Copyright © Flavio StanchinaJuly 22, 2009 at 00:48:28
|
Current package version: 8.19.10-1
(changelog)
This page explains how to build Debian packages from the ATI Linux driver for the following ATI products:
Important:
I do not work for ATI,
I'm only packaging their driver as a service to the community.
I do not have the full source code for the driver,
I'm just unpacking the packages available from the ATI website and
repacking them as Debian packages.
Therefore,
I can not fix bugs in the driver and
I can not help people with any problems
not related with installation of these Debian packages.
Although the ATI installer can create Debian packages based on my work,
I do not support those packages.
Please do not try to upgrade from the packages produced
by the ATI installer to the packages you'll find here
or to official Debian packages.
If you do and run into trouble, at least avoid filing bug reports in the Debian BTS.
You must completely remove the packages built by the ATI installer before installing packages from this page or the official Debian fglrx packages. |
Therefore, I've removed the prebuilt binary packages; if you really want to use these packages, I assume you are able to build binary packages from source as clearly explained in section 3a. Source. Everyone else, please use the official Debian packages if you're using sid, or the ATI installer packages if you're using sarge or Ubuntu.
Upgrading from these installer packages to the new Debian packages is supposed to be smooth and hassle-free. Pay attention to the control panel package: it has been renamed from "fglrx-control-qt3" to "fglrx-control", but hopefully I added the correct Conflicts and Replaces fields to make this a smooth upgrade. If the upgrade doesn't work as expected, please report any problems to me.
Going forward, my plan is as follows: I will maintain the old installer packages until we can add AMD64 support to the official packages. I will also try to build binaries for sarge (and etch, if the packages don't go into testing fast enough) as required, but I don't think there are many people that need them... if you are the kind of people that want the latest fglrx drivers, you probably left sarge even before it was released.
Have fun. ;)
apt-get remove "fglrx-4.*" apt-get install fglrx-driver fglrx-kernel-src
kernel-headers packages should suffice.
I didn't test if it's true. Feedback is welcome.
kernel-headers package.]
First things first: read the instructions below carefully before doing anything else!
This is not for the faint of hearth or the uninitiated. If you never tried to build a Debian package from source, learn how to do that before you come back here.
I don't mean to be offensive, but building kernels and modules correctly is difficult; pretending it isn't (or whining because it is) ain't going to help you in doing it. If at any time you don't feel confident with what you are doing, by all means stay with what you have and don't risk breaking your system.
apt-get and dpkg.
Please don't stay logged in as root any longer than necessary;
rather, I'd suggest to install sudo and use that.
The driver as a whole is made of three main components: the driver proper, a replacement libGL and a kernel module. The driver and libGL are in the fglrx-driver package, the kernel module's source code is in the fglrx-kernel-src package.
Note that the driver will work without the kernel module, but without 3D acceleration. If you're not interested in 3D acceleration, you can do without the kernel module.
The list of supported cards at the beginning of this page comes from the driver's release notes, but it is known to be somewhat incomplete. Here's a complete list of chipsets supported by the fglrx driver, as reported by the driver itself in the X server log file.
If your card is not supported, you can force the driver to recognize it as a different card by overriding the ChipID.
/usr/bin/nvidia-installer --uninstall
Of course, this is relevant only if you previously had tried to use a NVidia graphics card in your system.
You need to install the kernel source code that corresponds to your running kernel and
configure it with make-kpkg.
This is obviously easy if you already compile your own kernel,
but it's going to require a bit of work if you are using a prebuilt kernel.
Alternatively, if you are using a prebuilt kernel,
install the official kernel-headers or linux-headers
package that matches the kernel you're using and skip this section entirely.
However, you still have to check that the kernel you're using has all
the correct configuration options.
As a guideline, kernels for 686 and k7 CPUs should work fine,
but kernels for the venerable 386 might not include AGP, DRM and/or MTRR support.
/proc/version) and make sure you're using the same.
This is beyond the scope of this document and is left as an exercise;
if you can't figure it out, take it as an hint that this whole mess is beyond your abilities.
You don't need to actually compile the kernel, but you do need to configure it properly. If you don't want to compile and install your own kernel, just install the sources for the same kernel version that you are using and use the configuration that came with your kernel package:
apt-get install kernel-source-2.x.y cd /usr/src tar xjvf kernel-source-2.x.y.tar.bz2 cd kernel-source-2.x.y cp /boot/config-2.x.y-flavour .config make-kpkg --append-to-version "-flavour" --revision 2.x.y-z --config old configure
"-flavour" is supposed to be a placeholder.
