AMDGPU DC Gets More Raven Ridge Improvements, Audio Fixes

[phoronix]

Phoronix: AMDGPU DC Gets More Raven Ridge Improvements, Audio Fixes

Harry Wentland of AMD has sent out the latest batch of patches for the AMDGPU DC display code stack. Fortunately it lightens up the DRM driver by about six thousand lines thanks to removing some unused code...

http://www.phoronix.com/scan.php?page=news_item&px=AMDGPU-DC-More-Raven-Fixes

[Brisse]

Wonder if the HDMI/DP audio issues I had on my R9 Fury / MG279Q is fixed. I could only get 48khz, not 44.1khz.

[R41N3R]

Always nice to see all the improvements. On my RX480 amdgpu DC runs stable, not in any way worse than without DC. The only issue I have is that Plasma-Wayland freezes with a black screen as soon as the display turns off and I want to return to my session, not sure yet if this is a driver bug or not.

[aufkrawall]

Unfortuantely, with my RX 560 and my Qnix DL-DVI display without internal scaler, amdgpu DC only gives me a black screen. Also specifying a custom edid doesn't help.

[Space Heater]

I'm a bit out of the loop on AMD GPU patches, is there a precedent of these fixes being backported to previous kernels? In other words, will people have to wait until 4.17 for fixes like:

drm/amd/display: fix backlight not off at resume from S4

If so, waiting ~4 months seems a bit too long.

[Brisse]

Always nice to see all the improvements. On my RX480 amdgpu DC runs stable, not in any way worse than without DC. The only issue I have is that Plasma-Wayland freezes with a black screen as soon as the display turns off and I want to return to my session, not sure yet if this is a driver bug or not.

I've had similar troubles with GNOME when trying out AMDGPU DC. Crashes and/or random bad behaviour when trying to log back in from the lock screen.

[schmidtbag]

6000 unused lines of code? That makes me uncomfortable about the lack of attention or optimization in the rest of the code... Hopefully they can keep up with the cleanups.

[Tomin]

I'm a bit out of the loop on AMD GPU patches, is there a precedent of these fixes being backported to previous kernels? In other words, will people have to wait until 4.17 for fixes like:

If so, waiting ~4 months seems a bit too long.

I think the policy is that Greg may merge fixes from mainline to stable branches. So it might be less than that. Also the distribution of your choice may include those fixes even earlier.

[HeadsUpHigh]

6000 unused lines of code? That makes me uncomfortable about the lack of attention or optimization in the rest of the code... Hopefully they can keep up with the cleanups.

It doesn't work like that...

[cb88]

It doesn't work like that...

A lot of the code is generated straight from the same high level code they generate the chips from, so... if they aren't using some certain aspects of the chip or its just for debug during initial bring up, its just unused basically documentation code that the compiler will just eliminate out anyway.