[mvapich-discuss] Announcing the release of MVAPICH2 2.1 GA and MVAPICH2-X 2.1 GA

Panda, Dhabaleswar panda at cse.ohio-state.edu
Sat Apr 4 00:02:39 EDT 2015 

The MVAPICH team is pleased to announce the release of MVAPICH2 2.1
GA, MVAPICH2-X 2.1 GA (Hybrid MPI+PGAS (OpenSHMEM, UPC and CAF) with
Unified Communication Runtime).

Features, Enhancements, and Bug Fixes for MVAPICH2 2.1 GA are
listed here.

* Features and Enhancements (since 2.0.1):
    - Based on MPICH-3.1.4
    - Enhanced startup performance
    - Support for PMI-2 based startup with SLURM
    - Enhanced startup performance for Gen2/UD-Hybrid channel
    - Add -export-all option to mpirun_rsh
    - Checkpoint-Restart Support with DMTCP (Distributed MultiThreaded
      CheckPointing)
        - Thanks to the DMTCP team (http://dmtcp.sourceforge.net/)
    - Optimize size of buffer requested for control messages
      in large message transfer
    - Flexibility to use internal communication buffers of different size
      for improved performance and memory footprint
    - Enhanced communication performance for small/medium message sizes
    - Support for handling very large messages in RMA
    - Enhanced automatic detection of atomic support
    - Optimized collectives (bcast, reduce, and allreduce) for 4K processes
    - Optimize creation of 2-level communicator
    - Collective optimization for PSM-CH3 channel
    - Optimization of collectives for SDSC Comet system
    - Introduce support to sleep for user specified period before aborting
        - Thanks to Adam Moody at LLNL for the suggestion
    - Disable PSM from setting CPU affinity
        - Thanks to Adam Moody at LLNL for providing the patch
    - Install PSM error handler to print more verbose error messages
        - Thanks to Adam Moody at LLNL for providing the patch
    - Introduce retry mechanism to perform psm_ep_open in PSM channel
        - Thanks to Adam Moody at LLNL for providing the patch
    - Automatic detection and tuning for Haswell architecture
    - Tuning for IvyBridge architecture
    - Tuning for ConnectX-4 EDR adapters
    - Increase the number of connect retry attempts with RDMA_CM
    - Support for additional MPI-T performance variables (PVARs) in the CH3
      channel
    - Support for linking Intel Trace Analyzer and Collector
    - GPU support for MPI_Scan and MPI_Exscan collective operations
    - Link with libstdc++ when building with GPU support
     (required by CUDA 6.5)

* Bug-Fixes (since 2.0.1):
    - Fix failures with shared memory collectives with checkpoint-restart
    - Fix failures with checkpoint-restart when using internal communication
      buffers of different size
    - Fix issue with checkpoint restart when full path is not given
    - Fix undeclared variable error when --disable-cxx is specified
      with configure
        - Thanks to Chris Green from FANL for the patch
    - Fix errors with large messages pack/unpack operations in PSM channel
    - Fix for bcast collective tuning
    - Fix assertion errors in one-sided put operations in PSM channel
    - Fix issue with code getting stuck in infinite loop inside ptmalloc
        - Thanks to Adam Moody at LLNL for the suggested changes
    - Fix assertion error in shared memory large message transfers
        - Thanks to Adam Moody at LLNL for reporting the issue
    - Relocate reading environment variables in PSM
        - Thanks to Adam Moody at LLNL for the suggestion
    - Fix deadlock in ctxidup MPICH test in PSM channel
    - Fix issue with automatic process mapping
    - Fix automatic detection of support for atomics
    - Fix segfault seen during connect/accept with dynamic processes
        - Thanks to Neil Spruit for the fix
    - Fix issue in CUDA IPC code path
    - Fix issue with void pointer arithmetic with PGI
    - Fix corner case in CMA runtime detection
    - Fix error in large message (>2GB) transfers in CMA code path
    - Increase counter length for shared-memory collectives
    - Fix memory leaks in OFA-IB-CH3 and OFA-IB-Nemesis channels
    - Fix issues with optimizations for broadcast and reduce collectives
    - Fix hang at finalize with Gen2-Hybrid/UD channel
    - Fix issues for collectives with non power-of-two process counts
        - Thanks to Evren Yurtesen for identifying the issue
    - Make ring startup use HCA selected by user
    - Fix compilation warnings

MVAPICH2-X 2.1 GA software package provides support for hybrid
MPI+PGAS (UPC, OpenSHMEM and CAF) programming models with unified
communication runtime for emerging exascale systems. This software
package provides flexibility for users to write applications using the
following programming models with a unified communication runtime:
MPI, MPI+OpenMP, pure UPC, pure OpenSHMEM, and pure CAF programs as
well as hybrid MPI(+OpenMP) + PGAS (UPC, OpenSHMEM and CAF) programs.

Features and enhancements for MVAPICH2-X 2.1 are as follows:

* Features and Enhancements (since MVAPICH2-X 2.0.1):
    - Introducing CAF (Coarray Fortran) Support
      - Based on University of Houston CAF implementation 3.0.39
      - Efficient point-point read/write operations
      - Efficient CO_REDUCE and CO_BROADCAST collective operations

    - OpenSHMEM Features
        - Based on OpenSHMEM reference implementation 1.0h
        - Support for on-demand establishment of connections
        - Improved job start up and memory footprint

    - UPC Features
        - Based on Berkeley UPC 2.20.0 (contains changes/additions in
          preparation for upcoming UPC 1.3 specification)

    - MPI Features
        - Based on MVAPICH2 2.1 (OFA-IB-CH3 interface)

    - Unified Runtime Features
        - Based on MVAPICH2 2.1 (OFA-IB-CH3 interface).  All the runtime
          features enabled by default in OFA-IB-CH3 interface of MVAPICH2
          2.1 are available in MVAPICH2-X 2.1

Various performance numbers for MVAPICH2 2.1 and MVAPICH2-X 2.1 on
different platforms and system configurations can be viewed by
visiting `Performance' section of the project's web page.

For downloading MVAPICH2 2.1 and MVAPICH2-X 2.1, associated user
guides, quick start guide, and accessing the SVN, please visit the
following URL:

http://mvapich.cse.ohio-state.edu

All questions, feedback, bug reports, hints for performance tuning,
patches and enhancements are welcome. Please post it to the
mvapich-discuss mailing list (mvapich-discuss at cse.ohio-state.edu).

Thanks,

The MVAPICH Team



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20150404/62a3d23a/attachment-0001.html>

More information about the mvapich-discuss mailing list