10:30 AM-12:30 PM
Sandia Room
Geophysical estimation (or inversion) couples optimization and simulation to predict internal Earth structure from remote data. Modern simulators involve complex data structures, and therefore mate poorly with optimization packages written in structure-poor procedural programming languages such as Fortran. Thus the efforts of the numerical analysis community to provide high-quality software for optimization have had relatively little impact in geophysical inversion. Object-oriented programming promises a way out of this dilemna, in that it offers a natural means for connecting standard numerical optimization code, for instance, to simulators using data structures of essentially arbitrary complexity. While OO programming is now the dominant software style in many areas of computing, it has been slow to penetrate scientific computing, due to deficits in standards, portability, and performance. Many of these deficits have been overcome in recent years, and rapid development of object-oriented numerics for large-scale geophysical applications now appears possible. The speakers in this session will review recent work in geophysical OO numerics, including class libraries for optimization, mixed language programming, and emerging alternatives to C++ such as Java, Eiffel, and object-oriented extensions of Lisp. They will end with a panel discussion on the tools and techniques needed to get OO programming into general use in the geophysical community.
Organizer: William W. Symes
Rice University
GS97 Homepage | Program Updates|
Registration | Hotel Information | Transportation | Program-at-a-Glance | Program Overview