Computational & Technology Resources
an online resource for computational,
engineering & technology publications |
|
Civil-Comp Proceedings
ISSN 1759-3433 CCP: 50
INNOVATION IN COMPUTER METHODS FOR CIVIL AND STRUCTURAL ENGINEERING Edited by: B.H.V. Topping and M.B. Leeming
Paper X.2
Parallel and Distributed Dynamic Relaxation Form-Finding P. Ivanyi and B.H.V. Topping
Department of Mechanical and Chemical Engineering, Heriot-Watt University, Edinburgh, United Kingdom P. Ivanyi, B.H.V. Topping, "Parallel and Distributed Dynamic Relaxation Form-Finding", in B.H.V. Topping, M.B. Leeming, (Editors), "Innovation in Computer Methods for Civil and Structural Engineering", Civil-Comp Press, Edinburgh, UK, pp 157-165, 1997. doi:10.4203/ccp.50.10.2
Abstract
The design of a tension structure is a complex procedure
consisting of many parts which results in an iterative process.
This process may include: conceptual design; physical
modelling; computer modelling; formfinding; analysis;
and cutting pattern generation. To ensure the flexibility
required by the designer throughout this process and to
shorten the formfinding-analysis-cutting pattern generation
subtasks efficient parallel dynamic relaxation can be
used. But in parallel computations, processors must generally
exchange data. This exchange of data has usually a
great influence on the efficiency of the parallel programs
because it introduces communication delays during the
execution time. By contrast most commercial codes are
sequential and hence it is recognised that the parallelisation
of these codes must be simple and cost-effective.
This fact implies that for parallel processing there may
be a need for high-level subroutine libraries, to implement
the interprocessor communication schemes. In this paper,
some communication schemes for computational mechanics
problems will be discussed and assessed which may
be included in a communication subroutine library. With
this library a sequential dynamic relaxation program can
be transformed into a parallel one. Some form-finding
results calculated using the parallel program will be presented.
purchase the full-text of this paper (price £20)
go to the previous paper |
|