Mesa Devs Discuss Potentially Dropping Non-Gallium Drivers Or Forking Code For Gallium

[9Strike]

It's a complicated situation. Just droping i965 isn't a solution, but I think it doesn't need to be that performant or super actively developed, since it only applies to old Intel iGPUs. So either really create a non-performant Gallium replacement (still unrealistic) or just seperate i965 as an own driver and remove in that driver all the Gallium stuff.

[entropy]

i965 are 2014 macbooks. Still used in production environments (I know because I'm in this case). Nowadays the hardware renewal is 5+ years.
And as a matter of fact a 10 year old desktop with an r600 GPU is still my main home dev machine... Hardware simply ages well these days.

I'm also using a Thinkpad Haswell notebook of that time which serves me astonishingly well.
I'd be very disappointed if it wouldn't work well on newer distributions anymore and I have to stick to some LTS one with an old Mesa version.

[khnazile]

They should drop any hardware older than 3-4 years, not just those with non-gallium drivers.

[ihatemichael]

Dropping working drivers for old hardware that still works is fucking moronic.

What the hell is wrong with you people, this is not Windows.

[geearf]

I think the title is not as clear as it could be, as they are not planning on dropping all non-gallium drivers, but only the classic ones. For instance radv and anv are non-gallium drivers too and I'm sure won't be dropped.

They should drop any hardware older than 3-4 years, not just those with non-gallium drivers.

I hope you never get in charge of any driver I depend on.

[M@GOid]

Dropping working drivers for old hardware that still works is fucking moronic.

What the hell is wrong with you people, this is not Windows.

I believe you misunderstood the proposal. These non-Gallium drivers will not be dropped and disappear, but will live on maintenance mode on a different branch of Mesa. They will be available on your preferred distro, just like today.

[Space Heater]

Dropping working drivers for old hardware that still works is fucking moronic.

What the hell is wrong with you people, this is not Windows.

The actual mailing list post is talking about "dropping" them from mainline mesa, and putting them in a separate branch. Given these devices don't really receive useful updates, it's not crazy to put them in a legacy/stable branch, to avoid breakage from improving drivers for newer hardware. So even if this change were to be implemented, it's not as if your device would stop working, I'd imagine distributions would auto-detect which mesa branch your hardware needs.

[baka0815]

Right, and I'm sure that people use the graphics silicon on those CPUs.

To watch YouTube videos.

Which, let's admit, isn't exactly the pinnacle of performance-centric use-cases.

If it's become a maintenance burden, it probably needs to be split off.

Well, I'm using my Haswell in my HTPC and it serves me well to watch HD videos and Prime. I hope that they will support that APU for some time and don't just let it die.
How they're doing it is not mine to decide.

[starshipeleven]

Look up the hardware surveys. 7-9yr/old CPUs are fairly common.

We are talking of GPUs here

[jrch2k8]

Look up the hardware surveys. 7-9yr/old CPUs are fairly common.

The point is no matter what happens with gallium3D this old drivers won't pickup new features because most cases the silicon is not even there hence the bigger probability is getting breaks and bugs while impeding performance optimizations on actual modern hardware, is a loss/loss situation.

What Marek is asking is simply this:

1.) Mesa LTS with only(or mostly) legacy drivers in maintenance mode(which i think those people using old hardware will prefer instead of the constant breaks they have now)
2.) Mesa for modern hardware with Nir and Gallium3D.

also remember thanks to libglvnd you can install both and runtime will load the correct version for your hardware, hence for the regular users is transparent as long as your distro include both packages in some form or shape