The return of Talisman concepts

[CoolAsAMoose]

One neat feature of the Talisman architecture proposed by MS several decades ago (or at least back in -96 or something) was the inter-frame reuse of rendered layers (which were also compressed). Although I don't see the layer solution coming back, the reuse of the parts of the frame that are not changed between frames would make sense in all architectures that subdivides the screen in tiles.

I believe I have seen this concept in some presentation of PowerVR as well, where only the tiles that have changed are re-rendered. But is it actually used? Does it need application support? Could "traditional" architectures use this method?

 

[Dave B(TotalVR)]

Is there really any need for it, the onyl time they aren't gonna change is when you aren't moving and thats when FPS matters the least.

 

[mboeller]

If You want further info's about the Talisman architecture and rendering in Layers then this is an good page :

http://research.microsoft.com/MSRSIGGRAPH/


microsoft has done a lot of research for layered rendering. You can find most of it on the subpages.

 

[Qroach]

Dave,

It went farther then just reusing the rendered image over again. it was basically saving you from redrawing something unless there was a severe change from frame to frame. So basically you'd morph the shape of what was rendered, if something fell within a few pixel difference from frame to frame. it would save on both a scene that doesn't move and a scene that does move.

 

[duffer]

Talisman was even more flexible than that. It could do affine transformations on the pieces of the scene, so you could continue to reuse elements even if they moved in 3D.

Think of Talisman as sprites on steroids.

I think we'll certainly continue to see sprites being used in games, but they don't seem to be the best primitive to base your whole 3D rendering architecture on.

 

[Qroach]

Well, it's sort of like sprites, but I think it may be closer to rendering to a texture...

 

[Nexus]

Trespasser would have been the game for Talisman hardware. It used dynamic sprites for objects far away. The engine rendered an object from the viewpoint of the player into a texture and then used it as a sprite as long as the viewpoint was still at least partially correct. If it wasn't considered to be still correct/tolerable the texture was recalculated or the object was drawn in full 3D if it was close enough. It was a very interesting system for the time. Even Comanche4 uses something similar today, but in general I doubt we will see many games with it, todays hardware has so much power, we shouldn't need sprites anymore.

 

[Joe DeFuria]

Trespasser...ack...I still have nightmares about trying to something as mundane as stacking boxes.