Charm++ parallel programming system

[Phil Miller <mille121 AT illinois.edu>]

Hello everyone,

We have just tagged a release candidate for the next stable release, version 6.8.1, of Charm++. This version has passed all of our automated tests, and has been tested by various Charm++ users already. If there are no major regressions found in it, we will tag the 6.8.1 release next week. You can check out the release in the charm git repo from the git tag 'v6.8.1-rc1'.

This is a backward compatible patch / bug-fix release of the Charm++ system containing about 100 changes since 6.8.0.

Please test this code with your applications, and inform us of any identified failures.

Highlights of those changes are as follows:

General System Improvements

- Enable network- and node-topology-aware trees for group and chare

  array reductions and broadcasts

- Add a message receive 'fast path' for quicker array element lookup

- Feature #1434: Optimize degenerate CkLoop cases

- Fix a rare race condition in Quiescence Detection that could allow

  it to fire prematurely (bug #1658)

  * Thanks to Nikhil Jain (LLNL) and Karthik Senthil for isolating

    this in the Quicksilver proxy application

- Fix various LB bugs

  * Fix RefineSwapLB to properly handle non-migratable objects

  * GreedyRefine: improvements for concurrent=false and HybridLB integration

  * Bug #1649: NullLB shouldnt wait for LB period

- Fix Projections tracing bug #1437: CkLoop work traces to the

  previous entry on the PE rather than to the caller

- Modify [aggregate] entry method (TRAM) support to only deliver

  PE-local messages inline for [inline]-annotated methods. This avoids

  the potential for excessively deep recursion that could overrun

  thread stacks.

- Fix various compilation warnings

Platform Support

- Improve experimental support for PAMI network layer on POWER8 Linux platforms

  * Thanks to Sameer Kumar of IBM for contributing these patches

- Add an experimental 'ofi' network layer to run on Intel Omni-Path

  hardware using libfabric

  * Note that support for '+partitions' (also used as '+replicas' by

    NAMD) is known to be broken, currently (bug #1675)

  * Thanks to Yohann Burette and Mikhail Shiryaev of Intel for

    contributing this new network layer

- The GNI network layer (used on Cray XC/XK/XE systems) now respects

  the ++quiet command line argument during startup

AMPI Improvements

- Support for MPI_IN_PLACE in all collectives and for persistent requests

- Improved Alltoall(v,w) implementations

- AMPI now passes all MPICH-3.2 tests for groups, virtual topologies, and infos

- Fixed Isomalloc to not leave behind mapped memory when migrating off a PE