If ATi opens up its source code, many Linux users will be happy. For me though... I don't ask for open source development. I ask, first and foremost, for functional and stable drivers regardless of either open or close source code. I just think opening the source code might be a quicker means of arriving at the stability I like.
Just provide the goddamn Data Sheets, so good drivers like "radeon" has been for so long may be included in every mainstream distribution, and I strongly believe ATI's love-o-meter will skyrocket for users of GNU/Linux in a few months.
Heh... If only it that were THAT simple...
I won't argue with Mr. Tippet's (Of ATI's OpenGL driver fame...) assertion that it's one of the more complicated tasks, writing code for OpenGL acceleration. He's pretty much dead to rights on that one. Not everybody can do it, especially with things like programmable shaders, etc.- it's not JUST about registers, it's about telling the extremely high performance dedicated processing units in the GPU what to do to render scenes in the right way nowadays... It can be done, but it's not as simple as the work that was done with the fixed path devices of the earlier eras like the RagePRO, Rage128, i810, or G200/400/500.
The big problem with this all is that companies like ATI have "Intellectual Property" attorneys that have indicated to them that info of this nature may impair patentability, etc. Right or wrong, they think that leakage of any of this material will cost them a business edge or some valuation in their company.
I think we all know what our opinions and thoughts are on this. They think they're being careful, but they use Taiwanese and Chinese fab partners, which have NO qualms about cribbing from them to make their own stuff, albeit poorer quality- if they so chose. There's really nothing much to "protect" except the income of the IP Attorneys...
Intel's fired the first shot across the bow with this one. The current drivers for the 3000 series are a mixed bag, some things it pastes ATI's drivers with (Though not NVidias...Go figure...). Others, it just gets pasted and badly at that. But they're largely open and they're improving as we discuss this. I can only hope AMD figures that one out with their new purchase. Intel's trying to one-up them, and it's up to AMD to decide if they want to steal that thunder or not.
Not sure whom to believe, though. The OpenDesktop people indicate that their contacts within AMD indicate that this is pretty much a certainty- that there will be some sort of genuine opening up of things. From my perspective (can't say how I know all of this- just let it suffice to be that I DO know what I'm talking about here...), I just don't see that happening right at the moment.
You may have to change the red blog to the green blog soon Michael. I think the ATi brand is going.
Heh...I don't think...I pretty much know...
It's kind of obvious, especially in light of the statements of a desire to move the bulk of GPU computational ability into the x86 ISA- that means ON CPU, like MMX/SSE/SSE2- hopefully the new stuffle will do better than the prior attempts. Biggest kill in the other previous attempts wasn't that it wasn't performant- it's that the poor damn CPU ends up stalling at start and stop points for all the "extension" instruction sets; so it's of limited usefulness in providing computational boost for 3D graphics or anything else.