ATI Committed To Fixing Its OSS Problems

Sits writes "Chris Blizzard blogged from the Red Hat summit that an ATI marketing spokesman said, from the stage, that ATI knows it has a problem with open source and is committed to fixing it. Does this mean ATI will finally resolve alleged agpgart misappropriation, and fast track the release of open source 2D drivers on its latest cards while releasing specifications for its mid-range cards? Or is ATI only concerned with fixes to its binary driver to maintain feature parity with competitors?"

Open Source supporters within ATI

[plcurechax]

I know from talking to them at the Ottawa Linux Symposium a couple of years ago that the technical people within ATI were keen to support Linux the best that the could, but said they were mainly limited by management / legal to aim for competing with whatever nVidia offered the Linux community. If nVidia offered a complete open source driver, they would be pressured to do the same.

In other news

[diegocgteleline.es]

Announcing free software drivers for the new Intel 965GM Express Chipset [marc.info]

ATI, NVIDIA: fuck you. Open source graphic drivers are possible, period.

Re:Two words

[Anonymous Coward]

few more words:

Omega Drivers = Windows != Linux = OSS

Re:Does this mean hardware hacking is dead?

[TheRaven64]

Apparently (rumour alert), the newer Radeons are very similar, in terms of interface, to earlier ones. Porting the drivers over requires about a 100-line diff; most of the changes are register locations, the actual semantics are similar. Unfortunately, the person who wrote the diff did so under NDA, and ATi didn't allow him to release it.

Re:Likely binary drivers only.

[MightyPez]

Right, because no other vendor has ever been accused of that! [extremetech.com]

Open? Does not play with DRM, so forget it.

[Bearhouse]

Ever heard of, "Certified Output Protection Protocol (COPP), Protected Video Path Output Protection Management (PVP-OPM),
Protected Video Path User Accessible Bus (PVP-UAB) and Protected Broadcast Driver Architecture (PBDA..."

All lovely things that Microsoft and ATI (will/do) use to piss you off, and make connecting all of your expensive new PC & AV kit virtually impossible.

Better binary drivers? Maybe.

Genuinely 'open' architecture that would enable the OSS community to bypass (more easily) current and future DRM, while still being able to view the result on the lastest hardware? No way.

Re:Ok I am stupid ...

[644bd346996]

Short answer: no.

Long answer: No. X11+GLX is very different from GDI+DirectX. In almost all cases, it would be easier to reverse-engineer the hardware, rather than wrap the driver api. Also, it would probably be impossible to use windows graphics drivers in a secure manner. And the extra translation layer would kill performance. If you are going to reverse-engineer the drivers, you might as well look at the hardware info, and not the software api.

Note that in some cases, it is possible to use Windows drivers on a *nix operating system. The NDIS network card driver api is well documented, and is supported by projects for Linux and FreeBSD.

Re:Marketing?

[Svartalf]

Oh, never question that the developers know that there's a serious problem. Trust me on that one. It's just I seriously doubt that middle and upper management have a single clue as to how to honestly fix it.

Re:Does this mean hardware hacking is dead?

[Svartalf]

In reality, they're not. The API driving them IS. They're just multiprocessor stream engines these days. Pretty much little more than a 4, 8, 16, etc. way SMP machine for all intents and purposes. I should know, I worked on driver work for one of those "complex" beasties- the driver's feature set is what is really complicated and much of that has been done by the DRI crowd and just needing refining.

Re:Does this mean hardware hacking is dead?

[crush]

Fuck "hacking the hardware" and "whining about it". The "Open Source Community", specifically Dave Airlie, have already written Free drivers [livejournal.com] for the ATI cards, that ATI WILL NOT ALLOW THEM TO RELEASE. Note that that blog entry is from the middle of last fucking year.

At this stage, to buy anything except an Intel GMA X3000 is madness. Intel are delivering fully Free/Open solutions that are powerful enough for anyone except hard core gamers (and said games don't exist on Linux anyway) and CAD people.

They're also significantly cheaper and more power efficient than the stuff being put out by Nvidia and ATI.

Re:Open Source supporters within ATI

[LWATCDR]

Only problem is that I have never seen an Intel video board. They all seem to be integrated on the motherboard. So if you have an AMD system you are stuck.

Re:In other news

[LWATCDR]

Well sort of. Intel didn't release all the specs of the chip because of DRM restrictions. They work well enough but Intel didn't provide full documentation.

R300 opensource drivers

[DrYak]

There are open-source R300 [sf.net] drivers that cover the 9500 up to x850 range of cards. I've used it for on old 9600XT AGP card (AGP chip) and HIS-overclocked X800 AGP card (PCIe chip with PCIe-to-GP bridge). The performance seem to be acceptable for my needs - which is surprising, knowing that R300 driver was completely developed from reverse engineering.

Recently the driver has been included in the official DRI tree. Most distro use it to provide open-source 3D acceleration. It is the default drivers for near every GPL-compatible Beryl/Compiz LiveCD (like Kooraa, for exemple) and function well enough with them (the same can't be said for official binary drivers).

As usual you should stop focusing on the hardware maker - who doesn't { have the possibility to / want to } throw resources at an OS that represents only a smaller fraction of their market share.
You should instead seek what has been produced by the OSS community - through large-scale collaboration they often manage to put out some marvels.

There no way one could except ATI to open-source drivers. They may have problems with code in their drivers that wasn't produced in house and that can't be opened cheaply.
BUT what AMD/ATI realy need to do is to help the DRI/FreeDesktop guys develop their own driver, and for that they need to document a little bit their chips. The best thing could do to the OSS community isn't trying to make their BLOB drivers less borked. The best thing would be to provide list of registers and samples so the community could write a R500 driver.

Up to R4xx not afterward.

[DrYak]

Apparently (rumour alert), the newer Radeons are very similar, in terms of interface, to earlier ones. Porting the drivers over requires about a 100-line diff; most of the changes are register locations, the actual semantics are similar.

Up to R4xx (Radeon X8x0. The R5xx family (X1300 and up) is radically different. It's still called radeon, but it doesn't share the radeon core. In fact it doesn't have a 2D core at all. It is a purely 3D chips that use triangle operation and similar to do 2D blits.

The open source R300 [sf.net] driver had been adapter to function with R400 cards too (up to X850) (and has been included in the mainstream DRI on freedesktop since then) - it works well, that's what I use.
But an R500 driver would require writing a new (3D-only) driver from scratch. Which is difficult and slow because ATI doesn't provide any documentation at all for their hardware, not even under NDA.