Charm++ parallel programming system

[Phil Miller <mille121 AT illinois.edu>]

Charm 6.2.1 is now released. This integrates many optimizations and
bug fixes from the development version into the stable branch based on
6.2.0. We encourage all developers and users to build their
applications against this version.

Some significant enhancements in this release include
- Net builds support the use of clusters' mpiexec systems for job
launch, via the ++mpiexec option to charmrun.
- LAPI builds support Linux and AIX now, and should have better
performance with the SMP build option
- Better support for multicasts on groups
- POWER7 support added
- Topology information gathering has been optimized
- Converse (seed) load balancers have many new optimizations applied
- CPU affinity can be set more easily using +pemap and +commap options
instead of the older +coremap
- HybridLB (hierarchical balancing for very large core-count systems)
has been substantially improved
- Load balancing infrastructure has further optimizations and bug fixes
- Object mappings can be read from a file, to allow offline
topology-aware placement
- Projections logs can be spread across multiple directories, speeding
up output when dealing with thousands of cores (+trace-subdirs N will
divide log files evenly among N subdirectories of the trace root,
named PROGNAME.projdir.K)
- AMPI now implements MPI_Issend
- AMPI's MPI_Alltoall uses a flooding algorithm more agressively,
versus pairwise exchange
- Virtualized ARMCI support has been extended to cover the functions
needed by CAF

This version can be obtained as a tarball [1] or from the Charm git
repository [2]. A list of changes since the 6.2.0 release can be found
through the repository browser [3].

[1] http://charm.cs.illinois.edu/distrib/charm-6.2.1_src.tar.gz
[2] git clone git://charm.cs.illinois.edu/charm.git; cd charm; git
checkout charm-6.2.1
[3] 
http://charm.cs.uiuc.edu/cgi-bin/gitweb2.cgi?p=charm.git;a=shortlog;h=charm-6.2.1