PDA

View Full Version : XSI CLOTH/PARTICLES - How good and flexible are they???


Mike Pauza
08-13-2004, 10:03 PM
XSI Dynamics People:



Just heard about the jaw dropping price drop on foundation and wanted to find out more about XSI Dynamics. Specifically:

1) Does "cloth" simply refer to simulating flexible shells (like actual clothing), or is it a full blown softbody dynamics simulator that can be applied to custom geometry for line-like & solid-like object simulation, etc?

2) How good is it's garment simulation?

3) If it is a full blown simulator, can you have dynamics higherarchies so that simple machines with keyframable controls can be built.

4) How robust is "cloth's" collision detection? Does it simple use naive collision checking (N^2)...i.e. does 100frames of self collision for a piece of geometry with 30,000 polys calculate in 30 minutes or more like a month?

5) Do particles and cloth influence each other?

6) Do particles have particle-particle interaction (smart), or are they "dumb" and only controllable by fields.

7) Can individual particles and or vertices be controlled with script? If so is the script limited to simple one line expressions, or is it a complete programming language?



Any answers are very much appreciated!

Thanks, -Mike Pauza :):):)

ThE_JacO
08-13-2004, 10:33 PM
1) Does "cloth" simply refer to simulating flexible shells (like actual clothing), or is it a full blown softbody dynamics simulator that can be applied to custom geometry for line-like & solid-like object simulation, etc?
advanced comes with Syflex technology cloth, the one in foundation tho is the old Softimage cloth.
it's a classic spring network solution that can work on anything suitable to generate the classic 4 triangles springs scheme.
ease of use is quite good as all you need to do is select anything, nurbs or poly doesn't matter, and apply the cloth sim to it, but it has many other shortcomings.


2) How good is it's garment simulation?
not too bad but it can be terribly slow and it's pretty old stuff by now anyway.


3) If it is a full blown simulator, can you have dynamics higherarchies so that simple machines with keyframable controls can be built.
not in foundation as it doesn't come with the open dynamic constraints.


4) How robust is "cloth's" collision detection? Does it simple use naive collision checking (N^2)...i.e. does 100frames of self collision for a piece of geometry with 30,000 polys calculate in 30 minutes or more like a month?
classic and simple raycasting from vertices I'd guess by the times and results of the sims.
can be daunting.


5) Do particles and cloth influence each other?
particles can be influenced by an object with a cloth simulaiton since it can be set as an actual shape collision, the other way around is not true tho.
particles are not as bad as some tell, but their lack of extra environmental mass is one of the most annoying limitations.


6) Do particles have particle-particle interaction (smart), or are they "dumb" and only controllable by fields.
interparticles interaction (collision, avoidance etc.) is in place and works well.


7) Can individual particles and or vertices be controlled with script? If so is the script limited to simple one line expressions, or is it a complete programming language?
yes, they have an events system (they can't be controlled in fine level by SCOPs, but for foundation owners this is cool since foundation doesn't have SCOPs) that exploits the same SDK you'd use in C++ only with a slightly different syntax to instance and manipulate objects (SDK objects that is), but not having SCOPs means you'll have to rely on predefined events to trigger(collision, spawning, age, distance, % and so on)+maths+manipulation

mind, that can get you mostly everywhere if you want to write your own simple forces or regular behaviors, but the annoying thing is that with no full fledged in-scene connections (normally provided by scops) you can't do something like writing your own raytracer and collision detector.

however if you move to compiled all these limitations fall and you find yourself with some pretty cool options.

ColinCohen
08-13-2004, 10:43 PM
however if you move to compiled all these limitations fall and you find yourself with some pretty cool options.
By this do you mean C++ compiled .dlls as opposed to scripts?

ThE_JacO
08-13-2004, 11:37 PM
yes, moving to the C++ or VB API rather then scripting.
beside being obviously faster, operators (resident scripts with connections to scene elements) can be of two kinds, compiled or scripted.
scripted operators (SCOPs for friends) aren't present in foundation and tend to limit you a lot when you need to write something heavy that requires casting and sampling, for the very simple reason that no interpreted language has a solid and direct handling of memory anywhere in the same league of C++ or even only visual basic.

Mike Pauza
08-16-2004, 03:18 PM
ThE_JacO:

Thanks for the info. -Mike

CGTalk Moderation
01-19-2006, 12:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.