RADV: A Community Open-Source Effort To Get Vulkan Working On Radeon

[boxie]

Yeah, there were some raised eyebrows about the "no communication" part - Alex, Jammy and I have all talked about it.

Maybe approach michael and ask to publish an "AMD State of Play" article that outlines what has been done, what is currently in the works and what the future plans (as much as you are allowed to say), that would give people a "source of truth" article

[boxie]

Does it need a reason?

Nope - only "This seems like an interesting itch to scratch"!

Go get em tiger!

[juno]

It's a bit sad... I'd rather have a solution for the DAL dilemma. Adaptive Sync and even HDMI/DP audio is not usable with newer GPUs. I also expected the transition of Vulkan/OpenCL go faster so we could at least see something at this point. I considered switching Hawaii for Ellesmere b/c of efficiency, more VRAM and better software support but at this point it is 'only' good, but not good enough to justify this 'sidegrade'. Maybe have to wait and hope for Vega or just having too high expectations...
I mean, the launch was great. Way better support than before and everything. Great. But Linux users are still 2nd class customers...

Anyway, to get back to topic: thanks airlied for moving something. It is higly appreciated, even when considered useless work by some. If AMDs Vulkan driver is in a similar stage as DAL is, there is a long way to go...

[schmidtbag]

The GPUs need to be fast enough that games are regularly CPU-bound, for it to matter. Seems reasonable to implement Vulkan on recent-gen GPUs in my mind.I also suspect it was just what was in airlied's workstation. He doesn't work for you, I don't see why he should strive so much to make his work initially useful for you.

I agree, but you're missing the key point here - the GPUs being targeted at first already have support. I personally don't care about nouveau. When I use nvidia, I use the closed source drivers. When I use Intel, I don't play games and OpenGL is more than fast enough for my needs. I have no problem using AMDGPU-PRO either. So this isn't about me, this is about benefiting the community as a whole in the best way possible. Somehow, people think I'm the asshole here because I'm sticking up for people who don't have a Vulkan option. Not saying you specifically, just in general. And yes, I get it, Dave isn't obligated to do anything I say. I never said he was. I'm just giving a suggestion that would be of a greater impact.
EDIT:
I'm aware of Dave's hardware situation. So, I don't really have much of an interest in pushing this further. I appreciate his efforts regardless and I don't mean to be discouraging by any means.

Anyway, to get back to topic: thanks airlied for moving something. It is higly appreciated, even when considered useless work by some. If AMDs Vulkan driver is in a similar stage as DAL is, there is a long way to go...

Assuming you're referring to me, I never stated the efforts were useless. I said they're a poor priority, because there are Vulkan compatible GPUs out there that need more attention.

[ElectricPrism]

I'm glad airlied is working on this.

I look forward to buying a AMD card soon to take advantage of the open source stability benefits for Workstation & Gaming.

Keep it up dude, I can't wait to see it come downstream.

Too bad this sold out I might have jumped ship today:

http://www.overclock.net/t/1606325/newegg-us-sapphire-nitro-radeon-r9-fury-tri-x-oc-549-99-250-00-299-99

[ElectricPrism]

I'm glad airlied is working on this.

I look forward to buying a AMD card soon to take advantage of the open source stability benefits for Workstation & Gaming.

Keep it up dude, I can't wait to see it come downstream.

Too bad this sold out I might have jumped ship today:
http://www.overclock.net/t/1606325/n...-250-00-299-99

[bridgman]

I agree, but you're missing the key point here - the GPUs being targeted at first already have support.

The dominant point I saw in Dave's response was that he felt it made more sense to do the initial implementation on hardware he understood... and given how much fiddly HW-layer work a Vulkan driver involves (it's pretty much all HW Layer) it's hard to argue with that.

[Veerappan]

I agree, but you're missing the key point here - the GPUs being targeted at first already have support. I personally don't care about nouveau. When I use nvidia, I use the closed source drivers. When I use Intel, I don't play games and OpenGL is more than fast enough for my needs. I have no problem using AMDGPU-PRO either. So this isn't about me, this is about benefiting the community as a whole in the best way possible. Somehow, people think I'm the asshole here because I'm sticking up for people who don't have a Vulkan option. Not saying you specifically, just in general. And yes, I get it, Dave isn't obligated to do anything I say. I never said he was. I'm just giving a suggestion that would be of a greater impact.
EDIT:
I'm aware of Dave's hardware situation. So, I don't really have much of an interest in pushing this further. I appreciate his efforts regardless and I don't mean to be discouraging by any means.

Assuming you're referring to me, I never stated the efforts were useless. I said they're a poor priority, because there are Vulkan compatible GPUs out there that need more attention.

Vulkan support is currently from the closed-source driver, which means that even if I had a VI card, I wouldn't be using it. I'm not against closed-source/proprietary drivers, but I honestly have had a better experience with radeonsi than I had with fglrx previously.

I've got an SI (7850/Pitcairn), and I have absolutely no issues with Dave/Bas focusing on VI cards for now. The AMDGPU kernel driver will eventually probably support SI in non-experimental form, which would mean that they only have to target one winsys instead of two. Patience will probably get us cleaner code in the long-run. Would I like Vulkan support now, sure... but there aren't any games that I want to play on Linux that only support Vulkan... They all still have a GL renderer that I can use for now. We've waited for a while, a little longer won't kill us.

And for now, I'll probably be poking at what needs to be done to try to get the SI/AMDGPU experimental kernel driver going.... (once the TdF is done and I get my desktop back from streaming duties)

[smitty3268]

Doesn't Baz work for AMD? He's already making some commits so it looks like AMD is interested in this code.

Anyway, awesome work Dave!

[bridgman]

Doesn't Baz work for AMD?

Not as far as I know (and I'm sure someone would have told me). Believe Dave and Bas have been working together on the project.