Minutes of the MPI working group

Attending: Patrick Brady, Jolien Creighton, Alan Wiseman, Warren Anderson, Duncun Brown, Masha Barnes, Kent Blackburn

Next Meeting: Thursday, March 30th

Schedule of software developments:

Patrick posted a UWM timeline on the groups webpage outlining the timeline for UWM to deliver the 5 interface functions used by the wrapperAPI. This would look something like this:

1. filterInit() --- T. Creighton --- 3+7 days
2. indexFilters() --- T. Creighton --- 1+1 days
3. conditionData() --- J. Creighton/Brady --- 7+7 days
4. filterTemplates() --- Anderson/Brown/Wiseman --- 14+21 days
5. freeFilters() --- J. Creighton/Brady --- 1+N days

Issues raised are "how does the data get to each node" and documentation. A new requirements document spelling out the flow of the wrapperAPI will be posted to address the first. The second will follow the standards to be adopted by the software committee.

Review of MPICH 1.2:

The recent memory leak problems with MPICH were reported my Masha. She agreed to send Patrick email with the details of the patches needed so that UWM and CIT could work with identical patched versions of MPICh when code integration begins.

Model for testing wrapperAPI:

The model for how to test the wrapperAPI when all code was in place started to form in this meeting. The plane is to use 40m data in frames (version 3+). The data stream would consist of the GW channel and the calibration information (sweep sine). Data segments would be nominally ~300 seconds long, but shorter segments could be possible when the instrument was out of lock.

At the LSC meeting it was determined that there were too many questions about how a hierarchical search would be performed within in the wrapperAPI. The impact of this was important enough to warrant a distraction from the original goals of the project and pursue a study of how such a search could fit into the wrapperAPI framework. It was expected that a one week delay was in order to determine if any modifications should appear in the interface.

On March 23rd Patrick and Kent decided to post-pone this weeks meeting in order to allow more concrete analysis of the issues associated with the hierachical search to executed. Kent informed Patrick that he would have a new version of the wrapperAPI documentation posted that day with additional material outlining the parallel flow of the wrapperAPI. Patrick and Kent spoke again later to discuss the details of some of the steps in the flow of the code. Patrick said they would use this in their analysis and have something posted in a few days.

Organizational Details:

Software schedule now available on the webpage.

New version of wrapperAPI requirements document now available on webpage.

Minutes taken by Kent

Agenda for next Meeting:

1. Review schedule
2. Review changes to requirements document
3. Report on hierarchical search strategy