What next?

[craign]

I've replaced my NVIDIA graphics card with an ATI FireGL V8600 on my Fedora 9 system (Intel Dual Core) and have been playing around for about a week now. There are some clear improvements to my system: rotating the desktop cube, for example, is a beautiful thing.

However, in compiz mode, there is some substantial badness. Firefox's close/resize buttons have completely disappeared, and not infrequently the screen freezes for a couple of seconds. Video looks amazingly jaggy at times.

I've been searching the web, users groups, etc. for ways to improve this annoying, erratic behavior, but no luck so far. Questions to the more experienced than I:

1. Is what I'm experiencing bugs, or is there some magic configuration options that I don't know about?

2. If bugs, where do I post bug reports?

3. Should I conclude that I should have shelled out the massive extra cash and gone with an NVIDIA product? I'm naive at ATI, but it sure feels like it's behind the curve, Linux wise. :(

In essence, any suggestions on what next to do to have a more enjoyable compiz existence?

Many TIA,
Craig

[Dandel]

this is actually a known bug... the ATI driver currently does not handle XV and compiz properly at the same time. Try disabling compiz and see if it fixes the problem.

and as for your other questions...you have two places you can put bug reports up, namely the unofficial bug report system (http://ati.cchtml.com/) and the linux driver crew feedback. (http://support.ati.com/ics/survey/survey.asp?deptID=894&surveyID=508&type=web)

[craign]

Thanks super, Dandel, the behavior only appears when I choose compiz using the chooser widget, which is consistent with your explanation.

My question is, can I use compiz without XV? If so, how? Or am I stuck with this bug if I'd like to use compiz?

Many thanks,
Craig

this is actually a known bug... the ATI driver currently does not handle XV and compiz properly at the same time. Try disabling compiz and see if it fixes the problem.

and as for your other questions...you have two places you can put bug reports up, namely the unofficial bug report system (http://ati.cchtml.com/) and the linux driver crew feedback. (http://support.ati.com/ics/survey/survey.asp?deptID=894&surveyID=508&type=web)

[Melcar]

this is actually a known bug... the ATI driver currently does not handle XV and compiz properly at the same time. Try disabling compiz and see if it fixes the problem.

and as for your other questions...you have two places you can put bug reports up, namely the unofficial bug report system (http://ati.cchtml.com/) and the linux driver crew feedback. (http://support.ati.com/ics/survey/survey.asp?deptID=894&surveyID=508&type=web)


It's not a bug. It's a limitation of DRI and the current X.Org framework.

[craign]

So is it possible to turn off xv in Fedora and just use compiz? (If so, how? Been trying to figure that out & coming up empty.) Or is it not possible to have an unchoppy compiz with ATI?

It's not a bug. It's a limitation of DRI and the current X.Org framework.

[MikeEx]

If you absolutely must have compiz, You can set Gstreamer to No Xv.

Open a terminal and type in
gstreamer-properties
Go to the video tab
and set the Default Output Plugin to (No Xv)

You'll have to close any open videos for settings to take effect.

Pros: It works with compiz and videos don't flash any more
Con: It renders video via CPU

[craign]

Thanks super, but unfortunately the bad behavior remains.

The weird thing is, it's primarily when running Firefox.

If you absolutely must have compiz, You can set Gstreamer to No Xv...

[craign]

So am I correct then that the ghastly conclusion is that my new very expensive FireGL V8600 ATI card--the one that's so big it barely fits in my enclosure--is outperformed by the old NVIDIA card it's supposedly replacing, at least compiz wise? Was I an idiot to replace the old NVIDIA card? Be honest...

It's not a bug. It's a limitation of DRI and the current X.Org framework.

[Melcar]

Compiz should work fine. Some issues are present, like lack of mipmaps for certain cards (r6xx based cards as far as I can tell) and some performance issues on some of the more demanding effects. As I said before, not being able to play accelerated video streams is more of a DRI limitation than a fault of the driver; that can be "fixed" by either using plain ol' X for playback (both gstreamer and xine have options to set this, or the idividual player can do it as well) or run the streams on fullscreen mode with no other active window on-screen.

[craign]

Compiz should work fine...

Compiz does work until I'm using Firefox (which unfortunately I use often.) At that point, the window lacks the close/minimize/resize buttons which all of the other windows do have (weird) and when scrolling the mouse/window freezes, often for several seconds. If I'm not using compiz, these problems are not evident.

