Information Technology Reference
In-Depth Information
3.2.2.1 First Constraint: Targeting the Material and Operating System
to the Users
Most workstations are individual laptops; these are not supercomputers or CAD
stations and the post-processing of results is for the moment a poorly furnished
second order activity. The use of the API 3D standards 'Direc3D' (by Microsoft
)
or 'OpenGL' (by Khronos Group) is therefore strongly indicated! They allow
currently available graphic cards to be driven and use to be made of their internal
calculator thereby leaving the main processor free.
For the operating system, we decided on portability over the three commonly
used OS: 'Linux
', 'Windows
', 'MacOS
'. This constraint eliminates the API
'OpenGL', which is for the moment only supported by 'Windows'. GIDE there-
fore uses the API 'OpenGL' to drive the graphic cards.
3.2.2.2 Second Constraint: Displaying a Large Number of Elements
This constraint turns out to be difficult to respect. It is in contradiction with the
first. Without making use of a dedicated display machine a large number of bodies
cannot be displayed. The threshold of 10,000 is often the limit beyond which
manipulation of the display becomes jerky and even blocks for small configura-
tions. It is in the organisation of the bodies that we hope to overcome this con-
straint. To do this we use graphs of scenes, techniques developed initially for an
IRIS project [
9
] and then taken up by one of its authors in an OSG project
involving more general use [
10
].
The data from the discrete element calculation code are particularly well suited
to the construction of graphs of scenes since the latter are made up of vector
elements: circles, rectangles, spheres, cylinders, polyhedral, etc., and it is these
same basic elements that are used in simulations of the DEM type.
One of the major strengths of this approach lies in the elimination of calcula-
tions for masked bodies and other elements of the scene, called culling [
11
], and it
is this strength that is particularly interesting in our case. From experience, DEM
simulations have bodies organised in heaps, which means that most of the bodies
need not be represented as those on the surface mask them. For example, if we take
a cube of 50 bodies alongside, the simulation will count 50
3
= 125,000 bodies for
only three visible faces on a maximum of two rows, i.e. 50 9 50 9 2 = 15,000
bodies, which is the limit of manipulability on a laptop (Fig.
3.6
).
3.2.2.3 Third Constraint: Select the Bodies
The currently available offer of applications for displaying FEM is very rich,
unlike that for DEM, which is very poor.
What is the difference between the two, which justifies such a disparity?
Search WWH ::
Custom Search