You need to replace it with the correct value for your kernel,
for example something like "-1-686" for prepackaged Debian kernels.
It's the text that follows the kernel version in the output from uname -r
and in the file names in the /boot directory.
"2.x.y-z" is the version of the kernel-image package you have installed. It should also be the same version as the kernel-source package.
If you already have a configured kernel source tree, you can use it, but make sure that it's exactly the same configuration you are using.
The following kernel options might affect the fglrx module. This list shows the options I've been using successfully for the past two years; these options are correct for kernel 2.6.10, but they should be the same for any 2.6.x kernel and probably for 2.4.x kernels too.
| Kernel option(s) | Notes |
|---|---|
|
CONFIG_MODULES=y CONFIG_MODULE_UNLOAD=y CONFIG_KMOD=y | Make sure that CONFIG_MODVERSIONS is set correctly! |
| CONFIG_MTRR=y | The fglrx driver needs MTRR (Memory Type Range Register) support |
|
CONFIG_AGP=m CONFIG_AGP_chipset=m |
Enable any AGP chipset drivers you might need, for example
CONFIG_AGP_INTEL, CONFIG_AGP_NVIDIA, etc...
CONFIG_AGP_ATI is required only if you actually have an ATI AGP chipset on your motherboard;
it has nothing to do with the fact that your AGP video card is an ATI.
Do not set CONFIG_AGP=y if you want to use fglrx's built-in AGP support. There is a difference here between kernel 2.4 and 2.6:
The important part is having CONFIG_AGP=m. If you are absolutely sure that you always want to use the kernel's builtin AGP support, you can set CONFIG_AGP=y, but I do not recommend this as there are no advantages with respect to CONFIG_AGP=m. |
|
CONFIG_DRM=m or y CONFIG_DRM_RADEON=m | Do not compile the DRM modules into your kernel,
because they would conflict with the fglrx kernel module.
The best course of action is to enable them as modules
and let the user-space drivers load the modules they require.
Hardware detection tools might try to load the radeon DRM module automatically during startup; if that happens to you, either tweak the tools' configuration or disable the radeon DRM module. |
|
# CONFIG_FB is not set |
CONFIG_FB_VESA=y is supposed to work, but it appears that
it can cause
"mtrr overlapping" errors; CONFIG_FB_RADEON=y definitely causes problems, so disable it. |
| CONFIG_TMPFS=y | ATI says POSIX shared memory is required! |
If any of your kernel configuration options differ from the ones above and you are having problems, please try with the options exactly as above or be prepared to explain why your choice should make no difference.
You need some support for your AGP chipset. You have two choices:
the ATI driver's built-in AGP support or the kernel agpgart driver.
It's difficult to say in advance which one is better or will work at all,
because it depends on the exact driver and kernel versions you're using, but
the kernel driver should be your preferred choice if it works for you.
To use the kernel driver, add the following line to the "Device"
section of your X server config file:
Option "UseInternalAGPGART" "no"
The X server configuration is discussed in more detail below, but the AGP configuration is important so I'm mentioning this detail early because I'd like to make sure that it sticks in your memory somewhere.
unable to acquire AGP, error "xf86_ENOMEM"
in your X server log usually means that you are using
the built-in AGP support with an unsupported chipset.
Set "UseInternalAGPGART" to "no" and
load the kernel driver (see below for details)
which will hopefully support your AGP chipset.
If you want to use the ATI driver's built-in AGP support instead,
set "UseInternalAGPGART" to "yes"
(it's the default setting, so you can just omit it) and make sure that
the kernel driver is not compiled into your kernel:
either compile it as a module or disable it entirely.
Beware that discover (and possibly other hardware detection tools)
might try to load the kernel driver automatically during startup.
If that happens to you, you can tweak discover's configuration
or you can simply move the agpgart.* files
away from /lib/modules.
If neither one supports your chipset, try using the kernel driver
with the agp_try_unsupported=1 option on the kernel command line
or as a parameter when you load the module.
Note that in the 2.6 kernel the AGP chipset drivers are separate
from the agpgart driver itself, so if you build them as modules
you will need to load the agpgart module and a chipset module
such as via-agp, nvidia-agp, etc.
If you want to build the source packages,
make sure you have installed debhelper,
kernel-package (for make-kpkg),
libqt3-mt-dev, xlibs-dev, libxtst-dev,
fakeroot, rpm and wget.
apt-get install debhelper kernel-package libqt3-mt-dev xlibs-dev libxtst-dev fakeroot rpm wgetYou should also install
libncurses5-dev if you want
to be able to use make menuconfig in the kernel source tree.
