Parallel MuSES for Infrared Signature Modeling of U.S. Army Vehicles and Targets
Abstract
The objective for this work is the parallelization of the Multiservice Electrooptical Signature (MuSES) code, a commercially available computer code for infrared signature modeling of U.S. Army vehicles and targets. During an earlier phase of development in 2002, the progress in this work included the following: (1) profiling of the code for identifying potential segments of thermal and radiation computations for parallelization, (2) selecting a normalization loop within the selective overrelaxation (SOR) iteration loop for parallelization by OpenMP, and (3) some results for the ZSU Quad 23 AA vehicle. The results indicated a scalability performance of some 55% reduction in wall clock computation time on eight processors. Since then, the following parallelization enhancements have been added to the code: (1) partial direct solution computation loop within the SOR iteration loop, (2) tri-diagonal solution computation loop within the SOR iteration loop, (3) accelerated solution computation loop within the SOR iteration loop, and (4) routine ComputeRadNodeAveValues. The effectiveness of the enhanced code was evaluated on the previously mentioned ZSU tank, a Prius car, and the M2 Bradley Fighting Vehicle. As a result of the enhancements, the maximum performance improvement of the code increased to an average 72% reduction in the wall clock time on 16 processors for computing radiance in the IR range (3-4 micrometers, 8-12 micrometers). (3 tables, 5 figures)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 2004
- Accession Number
- ADA427266
Entities
People
- George Petit
- Raju R. Namburu
- Rama R. Valisetty
Organizations
- United States Army Research Laboratory