Information Technology Reference
In-Depth Information
see Fig.
1.13
. These are very well understood as important developments and thus
such research topics are continuously being funded within the European R&D
projects, which the author has been able to experience during the last two decades.
In the following chapter some of these projects are addressed, highlighting their
multidisciplinary visualization challenges.
1.2 European R&D Projects
In 1988, the author started the development of the SV system that was named
''CFView'' [
39
], and the author has established the Object-Oriented approach [
32
]
for developing SV systems. The object-oriented programming language C++ was
chosen, because of its high performance in number crunching and its capability to
support OOP constructs.
CFView [
15
] was designed to work with structured and unstructured mesh
types. CFView was found to be particularly valuable when treating the CFD
simulations of the European Hypersonic Database [
40
] and validating the simu-
lations against experimental data. CFView allows the simultaneous inspection of
fluid flow simulations with structured and unstructured grids; as it uses data
extraction algorithms such as plane-cutting, iso-surface, and particle-tracing for
both grid models, to uncover interesting features hidden in such large datasets.
These routines require large computational resources and tend to limit the inter-
active feedback of CFView. In order to overcome this problem, a heterogeneous
system named ''Parallel CFView'' was constructed which distributes the com-
puting load over several processors and permits real-time interactive visualization
of CFD data. The development of Parallel CFView [
41
] was part of the EU-funded
projects PAGEIN [
42
] and EEC/ESPRIT PASHA [
43
].
In the PASHA project, CFView visualization system was ported to MIMD and
SIMD platforms for comparative benchmarking. The performance of the MIMD
implementation was evaluated on two industrial test cases provided by the
EUROPORT-1 project. Both SIMD and MIMD parallel machines were used to
provide massive back-end processing power. A key achievement of the Parallel
CFView project turned out to be the interface for communication between
machines with heterogeneous architectures. An analysis was carried out to com-
pare the performance of such distributed computing environment consisting of
both: the SIMD and MIMD implementations and the sequential CFView system.
The results showed that the parallel implementation of the extraction algorithms
speedup the performance of the visualization process by a factor of 10, which
proved the viability of such visualization system configurations.
PAGEIN [
42
] demonstrated more effective resource utilization through the
support of distributed computing and collaborative engineering. The goal was to
seamlessly integrate scarce human expertise in domains like, supercomputers,
massively parallel computers, and data servers. PAGEIN enabled the engineers to
exploit the computing resources and reach the critical expertise mass without need
Search WWH ::
Custom Search