The X.Org Plans In Ubuntu 11.04, Again

[HokTar]

The question is, why not to stick?

Seriously, even if the r600g driver had achieved total feature and speed parity with the r600c, it would still lack an important advantage: Wider testing for a longer period of time.

Why throw away this advantage, if the gallium version doesn't offer anything significant over the classic version at the moment? Just to be cool and have the latest? This is the mark of the stupid, or the mazochist(<troll>in case of Fedora users</troll>).

I believe Ubuntu 11.10 will and Mesa of that time frame will use r600g by default, by that time it will be much more tested and feature complete, and it might provide advantages over the classic driver. Until then, sticking with the classic is sane.

This is a valid point.
However, I never preferred r600g just because it's newer. My post was maybe a bit inadequately worded in the sense that r600g is only about 3 months old but it is basically already on feature parity with classic, does not crash often (haven't in the last week or so) for me(tm).
So assuming that the speed of development won't slow down significantly it very well might bring obvious advantages (speed/opengl, etc) by next April.
Plus some guys are working on a video decoding state tracker for gallium/r600g.

Point is, that I thought they completely ruled out the possibility to default to r600g. However, as bridgman pointed out this is not the case, so *I take my word back*.

[bridgman]

For what it's worth, we are planning to only implement acceleration support for Northern Islands and future GPU cores in the r600g code base. This is obviously only a minor point since the community could backport support to the classic driver, but I believe the development community is strongly committed to the r600g code base.

[TemplarGR]

This is a valid point.
However, I never preferred r600g just because it's newer. My post was maybe a bit inadequately worded in the sense that r600g is only about 3 months old but it is basically already on feature parity with classic, does not crash often (haven't in the last week or so) for me(tm).
So assuming that the speed of development won't slow down significantly it very well might bring obvious advantages (speed/opengl, etc) by next April.
Plus some guys are working on a video decoding state tracker for gallium/r600g.

Point is, that I thought they completely ruled out the possibility to default to r600g. However, as bridgman pointed out this is not the case, so *I take my word back*.

The problem is, this work won't be complete by the time Ubuntu 11.04 development freezes.

Another thing to consider is that just because r600g advanced rapidly for a time this pace is not guaranteed, especially since from what i understand, it was mostly a port of existing code in r600c to gallium. Of course i have not studied gallium3d details so i may be wrong on this. Bridgman could clarify

[AdamW]

"While we may see Canonical use Wayland beginning with Ubuntu 11.10"

This is...ahem...unlikely.

See Mark's latest blog comment: "It’s highly unlikely that the default Ubuntu install in a year will be on Wayland. It’s possible that there will be versions of Ubuntu that use it, or proof-of-concept images, by then. More importantly, we won’t make it the default until it really is widely supported and supportable, by folks using a wide variety of hardware providers. And as for network connectivity, there’s sufficient time between now and then for these problems to get solved."

[nanonyme]

I guess I should mention that agd5f is at LPC too. In a perfect world I would be there too, and you wouldn't be hearing from me either

This begs the question, why can't you get to LPC?

[bridgman]

The problem is, this work won't be complete by the time Ubuntu 11.04 development freezes.

Another thing to consider is that just because r600g advanced rapidly for a time this pace is not guaranteed, especially since from what i understand, it was mostly a port of existing code in r600c to gallium. Of course i have not studied gallium3d details so i may be wrong on this. Bridgman could clarify

I thought the Ubuntu summary (will follow what upstream decides at the time) was the best anyone can do right now. I don't think anyone knows for sure what will happen over the next few months, although from the limited user testing so far the r600g driver does seem to be approaching the point where it could become the default. The chicken-and-egg problem is that most user testing comes as a consequence of including the driver in a distro, so "getting close and providing a fallback to 600c" will probably end up being the strategy of choice.

It seems likely that the recent level of development focus will drop back to roughly what was seen on r600c before the r600g push... the goal of the recent flurry of work was to get 600g to the point where it made sense for developers to work there instead of in the 600c code, and to the point where r600g was close enough to become the default after some testing and fixing.

My impression from random user comments is that both of those milestones have been reached, but that is hardly a definitive statement

[Pickup]

I'm quite tech-savvy but maybe not enough to get the point of this, I read different things every time I try to understand, so I ask this very stupid question once and for all:

Is (or will be) Wayland a replacement for X or not?

[Silverthorn]

I can't see the point of rushing towards r600g as default unless mesa somehow started supporting something more exciting than OpenGL 2.1.

[nanonyme]

I can't see the point of rushing towards r600g as default unless mesa somehow started supporting something more exciting than OpenGL 2.1.

The thing is, Gallium is about more things than just OpenGL. Not that the state trackers for other cool stuff like video acceleration are ready yet but having masses use Gallium instead of classic drivers pushes testers towards them which means that state tracker developers get tested drivers to test their implementations of extensions on. Hence yes, there is a point of rushing towards r600g as default eventually. Ubuntu seems to be rather reasonable in the matter though that they're following what upstream does instead of starting soloing around.

[Silverthorn]

The thing is, Gallium is about more things than just OpenGL. Not that the state trackers for other cool stuff like video acceleration are ready yet but having masses use Gallium instead of classic drivers pushes testers towards them which means that state tracker developers get tested drivers to test their implementations of extensions on. Hence yes, there is a point of rushing towards r600g as default eventually. Ubuntu seems to be rather reasonable in the matter though that they're following what upstream does instead of starting soloing around.

Yes, Gallium is versatile but it still lacks in usability for the end user. If the average end user can't decode video with it or play their shiny new OpenGl 3/4 game then we can't really expect them to report relevant bugs either. On the other hand, an average phoronix addict probably would consider filing bug reports and are probably already testing all the drivers available to them. They would probably not even shy away from using git themselves just satisfy some inner craving for the latest and greatest.