Hi, I'm currently rewriting my DX9 init stuff and therefore testing on some platforms. Using an IBM Thinkpad T60, which has a Mobile Intel 945GM Express Chipset I find, that it reports shader support VS0.0 and PS2.0, as well as no HWTRANSFORMANDLIGHT. I'm just curious, if anyone found similar caps on other devices and how this would be handled best? I don't think, emulating shaders in software is a real good choice. My current action would be to deny the app running on such a system, but hey: it comes with DX9c and is a new notebook! How can a Pixel shader be really useful if no vertex shader is present? I learned from several articles, that either using both shaders or no shaders at all is best practice. Thanks for some ideas, Herb

Emulating vertex shaders in software usually isn't that expensive. I seem to remember that's what the drivers of that chip does.

Quote:Original post by asp_
Emulating vertex shaders in software usually isn't that expensive. I seem to remember that's what the drivers of that chip does.

Seconded. Pre-GeForce2, all transform and lighting was done in software. Making a software shader isn't too expensive (Although obviously not as good as having a hardware shader), I suspect it might even be compiled into machine code by the driver so it can just execute the shader.

Thanks for your useful replies. I did some more search and found also interesting articles in ShaderX2 and ShaderX3 books about software vertex shader processing.
My code just failed because I have some test routines for shader caps in my compileShader... functions. Doing a quick hack and filling in a 2.0 in the caps field and using SW_VERTEXPROCESSING showed, that it runs with good performance, as you said.
Thanks again,
Herb