17:35 <CounterPillow> Alyssa isn't here but I guess asking the current maintainers of Panfrost would be more appropriate:

17:35 <CounterPillow> I pointed out that I don't believe random Debian users using PanVK and then reporting bugs for it would help the situation any (in fact I think having bug reports for a known broken driver would be less-than-useless).

17:35 <CounterPillow> So someone at Debian saw https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24873#note_2057934 "Bifrost-era PanVK is not being used." and thought they could help by enabling PanVK in the Debian mesa package. It's still gated behind the quite clearly named env variable thankfully.

17:35 <CounterPillow> Are you for/against/indifferent to Debian shipping Bifrost-era PanVK?

17:44 <CounterPillow> Since this decision seems to have been based on cherry-picking something Alyssa has said though, I am of course also very eager to receive her input on the situation, though judging by https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24943 I can guess she wouldn't want anyone using panvk :)

17:50 <daniels> CounterPillow: nope, it definitely shouldn't be shipped to users in its current state

17:51 <daniels> it's got developer attention and plans to pick it up from the initial work and shape it into something which _should_ be shipped to users and which can be supported long-term

18:03 <CounterPillow> Thanks, I let Debian know that they should disable it again.

18:22 <daniels> thankyou!

20:29 <alyssa> "So someone at Debian saw https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24873#note_2057934 "Bifrost-era PanVK is not being used." and thought they could help by enabling PanVK in the Debian mesa package"

20:30 <alyssa> That is an... odd interpretation of that comment

20:30 <alyssa> given it's immediately followed by not being developed + not being used

20:31 <alyssa> I don't get the general interest in bifrost vk, I suspect people have misconceptions about what VK support would actually entail

20:31 <alyssa> (It would not mean DXVK or VKD3D works. It would not allow higher GL versions via Zink. It would not likely improve performance on dual GL/VK workloads.)

20:32 <alyssa> That doesn't mean it's /useless/ ... if there's an app that's VK only but doesn't use any spicy features, then maybe panvk is enough

20:33 <alyssa> but in the (GNU/)Linux space ... I genuinely don't know of a single application that would benefit from the VK support given the expected limitations on the bifrost hw

20:33 <alyssa> At any rate, I don't have a horse in the race anymore

20:33 <alyssa> If people want to develop the driver further, cool. good luck to them.

20:34 <alyssa> In its current state at any rate, packaging it in distros would absolutely be a mistake

20:34 <alyssa> and again. even when bifrost panvk is 'finished'. I still don't know what Debian users would be doing with it.

20:40 <daniels> yes, it's pretty clear that packaging the current bifrost panvk for debian wouldn't benefit ~anyone

20:42 <alyssa> Wheeeee

20:43 <alyssa> 20:33 < alyssa> but in the (GNU/)Linux space

20:43 <alyssa> I should further qualify that as "stuff a community Debian user would be likely to run"

20:43 <alyssa> (in 2023, at least)

20:44 <CounterPillow> Thankfully I caught this while it was still just in experimental. I'm sure it was done with good intentions but yeah.

20:44 <alyssa> im just glad my life is free of bifrost again

20:44 <alyssa> midgard or valhall, none of this in between stuff :P

22:41 <diederik> Just as clarification: The "random Debian user" that requested the enablement of panvk is a DD. I know they contribute to MNT Reform but I don't know what specifically they have in mind for it

22:42 <diederik> Then there was another person who +1-ed the request and that person is also a DD and a mesa dev

22:44 <diederik> As apposed to CounterPillow, I do NOT think they will flood the panvk devs with useless bug reports

22:46 <diederik> And I shared that deliberately in a channel where I was reasonably certain that it only or mostly consisted of skilled users/devs

22:48 <diederik> Had it been proposed while a Freeze for the next Stable was imminent, I'd have at least made a point that it was likely not wise. But the next Freeze is ~15 months out.

22:49 <diederik> And when that point gets near and it is still considered broken, the argument can still be made to disable it so it won't end up in a Stable release

23:22 <anholt_> it's not useful to users, and user reports would not be useful to us. so it shouldn't have been shipped.

23:46 <daniels> it's fine, everyone did things with good intentions. it's just not going to help anyone to ship it now; when it is helpful, we'll let downstreams know

23:49 <diederik> I do not know why it was requested and +1-ed by DDs, so I can't comment on that. My problem was with characterizing them as (clueless/dump) random users. That's all I wanted to correct.

23:52 <diederik> o/

23:56 <robmur01> eh, "developer" can be a pretty broad brush - for instance I'm pretty intimate with the much of the panfrost kernel driver, but I am absolutely a clueless random user as far as the userspace side goes :D (even though I fixed that one thing in Mesa that one time...)

23:58 <robmur01> I think it's safe to assume no offence was meant