Release Notes
Version 2.1.1 (02/03/11)
Version 2.1.1 fixes bugs and is easier to use.
- Bug fixes
- Fixed: 2.1.0 would sometimes crash randomly during long runs.
- Fixed: Windows binaries used the same random seed every time.
- Fixed: Broken M0 and M7 model priors.
- Usability
- Read files in old macintosh format on UNIX and Windows.
- Build
- Fix build from source with system BOOST libraries.
Version 2.1.0 (07/07/10)
Version 2.1.0 adds some major new features:
- Major new features.
- Allow multiple separate genes (i.e. data partitions).
- The alignment of each gene may be either fixed or variable.
- Improve mixing when the tree has many tips (for MCMC)
- Improve initial search when the tree has many tips (for estimation)
- Use much less RAM.
- Allow multiple separate genes (i.e. data partitions).
- MCMC Speed and reliability
- Use slice sampling for continous parameters -- automatically detect step-size.
- Allow SPR moves to intelligently propose attachment sites.
- Improve CPU efficiency when the alignment is constrained.
- Spend more time resampling topology and alignment, versus substitution parameters.
- Models
- Default to more conservative gamma prior for branch lengths.
- New substitution models: C10 and C20.
- Usability
- Stop requiring --data-dir option: the information is now internal.
- Handle amino-acid code J.
- Warn on illegal characters.
- New analysis script bp-analyze.pl replaces GNUmakefile.
- Summarization tools
- Report the maximum posterior decoding alignment.
- Simplify computing the consensus tree and other summaries.
- Check mixing and effective sample sizes by comparing multiple runs.