GIDE: Graphic Interface for Discrete Element - Visual Computing: Scientific Visualization and Imaging Systems - page 63

3.2 Technology Choice
GIDE was designed to be as open as possible through recourse to recognised and
free libraries. We have been particularly attentive to the documentation of the code
and the development environment through collaborative work. The tools used are
all from open sources. The result is an application entirely uncoupled from EE
calculation code.
3.2.1 Data Format
3.2.1.1 First Version of the Data Format
In order not to be intrusive, the initial aim of GIDE was to be capable of
functioning without any modification of upstream software, such as MULTICOR
or other simulators, only via the transformation of data from these applications.
The transformation is based on an XML file describing the organisation of the data
and a conversion tool (included in the GIDE). Thus, all results data files are
transformed into the format HDF5 before use.
The format HDF5 was adopted as the native GIDE format; its advantages are:
• A data model allowing the representation of complex data,
• A portable file format,
• A library that can be used on different platforms, with interfaces in C, C++,
Fortran and Java,
• Optimised performance for access time and size of data,
• Tools for the display and processing of data in the format HDF.
Reading and writing are optimised in order to exploit various types of archi-
tecture: simple file on a standard file system, several files on a standard file system,
several files on a parallel file system and other situations. The first version of the
organisation of data in the HDF file offers the following architecture (Fig. 3.1 ).
Starting from the root '/' w access the simulation of global data, and then we
access each scene and finally the block of data for each type of discrete element. In
the HDF file there are sets of data called DATASET, which are gathered into
groups called GROUP. Each GROUP or DATASET has a name; it is the name of
the DATASET that determines the representation of a discrete element. The tree
representing the organisation of the data looks like Fig. 3.1 . Access to the first
DATASET is via path/scene/pas01/. The data are not only organised in the form of
a tree in the files, but also in the memory. OSG also organises the various elements
of a 3D scene in the form of a tree and even more in the form of a graph. GIDE
re-uses the mechanisms of OSG for this part; the discrete elements of GIDE are
specialisations of the GROUP of OSG. While data are being read, the tree is built
up of the scene for each step in the simulation, Fig. 3.2 .