For a long time now, I've grown used to using compiz to select windows and desktops--I'm highly visually oriented and to me it's the primary reason I don't use a Mac. So to have a system which now allows me to move around on my desktops but once I'm at an application I use commonly (Firefox), it's somewhat frustrating to have to then unswitch compiz.

Using gstreamer-settings to unset Xv didn't seem to make any difference.

Any other ideas?

Many TIA,
Craig

P.S. Another place where I can see quite a difference is when I'm running video in iTunes inside VMware. If compiz is on, even in fullscreen exclusive mode, the video is really choppy. Gstreamer-settings unsetting Xv didn't help. If I turn off compiz, the problem disappears.

Likewise, if I'm running AutoCAD 2007 inside VMware, I can't use it as the cursor freezes often with compiz on. Turning compiz off, the problem goes away.

Both of those applications don't bother me so much as watching video or using AutoCAD I'm inside that app for a long time, so turning compiz off then on is no big deal. But Firefox in Fedora is bothersome, as I'll often be moving between desktops (flipping around the desktop cube) cutting and pasting, etc.

[duby229]

Compiz does work until I'm using Firefox (which unfortunately I use often.) At that point, the window lacks the close/minimize/resize buttons which all of the other windows do have (weird) and when scrolling the mouse/window freezes, often for several seconds. If I'm not using compiz, these problems are not evident.

For a long time now, I've grown used to using compiz to select windows and desktops--I'm highly visually oriented and to me it's the primary reason I don't use a Mac. So to have a system which now allows me to move around on my desktops but once I'm at an application I use commonly (Firefox), it's somewhat frustrating to have to then unswitch compiz.

Using gstreamer-settings to unset Xv didn't seem to make any difference.

Any other ideas?

Many TIA,
Craig

P.S. Another place where I can see quite a difference is when I'm running video in iTunes inside VMware. If compiz is on, even in fullscreen exclusive mode, the video is really choppy. Gstreamer-settings unsetting Xv didn't help. If I turn off compiz, the problem disappears.

Likewise, if I'm running AutoCAD 2007 inside VMware, I can't use it as the cursor freezes often with compiz on. Turning compiz off, the problem goes away.

Both of those applications don't bother me so much as watching video or using AutoCAD I'm inside that app for a long time, so turning compiz off then on is no big deal. But Firefox in Fedora is bothersome, as I'll often be moving between desktops (flipping around the desktop cube) cutting and pasting, etc.

Your usage scenario is exactly the market that fglrx was designed for. What I'd do at this point is file a bug report with ATi. Whether it is the fault of the driver, or the fault of X is totally irrelevant. The problem exists and that is what matters. It's ATi's responsibility to allocate the resources necessary to fix that problem....

It's problems like these why Open Source development is --far-- superior to closed development, and exactly why I think fglrx should be dropped and replaced by a single open source driver.

[craign]

Your usage scenario is exactly the market that fglrx was designed for. What I'd do at this point is file a bug report with ATi...

Thanks, Duby, somehow, strangely, even less mondo $ for my ATI card which in some substantial ways is inferior to its inexpensive NVIDIA predecessor, your post heartens me. If I can be part of a concerted argument to make ATI better, at least that's something positive. Which place would be most effective to file my bug report?

TIA,
Craig

[craign]

...It's problems like these why Open Source development is --far-- superior to closed development, and exactly why I think fglrx should be dropped and replaced by a single open source driver.

Been thinking about this. Why do these hardware vendors hang on to their proprietary software? I would think that the hardware vendor that open sources their software would have a huge advantage in selling their hardware. Am I missing something?

[RobbieAB]

Yes, you are missing the fact that the hardware vendor may not own all the software in the drivers. You are missing the fact that there may be commercially valuable methods encoded in the drivers.

We all know how much difference a good driver can make, now are you asking ATI to let nVidia have a look at their drivers?

[craign]

...We all know how much difference a good driver can make, now are you asking ATI to let nVidia have a look at their drivers?

From my perspective as the consumer, I would think that would be in ATI's best interest. The reason why I bought an ATI board is it provided similar hardware to NVIDIA but at a lower cost. What I've experienced is that this board is outperformed by my old NVIDIA board in substantial ways that I will experience every time I use a web browser. So in two years, when I replace this board, I will remember (1) that it was a substantially greater pain to install the Fedora drivers than for NVIDIA, and (2) an inexpensive NVIDIA board outperformed a top of the line ATI one in ways that I'll likely still be experiencing at that time. My conclusion will be to buy an NVIDIA board when I will (invariably) replace my ATI board in two years.

