More AMDGPU DC Patches Posted As It Looks Unlikely It Will Land For Linux 4.13

[phoronix]

Phoronix: More AMDGPU DC Patches Posted As It Looks Unlikely It Will Land For Linux 4.13

Another set of 27 patches were posted today for AMDGPU's DC (formerly DAL) display stack, but it's looking increasingly unlikely that this code will be merged for Linux 4.13...

http://www.phoronix.com/scan.php?page=news_item&px=AMDGPU-DC-More-June-Work

[tildearrow]

while there won't be any Radeon RX Vega display support until DC lands,

Can I ask?
I've always heard that DC is needed for Vega, but without any explanation why. What does DC really have that is required for Vega output?

[Michael]

Can I ask?
I've always heard that DC is needed for Vega, but without any explanation why. What does DC really have that is required for Vega output?

The Vega support code is only wired up to made us of the DC display stack and doesn't have support for the 'legacy' (current) display code, just how they brought up Vega on the AMDGPU stack, for conserving resources.

[karolherbst]

Can I ask?
I've always heard that DC is needed for Vega, but without any explanation why. What does DC really have that is required for Vega output?

it's purely political. And saying it is _required_ basically is the same as taking the user hostage and saying "either you accept that or no features", which is an asshole move in my opinion. You just don't say stuff like this. There is always another way.

EDIT: not saying there are no good technically reasons for this, it's just the way this is handled is what I dislike about it.

[smitty3268]

it's purely political.

I don't think that's fair. It's a matter of manpower. They've chosen to concentrate on getting things to work with their new API, and backporting support to the old one would take time.

I agree that it's definitely something AMD *can* do, though, which is probably what you were trying to get at.

[microcode]

it's purely political. And saying it is _required_ basically is the same as taking the user hostage and saying "either you accept that or no features", which is an asshole move in my opinion. You just don't say stuff like this. There is always another way.

EDIT: not saying there are no good technically reasons for this, it's just the way this is handled is what I dislike about it.

I mean, they have failed if the goal was to take the users hostage. Nobody is clambering for the DRM maintainer to accept the DC patches, they know their best interests are at heart. For AMD, I think it is (as Michael said) a matter of resources. Display hardware is hideously complex these days, and a lot of the complexity is shifted out into drivers. If you look at how huge the DC code is, that's how much they'd also have to write to have it in non-DC AMDGPU.

[Mystro256]

it's purely political. And saying it is _required_ basically is the same as taking the user hostage and saying "either you accept that or no features", which is an asshole move in my opinion. You just don't say stuff like this. There is always another way.

I wouldn't say it's political, but rather, AFAIK, it was the original plan to have DAL in the kernel by now, in order to unify different OS driver development. Vega isn't exactly a walk in the park and I don't think this was re-scoped to have a contingency plan for Vega, like the dual path with the older cards. As Alex said, it would be more work to port Vega to legacy, than to get DAL (or rather DC) to an upstream-able state.

Considering the amount of projects being pushed out by AMD this year, I would say it's more of an unexpected bandwidth issue. Your use of "hostage" seems like a bit of a leap. Yes Vega could work with the legacy method, but AFAIK it would be a huge undertaking.

[zanny]

Is there an amdgpu-dc repo for Arch anyone knows of? I'd love to just try it, but the Arch wiki doesn't have one listed.

[starshipeleven]

Nobody is clambering for the DRM maintainer to accept the DC patches, they know their best interests are at heart.

Minor nitpick: Most people doesn't yet know. When their new stuff will fail to work with Ubuntu they will rage and flame in forums.

Developers in linux kernel won't see any of that.

[karolherbst]

I don't think that's fair. It's a matter of manpower. They've chosen to concentrate on getting things to work with their new API, and backporting support to the old one would take time.

I agree that it's definitely something AMD *can* do, though, which is probably what you were trying to get at.

yeah okay, but you could express such plans before already and see how the community thinks about your plans. You don't need to write tons of patches to figure out later there is a lot of controvery.

I wouldn't say it's political, but rather, AFAIK, it was the original plan to have DAL in the kernel by now, in order to unify different OS driver development. Vega isn't exactly a walk in the park and I don't think this was re-scoped to have a contingency plan for Vega, like the dual path with the older cards. As Alex said, it would be more work to port Vega to legacy, than to get DAL (or rather DC) to an upstream-able state.

Considering the amount of projects being pushed out by AMD this year, I would say it's more of an unexpected bandwidth issue. Your use of "hostage" seems like a bit of a leap. Yes Vega could work with the legacy method, but AFAIK it would be a huge undertaking.

okay true, I was over exaggerate here. Maybe they were simply optimistic and didn't think it would be such a big deal and now they kind of try to deal with the situation. It just that is has a bit of an odd touch and I don't really know how all this will work out in the end.