Radeon's Open-Source Linux GPU Driver Has Nearly Caught Up With Windows' Driver

[cj.wijtmans]

I have both an r9 390x and a 7970. I just tested Superposition 1080p extreme for the 7970 with this results:
ArchLinux mesa-git: 1735
Windows OpenGL: 1643
Windows DX11: 2100
This is amazing!
I would like to see a mesa control panel as well, or cli tool at least. Something official with all the features needed like tessellation amount, AO, AA and AF override etc.

go ahead and start a QT repo.

[gurv]

Everything that goes into the staging trees is reviewed and approved first, generally via the amd-gfx list, so if lack of review was your main concern you have that proper, reviewed and approved support today. Is the problem just that we are pushing the trees to agd5f's personal repo on FDO (the usual convention for open source developing) rather than something on amd.com ?

Agree that you shouldn't have to continuously build your own OS, and again the usual practice is that third parties (distros or individuals) do that building for you and maintain package repos you can use. I'll check in the morning (going zzz now) but I believe user mystro256 here is maintaining a repo of staging tree builds, as an example.

You may also be able to pick up the kernel driver portion of the AMDGPU-PRO driver packages today, of course, and we are planning to extend the AMDGPU-PRO install model to simplify updating the all-open stack as well. That said, all the time we spend on interim measures is time we are not spending on getting the display code upstreamed, so there is a bit of a tradeoff there.

Personally, I'm not complaining having to add some ppa or compile stuff myself as a stopgap measure while we wait for all this work to be upstreamed.
However it could be more straightforward: I recently compiled amd-staging 4.11 kernel to be ready to test a Freesync monitor I'm probably gonna buy in the next couple of months.

But I only discovered right here on this thread that it's not enough and that I need a patched ddx driver too! WTF?!
How come that is not upstreamed? Don't tell me you have some refactoring to do there too?

And where is this patched source code anyway? Do I have to hunt it down on some mailing list, grab the upstream source and cross fingers that it applies properly?!
Come on, this is not serious

[darkbasic]

You need a patch for the ddx to detect fullscreen 3D apps and call the freesync ioctl to enable it. It's not upstream yet since freesync is not upstreamable as is.

Will Freesync 2 ever by upstreamable?

[darkbasic]

They do today AFAIK... there are few exceptions but the list is shrinking fast. Is there something specific you have in mind ?

Last time I tried it Serious Sam 3 was still unplayable on anything but Nvidia.

[Sidicas]

That's an unfair assessment. If I people play a game on windows, and it runs in DirectX. then it should be benchmarked in DirectX. I know 0 people who play Shadows of Mordor on windows with opengl. But I know plenty of people who play shadows of mordor on windows with directx.

I agree, but app developers have got to be at least partially to blame.

I found this tidbit in a discussion about why the pcsxe2 OpenGL Renderer is so much slower than the pcsxe2 DirectX renderer..

--------------


The difference is easy. Dx is finished. Opengl is on-going but I'm working on another interesting project so it won't be finished soon. The others diff are
1/ opengl work on linux. It took me less than 1 hour to enable it on windows. So why not
2/ I could (potentially) fix some rendering issue. For example, shadow on Persona work without alpha stencil hack.
3/ I might able in the future (ie when I have time to code) to use advance GPU feature without dx12 and new MS OS

Conclusion just use Dx

From: http://forums.pcsx2.net/Thread-what-...9364#pid359364

[agd5f]

But I only discovered right here on this thread that it's not enough and that I need a patched ddx driver too! WTF?!
How come that is not upstreamed? Don't tell me you have some refactoring to do there too?

The upstream community wants a full vendor neutral solution rather than a vendor specific one.

[gurv]

The upstream community wants a full vendor neutral solution rather than a vendor specific one.

Argh ok I understand this one. I guess it's because Intel wants to support adaptive-sync at some point.
I don't think anyone is naive enough to think NVidia will adopt a standard anytime soon

And are there any plans toward that vendor neutral solution or has it not even started yet?
I get the feeling buying a Freesync monitor for gaming on Linux is a very risky move currently with the dc / ddx upstreaming.

[agd5f]

Argh ok I understand this one. I guess it's because Intel wants to support adaptive-sync at some point.
I don't think anyone is naive enough to think NVidia will adopt a standard anytime soon

And are there any plans toward that vendor neutral solution or has it not even started yet?
I get the feeling buying a Freesync monitor for gaming on Linux is a very risky move currently with the dc / ddx upstreaming.

We are thinking about it. We implemented a vendor specific solution so that we had a something to test with while we work out what the new common interfaces and use cases should look like. FWIW, this is vendor specific on other OSes as well at this point.

[nanonyme]

AFAIK we own it but it contains trade secrets primarily related to the Windows market. We have discussed opening it up several times but the conclusion each time was "no".

We have already moved other drivers onto the LLVM-based direct-to-ISA compiler logic we will use for the open sourced Vulkan driver, including OpenCL.

Right, I've just been multiple times bumping into claims on how the LLVM shader compiler isn't performant enough, hence was wondering

[gurv]

We are thinking about it. We implemented a vendor specific solution so that we had a something to test with while we work out what the new common interfaces and use cases should look like. FWIW, this is vendor specific on other OSes as well at this point.

Thanks for this information. Btw, even if I sometimes sound pissed, I really do appreciate the work you and everybody at AMD make
It's just that now that AMDGPU/RadeonSI is in good shape, it's only natural to want it to go even higher

In other news, I've succeeded compiling a git xf86-video-amdgpu with the Freesync patches from august 2016.
Do you think it should allow Freesync to work if I use it with amd-staging 4.11 kernel or are these patches obsolete by now?