We don't have the 3d engine running in the open source drivers yet, so the only acceleration available is shadowfb, but everything else should be supported well.
Announcement
Collapse
No announcement yet.
Fglrx 8.6 or 8.7 doesn't work with Intel AGP at all
Collapse
X
-
Originally posted by bridgman View PostWe don't have the 3d engine running in the open source drivers yet, so the only acceleration available is shadowfb, but everything else should be supported well.
I'm getting very exasperated with the lack of basic functionality such as accelerated 2D X so phrases like "everything else should be supported well" have a tendency to rile me up, especially given that the unfruitful time spent on my Radeon 3850 so far exceeds my life's sum total of time spent on all other graphic cards under linux!
Your explanation for poor 2D that you gave in another thread misses the mark, most linux users like myself are not die-hard gamers. We want solid 2D for starters and as full time working people getting accelerated 2D working can't be a month long research project. I have no idea what ATI management was possibly thinking, but if ATI products can't reliably deliver the basics without muss and fuss then ATI is doomed in this market. In this context fglrx is a total disaster.
Do you know of any drivers (I don't care about the source) that can produce satisfactory 2D for X for a Radeon 3850 ?
Thanks.
Comment
-
Originally posted by nbi1 View PostUnfortunately there does not appear to be an open source equivalent that supports the Radeon 3850 so it's either make this work or punt!
Originally posted by nbi1 View PostYour explanation for poor 2D that you gave in another thread misses the mark, most linux users like myself are not die-hard gamers.
Originally posted by nbi1 View PostWe want solid 2D for starters and as full time working people getting accelerated 2D working can't be a month long research project.
Originally posted by nbi1 View Postespecially given that the unfruitful time spent on my Radeon 3850 so far exceeds my life's sum total of time spent on all other graphic cards under linux!
Your conf file looks a bit funny -- wouldn't hurt to turn VideoOverlay off although I doubt that will make a difference to this problem.
Originally posted by nbi1 View PostDo you know of any drivers (I don't care about the source) that can produce satisfactory 2D for X for a Radeon 3850 ?
Thanks.Last edited by bridgman; 04 August 2008, 02:58 PM.Test signature
Comment
-
Originally posted by bridgman View PostI guess I was replying to you. Depending on what you meant by "equivalent" the answer could either be reasonable or wrong. From your subsequent post I'm guessing "wrong"
I'm drawing a blank here, can you give me a clue ?
Modern GPUs do not have 2d engines, so everything is accelerated with the 3d engine -- which in turn means that until the 3d engine is running we can't implement 2d, 3d or video acceleration.
Originally posted by bridgman View PostYou have been going down some interesting paths -- mixing open source drm with fglrx ddx driver is definitely a no-chance-of-success approach. I took a look at your bug report but didn't see the useful parts of the x log -- can you pls paste the full log there ?
Your conf file looks a bit funny -- wouldn't hurt to turn VideoOverlay off although I doubt that will make a difference to this problem.
Was there a problem with your original installation or did you just decide that the driver must not be working because specific 2d functions were slow ? EDIT - never mind, I found the answer in the second link.
Regarding drm you only have to google to find out why I'm at where I'm at. This has been kicked around on several forums without a definitive answer. The safest approach seems to be building the drm module to make it available as needed. If you can definitively tell me drm should not be built for fglrx use then I'd be happy to omit it. And next you can tell me how to eliminate the resulting "cant acquire AGP".
Are you telling me that accelerated 2D for X is an achievable goal in 8.7 or do I need to pursue another avenue?
In general I'm very amenable to working with vendors to resolve linux issues, but I've got to get the sense there's a good faith effort being put forth by the vendor. So far in the case of ATI and Catalyst 8.7 for linux I'm feeling horn swaggled.
Comment
-
Originally posted by nbi1 View PostThat's what I was referring to. This is well and fine from a technical perspective, but has implications for any vendor trying to address the needs of a market that is predominantly 2D! As someone else explained to you we'd rather have ATI put a label on the box saying "NOT SUITED FOR LINUX" than make misleading statements about a driver that doesn't even deliver the basics. I've read all the official AMD/ATI docs and at no time did I get the impression that I wouldn't get working 2D at a minimum. If we can't get that, then what use is the driver?
Originally posted by nbi1 View PostOriginally I followed the 8.7 Catalyst instructions to the letter. That gave me crap 2D so I turned to the backport which doesn't seem to support the 3850. That led me to decide I better try 8.7 again. But as you can see I've experienced many problems including the two all time leaders "perpetual mesa" and "can't acquire AGP". Given these two have plagued fglrx for a long time can't ATI dedicate the resources to lay these to rest?
Doing a Google search for "can't acquire AGP" gave me only 3 hits, one of them from your post. The other two seem related to a motherboard BIOS bug which allowed the AGP aperture to be allocated above 2^^32 which a lot of hardware apparently can't handle.
Originally posted by nbi1 View PostRegarding drm you only have to google to find out why I'm at where I'm at. This has been kicked around on several forums without a definitive answer. The safest approach seems to be building the drm module to make it available as needed. If you can definitively tell me drm should not be built for fglrx use then I'd be happy to omit it. And next you can tell me how to eliminate the resulting "cant acquire AGP".
Originally posted by nbi1 View PostAre you telling me that accelerated 2D for X is an achievable goal in 8.7 or do I need to pursue another avenue?
