about firefox ... / dont understand anything [radeon]

LarryT guess.who at freesurf.fr
Thu Mar 16 07:22:45 UTC 2006


Thank you Mike :)
Well, after i read what you wrote, i realize that i could have a look on 
the driver used by other os i have on my pc (centos 4 and 4.2 and fc4) !
Centos  uses radeon driver. So i have just swith to radeon driver. So 
far so good !
Gonna see what happens , now :)

thx anyway

larry

Mike A. Harris wrote:
> LarryT wrote:
> 
>>> A. Which graphics card are you using?
>>
>>
>> ATI Radeon 7000 RV6DL 32mo
>>
>>> B. Which drivers for the graphics card are you using?
>>
>>
>> Not sur to kown how to get this info :!
>> going through system/admin/display :
>> "ati - ATI Mach8, Mach32, Mach64 ad Rage XL card"
>> Here is the pout put of xorg.conf :
>> Section "Device"
>>         Identifier  "Videocard0"
>>         Driver      "ati"
> 
>                        ^^^
> 
> The "ati" driver wrapper module is a problem which should be avoided.
> 
> Ideally, in the perfect world, you should be able to use the "ati"
> wrapper module regardless of what ATI chip family and revision you
> have, and it should automatically load the proper actual driver for
> you, as all the "ati" module does, is scan your system and determine
> what (if any) ATI hardware is present, PCI, AGP, ISA, etc., and if it
> finds any, it looks up the device in a list which tells it which
> "real" driver to use.  The "real" drivers are "atimisc", "r128", and
> "radeon".  It then loads the proper driver, and unloads itself.
> 
> Or at least that is the intention of "ati".
> 
> Unfortunately, the real world doesn't quite line up with this ideal
> situation very well, and there are often PCI IDs which have been added
> to the "radeon" driver by one of the many upstream Radeon driver
> developers, which were not added to the "ati" wrapper's detection
> routines at the same time, resulting in the "ati" wrapper being
> unaware of a given chip, when in fact the chip is supported by the
> "radeon" driver.
> 
> The result is that the "ati" wrapper will be unable to detect some
> hardware and reroute it to the "radeon" driver, but if you use the
> "radeon" driver directly you avoid the middle man, and avoid that
> problem path entirely.
> 
> As such, it is very highly recommended to _never_ use the "ati"
> wrapper for Radeon hardware, as it is very highly likely to not
> be in perfect sync with the actual current Radeon hardware
> support.  There are other bugs/problems which have come up in
> the "ati" wrapper over time as well, and while they get fixed
> eventually, something else inevitably breaks.  It's just an extra
> layer of code which can cause unnecessary grief over just using
> the proper driver module to begin with:  radeon
> 
> If you use "radeon" as the driver, and your video card doesn't work,
> don't even bother trying to use "ati", because as I stated above, all
> "ati" does, is say "oh, you have a Radeon" and load the "radeon"
> driver.  Instead, if the "radeon" driver does not work, file a bug
> report in X.Org bugzilla and describe the problem in detail, and
> attach your X server log and config file.
> 
> Currently, our video detection should detect all supported Radeon
> hardware and automatically assign them to the "radeon" driver, all
> Rage 128 hardware and automatically assign it to the "r128" driver,
> and all Mach64, Rage, RageXL, and older hardware and assign it to
> the "ati" wrapper (although I suppose in light of what I said
> above, 'atimisc' might be more appropriate, but the problem case
> with "ati" has been Radeon hardware, rather than Mach64, so it's
> a moot point in practice).
> 
> On a final note, I should state that if you ask around, you will get
> very different advice about this from some other people perhaps.  Some
> people recommend using the "ati" wrapper for all hardware, and that
> conflicts with what I am recommending.  People who suggest to use
> the "ati" wrapper for all ATI hardware unfortunately have not had
> the pleasure of maintaining XFree86 and X.Org X11 in 12 operating
> system releases over a period of 5.5 years and discovering just how
> unreliable the "ati" wrapper is in the real world.  My advice comes
> from experience in action, rather than ideological best-case
> wishful thinking of those who recommend the wrapper.  ;o)
> 
> 
> 




More information about the fedora-test-list mailing list