shadow volume will be Large quantity usage in the future??

[wzh100]

I remember that someone said that JC would use shadow mapping in next game,but in 2002,JC said:"Shadow buffers make good looking demos with controlled circumstances, but when you start using them for a “realâ€￾ application, you find that you need absolutely massive resolution to get acceptable results for omni － directional lights, and a lot of the artifacts need to be tweaked on a per-light basis. While it is possible to do shadow buffers on GF1/radeon class hardware, without percentage closer filtering they look wretched. If we were targeting only the newest hardware, shadow buffers would have a better shot, but even then, they have more drawbacks than are commonly appreciated"
so i think that shadow volume will be Large quantity usage in the future.what do you tink of it??

 

[pcchen]

It has changed since then. There are many methods help reducing the aliasing/resolution problem. Of course, they are not yet perfect, but shadow volumes were also "perfected" only recently (in the past capping is a major problem).

Shadow volume are not very compatible to current GPU models: it needs CPU to compute the contour, therefore many things are restricted (for example, you'll have to use CPU for computing skinning). It's also not compatible with alpha textures. Another problem of shadow volumes is the heavy requirement of fillrate.

 

[wzh100]

I have read the passage about shadow volume which was written by 991060,as I know,the shadow is too Complicacy to realize,you need more and more alus,fp units,vertex,piplines,ROPs and some others to improve the z/stencil proformance.it is too hard for hardware,for shadow volume to say,the more you get,the worse you will,z-pass is not a good way to realize dynamic source/lighting,although we have the way to rectify.but shadow mapping is not a good way either,although we have PSM and ASM to improve the effect.so what is the best choose to shadowing? i really don't know.

 

[Reverend]

You really should try and structure your sentences better. I read about 4 or 5 words of your entire post and I give up reading the rest of it.

 

[wzh100]

You really should try and structure your sentences better. I read about 4 or 5 words of your entire post and I give up reading the rest of it.

I feel sorry,I am in diligent study English now,English is not my mother tongue.

 

[Scali]

Shadow volume are not very compatible to current GPU models: it needs CPU to compute the contour, therefore many things are restricted (for example, you'll have to use CPU for computing skinning).

Didn't we go over this 89403284932 times before?
You do NOT need the CPU, take 3DMark03 as an example. Its shadowvolumes and skinning are 100% GPU-based (which is why it is not CPU-limited at all, unlike Doom3, which does not use the GPU-based approach). On an R300+ or NV30+ this is a faster approach than the fastest CPU at this time, and chances are that GPUs will scale in speed much faster than CPUs in the next few years, as this has been the trend since the first GPU was introduced.

 

[Vifani]

Anyone know if Shadow volumes hardware accelerated 100% with penumbra wedges will be massively used ?

 

[Mintmaster]

Anyone know if Shadow volumes hardware accelerated 100% with penumbra wedges will be massively used ?

The volumes you need to construct for ordinary shadow volumes are quite a bit different from those needed in penumbra wedges. I should never say never, but it would be quite difficult to do this on the GPU. I think it is possible if you use a lot of degenerate triangles per edge. I counted 4 vertices and 10 triangles for each edge in the original model.

 

[Vifani]

Anyone know if Shadow volumes hardware accelerated 100% with penumbra wedges will be massively used ?

The volumes you need to construct for ordinary shadow volumes are quite a bit different from those needed in penumbra wedges. I should never say never, but it would be quite difficult to do this on the GPU. I think it is possible if you use a lot of degenerate triangles per edge. I counted 4 vertices and 10 triangles for each edge in the original model.

Yes, but the result is a perfect soft shadow without any aliasing (view in shadow mapping) and physically correct.

Do you agree ?

 

[Coriolis]

Anyone know if Shadow volumes hardware accelerated 100% with penumbra wedges will be massively used ?

The volumes you need to construct for ordinary shadow volumes are quite a bit different from those needed in penumbra wedges. I should never say never, but it would be quite difficult to do this on the GPU. I think it is possible if you use a lot of degenerate triangles per edge. I counted 4 vertices and 10 triangles for each edge in the original model.

Yes, but the result is a perfect soft shadow without any aliasing (view in shadow mapping) and physically correct.

Do you agree ?

It's a near-perfect soft shadow that is only perfect in certain controlled cases. There are a couple of limitations that make it imperfect in general use. One limitation is that it assumes there is a single silhouette as seen by all points on the light; this will always be wrong for any area light at certain placements, though in practice it isn't too bad if you don't make your lights too big. Another limitation is how it combines the shadow wedges for overlapping wedges; there's no way for it to know if the overlapping shadows should use the maximum of the two occlusion factors, the sum of the two occlusion factors capped at 1, or some fraction between the two. These limitations aren't exactly unique to this particular soft shadow technique.

I would expect this to show up in some games in the future, but personally I think that shadow maps are more likely to dominate. Shadow maps are just more scalable and handle binary alpha more easily.

 

[anaqer]

Anyone know if Shadow volumes hardware accelerated 100% with penumbra wedges will be massively used ?

I most certainly don't know, but have we not come to the conclusion that shadow volumes are a dead end in the long run, and thus dedicating any more transistors to hardware accelerating it is just a waste of good silicon?

 

[Ylandro]

Shadow volumes might be a dead end in the long run, because of the rendering characteristics.... But what's the alternative? Shadow buffers certainly aren't without problems.

It would be silly to dismiss a technique NOW, just because it isn't pratical to use in 10 years time...

It's all a matter of compromise to get your rendering done within the confinement of your rendering capabilities.