Tear-Free Acceleration For ATI EXA, Xv

[duby229]

OK, now you've lost even me...
- AMD has AtomBIOS, and are commissioning open drivers to suit. They're not ready yet, so if that's what you're annoyed about, I've no problem with you venting :)
- AMD have released/are releasing register-level specs for GPUs, which to me is as low-level an "API" as you can get before documenting sub-instruction algorithms
- AMD may have fancy sub-instruction algorithm stuff that they know how to optimise, but isn't apparent from register-level specs. This is one gap I hope will close in the future, because obfuscation is not ideal but may sorta work in the real world. This is also an area where I'd pick an open driver over a few more FPS, but to some it's a critical difference worth millions.
- AMD has bowed to market forces demanding proprietary crap like Blu-ray DRM, which also keeps the proprietary driver around, and I (like many other Linux users) don't care as long as there's an open API to offload video decoding tasks to hardware designed for it.

Atombios only covers modesetting, and a few other things. But even so, it's proven to be highly valuable. Performance may not be as good as it could, and there may be some compatibility issues, but for the most part is just works. My EET background is kinda limited, but personally I couldnt give two shits if I got a triangle drawn on screen by 30 lines of obfuscated code or 5000 lines of hand written assembly. It doesnt matter one tiny little bit to me as long as I have a triangle on the screen....

ATi is throwing a hissy fit, and moaning and complaining about, and generally dropping every excuse they can find about why they "need" a closed driver, when in absolute fact they just dont. They have patented hardware. They can protect those patents by using a firmware. Period.

And the most messed up thing about all of this is that if ATi made it modular enough, they could swap out older and slower portions of the firmware with newer optimized versions on the fly during run time if they wanted to. The technology exists, and is readily available today.

[ovoskeuiks]

personally I couldnt give two shits if I got a triangle drawn on screen by 30 lines of obfuscated code or 5000 lines of hand written assembly. It doesnt matter one tiny little bit to me as long as I have a triangle on the screen....

So why are you making such a fuss about drivers if you only care about the end result. You should try diverting some of the energy you're putting into arguing about all of this into contributing something constructive.

[duby229]

So why are you making such a fuss about drivers if you only care about the end result. You should try diverting some of the energy you're putting into arguing about all of this into contributing something constructive.

If you think advocating open source software in the face of corporate pressure isnt constructive, then you've still got a thing or two to learn.....

[ovoskeuiks]

If you think advocating open source software in the face of corporate pressure isnt constructive, then you've still got a thing or two to learn.....

I'm not suggesting that advocating of open source software isn't constructive, however might I humbly suggest you attempt to be a little less abrasive in your approach

We have a Walmart in my town, and right out in front by the big sliding doors there exists three concrete "stoppers" They are about 4 feet tall, and about 8 inches wide in diameter. Tell Novell to find one and sit on it, then once it's tightly entrenched in there ass, have them spin around on it over, and over, and over, and over, and over, and over, and over, and over.......

Comments like this are at best childish and most likely offensive to someone. I know/hope you're directing this at 'Novell' as a corporate entity and not the individuals who might work on the open drivers in your proposed grand scheme of ATI openness but if you want people to see things your way you might try to take the high road and focus on the facts and issues rather than resorting to statements like that.

[NaterGator]

Again, thanks for totally derailing this thread when there are plenty of others about the subject you guys are discussing.


Now that you guys are getting personal and basically just attacking each other, it'd be a good time to drop it in this thread and move it into a new one before you get this one deleted.

[duby229]

I'm not suggesting that advocating of open source software isn't constructive, however might I humbly suggest you attempt to be a little less abrasive in your approach



Comments like this are at best childish and most likely offensive to someone. I know/hope you're directing this at 'Novell' as a corporate entity and not the individuals who might work on the open drivers in your proposed grand scheme of ATI openness but if you want people to see things your way you might try to take the high road and focus on the facts and issues rather than resorting to statements like that.

I certainly respect your view. It has it's place and it's purpose. However so does does mine, so the least you can do is respect the views that I hold. You dont have to agree with them, but acknowledge that they have a valid place.

And by the way statements like that are --exactly-- what need to be said, and they need to be said directly to Novell by ATi right now. Today. And that is putting it mildly to say the very least.

[nphoronix]

So, when are we getting a functional driver with 3D support? I mean, since ATi is possibly never going to fix the flickering/tearing bugs in newer cards...

[bridgman]

Are you talking about the open source driver ?

[oblidor]

So, when are we getting a functional driver with 3D support? I mean, since ATi is possibly never going to fix the flickering/tearing bugs in newer cards...

