Misunderstanding?

Started by
4 comments, last by dawidjoubert 18 years, 4 months ago
Hi all, i've got a little question refering to OpenGL. I heard of many people that they are using OpenGL for motion and rotation of 3d-objects. But i'm wondering about the related functions in OpenGL. I know how and when to use matrices and transformation, but i can't imagine that a whole 3d application (maybe with a good engine) uses OpenGL for motion and rotation, since this API won't know anything about the format of my coordinates and vectors. A call to glVertex after a glTranslate on the modelview matrix won't really modify the coordinates (inside my code). So is it right that i also need to transform coordinates on my own? If I wan't to move an object and I'll use glTranslate i only modify the way it is displayed by OpenGL right? So if I would use collision detection too in my engine i need to move that object with my own methods, since a call to glTranslate won't modify the real coordinates. Am I right? Thanks!
Advertisement
Yes. Generally you keep for each object its position as the vector from the origin to world position. For rendering, you pass that same vector to OpenGL through glTranslate. For other things, you just use the vector as-is. Same goes for rotation and scaling.

Greetz,

Illco
The matrix known as MODELVIEW matrix of OpenGL is used to transform any geometry (say vertices in general) on-the-fly from the co-ordinate frame they are given in, to the view co-ordinates. That happens totally internally of OpenGL, and your local data isn't touched by that.

The question is what you mean with "So is it right that i also need to transform coordinates on my own?" In most cases it would be desastrous if OpenGL (or any other API) would alter your main copy of geometric data. Numerical resolution issues will introduce unstability over time. Your models will began to metamorphose more or less arbitrarily.

To avoid such things, the vertices of a model are usually defined in a _local_ co-ordinate frame. Inside that frame the model's vertices will be fixed (ok, not if bones or morphing is active, but in principle). Then glTranslate and glRotate and so on is used to tell OpenGL how the co-ordinate frame is currently given.

As another consequence, animating the model (e.g. moving it) does usually not mean to translate each single vertex but to translate the local co-ordinate frame. That is much more efficient.

So yes, in the case of collision detection you have to do some transformations yourself (in fact there're many more situations you need to do so). However, several optimizations exist to avoid doing transformation on a large scale. Colliding 2 models on vertex basis is actually a performance killer. So normally simpler shapes like bounding spheres, cylinders, boxes (object-orientated "OOBB" or axis aligned "AABB" or whatever) are used. So not all vertices but the bounding volumes are to be transformed.

Perhaps you're confused as to the nature of OpenGL - OpenGL actually has no features for animation whatsoever. It does not track 'objects' and draw them for you from frame to frame, it's just an API for drawing triangles. You have all the information about those triangles and you alter it as you see fit (and cause them to be transformed before drawing with the likes of glTranslatef) and then you tell OpenGL to draw them every once in a while to show the user what's going on.



AIM ME: aGaBoOgAmOnGeR
Quote:Original post by S0n0
Hi all,

i've got a little question refering to OpenGL.
I heard of many people that they are using OpenGL for motion and rotation of 3d-objects. But i'm wondering about the related functions in OpenGL. I know how and when to use matrices and transformation, but i can't imagine that a whole 3d application (maybe with a good engine) uses OpenGL for motion and rotation, since this API won't know anything about the format of my coordinates and vectors. A call to glVertex after a glTranslate on the modelview matrix won't really modify the coordinates (inside my code).

So is it right that i also need to transform coordinates on my own?
If I wan't to move an object and I'll use glTranslate i only modify the way it is displayed by OpenGL right? So if I would use collision detection too in my engine
i need to move that object with my own methods, since a call to glTranslate won't modify the real coordinates.

Am I right?

Thanks!
As the other posters have suggested, your intuition is correct. Although there are ways to keep this to a minimum, there are times when you need the transformed geometry for purposes other than rendering (notably collision detection), in which case you may have to perform the transformation yourself.
Quote:Original post by ShmeeBegek

Perhaps you're confused as to the nature of OpenGL - OpenGL actually has no features for animation whatsoever. It does not track 'objects' and draw them for you from frame to frame, it's just an API for drawing triangles. You have all the information about those triangles and you alter it as you see fit (and cause them to be transformed before drawing with the likes of glTranslatef) and then you tell OpenGL to draw them every once in a while to show the user what's going on.


I Think that is one FOR THE FAQ OpenGL, really all that OpenGL does is render polygons and their textures/shaders/blenders applied.

Although true it can render lines/points aswell.

----------------------------

http://djoubert.co.uk

This topic is closed to new replies.

Advertisement