Replacing the terrian entirely with a .dts mesh
by Roger Smith · in General Discussion · 01/29/2003 (10:07 am) · 5 replies
I was wondering if anybody knew if Torque would run faster using a dts mesh for terrain instead of the height-field??? Im not sure, becuase i cant code or understand C++. It was just an interesting idea i thought of. Replacing the terrain entirely isnt neccasary really, but i was mostly interested in the performance impact. Id like to model some LARGE terrian shapes like mountains.
#2
Virtual GIS
read the papers and watch the movice very enlighting.
01/29/2003 (4:54 pm)
This is one of those questions that should answer itself. It would be faster wouldn't that be the way there would have done it from the beginning. I do have some useful info to contribute, everything this you wanted to know and more about terrain optopmizations.Virtual GIS
read the papers and watch the movice very enlighting.
#3
Edit- Someone suggested to add more channels to the heightmap image so concave shapes could be generated.
01/29/2003 (5:13 pm)
I see. so the more dts in memory and being rendered and processed the more drastic impact compared to heightmaps. It seemed to me that an engine could organise and process dts pretty efficiently, and i was wondering if it was feasibl solution to the restrictions on the heightmap used for generating terrain.Edit- Someone suggested to add more channels to the heightmap image so concave shapes could be generated.
#4
I would suggest making like Tribes - use the terrain as the basis for your levels, then use interiors for any subterranean areas. That way you can do caves and such, without a huge performance hit.
01/29/2003 (11:29 pm)
That's fairly complex to do, especially when you're doing dynamic level of detail - which is the whole reason Torque is fast at terrains! :)I would suggest making like Tribes - use the terrain as the basis for your levels, then use interiors for any subterranean areas. That way you can do caves and such, without a huge performance hit.
#5
Terrains are custom built for a purpose. DTS models (I assume?) are not culled with any kind of special way, and even if they are it would not likely be optomised for Terrains.
The reason you can get away with a very high detail terrain with very low polygons is that in an average situation you will likely only be able to see 1/4 of the terrain at a time, and even more of that is not visible [oops! edit: because it is obscured] by the terrain itself. This will throw minimal polygons at the renderer, however a DTS file (I assume they are simply meshes?) will throw everything at the renderer.
01/30/2003 (12:55 am)
Quote: It seemed to me that an engine could organise and process dts pretty efficiently,
Terrains are custom built for a purpose. DTS models (I assume?) are not culled with any kind of special way, and even if they are it would not likely be optomised for Terrains.
The reason you can get away with a very high detail terrain with very low polygons is that in an average situation you will likely only be able to see 1/4 of the terrain at a time, and even more of that is not visible [oops! edit: because it is obscured] by the terrain itself. This will throw minimal polygons at the renderer, however a DTS file (I assume they are simply meshes?) will throw everything at the renderer.
Torque Owner Mz
If you want to know why, do some terrain research. www.vterrain.org is a good place to start.