Improving Computational Efficiency of VAST

PDF

Authors
  1. Jiang, L.
  2. Macadam, T.
Corporate Authors
Defence Research and Development Canada, Atlantic Research Centre, Halifax NS (CAN);MARTEC Ltd, Halifax NS (CAN)
Abstract
This report describes development and implementation of an in-core database in the VAST finite element program. The database was developed based on the C++ Standard Template Library (STL) generic data structures and a set of Application Programming Interface (API) functions were provided to allow access from Fortran. The arguments of the API functions were designed to be of a generic format, to minimize the number of functions and permit easy future expansion. Implementation of the database required significant restructuring of VAST code. A pre-processor module, named PREPR1, was developed to import all the input data and store them into the database. During the VAST executions, the database was not only used as the source of the finite element model data, but also used as the temporary storage of many of the intermediate results. This treatment of data flow eliminated a large amount of I/O operations in the original VAST program and resulted in very significant savings on the computation time. The resulting version of VAST has been extensively verified and benchmarked using test problems of different sizes and the benchmark results have indicated that by combining the new database and the new sparse solver, the overall speed of VAST was increased by a factor of five for engineering problems. The current API version of VAST has some limitations, such as limited element types and analysis capabilities as well as a size limit on the finite element model. Further tasks are recom
Keywords
Finite element analysis;engineering structures
Report Number
DRDC-RDDC-2014-C34;MARTEC-TR-13-42 — Contract Report
Date of publication
01 Sep 2013
Number of Pages
36
DSTKIM No
CA038975
CANDIS No
539164
Format(s):
Electronic Document(PDF)

Permanent link

Document 1 of 1

Date modified: