Testing The Open-Source "RADV" Radeon Vulkan Driver vs. AMDGPU-PRO

[Hi-Angel]

The answer you are looking for is in the post you already quoted: The legal review has not allowed AMD to open it up yet.

So… Do you know what do they review? Why is it taking so long? Why is this happening, given we're talking about a completely new driver which definitely was written with open sourcing it mind? Also, bridgeman mentioned some rewrites of the driver: again, what do they rewrite, they just created it!

That post tells nothing at all.

[bridgman]

so it is one developer for 8 years or half of developer for 16 years

Yep, or 2000 developers for one day, or 1/100th of a developer for 800 years, but if any of those were the plan I would have said that. Instead I said 8 devs for one year, because that's what the estimate was.

[adler187]

Sorry, I probably missed that, can anyone tell, or link a discussion, about why Vulkan driver wasn't open sourced yet?

The answer you are looking for is in the post you already quoted: The legal review has not allowed AMD to open it up yet.

[darkbasic]

bridgman already told you are off by 8 man-years

I told that it was easier to write a new Vulkan driver from scratch (using the existing OpenGL infrastructure of course) than opening the blob: thanks to RADVD this is under everyone's eyes. So what are you trying to say? I really don't understand your point.

[pal666]

Doesn't matter, my point still stands: the issue is the legal review.

bridgman already told you are off by 8 man-years

[pal666]

Most recent estimate was 8 developers for a year to open it up

so it is one developer for 8 years or half of developer for 16 years

[bridgman]

i said that if there is less than one developer is available to open closed vulkan, then it is less than one developer available to join radv

Yeah, that's why the question marks.

Why would you think "less than one developer" ? Most recent estimate was 8 developers for a year to open it up, although (a) we're looking for ways to reduce or at least break up the work, and (b) some of that work has been done now.

[xxmitsu]

airlied, BNieuwenhuizen I've understood what you are using NIR for inter-stage optimization in your RADV vulkan implementantion. Would it be possible to also use NIR when OpenGL API is used ? Somethink like NIR->LLVM ?

Would that provide speed benefits or at least some of the code reuse the one from intel's ?

[Hi-Angel]

Right... we knew at the start that any IP/Legal review would fail dramatically (and it did)... the work involved rewriting big chunks of the driver, and the driver itself is already much larger because it shares code across multiple OSes and multiple APIs.

Sorry, I probably missed that, can anyone tell, or link a discussion, about why Vulkan driver wasn't open sourced yet?

[darkbasic]

it is recompiled windows driver

Doesn't matter, my point still stands: the issue is the legal review.