Is ATi's temporal AA a D3D-only feature?

[GameCat]

We should put together a firing squad, I'm sure we would get lots of volunteers from these very forums.

If we do get a squad together, can it please, please kill all the people that use "shadow maps" for projected texture cookie cuts and "shadow buffers" for actual depth-based shadow maps as well? Thanks.

 

[Simon F]

We should put together a firing squad, I'm sure we would get lots of volunteers from these very forums.

If we do get a squad together, can it please, please kill all the people that use "shadow maps" for projected texture cookie cuts and "shadow buffers" for actual depth-based shadow maps as well? Thanks.

I should think that application of a cat-o'-nine-tails should be sufficient for that sort of thing.

 

[LeGreg]

should be called anti-temporal-aliasing.

Makes sense.

 

[Ailuros]

Then I suggest you take your complaint to ATI since they named the feature Temporal Anti-Aliasing, that's what they want people to call it, and that's what people are calling it.

That still doesn't justify the terminology abuse.

Sure, this is getting into semantics and whatnot, but two definitions of temporal are "Of, relating to, or limited by time" and "lasting only for a time" and each pattern only lasts for one frame, so that could be considered temporal.

Depends on the POV. Would you like a simple gaussian type of blur filter being called full scene antialiasing or a LOD hack anisotropic filtering?

It's more like sample dithering in the temporal dimension, yet it still doesn't mean that it has anything in common or results to what is usually known under temporal antialiasing. And it's not that real temporal AA doesn't exist in imagery in general either.

http://freespace.virgin.net/hugo.elias/graphics/x_motion.htm

Full precision means different things to different people/companies. Full precision means FP32 to nvidia, FP24 to ATI and >=FP24 to Microsoft.

That's a matter of implementation. FPxx no matter what value (16, 24, 32, more) is still internal precision.

Just because something is confusing to some doesn't make it invalid. Most people wouldn't know what you're talking about if you ask someone a question about ATI's spatial frame dependant anti-aliasing, but if you asked them about ATI's Temporal Anti-Aliasing they would have a much better chance of knowing what you're talking about, whether or not it is considered technically correct.

In these forums the majority of participants know what the difference is and I'd expect an experienced engineer to attempt to set the record straight too.

I've spend more than a couple of hours looking at the method in question in real time under different conditions and neither has the result anything to do with real temporal AA, nor has it any specific uses for me either. If I have enough performance to spare (an important conditional for the "tempAA" to actually work) I'd much rather pick a higher resolution with more AA samples and call it a day.