I just wanted to get some feedback on whether I should aim for a total re-write or should I try to use Xith3D as one component of my beefing up of the program.
This question is really hard to answer without any deeper knowledge of your source. As a rule of thumb I would say: If you're not happy with your coding, you should consider a rewrite. If you are happy with it, could try to port it. But I don't understand, what you mean about "one component". Xith3D should provide tools for nearly everything you will need for an application like this.
Had a look at the video. Looks like low poly.
Do you think Xith3D will produce noticeable FPS improvements over Java3D? Any guesses on the percentage (10%, 30%, 50%, 100%)?
Yes, I do definitely. But since I haven't directly compared the two in the last time, I cannot tell you any percentage numbers.
But I can tell you, that we did some comparing benchmarks about one and a half year ago with Xith3D vs. jMonkeyengine vs. Java3D, where Xith3D and Java3D were about equally fast and jMonkeyEngine was a bit faster. Well, jME has been tuned as well as Xith3D has. Don'T know about Java3D. But now Xith3D is by far faster than jME. So I would expect it to be even more faster than Java3D.
But you should first port or rewrite your program to/in Xith3D and then ask here on the board about some optimizations to be used/enabled, which will gain a lot of performance.
- Terrain. The program must be configurable to a WIDE variety of different real-life airports. Thus, each of my users will have a completely different set of taxiways, runways, gate aprons, and grassy island areas.
Well, Xith3D has a terrain implementation, which is already working quite well. But it is still work in progress, so the API will change to a more integrated one in the future.
I would like to provide texture's to the various components noted above. As an example, I would expect some combination of concrete, grass, snow, sand, and asphalt surfaces to be necessary.
Hmmm, this should be the smallest problem.
When triangulating these various terrain components will Xith3D hold up well to larger airports which perhaps have hundreds of individual taxiways, grass islands, etc.?
When the scene gets really big, you should consider using some advanced tree structure to improve performance, but that's something, that you should worry about at a later time.
Is there an alternative to triangulating a polygon and then applying a texture that I'm not familiar with which should be investigated?
Well, there's only some kind of "advanced triangulation", which uses strips of triangles (triangle-strips). This can also be combined with indexed geometry. If your geometries are good for being strippified, you should use it as well as indexing to save video memory.
Marvin
General Category
