Software Infrastructure for Computational Science

We seek to develop and improve the mathematical infrastructure for computational science both within NIST and in the science and engineering community at large. This is done in several ways. Firstly, we engage in research and development of advanced algorithms, software and related tools for key high performance computations. Second is the development of methodology and tools for the testing and evaluation of mathematical algorithms and software. Such work is a key ingredient in both judging the output of the research community in order to set agendas for future work, but also ultimately for use in the evaluation of commercial products and services. Finally, we also take a pro-active approach to the use of modern network technologies in providing the results of this research to the scientific community at large. The tasks and accomplishments listed below provide some examples of recent work in each of these areas. Several are components of larger NIST projects in information technology:

• The work on GAMS, in sparse linear algebra, in partial differential equations, and in parallel software tools are part of the NIST High Performance Computing and Communications (HPCC) program.
• Work on parallel adaptive multigrid methods and parallel Helmholtz solvers is being done in conjunction with the Scalable Computing Testbed under development by the High Performance System and Services Division.
• Work on the Matrix Market and the Software Testing Service for Special Functions is part of a joint ACMD/Statistical Engineering Division project on Reference Data Sets for Mathematical and Statistical Software.

• Information Systems for Math Software Research, Development and Reuse
  • The Guide to Available Mathematical Software
  • Moving Computer Science Research Journals Online
• Testing and Evaluation of Mathematical Software
  • The Matrix Market
  • Software Test System for Special Functions
  • Development of SLI Arithmetic
  • Measurement of Fortran Execution Time
• Mathematical Software for High Performance Computing
  • Object Oriented Numerical Software
  • Sparse BLAS Toolkit
  • Truncating the Singular Value Decomposition for Ill-Conditioned Linear Least Squares Problems
  • Approximating the Permanent
  • Large Scale Nonlinear Programming
  • Interior Point Methods for General Large Scale Quadratic Programming Problems
  • Adaptive Multi-level Methods and Software for Partial Differential Equations
  • Parallel Helmholtz Solvers
  • Unfolding Measured Distributions
  • Numerical Software for Special Functions
  • Development of Computational Geometry Algorithms
  • Parallel Application and Development Environment
  • Tools for efficient use of the NIST IBM PowerParallel System 2
  • A Multi-User, Interactive and Multimedia Modeling, Analysis, and Computing Environment for Concurrent Engineering Design
  • Cryptology and Number Theory