AMD Preparing For Another GPU Documentation Release

[seeker010]

Well, considering that the bulk of the critical OpenGL support is present in those already (Looks at his R200 class 3D accelerator in his P4M 1.7 GHz laptop that PASTES his more modern Athlon64 3200+ driven Xpress200m accelerator with most games that don't rely on OpenGL 2.X GLSL code...) it's not too bad. You're missing a few things like TV out, right? That is all we need from them at this point, really. Pretty much everything else seems to run rather well in comparison.

last I checked the fglrx driver still outperforms the r200 driver, although it's hard to do a comparison now because the fglrx is so old and doesn't compile nicely.

and also, I think this still holds true

. Among the areas that were missing from the specifications were Hyper Z programming information, Truform, DX8 multisampling/FSAA support, vertex shader, ATI_fragment_shader, TV-Out, MPEG decoder, BIOS table layouts for the external TMDS transmitter, errata and chip bugs, and finally performance counters and debug information. Fortunately, however, thanks to reverse engineering of the fglrx driver, Stephane Marchesin (now known for his Nouveau work) and Roland Scheidegger had reverse engineered the Hyper Z programming information. Still lacking, however, is information on the hierarchical Z feature. Scheidegger had also reverse engineered the R200 vertex shader while David Airlie had lent him a hand in getting the ATI_fragment_shader to work.

it would be nice to get n-patches support, mpeg decoding and some of the other stuff that hasn't been hacked yet. From the looks of it anti-aliasing isn't covered... and no EBMB extension either. well at least my radeon 9100 is sitting in a machine that only handles data traffic...

[bridgman]

Among the areas that were missing from the specifications were Hyper Z programming information, Truform, DX8 multisampling/FSAA support, vertex shader, ATI_fragment_shader, TV-Out, MPEG decoder, BIOS table layouts for the external TMDS transmitter, errata and chip bugs, and finally performance counters and debug information.

That sounds about right

I'm pretty sure MPEG will get covered, since most of those blocks haven't changed much and we do plan to enable some basic video acceleration. I'll know more about the rest after we get a bit further into the 3d doc effort for 5xx and 6xx.

[Thetargos]

IMO what has been the Open Source drivers killer is not lack of documentation, but third parties. For instance, what happened with VIA for the longest time, they had an Open Source driver for their Unichrome chips, but they lacked the MPEG-2 decoder (pipe, or whatever is called) code in them, then recently I was surprised that on my Fedora 7 install there was an update with this driver which actually in the package update information was stated that it was the first one to support the MPEG-2 totally open. Hopefully AMD will do the same and have the AVIVO bits released too.

[halfmanhalfamazing]

Any word on older families of chips?

R3xx and 4xx to my knowledge, OSS developers don't have any documentation on these at all.

There's r2xx, which is limited and under NDA, will these get full release?

And even older ATI chips, I'm sure there's still plenty of them out there. R1xx's, and Rages and so on. Is the entire portfolio eventually slated for full release?(for driver purposes obviously)

[greg]

(Looks at his R200 class 3D accelerator in his P4M 1.7 GHz laptop that PASTES his more modern Athlon64 3200+ driven Xpress200m accelerator with most games that don't rely on OpenGL 2.X GLSL code...)

Uh, you mean the X200M is faster?

[bridgman]

>>Any word on older families of chips? R3xx and 4xx to my knowledge, OSS developers don't have any documentation on these at all. There's r2xx, which is limited and under NDA, will these get full release?

I'm trying to re-release the original 1xx/2xx documentation without NDA. There will definitely be more info on 3xx/4xx as well, and some of *that* information will also fill in gaps on 1xx/2xx. In general, for the older ASICs which already have pretty good open source drivers we will focus on filling in gaps and helping with problems rather than creating all new documentation packages.

>>And even older ATI chips, I'm sure there's still plenty of them out there. R1xx's, and Rages and so on. Is the entire portfolio eventually slated for full release?(for driver purposes obviously)

I hadn't really thought about Rage parts -- honestly we weren't planning on going back *that* far. We can probably try to answer developer questions on those but I don't think we would do any kind of documentation package.

[Thetargos]

@Bridgman

Thank you very much for having the patience to answer our questions regarding AMD's specs release. It is indeed VERY nice to have that kind of feedback on the matter, from first hand experience and from one of AMD employees. Thank you.

[seeker010]

thanks bridgman for giving us the spiel. hopefully in the not too distant future we can expect ATI to become the preferred gpu vendor for those running linux

[halfmanhalfamazing]

>>Any word on older families of chips? R3xx and 4xx to my knowledge, OSS developers don't have any documentation on these at all. There's r2xx, which is limited and under NDA, will these get full release?

I'm trying to re-release the original 1xx/2xx documentation without NDA. There will definitely be more info on 3xx/4xx as well, and some of *that* information will also fill in gaps on 1xx/2xx. In general, for the older ASICs which already have pretty good open source drivers we will focus on filling in gaps and helping with problems rather than creating all new documentation packages.

Thanks for that. I kinda figured this was in the cards.

>>And even older ATI chips, I'm sure there's still plenty of them out there. R1xx's, and Rages and so on. Is the entire portfolio eventually slated for full release?(for driver purposes obviously)

I hadn't really thought about Rage parts -- honestly we weren't planning on going back *that* far. We can probably try to answer developer questions on those but I don't think we would do any kind of documentation package.

It's not that big of a deal to me personally, my card is r2xx class, I just ask as well as say what comes to mind, usually. And I hadn't seen anybody else ask it.... anywhere. This documentation release(s) is a big deal however, and thanks for your hard work, contributions, and all of that. :-) Thanks to AMD/ATI as well.

As others have been saying, it's nice to get feedback right from the source.

[olivermcfadden]

... We can probably try to answer developer questions on those but I don't think we would do any kind of documentation package.

I would certainly like to see a mailing list or similar setup so that us
developers could ask AMD/ATI engineers questions, where needed. Even if it's
private or read-only to the public.

I am more interested in R300 and upwards than the older hardware, though.

Anyway, thanks for the documentation released so far, and I'm looking forward to
the upcoming documentation and eventually 3D documentation. :-)