Re: [Gnu3dkit-dev] OGL 2.0 shading

[Gerard Iglesias]

Philippe C.D. Robert wrote:

> Hi,
>
> did someone play with the shader compiler of the OGL 2.0 proposal form 
> 3DLabs? If yes, could this be used as foundation or starting point for 
> our shader compiler? 

I am reading what I can around the shading language, especially I read 
this QA :
                                       
http://www.cgshaders.org/articles/interview_davidkirk02.php

And found this one :

* What's the difference between an "offline" shader language (such as 
RenderMan) and a "real-time" shader language such as Cg? Why can't the 
former just be compiled directly to assembly code for GPUs?*

Not only is Cg a "real-time" shader language, Cg has been designed to be 
hardware friendly and hardware efficient. This means that Cg should run 
well on most modern GPUs. Cg is like C - C for graphics, or C for GPUs. 
Cg specifies data types, data connections, and arithmetic and graphics 
operation primitives that exist in GPUs.

RenderMan is a still higher language and is custom designed for shading, 
not for optimal execution on GPUs. There are several universities and 
tool providers who are in the process of developing a compiler that 
translates from RenderMan shaders to Cg code. That provides the best of 
both worlds. You get the high level shader focus of RenderMan combined 
with the runtime compilation and efficient execution of Cg on GPUs.

Cg is also more than a shading language. Cg can be used to program any 
function that can be run on GPUs, including geometry, physics, collision 
detection, and also many non-graphics tasks.



The other things are interesting to read also....

And I find their work more visible and active than the one from 3DLabs, 
didn't see anything new from a long time, while NVIdia promise version 
for OSX soon.

What do you think about this?

  Sincerely

       Gerard