If, however, ATI works well on Linux in two years, I might consider another ATI board. From reading (that I should have clearly done before buying this board, mea culpa) about ATI's history of releasing Linux compatible drivers, my guess is that the only what ATI graphics will be competitive will be if they open source their driver code to accelerate their software development cycle.

Linux isn't going away, and the user share isn't getting smaller. A business plan for ATI based on closed driver software will result in losing me as a customer (and I can't imagine that I'm alone,) but one based on open source software will keep me and many like me using ATI products, and I can't imagine resulting in a loss for those who choose to use other OSs. So yes, I'm suggesting that ATI should let the entire world look at their drivers because their hardware will then be usable for an important part of their market share. I still don't see what I'm missing, other than a backward looking desire to keep something secret which seems like it should be secret, a very 1980s approach.

I'm also still wondering where the most effective place would be to submit my bug report to ATI, so if anyone has that URL, I'd be grateful.

[RobbieAB]

A few points:

1) Current top-end nVidia is having problems as well, and in some cases are (it might be were now) being out performed by older nVidia cards.

2) Drivers are critical, so if ATI release an open-source driver, what is to stop nVidia copying ATIs magic recipes and mixing them with nVidias secret ones? Result? nVidia out-perform ATI on the same price range across all platforms.

3) ATI don't own all the code in the drivers.

Nothing you have said answers either of points 2 and 3, so...

[craign]

1) Current top-end nVidia is having problems as well, and in some cases are (it might be were now) being out performed by older nVidia cards.


That may be, but it's a whole lot easier to install NVIDIA at this point on Fedora. While that may not be reason enough to choose NVIDIA from a company engineer's perspective, consumers tend to follow the path of least resistance.


2) Drivers are critical, so if ATI release an open-source driver, what is to stop nVidia copying ATIs magic recipes and mixing them with nVidias secret ones? Result? nVidia out-perform ATI on the same price range across all platforms.


This isn't software in a vacuum--it accompanies a hardware device. Nothing will stop NVIDIA from copying all of ATI's code, but as long as ATI provides comparable hardware at lower cost, I the consumer will buy it. And as the underdog, I would imagine ATI would want to do anything to keep my business.


3) ATI don't own all the code in the drivers.


I'll admit I don't know what ATI does and doesn't own (how could I in a closed environment?) but I'll bet the acceleration of software development in an open source environment would more than make up for the code ATI doesn't currently own.

Imagine a world where ATI is the preferred graphics hardware by Linux users because the Linux community is writing its drivers! Or perhaps you are an ATI software engineer worried about losing his/her job in such an environment--that's the only remaining reason I could imagine for this line of defense ;)

[Dandel]

firefox can be run in another method... by any chance is the bugs your having happen when you visit websites that use flash?

oh and also, note that flash has 2 different versions now, which run with different results.

Flash Player 10 (http://labs.adobe.com/technologies/flashplayer10/) ( currently beta, but fixes a lot of linux based bugs with flash. )

Flash Player 9 (http://www.adobe.com/products/flashplayer/) ( current stable release, and has a lot of problems on various websites, but still worth checking out. )

and as a quick note, you can find out which version of flash you have by loading this page. (http://www.adobe.com/shockwave/welcome/)

[craign]

firefox can be run in another method... by any chance is the bugs your having happen when you visit websites that use flash?

By "another method" do you mean turning off compiz? I can do that, but clicking compiz to KWin every time I change desktops gets old fast.

The disappearance of the close/minimize/restore window widgets happens in Firefox no matter if Flash is running or not, and the scrolling freeze and general freeze of the window and cursor happens on sites without Flash (just tested it.)

Thanks for trying to help,
Craig

[Dandel]

For now, i can tell you how to completely disable the special features of flash, all you need to do is install the openssh server and run it over that... as follows...

ssh -XC localhost firefox

running it like that will completely remove the accelerator that is enabled for flash.

