David Airlie Tackling RADV Vulkan Conformance

[pal666]

One may hope that RADV will eventually be faster, as Mesa OpenGL is significantly faster.

radeonsi has reasons to be faster than blob, radv doesn't have. radeonsi is faster now, radv isn't. and amd vulkan will not even be a blob, it could be improved just like radv

[airlied]

I didn't quite understand (either): You seem to be talking about your WIP branch, and the must-pass tests, when you say you pass 54%. So that sounds like you are only half-conformant (unless 100% refers to something else than the must-pass list).

The mustpass list means no Fails. not 100% coverage. NotSupported/Warning aren't considered fails.

Dave.

[duby229]

radv does not come for free. amd will not stop developing amdgpu-pro because of radv, intel will not stop developing anv because of radv. but airlied stopped developing mesa because of radv. so you have duplicate driver and lost man-year of improvements of something which people really need

Huh? Can you try to explain this a little bit more detail? I don't understand. I may be missing information about this. Dave has a very full plate and it is damn impressive what he is able to consume off it. I certainly don't believe at all that there are lost man-years, that's got to be total nonsense considering how much work airlied put out during the course of his career.

[pal666]

number of vulkan apps in existance using tessellation? 0.

btw, number of linux games which need vulkan? 0

[pal666]

Huh? Can you try to explain this a little bit more detail? I don't understand. I may be missing information about this. Dave has a very full plate and it is damn impressive what he is able to consume off it. I certainly don't believe at all that there are lost man-years, that's got to be total nonsense considering how much work airlied put out during the course of his career.

you seem to have lost context. he put out much work, but if he put it in non-duplicate driver it would be even better

[schmidtbag]

Huh? Can you try to explain this a little bit more detail? I don't understand. I may be missing information about this. Dave has a very full plate and it is damn impressive what he is able to consume off it. I certainly don't believe at all that there are lost man-years, that's got to be total nonsense considering how much work airlied put out during the course of his career.

I'm guessing his point is there's no guarantee Dave will keep up with RADV, but, it is AMD's obligation to maintain their drivers. Considering he stopped his mesa work for an effectively redundant driver (as of today anyway), that's a "deficit" to the community.

[duby229]

I'm guessing his point is there's no guarantee Dave will keep up with RADV, but, it is AMD's obligation to maintain their drivers. Considering he stopped his mesa work for an effectively redundant driver (as of today anyway), that's a "deficit" to the community.

It's not redundant, there is no other open source AMD Vulkan driver. And Because of Dave there now is one. That's definitely not a deficit.

[duby229]

you seem to have lost context. he put out much work, but if he put it in non-duplicate driver it would be even better

What duplicate? Where are there -any- other open source AMD Vulkan drivers?

[airlied]

you seem to have lost context. he put out much work, but if he put it in non-duplicate driver it would be even better

radeon/r600 duplicated fglrx, virgl pretty much duplicated vmware gfx, I spent a lot of time duplicating closed source things, until AMD open source their vulkan driver, this will be no different. Once they actually open source, then we can address the duplication situation, until then there isn't really any duplication.

[schmidtbag]

It's not redundant, there is no other open source AMD Vulkan driver. And Because of Dave there now is one. That's definitely not a deficit.

I said "effectively redundant". It isn't literally redundant for the reasons you specified, but for 99% of users who will not compile or modify the code themselves, it's ultimate functionality is redundant compared to the closed drivers. If/when AMD does open their current drivers (they claim they will), then RADV could be wholly redundant.

Considering other things that need attention (particularly things Dave used to work on and/or has the skill to work on), I would consider redundancy to any degree a deficit.