Decompression of these kind of simple quantized formats has been part of VS specification since VS1.0, and was already made standard fare by architectures predating VertexShaders by 2 years.Robbz said:The thing I was a bit unsure about was if you actually can decompress such things properly in the vertex shader.
It's also pretty much mandatory to use if you want to get anywhere near optimal performance on any current console architecture.
And it'd be even nicer if we had infinite memory, bandwith, and computational power. Sadly there's no such thing as a free lunch in realtime processing - there will always be limitations.Still, it's a pity that one have to do this "extra" work because of memory BW considerations. It'd be a lot nicer if the BW would be "enough" to sustain the maximum precision
PS2 automatically compresses output vertex data for you in this kind of fashion (it's part of the bus arbitration circuit), and we're talking about 6 year old architecture. Obviously it can't be all that hard to do can it?I was just not convinced that it would be a good thing for realtime tesselated stuff.
Btw, as far as base storage goes, scalar/vector quantizations was what we used as baseline for the last 5 years, I am expecting to do much more then that this time around.