[openib-general] Announcing the release of MVAPICH2 0.6.5 with uDAPL support

[Dhabaleswar Panda]

The MVAPICH (MPI over InfiniBand) team at the Ohio State University is
pleased to announce the release of MVAPICH2 0.6.5 for multiple
platforms (EM64T, G5, IA-32, IA-64, and Opteron) and network
interfaces (PCI-X and PCI-Express-including the new mem-free cards).

In addition to the optimized design over the VAPI interface for
InfiniBand, MVAPICH2 0.6.5 also supports optimized design over the
uDAPL interface so that users can take advantage of MPI-2
functionalities available in MVAPICH2 on networks supporting uDAPL
provider library.  The uDAPL interface of MVAPICH2 0.6.5 has been
tested with InfiniBand (IBGD uDAPL), Myrinet (DAPL-GM beta), and
Ammasso GigE (Ammasso uDAPL). Through the uDAPL interface, it provides
portability across different networks while delivering high
performance.

MVAPICH2 0.6.5 is being distributed as a single integrated package
(with the latest MPICH2 1.0.1 and MVICH). It can be downloaded with a
`single click' and installed. It is available under BSD license.

MVAPICH/MVAPICH2 software is being used by more than 230 organizations
world-wide (in 27 countries) to extract the potential of InfiniBand
networking technology for designing high-end computing systems and
servers. It is also being distributed by many IBA vendors in their
software distributions.

This new release has the following features:

      - MPI-2 functionalities (one-sided, collectives, datatype)
      - all MPI-1 functionalities
      - optimized one-sided operations (Get, Put, and Accumulate)
      - support for active and passive synchronization
      - optimized two-sided operations
      - uDAPL support (tested for InfiniBand, Myrinet, and Ammasso GigE) 
      - scalable job start-up
      - optimized and tuned for the above platforms and different 
        network interfaces (PCI-X and PCI-Express)
      - single code base for all of the above platforms
      - memory efficient scaling modes for medium and large clusters

Other features of this release include:

- Excellent performance: For two-sided operations, MVAPICH2 0.6.5
  with VAPI interface delivers 5.0 microsec latency (with the switch), 
  up to 965 MB/sec unidirectional bandwidth, and up to 1725 MB/sec
  bidirectional bandwidth on EM64T system with PCI-Express. For
  one-sided Put operation, MVAPICH2 0.6.5 delivers 6.98 microsec
  latency (with the switch), and up to 972 MB/sec unidirectional
  bandwidth on the above platform.

- With uDAPL interface, MVAPICH2 0.6.5 delivers latency for small 
  messages very close (within 1.0 microsec) to that of the uDAPL
  library. It delivers bandwidth very close to that provided by the
  VAPI interface.

- Detailed performance numbers for two-sided and all one-sided
  operations on various platforms and interconnects using VAPI and uDAPL
  interfaces are available on the project's web page.

- A set of benchmarks to evaluate one-sided operations (Put, Get, and 
  Accumulate)

- An enhanced and detailed `User and Tuning Guide' to assist users: 

       - to install this package on different platforms 
            with both interfaces (VAPI and uDAPL) and different options

       - to vary different parameters of the MPI installation to 
            extract maximum performance and achieve scalability, 
            especially on large-scale systems.

You are welcome to download the MVAPICH2 0.6.5 package and access
relevant information from the following URL:

http://nowlab.cis.ohio-state.edu/projects/mpi-iba/

Our upcoming release will include a high performance design 
of MVAPICH2 with MPICH2 1.0.2 and uDAPL support.

All feedbacks, including bug reports and hints for performance tuning,
are welcome. Please send an e-mail to mvapich-help at cse.ohio-state.edu.

Thanks, 

MVAPICH Team at OSU/NBCL 

----------

PS: If you would like to be removed from this mailing list, please end
an e-mail to mvapich_request at cse.ohio-state.edu.