|
|
Constructive geometry vs collision primitives
#13243
08/08/03 12:55
08/08/03 12:55
|
Anonymous
OP
Unregistered
|
Anonymous
OP
Unregistered
|
Most (not all) of the commercial game engines I've worked with use collion shapes (solids) and visual shapes which are distinct from each other.
For environmental objects the Conitec engine seems to favor a constructive geometry approach where the environment is build out of primitives which are used for both the visual and the collision database.
Does anyone know if there is a performance penalty associated with constructing collision solids out of primitives and making them invisible and overlaying entities which supply the visuals, but which are non-collidable? In addition to being a more familiar way to work this approach would allow me to make the visual shapes more complex and LOD appropriately without requiring more complex collision solids.
As a new user (though an experienced real-time technical artist) I'd be curious about any other comments on the merits and shortcomings of these two approaches.
Thanks
|
|
|
Re: Constructive geometry vs collision primitives
#13244
08/08/03 13:17
08/08/03 13:17
|
Joined: Feb 2003
Posts: 2,826 Margaritaville (Redneck Rivier...
myrlyn68
Senior Expert
|
Senior Expert
Joined: Feb 2003
Posts: 2,826
Margaritaville (Redneck Rivier...
|
You can do this quite easily if you would like - the only thing that would be of concern is the additional faces which would need to be drawn in order to realize it. Even if you where to make the level geometry invisible, the engine would still need to determine where it is in relation to the character in order to allow for proper collision detection. So basically, you take a hit in your overall scene budget for both the visible and invisible faces in this situation.
There are also other considerations in regards to the way shadows and lights interact with CSG geometry in A6, in comparrison to how they interact with normal mesh objects.
Depending on your timeline you might not have to worry too much though - they are planning on moving away from the current method in the future (near or far is unknown as of yet). Unfortunately it is a "loose" schedule, so it would be hard for anyone but JCL to say if it is 3 or 6 months away or even more.
Virtual Worlds - Rebuilding the Universe one Pixel at a Time.
Take a look - daily news and weekly content updates.
|
|
|
Re: Constructive geometry vs collision primitives
#13245
08/08/03 15:23
08/08/03 15:23
|
Joined: Jul 2000
Posts: 11,321 Virginia, USA
Dan Silverman
Senior Expert
|
Senior Expert
Joined: Jul 2000
Posts: 11,321
Virginia, USA
|
quote: There are also other considerations in regards to the way shadows and lights interact with CSG geometry in A6, in comparrison to how they interact with normal mesh objects.
This all depends on how you go about building your level. If, for example, you build your meshes, light them and BAKE the lighting/shadows into the skin/texture, then you can create very realistically lit areas with a lighting that is better than the native lighting on level geometry.
|
|
|
Re: Constructive geometry vs collision primitives
#13247
08/08/03 17:05
08/08/03 17:05
|
Joined: Jul 2000
Posts: 11,321 Virginia, USA
Dan Silverman
Senior Expert
|
Senior Expert
Joined: Jul 2000
Posts: 11,321
Virginia, USA
|
quote: But, there could be penalties with an oversaturation of sprites if one goes overboard on quantity, size and with respect to portal counts.
Portal counts have nothing to do with sprites or models. Only with level geometry.
|
|
|
|