In-game rendering speed for objects
by James · in Artist Corner · 03/06/2006 (9:28 am) · 5 replies
Would it be safe/correct to say the following:
DIF - slow to render (compared to DTS) but fast for collision detection
DTS - fast to render (faster than DIF) but slow for collision detection
Terrain - faster than DIF but slower than DTS to render. Fast collision detection
Need this bit of knowledge to make intelligent decisions on creating a map :)
Thanks in advance.
DIF - slow to render (compared to DTS) but fast for collision detection
DTS - fast to render (faster than DIF) but slow for collision detection
Terrain - faster than DIF but slower than DTS to render. Fast collision detection
Need this bit of knowledge to make intelligent decisions on creating a map :)
Thanks in advance.
#2
What I'm asking is my generalization accurate?
Your feedback seems to suggest so (shadows dynamic or otherwise are not a concern here).
A usage example would be: do I go to the trouble of making buildings in DIF format when nothing will collide with them (background) or to speed things up (rendering) would it make sense to use DTS?
I just want to understand how the formats perform when compared on two basic levels: render vs collision detection :)
03/06/2006 (1:04 pm)
Stefan, I know they are made for their respective usage :)What I'm asking is my generalization accurate?
Your feedback seems to suggest so (shadows dynamic or otherwise are not a concern here).
A usage example would be: do I go to the trouble of making buildings in DIF format when nothing will collide with them (background) or to speed things up (rendering) would it make sense to use DTS?
I just want to understand how the formats perform when compared on two basic levels: render vs collision detection :)
#3
Dts has bounding box collision or user specified collision (i.e you make an invisible collision mesh) but is obviously better for complex objects. No convex shape restriction for the normal mesh (though the invisible collision mesh, if used, still does).
terrain has accurate collision and can do curvy shapes, it exchanges the convex shape restriction of dif for a rolling landscape one (try making a perfect 90 degree angle!).
for your specific example :
03/06/2006 (1:47 pm)
Dif has accurate collision but if you are planning on making something complex avoid dif, it's also restricted by having to use convex shapes to make things (i.e boxes and wedges).Dts has bounding box collision or user specified collision (i.e you make an invisible collision mesh) but is obviously better for complex objects. No convex shape restriction for the normal mesh (though the invisible collision mesh, if used, still does).
terrain has accurate collision and can do curvy shapes, it exchanges the convex shape restriction of dif for a rolling landscape one (try making a perfect 90 degree angle!).
for your specific example :
Quote:do I go to the trouble of making buildings in DIF format when nothing will collide with them (background) or to speed things up (rendering) would it make sense to use DTS?i would use dts. As you said, it isnt going to need collision and you can make way more complex looking buildings with dts then dif. Have you used quark or radiant? ...
#4
The main restriction with difs are their static nature the fast collision and rendering is gained by pre-processing prior to use (building the binary space partition), so it cannot be changed dynamically and still have the optiimzed performance. In addition, it is severely restricted to convex requirements (brush based, or csg editing), while dts shapes provide more flexibility and dynamic modification.
DIF's do lose a lot of their performance benefits when viewed externally, because you are outside the shape, and therefore lose many of the culling advantages of the BSP structure.
03/06/2006 (6:05 pm)
Actually, dif has very fast rendering, and when viewed from inside the structure, extremely fast rendering due to the ability to cull out non-visible surfaces (see portalization).The main restriction with difs are their static nature the fast collision and rendering is gained by pre-processing prior to use (building the binary space partition), so it cannot be changed dynamically and still have the optiimzed performance. In addition, it is severely restricted to convex requirements (brush based, or csg editing), while dts shapes provide more flexibility and dynamic modification.
DIF's do lose a lot of their performance benefits when viewed externally, because you are outside the shape, and therefore lose many of the culling advantages of the BSP structure.
#5
Thanks everyone for your responses, it's all been very much appreciated :)
03/06/2006 (6:43 pm)
Thanks Stephen, that's very helpful and what I needed to understand.Thanks everyone for your responses, it's all been very much appreciated :)
Torque Owner Stefan Lundmark
I would put it more like this:
DIF - slow but accurate collisions, static shadows
DTS - inaccurate collision detection, less detailed but dynamic shadows
Terrain - bla bla :p
They are simply made to fit their respective usage. A terrain is meant to have large portions of it visible at one time, while LODing the parts which are far away.. while a DIF can have portals to speed up FPS, something a terrain does not need.. etc.