[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: VGA cards



On Tue, Dec 15, 2009 at 14:39, Camaleón <noelamac@gmail.com> wrote:
> On Tue, 15 Dec 2009 14:12:58 -0800, Kelly Clowers wrote:
>
>> On Tue, Dec 15, 2009 at 13:48, Camaleón wrote:
>
>>>> So what, you want AMD to hire a bunch of devs to write a complete open
>>>> source driver? They can't afford to do that.
>>>
>>> They couldn't do that... although they wanted.
>>>
>>> AMD has not released the full 3D specifications for their drivers, only
>>> *partial documentation* for some of their chipsets, so even with 100
>>> dedicated engineers working on ATI drivers, we still have an incomplete
>>> open source driver.
>>
>> What? Where did you get that? 100 dedicated devs would be absurd for any
>> driver, and their sure aren't that many in the case of ATI.
>
> That "100 people" was just a "supposition", sir :-)

Hmmm, then that sentence needs some qualifiers or something:

"so even *if we had* 100 dedicated engineers working on
ATI drivers, we *would* still have an incomplete open source
driver."

Which probably isn't true anyway, with that much in the way of
resources, you could reverse engineer it in short enough order.

>> There are a
>> few (like 3 or 4 I think) people employed to do work on ATI drivers, I
>> don't know that those are their only duties. And then there are a
>> handful of people that do some work in their free time.
>>
>> Arguably even more than docs, a shortage of developer-hours is a major
>> problem for all Xorg related work.
>
> Just 3 or 4 people should be enough if they could get access to the full
> specs for the hardware. But that is not the case.

3 or 4 people, if they had enough time (where enough = lots). Writing 3D
drivers is no trivial task even with full documentation, which they have (or
really close to full at least).

>>> >From AMD website [1]:
>>>
>>> ***
>>> Is complete driver source code available?
>
> (...)
>
>> That is simple referring to their closed source drivers.
>
> Which are the only ones having full 3D support and capabilities.
>
>> The only thing
>> I know of that has actually been held back is some of the MPEG
>> acceleration stuff, since ATI implements it by licensing patents from
>> MPEG-LA . AFAIK, anything else that has not been released is due to the
>> docs not actually existing in a coherent form, or they are still being
>> cleared by the legal department.
>
> And do you find that a "linux-friendly" approach?

Yes.

> The status of the supported capabilities by "radeon" and "radeonhd"
> drivers is as follows:
>
> http://xorg.freedesktop.org/wiki/RadeonFeature
>
> http://xorg.freedesktop.org/wiki/radeonhd%3Afeature
>
> Does not look so good.

I am very familiar with those pages. There is still more work to do of
course, but all in all, it looks like very good progress.

>>> I am not seeing here nothing but the same "arguments" provided by
>>> nvidia and other supposed "linux-friendly" hardware vendors out there.
>>
>> I don't know that anyone has called Nvidia "Linux friendly".
>
> No, but at least nvidia clearly says what is their thinking about this.
>
>>>> What they have done is fantastic and really about the very best that
>>>> can reasonably be expected.
>>>
>>> Well, yes.
>>>
>>> But AMD could do more for its Linux users and in fact, does not :-(
>>
>> Yes and no? What is this, a quantum superposition?
>
> I agree that Xorg people have done a very good job (by their own) with
> radeon/radeonhd drivers.

I wasn't speaking of the independent xorg devs (although
they also do a good job), I was saying AMD is doing a very
good job.

> But I have to disagree in regards AMD/ATI. It's not a linux-friendly
> company and has not released the full specs for their vga cards. Just
> some papers. In these days, that's not enough.

Just some papers? What else would they release?
What specifically do you think they need to release?

Look at the r600 docs for example: a 342 page Instruction Set
Architecture guide, a 43 page 3D acceleration guide, a 166 page
3D register guide and sample code for manipulating the atom
command processor.

Further, every time I read a blog from an xorg dev or an interview
with one, they have nothing but praise for what AMD has done.


Cheers,
Kelly Clowers


Reply to: