Speed Enhancement of Dynamic Bem on Inexpensive Hardware with Parallel Fortran 2008

Abstract eng:
The Boundary Element Method (BEM) has been extensively used in dynamic problems in structural engineering. The method is computationally intensive and traditionally depends on raw computer processing power to solve larger structures. However, advances on processor speed have stagnated due to physical limits. Instead, processing power is now based on parallel computing. The BEM can benefit from parallel computing, in dynamic problems in the Laplace transform domain, as the computation for each (complex) frequency is independent to each other, and can be executed concurrently by a different core or processor. The volume, in bytes, of the outcome of each frequency computation is small, so that it can be sent through a network with negligible overhead. Thus, the method does not need special hardware and it can be executed equally well by an ordinary, off-the-shelf, multi-core and/or multiprocessor computer, as well as by a typical computer cluster of low bandwidth. The method is implemented using coarrays as described in the upcoming Fortran 2008 standard. The biggest advantage of Fortran 2008 is that remote memory is referred to directly, like an array with the instance identity as index, instead of having to call a subroutine for loading and storing data. A small but sufficient set of easily learned statements and constructs are defined, which make the program much clearer. The method is tested with typical dynamic problems and shows promising results.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket