Unsorted Notes
MVAPICH2 and Intel-MPI on Ekman
A short performance comparison between MVAPICH2 and Intel MPI on Ekman.
The picture shows the performance of a simple PingPong benchmark on MPI for MVAPICH2 V. 1.7 and Intel-MPI V.4.0 Update 3. Both libraries have been used with the Intel Fortran Compiler V.12.0.5.
Summary: The performance of both libraries between different nodes is almost equal. MVAPICH2 provides the overall stronger performance due to its optimised installation for communication within nodes using shared memory and multi-core processors.
2011-12-04 – Category: programming – Tags: ekman mpi pdc