For those of you interested in MPI and Infiniband, Dhabalaswar Panda and his team of MPI engineers from Ohio State University have announced the latest release of MVAPICH. Version 1.2RC1 includes the following new features:
- New Features for OpenFabrics/Gen2 Interface
- Network Fault Resiliency (NFR) for tolerating transient network failures. Using this feature, long running MPI codes can tolerate network failures. InfiniBand HCAs are automatically reset and MPI programs can continue computation without having to restart.
- Support for RDMA over Ethernet (RDMAoE) for ConnectX-EN adapter.
This version also has quite a list of bug fixes and enhancements from the 1.1 release. This version will be made available through the OFED 1.5.1 source tree or you can grab the package directly from the MVAPICH team here.