AMD Puts Out Radeon HD 6000 Open-Source Driver

An anonymous reader writes "AMD has just released their open-source driver for the Radeon HD 6000 series graphics cards (sans the Cayman GPUs) with KMS, 2D, and 3D acceleration."

Re:VAAPI Acceleration?

[Evanisincontrol]

Phoronix released a followup article [phoronix.com] today. A driver supports X video acceleration, but no VAAPI yet. Like you said, these cards seem to be aimed at gamers, but with the prevalence of HD video content these days, I'd be surprised if VAAPI wasn't a high priority for this driver.

Re:VAAPI Acceleration?

[CynicTheHedgehog]

I think the issue with the open source radeon drivers isn't that the developers don't know how to implement stuff, or that it's a closed API or whatever. The problem is one of manpower. Reading the Phoronix forums it looks like there's maybe 2 guys that do the majority of the commits (for the really hard stuff anyways) and they just have too many things to do. I recall one post explaining that some kind of compiler was needed, and so one of the developers stubbed out a very simple one to be able to move forward on something else, with the intention of coming back and replacing the stubbed out compiler later, when the more critical issues were addressed. The compiler was a key component to all sorts of 3D stuff and was a critical factor in performance, but functionality and performance take a back seat to compatibility and stability.

My hope is that with the release of these open source drivers a lot of that boilerplate stuff will come with it, so that the community can truly focus on implementing newer APIs and such, although I don't know enough about video driver development to know whether that's necessarily the case.

Patent-covered algorithms?

[fph il quozientatore]

I thought the biggest stopper in building OSS 3D drivers were that they usually contain technologies covered by software patents. What happened to the patented parts? Have they been stripped out of this OSS version, with the effect that it is now slower than its closed-source counterpart? Or did they find another way out?

Re:Patent-covered algorithms?

[Kjella]

The closed source drivers remain closed for many reasons, but I think licensed code is a much bigger part of that than patents. The open source drivers are built separately from scratch and are in general much slower yes. They have estimated - and don't take this as an official AMD statement but guesswork from the people working on it - that the open source driver could reach 60-70% of the closed driver on average using the simple architecture they've chosen. Simply because the closed source driver has a ton of code paths and optimizations for various situations, the OSS team is much, much smaller than the closed source team and can't possibly replicate that anyway.

Re:Put your money where your mouth is

[Carewolf]

NVidia's linux drivers have been incredibly buggy for years now. I recently changed the open source nouveau drivers which easily solved all performance problems I had. NVidia is both poorly implemented, has poor support for moderen X11 extensions (XRandr), the recent 26x series also has VERY serious memory-leaks which slowly brings your computer to a crawl.

Btw. I am not stating this as a user, I am stating this as a develop who has been working on getting KDE to work better on NVidia hardware.