The Monte Carlo simulation process is parallelized with use of the message passing interface [1]. It can be assumed that the most time-consuming part of the solution is structural analysis of the models using the finite element method.

The simplest possible approaches have been adopted: the given number of simulations is divided between several concurrently running processes. Initial data are distributed to processes before the solution and then only realisations of random variables are sent. The SPRNG [2] parallel random generator library is used for this purpose.

In the simplest case there is no need for further interprocess communication until the final collection of results data. However, in most cases it is necessary to store results at least as bounded histograms. Application of histograms better describes the distribution of resulting random variables compared with a simple statistics application while reducing the amount of data to be stored.

It is shown that utilisation of parallel processing can considerably shorten computational time even if only a relatively small number of processing units are available (a small network of PC workstations has been used here). The LAM-MPI [3,4] implementation of the MPI standard has been used.

The resulting performance of the stochastic static analysis of reinforced concrete was good if up to four processors were utilised. Although the performance did not improve until the seventh processor was added. Speedup with four to six processors was about 3.8. The seventh processor improved the speedup to 5.5.

The speedup of the seismic analysis for the frame model of a precast panel building was compared to the application of two processors. Four processors made the computation 1.9 times faster while eight processing units computed 2.9 times faster than two units.

1

MPI Forum, "MPI: A Message-Passing Interface Standard." Version 2.2, September 4th 2009, http://www.mpi-forum.org

2

M. Mascagni, A. Srinivasan, "Algorithm 806: SPRNG: A Scalable Library for Pseudorandom Number Generation", ACM Transactions on Mathematical Software, 26, 436-461, 2000. doi:10.1145/365723.365738

3

G. Burns, R. Daoud, J. Vaigl, "LAM: An Open Cluster Environment for MPI", Proceedings of Supercomputing Symposium, 379-386, 1994. http://www.lam-mpi.org/download/files/lam-papers.tar.gz

4

J.M. Squyres, A. Lumsdaine, "A Component Architecture for LAM/MPI", Proceedings, 10th European PVM/MPI Users' Group Meeting, 379-387, Venice, Italy, Springer-Verlag, 2003. doi:10.1007/978-3-540-39924-7_52

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £85 +P&P)