Multi-disciplinary Research AreaComputational Science[Overview] [Faculty] [Projects] [Courses] Computational science is an interdisciplinary mode of scientific investigations and engineering design through simulations. Large scale simulations of interest are often computationally-intensive and/or data-intensive, requiring advanced computational resources. The computational science program in the computer science department focuses on the development and application of algorithms, software, and systems for enabling such simulations.FacultyNancy Amato (Motion Planning, Robotics, Computational Geometry, Virtual Reality, Computational Biology/Chemistry, Parallel and Distributed Computing, Parallel Algorithms, Performance Modeling)
Vivek Sarin
(Numerical Methods, Parallel Algorithms, Computational Science)
Valerie Taylor
(Performance analysis and data partitioning for parallel and
distributed applications)
Glen Williams
(Computer Graphics, Scientific & Engineering Applications,
Computational Mathematics) ProjectsSeismic Ray Tracing (Amato) Our lab has worked on the parallel computing and performance modeling aspects of the project. We will be working on developing an efficient parallel algorithm for Seismic Ray Tracing. The second phase of the project deals with implementing the algorithm in a parallel machine using the Standard Template Adaptive Parallel Library (STAPL) which is being developed at Texas A&M. The next phase of the project deals with incorporating self tuning elements in the code. Robust Preconditioners for Sparse Linear Systems (Sarin) We are developing robust, parallelizable preconditioning techniques for large, sparse linear systems arising in scientific simulations. Modeling and Simulation of Sub-Micron VLSI (Sarin) We are developing novel algorithms for fast and accurate estimation of the parasitic RCL of large VLSI circuits. A software package is being developed to provide the capability of fast parasitic extraction on a variety of parallel computers. Prophesy: Performance Analysis and Modeling Infrastructure (Taylor) Prophesy includes three major components: automatic instrumentation, databases for archiving performance data, and a model builder. The model builder can use one of three techniques (curve fitting, parameterization, or coupling) to develop performance models. Prophesy can be used to identify performance trends, performance tune applications, or provide predictions that can be used by other systems such as resource schedulers. (URL: prophesy.cs.tamu.edu) Modeling and Simulation of a Very High Frequency Spring (Williams) We are developing algorithms for the animation of a slinky. We have completed the first phase, a non-contact, unconstrained model, and have produced a movie of various motions. The equations and the associated computational techniques for the constrained motions are currently under development. Failure of Earthen Embankments (Williams) We are developing numerical techniques to determine the failure surface of an earthen embankment, such as a dam. The embankment medium is a non-homogeneous, multi-layered, discontinuous half-space which may include sheet pile walls, tension cracks, or other nonlinear characteristics. View Dependent Radiosity (Williams) We are developing algorithms to enable the basic radiosity algorithm to be implemented in a view dependent manner. Currently view independent and requiring a solution of n simultaneous equations for a complete image with n polygons in a scene, prioritized viewing techniques will allow the radiosity algorithm to be used in a more computationally efficient manner.
|
| ||||