Originally posted by nbi1 View PostIn general I'm very amenable to working with vendors to resolve linux issues, but I've got to get the sense there's a good faith effort being put forth by the vendor. So far in the case of ATI and Catalyst 8.7 for linux I'm feeling horn swaggled.Last edited by bridgman; 04 August 2008, 04:11 PM.Test signature
Comment
-
Originally posted by bridgman View PostI don't really understand you here. Accelerated 2D and 3D was available from roughly the time the card launched. As near as I can figure from your posts something went wrong with your installation so you didn't get acceleration. In your case I think you also picked up some bad advice from the internet and wasted a bunch of time as a result, which is unfortunate but does sometimes happen.
Since "perpetual mesa" means "something went wrong with the installation on my particular combination of hardware, distro, updates and drivers" I doubt we will ever get rid of it completely, but it is happening a lot less these days now that the installer has been opened up to allow independent maintenance of distro-specific packaging scripts.
Doing a Google search for "can't acquire AGP" gave me only 3 hits, one of them from your post. The other two seem related to a motherboard BIOS bug which allowed the AGP aperture to be allocated above 2^^32 which a lot of hardware apparently can't handle.
I can definitively tell you drm should not be built for fglrx. re: getting rid of the error message, please post your full x log to the bugzilla ticket so we can all see what is going on.
Yes.
Why do you say that ?
Ok, thanks. Will rerun without drm and post the X log to my bugzilla ticket.
It's true that I may have picked up some bad advice, but it begs the question why so many people are thrashing around with these problems. Getting working 2D out-of-the-box for linux should be as easy as for Windoze. Especially in the case of Debian Etch one of the most stable distros on the planet.
Why do I say that I feel horn swaggled? Because ATI has generated the impression that they make a usable driver available via package generation. Unless there is a 100% distro neutral way of distributing the driver ATI cannot absolve itself of the responsibility of testing the results of the distribution mechanism they provide, i.e. test the results of generating packages. If that doesn't yield a problem free install then for all intents and purposes it's the equivalent of a buggy driver. It just shouldn't be that difficult!
For instance, the packages could include pre and post install scripts that validate the driver environment (libraries, kernel source, tools, etc.). Frankly I'm surprised that this seems to be missing from the generated Debian packages as the Debian packagers tend to be very thorough on that count. What exactly have they made available to ATI in the way of packaging?
Despite my irritation with the driver (or lack thereof) your help is greatly appreciated. I'll post my X log ASAP after I get home from work.
Comment
-
Why do I say that I feel horn swaggled? Because ATI has generated the impression that they make a usable driver available via package generation.
Unless there is a 100% distro neutral way of distributing the driver ATI cannot absolve itself of the responsibility of testing the results of the distribution mechanism they provide, i.e. test the results of generating packages.
Operating Systems Distributions Supported
The latest version of the ATI Catalyst™ Linux software suite is designed to support the following Linux distributions:
Red Hat Enterprise Linux suite
Novell/SuSE product suite
--------------------------------------------------------------------------------
Note: The ATI Catalyst™ Linux software suite may install on a number of other Linux distributions. Refer to the Package Generation installation instructions for more information.
--------------------------------------------------------------------------------
--------------------------------------------------------------------------------
Note: AMD has accepted contributed packaging scripts to allows creation of other packages, but does not necessarily test, verify or warrant the reliability. Currently Red Hat Enterprise Linux suite and Novell/SuSE product suite are supported Linux distributions
--------------------------------------------------------------------------------
Thanks,
JohnTest signature
Comment
-
Originally posted by bridgman View PostNot sure if I asked this, but have you used the radeonhd (or radeon) driver with shadowfb acceleration ? If you haven't, you'll find it *much* faster for 2d than fglrx running without acceleration -- definitely worth a try.
Anything is bound to be "much" faster than fglrx without 2D acceleration. That isn't saying anything of value. My ancient Matrox Millenium runs circles around the crippled fglrx in 2D. Unless the 2D performance is on par with the 2D performance of the Catalyst Windoze driver I would not consider it an acceptable alternative.
Having indicated my willingness to help get 2D working for fglrx are you now telling me I should abandon that effort? Perhaps for reasons not shared with me nor anyone else in the linux community?
Don't mean to sound paranoid, but the timing of your comment strikes me as odd coming right after my agreement to run fglrx without drm and to post the resulting X log. Having invested as much time as I have I'd really like closure on 2D using fglrx.
Comment
-
Originally posted by nbi1 View PostNo, I have not used radeonhd. Don't know how to enable shadowfb.
Originally posted by nbi1 View PostAnything is bound to be "much" faster than fglrx without 2D acceleration. That isn't saying anything of value. My ancient Matrox Millenium runs circles around the crippled fglrx in 2D. Unless the 2D performance is on par with the 2D performance of the Catalyst Windoze driver I would not consider it an acceptable alternative.
Originally posted by nbi1 View PostHaving indicated my willingness to help get 2D working for fglrx are you now telling me I should abandon that effort? Perhaps for reasons not shared with me nor anyone else in the linux community?
Seriously, your comments about prioritizing 2D over 3D made me think you would have been pretty happy with the open source drivers, since 3D is really where fglrx has a big lead over the open source drivers.
Originally posted by nbi1 View PostDon't mean to sound paranoid, but the timing of your comment strikes me as odd coming right after my agreement to run fglrx without drm and to post the resulting X log. Having invested as much time as I have I'd really like closure on 2D using fglrx.Last edited by bridgman; 04 August 2008, 07:23 PM.Test signature
Comment
Comment