Synchronizing the differences between point estimation and histogram estimation for free energies
A contributing factor with Sander's last minute large commit for free energies in 4.6 was that he and I weren't talking enough about how to formalize and synthezie free energy analysis in Gromacs. I've got the best performance with taking the output energies and postprocessing them with MBAR (using my pymbar implementation), whereas Sander has been using g_wham and g_bar to perform the analysis. We should investigate ways to make sure the different analysis visions are synthesized in future versions of gromacs, which will help simplify the code. I like the MBAR formalism, since BAR and WHAM are special cases of MBAR, but haven't had time to get around to making a gromacs version of MBAR (since pymbar works just great, there hasn't been a huge need).
I don't have a task list at this point, just something to resolve eventually -- preferably for 5.0, but it may or may not get done at that point.
Note that this issue can be treated completely independently of the actual calculation of free energy information -- both approaches use the exact same information. It's just a matter of how the information is stored for analysis. The main components are
1) how to calculated uncorrelated samples that are analyzed
2) whether save binned data or save data points to process at the end
3) how to output the data
I'd put this a little lower on the developmental priority list as things are working now, but it will make things simpler for the future to synthesize these approaches.