AMD Posts Initial AMDGPU Hybrid Driver With Vulkan Support

[duby229]

Well, I modified the file tree to match what Gentoo has and put in symlinks to my tree in /opt/amdgpu-thingy and I only managed to get opencl to respond, reporting only the graphics card and not the cpu. Couldn't get X to start with amdgpu on a 4.5.0 kernel and the linux-firmware. I then removed everything and broke my install. It's taken me about 3-4 hours to get X back working again with Catalyst, don't ask me why, but it just wouldn't start X with that either.

Gonna wait for official Gentoo ebuilds.

I gave up on getting it to work yesterday. And I'm not willing to install Canonical's mess to play around with it. So I'm just kind of waiting for now.

[Emanem]

Yes... three closed-source components running on top of slightly tweaked copies of open source kernel driver, X driver, libdrm and multimedia drivers.

(also sent you a pvt message - maybe it's more interesting for the tech audience so posted here too)

Hi bridgman,

I'm an enthusiast Linux gamer - I'm also into software development, especially C/C++ and interested in Linux kernel.
First let me say "Thank you AMD!". Open source is the way forward, and you folks are way better than nVidia!

Now, I understand that the open and closed drivers are like shared objects which are loaded at runtime and use a common interface.
Is my understanding correct? If yes, can you please point me to any github repo where I can find that interface (and also the main entry point if possible)?
Can you point me to some low level technical docs (not like the slides from XDC 2015 - those are too high level)?

I currently own a nVidia 980 GTX, but would like to switch over to AMD over later this year or the next. I understand a lot of focus is on Vulkan, but what about OpenGL and supporting existing (Steam) games?
Are you going to improve the OpenGL driver (at least the closed source) for a while? Or are your resources mainly focused on the Vulkan implementation?

Thanks again for all your hard work, keep on!

Cheers

[lockheed02]

bridgman & team: Thank you folks!

Lockheed

[ermo]

I think we understand the desire well enough that more voices of support isn't going to make any real difference (I've never believed they make much difference at the best of times personally). Not sure if we'll be able to put enough time into it to do it all ourselves but I expect that between us and community it should be able to happen. We're trying to formalize some of that now.

Fair enough.

Since I don't plan on buying any new gfx cards before I've seen the performance of Polaris vs. Pascal, is there any avenue where one could help fund the man-hours needed to include GCN 1.0 support? I'm certainly willing to put my money where my mouth is on this one.

In fact, I could see a lot of people shelling out $10-$50 to increase the lifespan of their gfx card investment, but I could be wrong of course. =)

[fuzz]

Fair enough.

Since I don't plan on buying any new gfx cards before I've seen the performance of Polaris vs. Pascal, is there any avenue where one could help fund the man-hours needed to include GCN 1.0 support? I'm certainly willing to put my money where my mouth is on this one.

In fact, I could see a lot of people shelling out $10-$50 to increase the lifespan of their gfx card investment, but I could be wrong of course. =)

I would also like to help. Money, hours, testing. I have a 390x and am even interested in getting back into C development...

[Berniyh]

Is it really that important? I don't think it's worth it and I own a GCN 1.0 card for my gaming PC, which also use for Linux gaming.

[fuzz]

Is it really that important? I don't think it's worth it and I own a GCN 1.0 card for my gaming PC, which also use for Linux gaming.

I personally didn't mean to say I would only like to help for GCN 1.0... as I don't even have GCN 1.0 :P

[vkrastev]

I'm thinking the initial support will be some combination of us helping with initial code and debugging into the closed components if needed, community devs/users help with testing & bug fixing so we don't have to go find a big pile of QA and dev resources we don't have today. Something like that.

Ideally in a usable state before the current round of 14.04.x HWE distros that Catalyst supports hit EOL in August, but that's tight.

We are trying to put a more detailed plan together now - I was hoping to finish this week but the hybrid driver launch interfered with that a bit. That's why I've been saying "ask again in a week".

Any news about your SI-support roadmap?

[calexil]

bridgman So I was certainly not going to wait all year for my old GCN1.0 card to be supported, so I sold it and bought an r9 380x. installed kernel 4.2, and the amdgpu-pro driver and... holy shit this thing is fast, I am getting 20-30% benchmark gains over crimson. One huge problem though... and this is a deal breaker, no amount of compositing can fix the wild tearing I have going on, :scrolling webpages, watching youtube/mp4/etc, non vsync games... all have insane tearing and since the driver is so new... there is no documentation on how to remedy it. My os is Linux Mint 17.3(Trusty based) kernel 4.2.0-35, latest amdgpu-pro beta.

Also as a little fyi, ocl-icd-libopencl1:i386/amd64 from the wine 1.6-7 dep list has a conflict that causes a driver install failure. This is easily remedied by removing those two packages, but I think wine 1.6 is incompatible with amdgpu-pro

[haagch]

Have you tried tearfree?

https://wiki.archlinux.org/index.php/ATI#Driver_options

tl;dr:
/etc/X11/xorg.conf.d/20-radeon.conf

Section "Device"
Identifier "Radeon"
Driver "amdgpu"
Option "AccelMethod" "glamor"
Option "DRI" "3"
Option "TearFree" "on"
EndSection