RADV Is Now Considered A Vulkan Conformant Driver

[schmidtbag]

Talking about about AMD's promised open sourcing of it's proprietary Vulkan driver? The -only- single effort they made was an empty promise years ago. And I'm sorry, but that's not an effort at all.

I really don't think AMD as a company or their developers are to blame here, but rather the lawyers who are just taking their sweet time making sure everything is ok to open source. The main fault AMD had was not open-sourcing it all in the first place - shortly after releasing the drivers, they said they were going to open them, so that wasted a lot of time.

I also think RADV might have slowed things down dramatically. If it didn't exist, maybe AMD's drivers would've been available a long time ago. But since RADV was making so much progress in so little time, they were probably like "meh, we have bigger things to worry about".

[duby229]

AMD has delivered on its previous promises of open sourcing parts of their stack, and that for almost a decade now.
So it is pretty sure they will open source it if they said so.
In that field things are complex technically and legally. So give them time.

Yeah, I get it, they have to leave it up to lawyers to get it to a position where programmers can release code. I'm fully aware. But that process has been anything else but transparent. There is literally zero information about it. None at all. No evidence what-so-ever.

[Ziple]

Talking about about AMD's promised open sourcing of it's proprietary Vulkan driver? The -only- single effort they made was an empty promise years ago. And I'm sorry, but that's not an effort at all.

AMD has delivered on its previous promises of open sourcing parts of their stack, and that for almost a decade now.
So it is pretty sure they will open source it if they said so.
In that field things are complex technically and legally. So give them time.

[duby229]

Until RADV consistently matches or exceeds the performance of AMDGPU-PRO, I don't think AMD should be so quick to ditch their open-source efforts. But, I do agree that contributing toward RADV would probably be the best route.

Talking about about AMD's promised open sourcing of it's proprietary Vulkan driver? The -only- single effort they made was an empty promise years ago. And I'm sorry, but that's not an effort at all.

[schmidtbag]

Until RADV consistently matches or exceeds the performance of AMDGPU-PRO, I don't think AMD should be so quick to ditch their open-source efforts. But, I do agree that contributing toward RADV would probably be the best route.

[Twysock]

Soon ®

But when will then be now?!

[M@GOid]

...so AMD's official open source Vulkan driver is still coming?

Soon ®

[airlied]

Do does that mean vkcube runs without crashing now? I never managed to get that demo to run on RADV with either my R9 270 or even RX Vega56.

The simple vkcube demo from krh? Does it pass validation? I'm pretty sure it is just a broken app.

[LeJimster]

Do does that mean vkcube runs without crashing now? I never managed to get that demo to run on RADV with either my R9 270 or even RX Vega56.

[frosth]

wow, it's incredible thanks you both guys.

btw. look at Kronos "conformant-products" especially for "AMD Radeon R9 285".

https://www.khronos.org/conformance/adopters/conformant-products

in some tests it is GCN 1.2 family but in the others it's GCN 1.3