Shadow buffers + stencils for JC's next engine

[Reverend]

Posted this confirmation quite a while at my temp site.

Does that mean the next game from id Software wouldn't be strictly indoors? How is he "mixing it up" (I didn't really ask him after getting his email)? Maybe only for certain game properties? Like what? And damn, he didn't answer go into my question about soft shadowing.

 

[Cryect]

First did he say he was just using the stencil buffer or explicitly say he was using stencil shadows?

 

[Blazkowicz]

UE3 use 'em both as well, doesn't it?

 

[KimB]

Well, considering the last game from id Software wasn't strictly indoors....

 

[ERP]

There are some interesting hybrid algorithms that use both to draw a single shadow.

 

[Richard]

Cryect, at QCon 2004 he said he was using stencil buffers but not for shadows.

I select out the areas that are going to be receiving shadow calculations on there, which I actually use stencil buffer tests (all the work with stencil buffers and the algorithms for that is still having some pay-off in the new engine, even though we’re not directly using that for shadowing)

 

[991060]

Basically, if you want soft shadow, shadow mapping is the way to go, no matter your scene is indoor or outdoor.

 

[991060]

There are some interesting hybrid algorithms that use both to draw a single shadow.

like this?
http://graphics.csail.mit.edu/~ericchan/papers/smapSV/

 

[Dave B(TotalVR)]

There are some interesting hybrid algorithms that use both to draw a single shadow.

I imagine, if properly lined up, you could save time on the shadow buffer rendering by rendering an outline only, then use it at the same time as a stencil shadow - which could produce shadows with stencil accuracy and a nice cheap softness.

 

[KimB]

Doesn't seem likely, Dave. You still have to calculate, by some means, the location of the shadow edges.

Sounds to me that what JC basically is doing is normal shadow map rendering, but when the shadow maps are applied to the scene, a stencil buffer is used to determine which pixels are at the edge of the shadow, and then extra filtering is done at those positions, to improve the look of the shadow edge.

This is different, by the way, from the paper that 991060 referenced, which was basically a way to use a low-res shadow map to figure out where the shadow edges in the scene are, then just doing shadow volume rendering in those places near a shadow edge to improve shadow volume rendering performance.

 

[Dave B(TotalVR)]

"Doesn't seem likely, Dave. You still have to calculate, by some means, the location of the shadow edges."


You have to calculate that for normal shadow maps, apply the same calculation but when you are inside the stencil volume you wont need to do the blending so hence it would be slightly faster than normal shadow mapping - except that you ar also rendering a stencil volume.

Sounds to me that what JC basically is doing is normal shadow map rendering, but when the shadow maps are applied to the scene, a stencil buffer is used to determine which pixels are at the edge of the shadow, and then extra filtering is done at those positions, to improve the look of the shadow edge.

That also sounds very plausible.

 

[Ostsol]

You probably still need to do the blending, actually, if you want soft shadows. Also, the outer edges of the penumbra will extend beyond the boundry defined by a basic shadow volume.

 

[Intel17]

Sounds to me that what JC basically is doing is normal shadow map rendering, but when the shadow maps are applied to the scene, a stencil buffer is used to determine which pixels are at the edge of the shadow, and then extra filtering is done at those positions, to improve the look of the shadow edge.

I agree with you Chalnoth. This sounds like what he mentioned at QCon2004 (as Mordenkainen kindly quoted for us).

I seriously doubt he'd use multiple shadowing algorithms in the same renderer because, he was a heavy proponent of unification and doing everything the same way.

EDIT: I've just emailed JC, if I get a response, i'll post it here.