John Carmacks "100 passes per polygon" achievable

[Fafalada]

You don't control much with displacement mapping.
Nor the necessary density (detail concentration) ,nor polycount.So efficiency is not there, things can get messy and very "Dirty work".Also the only really good displacement mapping i know is based on micropolygon.

As far as realtime displacement mapping goes, one of the primary benefits is that hardware gives you automatic smooth LOD through texture filtering+mipmaps, at least, that's the idea.

I'm also not sure how you can make and solve collisions based on a displacement bitmap (?).

Why would you want to?
Like I said above, you'd usually use something in order of 100:1 less detailed for collision mesh then display mesh, assuming you actually use polygon collision in the first place (many games still do well enough using boxes, or even spheres, for entire objects/characters).

If you ever write a CD scheme you'll understand why - it's O(n^2) with regard to triangle count.

That would only be with brute force approach though, no?

 

[_phil_]

As far as realtime displacement mapping goes, one of the primary benefits is that hardware gives you automatic smooth LOD through texture filtering+mipmaps, at least, that's the idea.

I don't see this much practical ,since, to feed displacement you have to tesselate even where details aren't needed ,to get sufficient polygons density where details ARE needed.This is where i see inefficiency.

Obviously ,we then only need better ,clever tesselation algorythms ,able to put details (polys ) only where needed.
That probably already exists.Some light on this ?

 

[Laa-Yosh]

There's a good quote in "Advanced Animation and Rendering Techniques" (Watt and Watt) which basically says you rarely need displacement mapping.

I'm interested in a little bit more detailed explanation as well.
Although we're doing CGI and not realtime content, but we'd have insane problems without displacement. Then again, we also have PRMan which can render displacement pretty fast and nice...

About normal maps: those won't help you with shilouettes, especially above 1K resolution. If a view dependent, preferably micropolygon-based tesselation and displacement mapping combo can be implemented, then it is a must have as soon as possible.

 

[pc999]

See the reports for UnrealEngine3 lots of detail , landscape and lightning all in one . Also JC will not put anything new before than 2007/8 ( if he only tooks 3/4 to do a new engine ) so you can relax

 

[KimB]

What I would like to see is a greater emphasis on polygon power and not pseudo stuff like bump mapping.

Very good bump mapping can look mighty impressive. See the Unreal engine screenshot here.

Notice how good that statue looks? Now look closely at the edge of its head. There are far fewer polys there than one would first think.

Anyway, of course polycounts should continue to increase, but it's really all about how much visual quality you can get for the performance. And there are bump mapping techniques that can coax tremendous amounts of image quality without impacting performance much at all.