[Discuss] open source graphics drivers

[Richard Pieri]

Martin Owens wrote:
> We don't get to act on that. We can hope on it, but writing better
> drivers is more convincing to a company than sitting on ones hands.

If that assertion were true then nVidia and ATi would have opened up 
their architectures ten years ago. Instead, the trade secret mentality 
in both companies not only kept most things closed and proprietary but 
added layers of obfuscation to make it extremely difficult to reverse 
engineer their products.

> That's what reverse engineering is about. Learning how the hardware
> works. It's not perfect, but with enough resources it can be done.

Given what's been discovered from the documents nVidia has released so 
far? It's no where near good, never mind perfect. There's a lot about 
the Kepler architecture and its various implementations that the reverse 
engineers got wrong. Example: the red screen bug which has been 
frustrating developers for something like two years. Not a bug at all. 
It's a debug setting left enabled on some cards at the factory.

> I wonder how much of the softening has been Valve's John Carmack poking
> their CEO in the ribs about his awful yet good proprietary drivers.

Certainly a contributing factor, but I figure Android and CUDA/OpenCL 
have more to do with it. Steam Machines may be revenue in the future but 
Android and CUDA are revenue today.

By the way, it's Gabe Newell at Valve. John Carmack is over at iD 
Software and Armadillo Aerospace.

-- 
Rich P.