Next: 7 Conclusion Up: Implementing Baraff & Witkin's Previous: 5 Discussion
To test out the simulator, I used a variety of cloth sizes and constraints. These can be seen at the Freecloth website http://davidpritchard.org/freecloth, in the screenshots section.
The simulator's performance is reasonable, with a
cloth typically
requiring about thirty to forty minutes of computation per second of animation
on a 3 GHz Pentium IV system using a step size of 
. Smaller cloth sizes
are much faster, and there is still a fair bit of room for optimisation,
especially in terms of adaptive timesteps.
My implementation does exhibit some instability, but almost always works with a step size of 0.01 seconds; sometimes, a smaller stepsize is needed.