Recently I installed OpenMPI 1.4.3 on my cluster. I found that
SYSTEM CPU is higher than older versions of 1.4.X, when I ran
our FEM program. Further more, OMPI_1.4.3 is a little bit slower
than 1.4.2. What is the differnece between these two versions, or
what affects the "SYSTEM CPU" and execution speed?
USER CPU SYSTEM CPU
OpenMPI 1.4.2 99% 1%
OpenMPI 1.4.3 95% 5%
Information of my cluster is as follows:
Nodes: 8 nodes
Cores of a node: 8 cores (AMD Opteron)
Network: DDR Infiniband
OS: CentOS 5.3
Compiler: PGI 10.5 with ACML 4.3
Typical command line:
mpirun -host n1,n2,n3,n4 -np 8 -x OMP_NUM_THREADS=4 ./Solver
Best regards,
tmishima
