2013: A Good Year For Open-Source AMD?

[bridgman]

Hows that possible? Canadian types are the most agreeable people on the planet. I could spit in your face and then you would certainly thank me for the human contact.

I'm half German and half Irish. With a bit of effort, I can disagree with anyone on the planet

[TobiSGD]

Don't buy high-end hardware?

"I've said this a few times over the years so hopefully this won't come as a surprise, but I wouldn't be buying high end cards yet if I was only planning on using them with the open source drivers, even if I could magically run the Intel open source stack on AMD or NVidia hardware."

This must be a joke. I can't think anything else.
Of course I understand that with limited developing capacities you can't have any feature of new graphics cards at release time, at least currently. May be better in the future.
But this comment is a slap in the face of anyone that owns, like me, a laptop with the 2008 released HD3200. I am still not able to use the open source drivers on this four years old hardware, which is neither high-end nor nor a recently released GPU, just because it overheats my machine and sucks out the battery in no-time. I don't care about performance or feature completeness, as long as something that is such basic for a mobile machine as power-consumption has issues that are known for a long time and are still not resolved.
Basically this means to me: If you want to buy AMD and use open-source you have to find the few GPUs that are working correctly. That is a serious show-stopper.

[89c51]

This must be a joke. I can't think anything else.
Of course I understand that with limited developing capacities you can't have any feature of new graphics cards at release time, at least currently. May be better in the future.
But this comment is a slap in the face of anyone that owns, like me, a laptop with the 2008 released HD3200. I am still not able to use the open source drivers on this four years old hardware, which is neither high-end nor nor a recently released GPU, just because it overheats my machine and sucks out the battery in no-time. I don't care about performance or feature completeness, as long as something that is such basic for a mobile machine as power-consumption has issues that are known for a long time and are still not resolved.
Basically this means to me: If you want to buy AMD and use open-source you have to find the few GPUs that are working correctly. That is a serious show-stopper.

I will remind here that AMD never ever promised that will write the open-source driver by itself. They just provide docs. Its cards are meant to be used with Catalyst to fully use them. If catalyst doesn't work feel free to holler at them or buy an nvidia. But nagging about open-souce support is useless.

[Linuxhippy]

Gallium3D

bridgman: If the 2D DDX based on Gallium3D works well, are there plans to backport it to VLIW4 (Trinity) based architectures?
Do you plan to take advantage of Trinitys UMA based nature?

Also, could you please elabortate a bit what the current state of 2D acceleration of Catalyst is?
There have been some articles on phoronix about some magical switch to turn on a new acceleration architecture, but beside some disappointing benchmark results and rendering corruptions, nothing new has been posted here on phoronix.

[bridgman]

I am still not able to use the open source drivers on this four years old hardware, which is neither high-end nor nor a recently released GPU, just because it overheats my machine and sucks out the battery in no-time. I don't care about performance or feature completeness, as long as something that is such basic for a mobile machine as power-consumption has issues that are known for a long time and are still not resolved.

Are you sure you are enabling power profiles ? Which profile are you using -- either mid or low should be OK on most HD3200 systems.

http://www.x.org/wiki/RadeonFeature#KMS_Power_Management_Options

[TobiSGD]

Are you sure you are enabling power profiles ? Which profile are you using -- either mid or low should be OK on most HD3200 systems.

http://www.x.org/wiki/RadeonFeature#...gement_Options

I have tried it with the both profiles, but that doesn't help a single bit, the machine still is hot and power-consumption is far to high for mobile use. I wouldn't care, since it works fine with Catalyst, but AFAIK the support from Catalyst for this card will be dropped in the near future, basically rendering this (2 1/2 years old) machine useless.

[bridgman]

bridgman: If the 2D DDX based on Gallium3D works well, are there plans to backport it to VLIW4 (Trinity) based architectures?

I guess it depends on how well it works

Seriously, other than maybe some bug fixing I don't *think* there should be any backporting required. We haven't really discussed it internally but I expect we would want it to work on Trinity as well.

