PDA

View Full Version : just saw the half life2 video..



adboy
05-21-2003, 02:33 AM
and could someone explain to me how in the GAME engine they can have REALTIME hard and soft body dynamics and in lightwave we get next to nothing unless we are willing to wait a long time

come on newtek please do something about this

i understand that it may not be 110% acurate but its plenty good enough for the majority of stuff

Mylenium
05-21-2003, 02:50 AM
That can be explained quite easily.

a) They use Havoc. This engine has been under development for years and is highly optimized.

b) They use only rather few objects (in comparison to classical 3D) or bounding boxes and even then a lot of the stuff is already precalculated (to a certain degree).

c) They use few iterations. Normally only the primary and secondary collisions are considered at all.

As for LightWave: Dynamics? Yes, but then real integrated dynamics (hard, soft, particles, cloth, ropes etc.) with correct simulation. A simplified engine would be useless to me, since I do mostly "serious" stuff. However it should be tweakable so it can be set up to be faster and less precise.

Mylenium

adboy
05-21-2003, 05:21 AM
define "serious" :)

"a) They use Havoc. This engine has been under development for years and is highly optimized. "

dont understand why newtek isnt optimising and developing theres, the dynamics are just part of this engine, it has alot of other stuff to do and its still pretty damn quick so why can't we have development on the hard and soft body side of layout

"b) They use only rather few objects (in comparison to classical 3D) or bounding boxes and even then a lot of the stuff is already precalculated (to a certain degree)."

hmm in the example i saw they where using more objects than i have used in any hard body simulation in lightwave because there is no hard body simlation in lightwave :)

"c) They use few iterations. Normally only the primary and secondary collisions are considered at all. "

not exactly sure what u mean :)

Mylenium
05-21-2003, 08:40 AM
Well, here goes:

serious = 3D other than for games. Even though the games industry is becoming more and more important, I'm not a real fan of games, especially these stupid shooters or so called strategy games - it's all about killing and war. So much for a better world.

Developing a physics engine is no trivial thing. This is mainly to the level of mathematical abstraction required. Also NT seem somewhat unwilling in this area. Both MD and PFX are third party plugins they just integrated into LW (more or less). Also the development team is rather small which may be another reason.

As for iterations/ evaluations: Let me give you an example there. Think of a cube that is falling on to the ground. As you would expect, the physics engine will calculate the falling of the cube and the initial collision with the ground. It will do so until it is settled. At first sight, there may not be obvious differences let's say with Havoc and another physics engine. However, internally this is a different story. First difference is temporal oversampling. While Havoc will only calculate every visible frame, a more sophisticated engine will calculate inbetween the frames (oversampling just like with Motion Blur). Also the latter will do more precise evaluations if e.g. you cube tumbles before settling - Havoc may cancel after a few iterations/ evaluations and call it done. Due to that Havoc may not expect further dynamic action and thus produce somewhat strange results if for instance the plane/ ground begins to tilt over.

You see these are the things that make the difference. However, this is all very theoretical and subjective. I guess we'll just have to see what NT do about it.

Mylenium

amorano
05-21-2003, 05:32 PM
Originally posted by Mylenium
Well, here goes:

Developing a physics engine is no trivial thing. This is mainly to the level of mathematical abstraction required. Also NT seem somewhat unwilling in this area. Both MD and PFX are third party plugins they just integrated into LW (more or less). Also the development team is rather small which may be another reason.


Trival no, possible even with one person to make something very good, yes. I have done this on two game projects. Source is abundant, help is abundant, there are tons of free engines to source through and plugin to LW also. Not something that should be overlooked by NT.



As for iterations/ evaluations: Let me give you an example there. Think of a cube that is falling on to the ground. As you would expect, the physics engine will calculate the falling of the cube and the initial collision with the ground. It will do so until it is settled. At first sight, there may not be obvious differences let's say with Havoc and another physics engine. However, internally this is a different story.
Mylenium

You refer to Euler's method, i.e. smaller step sizes = more iterations. Doing Taylor expansions is where it becomes slow. This is not major math, but how exact you want to get means more time.

There are others which are faster (Runge-Kutta, Improved Euler, Adpative Euler, etc, etc) if you have a copy of maya you will see you get to choose which method depending on the amount of precision (along with setting the intergrands). So.... again, not something to be overlooked.

harhar
05-21-2003, 09:23 PM
newtek probably has no mathematicians, just programers implement siggraph papers.