hybrid-graphics-linux team mailing list archive

[Bryce Harrington <bryce@xxxxxxxxxxxxx>]

On Mon, Jun 11, 2012 at 05:30:39PM -0400, Eric Appleman wrote:
> Per my understanding the decision of X server 1.12 vs 1.13 is undecided.
> Assuming 1.13 could ship (meaning AMD does a special Catalyst like they
> always do), DRI2 offloading without dynamic switching could be supported
> for Intel UXA + Nouveau setups.
> 
> In terms of exposure to the user, David Airlied put forth the following
> comment [1][2].
> 
> "the first implementation will just involve setting an env var, like
> DRI_PRIME=1, then I'm hopefully going to add a gnome-shell extension to
> make the same possible from a launcher or some sort.
> 
> After that people can do what they like ;-)"
> 
> X server 1.14 should have dynamic switching and my assumption is that
> 13.04 will ship with it, but the opportunity exists to have some level
> of hybrid graphics support for 12.10.
> 
> However, the plans to ship Intel SNA kinda interferes with this since
> Prime is UXA-only at the moment [3]. I have no idea if SNA Prime is even
> feasible for 13.04.
>
> If SNA is not deemed ready for 12.10 due to bugs or a perceived
> immaturity, UXA Prime will ship (assuming 1.13).

Thanks for the head's up on UXA Prime.

Offhand, both SNA and Prime sound like they're promising higher
performance for users.  SNA presumably would benefit a wider scope of
users since it applies to all hardware, however Prime would enable
hardware that currently is bordering on unusable, even if that
represents a smaller scope in the market.  Conversely, the regression
risk on SNA is a lot higher, so I still think it's the most fruitful
thing to test near-term.

But, a large part of why we're going to try enabling SNA now is to help
bring issues - like this one - to light.  I'm hoping if we do end up not
going forward with SNA this cycle, that at least it'll give us some
tangible feedback for upstream, so we can have a better chance in the
future.

Bryce