Do you plan to take advantage of Trinitys UMA based nature?

Yes and no. Yes in the sense that there is probably the opportunity to eliminate some copies by taking advantage of UMA.

No in the sense that (a) I don't think we have done enough experimentation to have a good feeling for added costs of cache flushing and remapping vs saved costs of copying and (b) I don't know if there is any unreleased info required to take advantage of it so (c) we don't have a priority or plan for it yet.

I think it's fair to say that at minimum we would do some testing, figure out if any additional programming info is required, and try to get that info released. If the initial testing goes well we would probably try to push out finished code.

Note that one of the devs may already be on top of this; it just hasn't bubbled up near the top of the priority stack yet.

Also, could you please elabortate a bit what the current state of 2D acceleration of Catalyst is? There have been some articles on phoronix about some magical switch to turn on a new acceleration architecture, but beside some disappointing benchmark results and rendering corruptions, nothing new has been posted here on phoronix.

That's a harder question. I haven't had time to really talk to the Catalyst devs at all in the last few months so most of what I know is what I read online. What I do know (or at least think I know ) is :

- the acceleration related to the magical switch (AAA, sometimes referred to (incorrectly) as Direct2D because some of the low level routines were shared with Direct2D) has been enabled by default for a while, maybe a year or more

- some additional accel functions have been added for browser accel, but I don't know any details

- the variability in results seems to be related to conflicts between triple-buffering (Tear-free desktop) code added to the Catalyst driver and similar functionality in Compiz and other compositors -- too many things were turned on by default so the result was really slow

I don't know current state, ie whether compositors or Catalyst have changed defaults or whether tweaking is still required in some cases to get everything playing nice together and not being too conservative. I have seen some "ohmigod it just works" posts but not a huge number so guessing that tweaking is still required at least with some common combinations.

[LuckyCyborg]

@bridgman

What happened in MUX A+A PowerExpress support beyond Linux Catalyst 11.8?

Is there any chance that that PowerExpress support once again become functional soon?

I have a laptop where sits unused one HD5650 video card, and I am therefore obliged to use the on-board HD4200. Not really funny because that HD5650 wasn't gratis...

[TobiSGD]

I will remind here that AMD never ever promised that will write the open-source driver by itself.

I never said that. I replied to a statement made from one of the developers of the radeon drivers. The reply would be the same if it were not one of the developers paid by AMD.

Its cards are meant to be used with Catalyst to fully use them.

Again, I don't talk about feature completeness, I talk about a very basic function for a mobile device, that is not working as intended which makes the driver useless to me. By the way, Catalyst will drop support for that GPU soon, AFAIK.

If catalyst doesn't work feel free to holler at them or buy an nvidia.

I already complained about the inability of the Catalyst developers to implement a simple patch into the installer to make Catalyst work on kernels >3.2.8 on 32 bit systems in a different thread, so no need for that here.

I knew that there were major issues with ATI/AMD drivers some years ago, but from what I experienced and have read in forums the driver quality seemed to be much better in the near past. Now I begin really to regret that I a) recently have replaced my outdated Nvidia with an AMD card and b) that I recommended to buy AMD to other Linux users.
For me and my machines it seems that the drivers (both radeon and Catalyst) are down a step (or two) in comparison to Intel and Nvidia.

[bridgman]

I have tried it with the both profiles, but that doesn't help a single bit, the machine still is hot and power-consumption is far to high for mobile use. I wouldn't care, since it works fine with Catalyst, but AFAIK the support from Catalyst for this card will be dropped in the near future, basically rendering this (2 1/2 years old) machine useless.

Hmm, that doesn't sound right. There should be significant differences between profiles.

There are some systems where the vendors only added a single power setting in the VBIOS (which means more complex code is needed in the kernel driver) but I didn't think we had those with HD3200. Might ask you to file a bugzilla ticket with a VBIOS dump, will see what agd5f suggests.

Power management doesn't seem to be one of those areas that community developers work in for fun (as libv pointed out recently) so we might have to go back and give that code another kick.