If you are talking about flickering when watching a DVD, make sure you are not running Compiz and using opengl to render the video. Compiz and opengl applications don't work yet for any card (Nvidia or ATi etc...)...

[nphoronix]

If you are talking about flickering when watching a DVD, make sure you are not running Compiz and using opengl to render the video. Compiz and opengl applications don't work yet for any card (Nvidia or ATi etc...)...

Yes, my point indeed was about my older NVIDIA 8800GTS having compiz enabled AND playing videos without flicker/tear and stuff, but my 3870 fails to do that, even with opengl (although at least opengl plays as it should with compiz disabled). However, I bought ATi for a reason, and I'd really like that xv extension finally supported.

And yeah, I was asking about the state of the open source drivers, since ATI doesn't seem to want to fix it anytime soon. From what I remember, the R600 3D documents should have been handed in by now, right?

[Vighy]

And yeah, I was asking about the state of the open source drivers, since ATI doesn't seem to want to fix it anytime soon. From what I remember, the R600 3D documents should have been handed in by now, right?

Bridgman should be one of those working on it... he should know at what point they are :D

[bridgman]

The first two passes through IP review for 6xx/7xx documentation failed. I can't really go into the details because that would kinda defeat the purpose of keeping some stuff secret, but potential risks were raised which we did not have answers for at the time.

I believe we have worked through all of those issues now and I am going to kick off another review pass this week.

[lucky_]

The first two passes through IP review for 6xx/7xx documentation failed. I can't really go into the details because that would kinda defeat the purpose of keeping some stuff secret, but potential risks were raised which we did not have answers for at the time.

I believe we have worked through all of those issues now and I am going to kick off another review pass this week.

How long lasts one of these reviews ?
One week, one month ?

[duby229]

The first two passes through IP review for 6xx/7xx documentation failed. I can't really go into the details because that would kinda defeat the purpose of keeping some stuff secret, but potential risks were raised which we did not have answers for at the time.

I believe we have worked through all of those issues now and I am going to kick off another review pass this week.

Thats interesting. Thanks for working so hard on this documentation. I'm glad to see progress is still being made.

[rbmorse]

How long lasts one of these reviews ?
One week, one month ?

That's hard to predict. Depends on the issues, and whether, for example, there are things like third-party interests involved.

Some of these things can take a long time only to find they can't be fixed, especially where there is third-party involvement. I was involved in one where negotiations went on for a year before the other side decided they didn't want to play, during which our own project was on largely on hold pending resolution.

They offered an NDA that would have let us continue to develop pending resolution of release issues, but if we didn't get permission in the end that effort would have been lost and our devs tainted by exposure to the source so we declined and put the devs onto other projects. After awhile I got fired^h^h^h^hlaid off because without the project I didn't have a function.

It would have been cheaper to develop an independent solution, but the lawyers kept telling us an agreement with the other was imminent. I now believe they were just stringing us along in an attempt to get us to abandon the project. It worked.

[lucky_]

That's hard to predict. Depends on the issues, and whether, for example, there are things like third-party interests involved.

Some of these things can take a long time only to find they can't be fixed, especially where there is third-party involvement. I was involved in one where negotiations went on for a year before the other side decided they didn't want to play, during which our own project was on largely on hold pending resolution.

They offered an NDA that would have let us continue to develop pending resolution of release issues, but if we didn't get permission in the end that effort would have been lost and our devs tainted by exposure to the source so we declined and put the devs onto other projects. After awhile I got fired^h^h^h^hlaid off because without the project I didn't have a function.

It would have been cheaper to develop an independent solution, but the lawyers kept telling us an agreement with the other was imminent. I now believe they were just stringing us along in an attempt to get us to abandon the project. It worked.

MMh let's hope it won't be that crappy...

[Vighy]

The first two passes through IP review for 6xx/7xx documentation failed. I can't really go into the details because that would kinda defeat the purpose of keeping some stuff secret, but potential risks were raised which we did not have answers for at the time.

I believe we have worked through all of those issues now and I am going to kick off another review pass this week.

so... let's just hope.. right? :D or you think it will be possible that this time it will be ok?

[bridgman]

I think chances are good that this will be the one, but I don't want to sound too confident and jinx things. We've been a couple of weeks away for quite some time now :mad:

[d2kx]

At least there will be some R6xx/R7xx support already at the time the documents are being released. The CS-branch of RadeonHD looks like it'll be merged to master anytime now, too (can you comment on that one, bridgman?).

[agd5f]

THe CS branch is basically a clean up and reorganization of the command submission code that was inherited from radeon. It's not specific to any particular chip.

[Vighy]

THe CS branch is basically a clean up and reorganization of the command submission code that was inherited from radeon. It's not specific to any particular chip.

will radeon merge it back?