| NAME | ORGANIZATION | EMAIL ADDRESS |
| Anna Rounbehler | Raytheon | Anna_C_Rounbehler@res.raytheon.com |
| Richard Steinberger | Lockheed Martin GES | rsteinbe@motown.lmco.com |
| Bob Graybill | Lockheed Martin GES | rgraybil@motown.lmco.com |
| Rick Pancoast | Lockheed Martin GES | rick.pancoast@lmco.com |
| Myra Jean Prelle | Mercury | mjp@mc.com |
| Richard Games | MITRE | rg@mitre.org |
| Marvin Kuzmitz | NUVC/Npt | kuzmitz_marvin_l@code21.npt.nvwc.navy.mil |
| Arkady Kanevsky | MITRE | arkady@mitre.org |
| Ken Cain | MITRE | kcain@mitre.org |
| Paul Harmon | Licoln Laboratory | harmon@ll.mit.edu |
| Steve Paavola | Sky Computers | paavola@sky.com |
| Jon Greene | Mercury | greene@mc.com |
| Sharon Sacco | CSPI | ssacco@cspi.com |
| Thomas Robey | Khoral Research | trobey@khoral.com |
| Clay Taylor, Jr. | MPI Soft. Tech., Inc | cdtaylor@mpi-softtech.com |
| Anthony Skjellum | MPI Soft. Tech., Inc | tony@mpi-softtech.com |
| Arlan Pool | Mercury | alp@mc.com |
| Randall Judd | SPAWARSYSCEN S.D. | judd@spawar.navy.mil |
| Dennis Cottel | SPAWARSYSCEN S.D. | dennis@spawar.navy.mil |
| Michael Grieco | (ASEC) NSA K41 | mjgrieco@jswg.org |
The meeting came to order at 9:00am
Skjellum
Introduction to the meeting, discussed meeting rules, some Data-Reorganization philosophy, and the agenda for the day.
[Not yet available on the web page]
Rounbehler
Discussion on MPI/RT based approach.
[Not yet available on the web page]
Games
Voiced concerns on how an individual parallel topology would be mapped to a particular problem's data space. How will the developer use an API to deal with the problem of a system with contiguous or distributed memory.
There was much discussion on this topic including input from Green, Cottel.
Green
Discussed mapping slices and tiles. Introduced the idea of irregular partitioning of the data cube.
Cottel
Pointed out cases of data independance between slices in a data cube.
Lunch break
Skjellum
Opened afternoon discussion.
Paavola
How will the data accessed, considered as contiguous or as distributed. Indicated that the required data space may not fit on one node, so how would gather operations be achieved.
More discussion, including the possibility of "out-of-core" but "out-of-core" is probably not in the scope of this meeting. Also methods of data transfer were discussed. There is a need to reference current research articles on the corner-turn operation.
Further discussion ensued about the possibility of bulk memories, SMP's, and other client-server type modes of data reorganization, in addition to the clique and bi-partite graph models that everyone is most familiar with. It was discussed that the scope would ultimately have to be limited, but no decisions were made as yet.
It was agreed to have several further meetings, up to the deadline of September 30, 1998 for completion of report. Meetings were tentatively scheduled for January (January 13, Albuquerque is now final), March/April, May, July, September, with interest in holding one of the meetings at Mercury Computer Systems (they volunteered), and the final meeting in the DC area (eg, STA) in order to provide feedback to the sponsors and others in the DC area. The MITRE host agreed to host the other meetings in the Boston area. Several would be juxtaposed to MPI/RT meetings for convenience and cost savings.