AMD Publishes Initial Open-Source Driver Code For Next-Gen Polaris

[hoohoo]

Great news and good work. AMD is looking more and more like my choice of GPU in the fall.

[tomtomme]

FWIW, we've had pre-launch support for every family of APUs I think.

off topic: are you holding onto the "X.Org ATI Driver Developer" because of nostalgia? :P

[tomtomme]

We're excited inside AMD as well -- we've always had public launch-time support for APUs but this is the first time for dGPU.

so definitely a Polaris for me in September. Thank you AMD. I hope Freesync will make it until then. Any chance Zen might be out the door until the 20. september too? I really need a new machine then...

[tomtomme]

You AMD guys are the worst, I want to buy a Polaris card now to show support for all this great work, but you keep making my 290 faster every release making me content with what I have!

and you even get vulkan, bah. AMD are such badasses...

[nanonyme]

It seems that airlied is still not happy with it though, he replied to alex's post about an hour ago: https://lists.freedesktop.org/archiv...ch/103468.html =(

Although he did suggest it might be accepted in staging as long as there is agreement that DAL will be evolved to meet the kernel's requirements.

I'm not familiar with how distributions deal with staging; are distributions known for enabling staging stuff on release kernels? for instance, can we expect Arch and Ubuntu to enable DAL by default if it makes it into staging, or will we have to compile our own kernels regardless?

It depends case by case. I think Fedora at one point had some staging drivers enabled but not all obviously. Anyway, it's starting to feel to me that there's a high risk of delays with this DAL strategy. In theory they could let it mature out of tree like DRM used to be but there's maintenance concerns wrt supporting changing kernel API's that route

[juno]

The initial Baffin PCI IDs include 0x67E0, 0x67E1, 0x67E8, 0x67E9, 0x7EB, and 0x7FF. The Ellesmere IDs are 0x67C0 and 0x67DF.

https://www.phoronix.com/scan.php?pa...ourced-Polaris

Those don't make sense. Where do they come from, anyway?
0x67e0/0x67e1 could be one chip, 0x67e8/0x67e9 another (XT/PRO variants). It used to be like this for some years. 0x07eb/0x07ff are like out of range.

[CatMerc]

Yep, but the hybrid driver should allow day-one support since it can run ahead of upstream.

So let me get it straight, the fully open source stack will not be ready for Polaris launch (most likely), but the hybrid stack using AMDGPU + GPU-PRO will have support?

[CatMerc]

Edit: Never mind, thought Phoronix ate my previous post.

[FLHerne]

However the upstreaming process works out, I'm impressed by the time and resources AMD have clearly put into this. Finally we have day-one open-source support for a proper GPU!

For the first couple of years of having my HD7770, I promised myself to go Nvidia next time - the choice between unstable, closed-source FGLRX and unstable, slow RadeonSI was useless. Now you've got a stable, fast, open-source driver without any of the tradeoffs.

One more happy customer. :-)

[ernstp]

There were still a lot of questions why the DAL was necessary at all, why things couldn't be fixed in core drm for all drivers. For example "4K @ 60"... ?