CPM

Version 1.2.0 (Revision 308)
----------------------------
A - tutorial on implementing model-based collectives
A - support for hierarchy-based input files and hierarchy-based model estimation, which can drastically reduce the model parameter estimation
A - supplementary tools for generating hierarchical input files
M - model can now use modified p2p communication in MPIBlib
M - package structure and documentation improved

Version 1.1.0 (Revision 226)

A - a LinInterp model - linear interpolation of p2p observations
A - generic & model-specific collective algorithms
A - tree builders based on p2p prediction (BFS/DFS/UCS binomial with slow/fast link selection, model-based modification of Traff)
A - generic model-based bcast/reduce, scatter/gather, scatterv/gatherv + implementation for all supported models
M - linear algorithms renamed to flat
A - models static library and collectives shared library
A - tests and tools