Pixel Shader 2.0+ How hard would it be?

[Shalmanese]

Okay, by now everybody should have seen the NV30 previews so I was wondering exactly why nVidia chose to go with Pixel Shader 2.0+. Initially, I thought it was just a stupid move since no game will come out which can only be run on NV30 cards and not R9700 cards since it would risk losing a huge proportion of the market (especially when DX 10 & 11 will already be out so they might as well go DX 10). But, how hard would it be to actually implement PS2.0+ as opposed to just plain vanilla DX9 spec? It seems that nVidia were able to throw around a lot of very big numbers which would inevitableby be useless anyway due to bottlenecks elsewhere in the system.

 

[Mavrick]

Really not sure about this, but since nVidia's Cg compiler fully supports Pixel Shader 2.0+, I bet it would be quite easy for a developer to just program in Cg and then compile the code in PS 2.0+. Of course, there would also have to a plain DirectX 9 version for the non-FX owners.

But since most games today come out with optional DirectX 8 shaders (which can be turned of on pre-GeForce3 cards), I guess it would not be to hard

 

[FuriousBroccoli]

As a game developer I can tell you flat out, 2.0 will not get used for games, ever. It'll be really cool for high-end offline rendering, but no developers are going to spend ages coding duplicate code for the NV30.

CG/HLSL won't help, because the advantages of 2.0+ (it's officially called 2.x by MS), are mainly program length, so if you go over the R300 length, a compiler can't do anything about that at all, so you either stick to 2.0 or write two shaders for every effect.

The final factor is that of practical usage. If you fully used 2.x, you'd get 2fps on an average game scenario. 2.0 features are the sweet quality/performance spot that developers will be writing for.