An Open-Source Mobile Linux Graphics Stack?

Printable View

08-20-2011, 04:20 PM
phoronix

An Open-Source Mobile Linux Graphics Stack?

Phoronix: An Open-Source Mobile Linux Graphics Stack?

The mobile device landscape, particularly for those devices running Linux, is quickly evolving. Just in the past few days, Google bought Motorola, Qualcomm open-sourced the remainder of their Gobi API for controlling modems, and HP ended off all their webOS devices, among other changes. But will the future mobile Linux device landscape deal with more open-source drivers, particularly when it comes to graphics?..

http://www.phoronix.com/vr.php?view=OTgyMA
08-20-2011, 06:22 PM
smitty3268

Android x86

I was surprised you didn't mention that Mesa is currently adding Android x86 support.
08-20-2011, 08:54 PM
bridgman

Quote:

Originally Posted by smitty3268

I was surprised you didn't mention that Mesa is currently adding Android x86 support.

Yep. Mesa has been running on Android for a while but you have to build it in a non-standard way -- IIRC a lot of the current work is adding Android build support and build targets to the native Mesa build framework. That's what it looks like anyways ;)
08-21-2011, 03:25 AM
89c51

Quote:

Originally Posted by bridgman

Yep. Mesa has been running on Android for a while but you have to build it in a non-standard way -- IIRC a lot of the current work is adding Android build support and build targets to the native Mesa build framework. That's what it looks like anyways ;)

so is there a chance that android will use a full open stack in the future?? (linux--> G3D --> maybe wayland -->toolkit)
08-21-2011, 09:08 AM
bridgman

I would say "yes with the usual caveats" - if there is an open stack for hardware on regular Linux you'll see a corresponding open stack on Android for sure. On the other hand if there is no open stack today I wouldn't expect everyone to open up just for Android.

The harder question is what will go into high volume production, depending on things like video decode (it's a lot easier to "do it in software" on a quad core 2.9 ghz desktop than on a dual core 800 mhz tablet) and 3D performance (will mobile hardware vendors end up getting picked based on benchmarks or will there be a "fast enough" mindset ?).
08-21-2011, 10:14 AM
V!NCENT

Quote:

Originally Posted by bridgman

will mobile hardware vendors end up getting picked based on benchmarks or will there be a "fast enough" mindset ?

I think battery life is a big factor. Decode on a chip or in software... I think it's not going to happen soon, especialy because that iPad 2 gets their ten hours of battery life because of some special dedicated ARM modification sauce (offload common OS tasks with dedicated CPU circuitry), as rumored.
08-22-2011, 10:18 AM
patrik

What about forcing openness?

If one would reverse engineer an embedded GPU and write an open driver (I know it's not a trivial thing), then how badly sued could you get?

From what I've heard, the IP holders don't wanna release specs because they fear getting sued for patent infringement by another IP holder. But on the other hand they most likely all infringe on each other. So a matter of gently forcing one of them open would perhaps change the game entirely.
08-26-2011, 03:29 AM
Yfrwlf

Quote:

Originally Posted by patrik

If one would reverse engineer an embedded GPU and write an open driver (I know it's not a trivial thing), then how badly sued could you get?

From what I've heard, the IP holders don't wanna release specs because they fear getting sued for patent infringement by another IP holder. But on the other hand they most likely all infringe on each other. So a matter of gently forcing one of them open would perhaps change the game entirely.

Who cares? If a company was willing to sue a user of their product for improving interoperability with it, they should be outed anyway. Hopefully the damage would be substantial and shine light on how ridiculous patents are.