Game Development Reference
the renderer. Therefore, we do not need to update the surface mesh each time we
update the volumetric mesh.
Storing ξ requires four floating-point values per surface-mesh vertex. If the
surface mesh is dense, a tetrahedron may have many associated vertices. In this
case, we can save memory by computing the transformation T ( p ) of a point p ,as
shown in Equation (10.19).
Therefore, each time we want to update the surface mesh, we compute T
once for each tetrahedron e and the new position T ( p ) of all the surface vertices
associated with e using Equation (10.19). It should be noticed that C e changes
in every iteration step and P − e can be precomputed and stored. Now we have
to save 16 floating-point values per tetrahedron instead of four values per surface
Thanks to the nature of the operations used by the two processes described
above, we can dramatically speed-up our system if we perform all the operations
in the GPU. Due to space constraints, we are not going to describe the GPU algo-
rithm, but it can be obtained quite straightforwardly from the description made in
[Etzmuss et al. 03] O. Etzmuss, M. Keckeisen, and W. Strasser. “A Fast Finite
Element Solution for Cloth Modelling.” In Proceedings of 11th Pacific Con-
ference on Computer Graphics and Applications , pp. 244-251. Washington,
DC: IEEE Computer Society, 2003.
[Garcia et al. 06] M. Garcia, C. Mendoza, A. Rodriguez, and L. Pastor. “Opti-
mized Linear FEM for Modeling Deformable Objects.” Comput. Animat.
Virtual Worlds 17: 3-4 (2006), 393-402.
[Muller and Gross 04] M. Muller and M. Gross. “Interactive Virtual Materials.”
In Proceedings of Graphics Interface 2004 , pp. 239-246. Waterloo, Ontario:
Canadian Human-Computer Communications Society, 2004.
[Muller and Teschner 03] M. Muller and M. Teschner. “Volumetric Meshes for
Real-Time Medical Simulations.” In Bildverarbeitung fur die Medizin 2003 ,
CEUR Workshop Proceedings, 80, pp. 279-283. Aachen, Germany: CEUR-
[Press et al. 07] William H. Press, Saul A. Teukolsky, William T. Vetterling, and
Brian P. Flannery. Numerical Recipes: The Art of Sientific Computing .New
York: Cambridge University Press, 2007.