Jump to content

View more

Image of the Day

 #indiedev  #indiegame #screenshotsaturday https://t.co/xKIzRQMdHm
IOTD | Top Screenshots

The latest, straight to your Inbox.

Subscribe to GameDev.net's newsletters to receive the latest updates and exclusive content.

Sign up now

Localspace to Worldspace

2: Adsense

Old topic!

Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
2 replies to this topic

#1 Anddos   Members   


Posted 19 January 2013 - 06:42 AM

I understand localspace is the cordinates of the 3d model and worldspace is the left hand cordinate system , i want to know what is happening with the vertices postions when you use SetTransform WORLD , do the vertices stay the same or do they change?


#2 haegarr   Members   


Posted 19 January 2013 - 07:35 AM

Technically, a space in this sense is a reference system used to interpret co-ordinates. The shapes of objects as well as their spatial relations are constant even when you alternate the space.


Vertices in the sense of 3D APIs provide a couple of attributes. The geometric attributes of models, i.e. position, normal, and tangents, are often stored as co-ordinates w.r.t. a space that is called the model space (or local space, although any space that is not global one is a local one; so "model space" is somewhat more accurate). When you set the WORLD transform, you actually define the transformation of co-ordinates given in local space to be re-interpreted in global space. This doesn't alter the numbers stored with the vertices in main memory; instead, it tells how to compute temporary vertex co-ordinates during processing on the GPU (in fact, there are more space transformations involved than just the WORLD transform).


So, you supply the vertices in model space to the GPU, the GPU picks one vertex after the other, applies the WORLD transform (and others) to it to yield in temporary vertices with co-ordinates in world space (and others), and uses these temporary vertices for drawing. At any of these steps, when you look at the co-ordinates stores with the mesh you'll see the same numbers.


Notice please that model/world spaces are not the only ones in place. E.g. texture co-ordinates are (usually) given neither in model nor in world space; instead, they are given in texture space, even when being stored in a mesh. 

#3 GuyWithBeard   Members   


Posted 19 January 2013 - 10:08 AM

Consider the following scene:

- A parking lot area, the form of a square, with the world space origin in the middle of the parking lot area.
- A bike parked in the northeast corner of the parking lot.
- A camera which someone has left on a bench in the southwest corner of the parking lot.

Both the camera and the bike have their own local spaces, and any position within their local space can be transformed into world space. First consider the bike. If you were to sit on it, the front of the bike would be facing "forward", and back "backward". You would grab the right part of the handlebar with your right hand, etc. All of these things remain the same regardless of where in the world the bike is. Whether it is upside down or lying on the ground. Its local space remains the same.

To tell your game where in the world the bike is, its vertices have to be transformed, "moved to", world space. At this point it matters where the bike is and how it is orientated. If you move the bike from the north east corner to the northwest corner, the world position (and possibly orientation) of the bike seat changes. However, the local position of the seat is still the same (ie. it has not moved in regard to the other parts of the bike).

Now consider the camera. It is not in the same place as the bike, so moving a position from its local space to world space is a different kind of transform, but the idea is the same.

Now, if you would want to render the bike from the point of view of the camera, you would need to transform all of its vertices (in bike local space) to the space the camera sees them from (the camera local space, sometimes called view space). To do this, you would transform all your vertices like this:

Bike local space -> world space -> camera local space.

You need to pass through world space since you need to consider "where in the world" the bike is. The GPU actually requires you to transform the camera local space coordinates into what is called clip space. This usually happens in the vertex shader.

The different spaces have nothing to do with handedness. Usually you use either left-handed or right-handed coordinate systems, and thus all your different spaces would be left-handed or right handed. This is of course not enforced, but that's usually how it is.

Old topic!

Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.