At developers meeting on 10/27 we discussed the following:
- The group discussed the state of the effort to run OpenMx on the CRAN platform. Mike Wilde and Tim Armstrong are currently addressing some straight forward issues to get OpenMx to run.
- Michael Spiegel and Tim Armstrong have both observed a bottle neck of serializing and unserializing data. Spiegel has submitted a patch to the R Developers' group to address this bottleneck. Spiegel has seen a 2.4x benchmark on a micro-benchmark that just serializes and unserializes data. He has seen a 10% on a real benchmark.
- The group discussed how constrained OpenMx is by physical memory. This issue was also discussed in last week's meeting.
- The group discussed a possible funding opportunity with the Census Bureau for calculating combined likelihood behind firewalls.
- The group discussed the effort to debug the parallel hessian calculation with good progress. Several people have experienced trouble using gdb under mac. In the future the developers will use gdb under linux.
- The group discussed the claim that ClassicMx is running faster than OpenMx. The group is planning to run a "bake-off" of LogH vs. ClassicMx vs. OpenMx. The models within models/nightly/ will be used as the subject programs. Several programs may be added to models/nightly.
- The group continued to discuss the error identified by Timo last week. The model is identified so that is not causing the standard error issues. Furthermore, the hessian calculation is also incorrect. Tim and Timo are working on debugging this issue.
- The group discussed making the error message for mxOption more explicit.
- The group discussed which omxFunctions to document on the wiki and the man pages. It was decided that omxParallelCI and omxRAMtoML were the most important to document. Their will be man pages and wiki entries for each. The others can be documented by users and developers but links will need to be added to the wiki and the documentation will need to be supplied at the added link.