Some Parallel Extensions to Optimization Methods in OPT++
Abstract
OPT++ provides an array of optimization tools for solving scientific and engineering design problems. While these tools are useful, all of the code is serial. With increasingly easy access to multiprocessor machines and clusters of workstations, this results in unnecessarily long times to solution. In order to correct this problem, we have implemented a number of parallel techniques in OPT++. In particular, we have incorporated a speculative gradient algorithm that drastically reduces the time to solution for standard trustregion and line search algorithms. In addition, we have implemented a new version of the TrustRegion Parallel Direct Search (TRPDS) algorithm of Hough and Meza that yields a significant reduction in solution time for problems with expensive function evaluations.
 Authors:
 Publication Date:
 Research Org.:
 Sandia National Labs., Albuquerque, NM, and Livermore, CA (US)
 Sponsoring Org.:
 US Department of Energy (US)
 OSTI Identifier:
 766611
 Report Number(s):
 SAND20008877
TRN: AH200038%%313
 DOE Contract Number:
 AC0494AL85000
 Resource Type:
 Technical Report
 Resource Relation:
 Other Information: PBD: 1 Oct 2000
 Country of Publication:
 United States
 Language:
 English
 Subject:
 99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE; ALGORITHMS; ARRAY PROCESSORS; DESIGN; OPTIMIZATION; PARALLEL PROCESSING; O CODES
Citation Formats
V. E. Howle, S. M. Shont, and P. D. Hough. Some Parallel Extensions to Optimization Methods in OPT++. United States: N. p., 2000.
Web. doi:10.2172/766611.
V. E. Howle, S. M. Shont, & P. D. Hough. Some Parallel Extensions to Optimization Methods in OPT++. United States. doi:10.2172/766611.
V. E. Howle, S. M. Shont, and P. D. Hough. 2000.
"Some Parallel Extensions to Optimization Methods in OPT++". United States.
doi:10.2172/766611. https://www.osti.gov/servlets/purl/766611.
@article{osti_766611,
title = {Some Parallel Extensions to Optimization Methods in OPT++},
author = {V. E. Howle and S. M. Shont and P. D. Hough},
abstractNote = {OPT++ provides an array of optimization tools for solving scientific and engineering design problems. While these tools are useful, all of the code is serial. With increasingly easy access to multiprocessor machines and clusters of workstations, this results in unnecessarily long times to solution. In order to correct this problem, we have implemented a number of parallel techniques in OPT++. In particular, we have incorporated a speculative gradient algorithm that drastically reduces the time to solution for standard trustregion and line search algorithms. In addition, we have implemented a new version of the TrustRegion Parallel Direct Search (TRPDS) algorithm of Hough and Meza that yields a significant reduction in solution time for problems with expensive function evaluations.},
doi = {10.2172/766611},
journal = {},
number = ,
volume = ,
place = {United States},
year = 2000,
month =
}

Parallel and sparse methods for largescale numerical optimization
This report discusses progress in parallel and sparse methods for largescale numerical optimization of linear programming problems. (LSP) 
Parallel optimization methods for agile manufacturing
The rapid and optimal design of new goods is essential for meeting national objectives in advanced manufacturing. Currently almost all manufacturing procedures involve the determination of some optimal design parameters. This process is iterative in nature and because it is usually done manually it can be expensive and time consuming. This report describes the results of an LDRD, the goal of which was to develop optimization algorithms and software tools that will enable automated design thereby allowing for agile manufacturing. Although the design processes vary across industries, many of the mathematical characteristics of the problems are the same, including largescale,more » 
A class of trustregion methods for parallel optimization
The authors present a new class of optimization methods that incorporates a Parallel Direct Search (PDS) method within a trustregion Newton framework. This approach combines the inherent parallelism of PDS with the rapid and robust convergence properties of Newton methods. Numerical tests have yielded favorable results for both standard test problems and engineering applications. In addition, the new method appears to be more robust in the presence of noisy functions that are inherent in many engineering simulations. 
FORTRAN M. FORTRAN Extensions for Modular Parallel Processing
FORTRAN M is a small set of extensions to FORTRAN that supports a modular approach to the construction of sequential and parallel programs. FORTRAN M programs use channels to plug together processes which may be written in FORTRAN M or FORTRAN 77. Processes communicate by sending and receiving messages on channels. Channels and processes can be created dynamically, but programs remain deterministic unless specialized nondeterministic constructs are used.