[craign]

For those naive shoppers such as myself two weeks ago, I don't want to imply that I'm unhappy with this ATI board. If using the web wasn't such a big deal and Firefox such a great browser, flipping around the desktop cube, moving around the windows, etc. is truly a beautiful thing on Fedora 9 with an ATI FireGL V8600 board. I'm OK with clicking out of KWin when I'm using VMware for obvious reasons.

I'm holding out hope that somehow my Firefox problems will be fixed in the not to distant future, and ATI will remain a choice when I'm again shopping for a new board, which seems to happen every two years. I happen to believe that is more likely if ATI open sources their driver software, but I'll take any solution that works.

[craign]

ssh -XC localhost firefox


Tried that, no difference. Behavior remains:
(1) No close/minimize/resize window widgets
(2) Intermittently (but frequently enough to be a usability issue) on scrolling or moving mouse, screen and cursor freezes for several seconds

Thanks again,
Craig

[Dandel]

hmm... weird, can you give a list of plugins that are enabled on compiz, along with what is disabled?

[craign]

hmm... weird, can you give a list of plugins that are enabled on compiz, along with what is disabled?

Weird is apparently my computer's current middle name :)

Is there an easy way to generate a complete list?

[craign]

hmm... weird, can you give a list of plugins that are enabled on compiz, along with what is disabled?

Does this help?

~/.gconf/apps/compiz/plugins $ ls
cube decoration %gconf.xml scale

$ cat cube/allscreens/options/%gconf.xml
<?xml version="1.0"?>
<gconf>
<entry name="index" mtime="1220130888" type="int" value="11">
</entry>
<entry name="abi" mtime="1220130888" type="int" value="20080424">
</entry>
</gconf>

$ cat decoration/allscreens/options/%gconf.xml
<?xml version="1.0"?>
<gconf>
<entry name="command" mtime="1220133404" type="string">
<stringvalue>emerald --replace</stringvalue>
</entry>
</gconf>

$ cat scale/allscreens/options/%gconf.xml
<?xml version="1.0"?>
<gconf>
<entry name="index" mtime="1220130859" type="int" value="13">
</entry>
<entry name="abi" mtime="1220130859" type="int" value="20080301">
</entry>
</gconf>

[duby229]

A few points:

1) Current top-end nVidia is having problems as well, and in some cases are (it might be were now) being out performed by older nVidia cards.

2) Drivers are critical, so if ATI release an open-source driver, what is to stop nVidia copying ATIs magic recipes and mixing them with nVidias secret ones? Result? nVidia out-perform ATI on the same price range across all platforms.

3) ATI don't own all the code in the drivers.

Nothing you have said answers either of points 2 and 3, so...

License it under the GPL then... The GPL specifically protects against these kinds of violations throught the use of what is now known as a copy-left agreement. In other words it means that if you use code that is licensed under this agreement, then any derivative works must also be licensed under this agreement. Which in turn means that any improvements that nVidia makes will also be available to ATi. If you think about it, it makes it possible for the two companies to compete more directly, and the advantage will go to who-ever has the best hardware. Given current architectures, and the current trend to rely more heavily on the 3D engine (you know eliminating all that special purpose hardware, and moving those functions to the 3D engine, and making the 3D engine more general purpose), ATi has the advantage.

You need to give the GPL a little more credit, we arent totally stupid. The whole point of the license is to protect your code.

It's actually worth reading. You should take an hour and a half or so and skim through it. You can get most of the important data in an hour and a half or so..

[duby229]

That may be, but it's a whole lot easier to install NVIDIA at this point on Fedora. While that may not be reason enough to choose NVIDIA from a company engineer's perspective, consumers tend to follow the path of least resistance.



This isn't software in a vacuum--it accompanies a hardware device. Nothing will stop NVIDIA from copying all of ATI's code, but as long as ATI provides comparable hardware at lower cost, I the consumer will buy it. And as the underdog, I would imagine ATI would want to do anything to keep my business.



I'll admit I don't know what ATI does and doesn't own (how could I in a closed environment?) but I'll bet the acceleration of software development in an open source environment would more than make up for the code ATI doesn't currently own.

Imagine a world where ATI is the preferred graphics hardware by Linux users because the Linux community is writing its drivers! Or perhaps you are an ATI software engineer worried about losing his/her job in such an environment--that's the only remaining reason I could imagine for this line of defense ;)

If ATi fires there Linux devs, then they are truly retarded. There are only a small handful of developers that know anything about graphics drivers on linux. ATi has some of them on staff right now. Firing them would be the single dumbest thing ATi could do, but they definitely need to be re-allocated to the team developing the open source drivers... It's unfortunate, but right now that team is a one man army....

