PathScale Modified OSU MPI Bandwidth Test
(OSU Version 2.2, PathScale $Revision: 1.1 $)
# Running on 4 procs per node
# Size
Aggregate Bandwidth (MB/s) Messages/s
1
8.150462
8150461.697283
2
16.693747
8346873.631841
4
33.086567
8271641.814960
8
66.733488
8341686.016159
16
108.097082
6756067.602089
32
213.733132
6679160.388156
64
389.544112
6086626.744516
128 569.671531
4450558.832794
256
725.826904
2835261.345093
512
839.450014
1639550.807757
1024
913.428063
892019.592596
2048
954.482747
466056.028717
4096
954.474461
233025.991467
8192
954.452712
116510.340772
16384
954.496729
58257.857017
32768
954.547225
29130.469523
65536
949.074433
14481.726572
131072
951.786548
7261.555084
262144 952.193849
3632.331272
524288
952.391830
1816.543255
1048576
952.490368
908.365600
2097152
952.539382
454.206172
4194304
952.566591
227.109573
Searching for N/2 bandwidth. Maximum Bandwidth of 954.547225
MB/s...
Found N/2 bandwidth of 476.993060 MB/s at size 94 bytes
This microbenchmark is available and can be downloaded from the QLogic website: