RadeonHD 1.2.5 Driver Released

[rbmorse]

It's always been a good question. And, it's always had the same bad answer.

Living with/working around developer ego is the price of doing business in the open source development world.

There's probably a good case study for some MBA candidate hiding in all this, assuming the numbers were available.

[bridgman]

Before everyone gets too happy, I should mention that you *will* probably have to download and build new drm code to support 6xx/7xx 3d acceleration code.

[ben0]

Do the x-video changes now make radeonhd tear-free as well, or is radeon still the only driver to support it?

[bridgman]

I think radeonhd also has tear-free Xv these days but am not 100% sure.

[mattst88]

The two drivers are still fairly different in terms of display/modesetting code and overall structure. The 2D/video acceleration code is somewhat different for earlier GPUs but pretty close for 6xx/7xx.

.... *Dodge*

[bridgman]

These aren't the droids you're looking for...

[MaestroMaus]

These aren't the droids you're looking for...

Should have been: I find your lack of fate disturbing...

[kensai]

Nice, since AMD is concentrating on their open source effort, at this pace we will have good open source drivers from AMS in 5 years, in the best case scenario. So in 5 years I can choose between bad catalyst 14.4 and good open source drivers.

[bridgman]

Should have been: I find your lack of fate disturbing...

Sorry, bad joke. I was trying to "use a Jedi mind trick" to avoid getting dragged into yet another discussion about why we have two active drivers

http://upload.wikimedia.org/wikipedi...nmindtrick.JPG

The serious answer (which I have given before) is that the two development groups have discussed a number of approaches for getting to a single code base and the conclusion was that we should keep both drivers going for a while, then look at moving to a single code base once kernel modesetting was in use on all the enterprise distros. That may be as simple as a "use KMS for modesetting, use Gallium3D for acceleration" driver but nobody knows for sure right now.

Top priority for most of the distro developers is making sure that their enterprise customers are well supported. Those distros will be running pre-KMS kernels for quite a few years, and so there is a real business need to keep the current drivers going forward for a while longer.

At some point the developers will decide to converge on a single code base and we'll need to find another "competition" to watch instead; in the meantime the cost of supporting two code bases isn't very high. Most of the new work is being done in mesa and drm, and each of those only has a single tree anyways.

[yangman]

FWIW, nobody would have noticed or even cared that the LUT has 10-bit precision were it not for the initially broken implementation in radeonhd. Because of that and one unsatisfied hardware calibrator user, we now have a better implementation in both drivers. I'd say that's a win.

I think we all agree that diversity and choice are important to a healthy OSS culture. Just because it hasn't been the norm doesn't mean the same principals can't apply to drivers as well.