AMDGPU DC Code Lands For Linux 4.15 Kernel

[mibo]

So, where is the Linus rant that we were promised?

[Adarion]

Yes, what did Linus say?

That was good news this morning when I went to help out at a local Linux Presentation Day. It gave me energy!

[YamashitaRen]

CIK and Si support will not polish unless you are using it and making bug reports. I did use the amdgpu driver with A8-7600 years.

Unfortunately there don't seem to be powerplay SI support in amd-staging-drm-next either...

[M@GOid]

Linus silence is disconcerting. I hope he did not merge the thing while being threatened. Where was Bridgman and his boys at the time of the merge?

[Krejzi]

Linus silence is disconcerting. I hope he did not merge the thing while being threatened. Where was Bridgman and his boys at the time of the merge?

Looks like a hostage situation or gun point, no?

Joking aside, Linus and AMD devs had a discussion on the main pull request about pre-generated headers which would be the DC's greatest annoyance, since most of the abstraction layer is being thrown out and shared linux code is being reused.

[ElectricPrism]

It's not in perfect shape but it's the perfect time to merge.

Wairing longer wold stifle development since distros are already shipping amd dc.

[frosth]

Sure, but I didn't want to use the out of tree DAL/DC stack and I didn't need this feature. I think now it is stable enough, so I did the switch.

Oh, i belive it took 2 years not because DAL/DC is not stable enough, but (this was what I read here about) problem was about code condition. Look at post from @dwanger, this issue is not happened in amdgpu-pro-dkms kernel driver. And this problem appear because they had to change their code because of drm people didn't allow it to merge this "big monster" . And don't get me wrong, the best thing is get clean code and merge somthing fit to whole kernel style. But now we start testing DC. This is why polaris/VI are off by default, because DC is not perfect for older asics and they don't want listen constant complainning. They have to put VEGA on just because we need it for display.

PS. if you have HDMI+DVI monitors, using DC code is extremaly danger (don't know about 2xHDMI) after HDMI monitor turn off and again turn on, you lose all displays, and only reset helps. Even simply dpms or powersave (autooff) in TV/monitor settings trigger that.
I am really happy with DC in 4.15 but I wasn't call it stable by now and will still use amdgpu-pro-dkms driver + 4.9LTS for HDMI audio.

[Brophen]

Does anyone else smell a Steam Machines comeback?

Would be nice. I think Valve needs to just make their own, "The Steam Machine", powered by Raven Ridge so it's inexpensive. Then testing should be done against that particular machine to make sure all the games available play nice with it. If you're going for the Console crowd (I assume they wouldn't be going for the PC crowd?) it needs to cost like a console costs and "just work" too.

Here's to hoping

[oscar]

Ok so can anyone point me what currently nightly kernel branch to use for Vega GPU in Ubuntu from http://kernel.ubuntu.com/~kernel-ppa/mainline/ :

http://kernel.ubuntu.com/~kernel-ppa/mainline/daily

http://kernel.ubuntu.com/~kernel-ppa/mainline/drm-tip/?

http://kernel.ubuntu.com/~kernel-ppa...ine/drm-next/? (seems 3 days old)
kernel.ubuntu.com/~kernel-ppa/mainline/drm-intel-next/?
thanks..

[R41N3R]

Oh, i belive it took 2 years not because DAL/DC is not stable enough, but (this was what I read here about) problem was about code condition. Look at post from @dwanger, this issue is not happened in amdgpu-pro-dkms kernel driver. And this problem appear because they had to change their code because of drm people didn't allow it to merge this "big monster" . And don't get me wrong, the best thing is get clean code and merge somthing fit to whole kernel style. But now we start testing DC. This is why polaris/VI are off by default, because DC is not perfect for older asics and they don't want listen constant complainning. They have to put VEGA on just because we need it for display.

PS. if you have HDMI+DVI monitors, using DC code is extremaly danger (don't know about 2xHDMI) after HDMI monitor turn off and again turn on, you lose all displays, and only reset helps. Even simply dpms or powersave (autooff) in TV/monitor settings trigger that.
I am really happy with DC in 4.15 but I wasn't call it stable by now and will still use amdgpu-pro-dkms driver + 4.9LTS for HDMI audio.

Do you have a bug report for this? Or is it the same as the issue of dwagner? I had this with and without DC (especially on Plasma-Wayland), so DC didn't change anything for me.