[Dandel]

Does this help?
yes actually, try changing the command...
emerald --replace
with
kde-window-decorator --replace

and if this does not work, try disabling the compiz window decorator feature.

also, mind telling us which version of the desktop manager you have...
for compiz the command is easy, it's... compiz.real --version
and as for kde, it's the command, kde-config --version
and gnome, is the command gnome-session --version

[bridgman]

FWIW, GPL doesn't help at all. It's just a license that protects the actual code, not the ideas expressed within that code. It would take very little time to re-express the same ideas in different code which would not be subject to any of the GPL restrictions.

This was one of the approaches we looked at early on in the project, but unfortunately it doesn't help in this particular situation.

[duby229]

FWIW, GPL doesn't help at all. It's just a license that protects the actual code, not the ideas expressed within that code. It would take very little time to re-express the same ideas in different code which would not be subject to any of the GPL restrictions.

This was one of the approaches we looked at early on in the project, but unfortunately it doesn't help in this particular situation.

And is that a bad thing? You make it sound like it's terrible. Like it is the worst possiblre thing. But it isnt. First of all I dont think that nVidia could take your code and re-implement it. I know for a fact that is a GPL violation. The GPL specifically states that if they use your code they they need to abide the GPL... Period..... It really is that simple.

Check this site out... gpl-violations.org The information here is somewhat limited, but you can go through the list of GPL enforcements, and time after time you find that it protects against exactly the situation you just described The problem is that you make it sound like it is some kind of bad thing.

The truth is that it is not. In the end you'll be able to benefit from the work that every other GPU company is making in the free software community. Sure they will benefit from your work too, but --ONLY-- if they too adopt the GPL... In which case you'll win due to superior hardware.

[bridgman]

First of all I dont think that nVidia could take your code and re-implement it. I know for a fact that is a GPL violation. The GPL specifically states that if they use your code they they need to abide the GPL... Period..... It really is that simple.

Yep, if the code is used verbatim then the user would need to abide by the terms of the GPL. That's not what we are talking about here -- we are talking about the re-implementation of ideas in new code. Neither copyright nor GPL protect against that, only software patents (or secrecy ;)) can do so.

Strictly speaking it is copyright which protects the code and GPL defines the terms under which the authors of that code allow others to use it. If you use GPL-licensed code without obeying the GPL terms you are violating copyright unless the authors have given you permission through an alternate licensing scheme.

Check this site out... gpl-violations.org The information here is somewhat limited, but you can go through the list of GPL enforcements, and time after time you find that it protects against exactly the situation you just described The problem is that you make it sound like it is some kind of bad thing.

Harald Welte's project, right ? I think you will find that gpl-violations.org is going after something different - verbatim use of GPL-licensed code (eg. the Linux kernel) without abiding by the terms of the GPL for that code.

The truth is that it is not. In the end you'll be able to benefit from the work that every other GPU company is making in the free software community. Sure they will benefit from your work too, but --ONLY-- if they too adopt the GPL... In which case you'll win due to superior hardware.

If all of our major competitors in the workstation market were providing their commercial workstation drivers under a free software license I might agree with you. However, you are assuming that anyone adopting the *ideas* from our GPL-licensed code would be forced to release their code under GPL, but that is not the case.

[duby229]

Ok, so let me get this... Your not worried about protecting the code, instead your worried about protecting the idea that the code implements?

What gives you the right to tell someone that they arent allowed to implement an idea?(please dont tell me software patents) And what does code have anything to do with that? At least with the GPL your code is protected. Right now if a company wants to implement your idea's all they have to do is reverse compile your code, re-implement it for there hardware, and then recompile it.. And you'll never know the difference. Oh wait did I just let loose the (industry wide) dirty little secret?

EDIT: Lets me just say I think that is back asswards. First of all I think if everyone would just share there idea's to begin with we could live in a much more uniform and compatible universe. If ATi, nVidia and Intel would share there idea's and all developed complete open source strategies we could live in a world where video acceleration works across the board. Where new 3d acceleration features are supported immediately. Where font rendering actually looks good. Where 2D acceleration works -properly- with modern compositors. etc, etc, etc... The list goes on and on and on.

[rbmorse]

Why can't we all just get along?