GDC 2006 presentations from Microsoft

Wonder how quick the uptake on DX10 will be. Looks like the fallback path to DX9 won't be a viable approach in a lot of cases and that mind hinder adoption somewhat. Exciting stuff though. Those slides must scare some of the lower tier developers.
 
They say HLSL compiler will target SM 3.0/2.0x/1.x code - does it mean developers can just ship D3D10 renderer and do not worry about refactoring old shaders?
 
trinibwoy said:
Those slides must scare some of the lower tier developers.
An interesting effect of the no-caps approach in D3D10 might actually make Direct3D development easier for the "little guys" (especially indies).

The combinatorial explosion with D3D9 hardware caps (Several IHV's + Many GPU's + Many Drivers) can be a serious headache for both development and support. Testing can only be done on easy to find and common hardware - it's too impractical to test even AAA engines on every combination of hardware. Support obviously becomes an issue - distribute your application and someone emails/phones/posts a "It doesn't work on my PC, why?" and it boils down to their exotic hardware/software configuration...

Testing won't become optional, but the fixed-caps in D3D10 vasty reduce the number of combinations as well as the effectiveness of limited testing.

They say HLSL compiler will target SM 3.0/2.0x/1.x code - does it mean developers can just ship D3D10 renderer and do not worry about refactoring old shaders?
No, D3D10 is SM4, anything else (currently) is Direct3D 8 or 9. The current HLSL compiler is good, but from what I've been told it's still pretty simple as far as compilers go. It's getting super-charged for D3D10 and seems that they're putting lots of work into improving it. With FXC supporting down-level compilation it should mean that you can get some of those improvements for current shader models. Better constant packing, better optimization, faster compilation etc..etc..

hth
Jack
 
Back
Top