If you will build only the kernel module,
you need debhelper,
kernel-package and fakeroot.
Once you've sorted out all this, you can proceed to download the following package sources or the prebuilt packages.
These scripts will package the ATI driver version 8.19.10.
You can ask for a different version by setting the PVERSION
environment variable (see below).
The scripts will detect which version of the X server is currently installed
on your machine and download the matching driver package.
You can ask for a different version by setting the XTYPE and XVERSION
environment variables (again, see below).
Download the files above into a directory of your choice
(preferably /usr/src)
and unpack with
dpkg-source -x fglrx-installer_8.19.10-1.dsc
Then cd into the newly created fglrx-installer-8.19.10 directory and run
dpkg-buildpackage -b -rfakeroot -tc -uc -D
The build script will download the driver from the ATI site, then it will build the following packages in the parent directory:
Now you can install the packages and compile the kernel module.
If you want to build a specific driver version, use the following command:
PVERSION=8.14.13 dpkg-buildpackage -b -rfakeroot -tc -uc -D
The following driver versions are supported: 8.8.25, 8.10.19, 8.12.10, 8.14.13, 8.16.20, 8.18.6, 8.18.8, 8.19.10.
If you want to build for a specific X version, use the following command:
XTYPE=XFree86 XVERSION=4.3.0 dpkg-buildpackage -b -rfakeroot -tc -uc -D
The following X server versions are supported:
You can of course use
XTYPE,
XVERSION and
PVERSION at the same time.
No packages are available at the moment; look in the News section at the top of the page for more details.
I often get requests for packages of older driver versions: I would love to help, but I don't have enough time to do so. Really. I'm sorry but I have a life to live besides building packages.
You need to install
the "fglrx-driver"
and "fglrx-kernel-src" packages;
the "fglrx-driver-dev" package contains
header files and the libfglrx_gamma.a static library
and you need it only if you develop ATI-specific OpenGL applications;
the "fglrx-control-qt3" package installs
the control panel and can be useful for multi-monitor configurations;
the "fglrx-sources" package contains
the control panel source code and a sample application
and it is not required for normal use.
If you configured your /etc/apt/sources.list with one of my prebuilt package repositories,
install the packages with apt, aptitude, dselect or similar tools.
For example, with apt-get:
apt-get install fglrx-driver fglrx-kernel-src
If you downloaded the packages or built your own, install them with dpkg:
cd .. dpkg -i fglrx-driver_8.19.10-1_arch.deb dpkg -i fglrx-kernel-src_8.19.10-1_arch.deb
After you have installed the packages, you'll also need to install a suitable kernel module to enable 3D acceleration. I've prepared kernel module packages for the prebuilt kernel images currently available in the official Debian archive: if you are using one of the supported kernels, install a prebuilt package and skip to section 5. Configure X.
If you are using a custom kernel, you'll have to build the kernel module yourself;
change directory to /usr/src and unpack the module sources:
cd /usr/src tar xjf fglrx.tar.bz2
/usr/src/modules/fglrx* directory
after an upgrade of the sources package.
You can build the kernel module in several ways.
Using module-assistant is usually easiest
and is therefore recommended.
module-assistant (recommended)If you have never built modules with module-assistant before, you should prepare the build environment by running:
module-assistant prepare
In the usual environment, module-assistant should detect all requirements. Example (assuming that you build for the running kernel):
module-assistant a-i fglrx
module-assistant will unpack the sources for you, if necessary. Make sure to remove any old /usr/src/modules/fglrx directory after an upgrade of the fglrx package, because module-assistant will not do that for you.
If you prefer to build the packages with m-a from your own user account (with ~/src/linux being the kernel source):
mkdir -p ~/src/modass module-assistant -u ~/src/modass -k ~/src/linux install fglrx
Note: A detailed explanation of module-assistant is beyond the scope of this document. Please refer to the module-assistant documentation.
make-kpkg commandcd /usr/src/kernel-source-2.x.y fakeroot make-kpkg --append-to-version "-flavour" --added-modules fglrx modules_image
"-flavour" is supposed to be a placeholder. You need to replace it with the correct value for your kernel, for example something like "-1-686" for prepackaged Debian kernels. Refer to 2.1. Kernel source for more details.
This will build a new package, which you will have to install:
dpkg -i /usr/src/fglrx-kernel-2.x.y_8.19.10-1+kernelversion_arch.deb
Important:
If you build your own packages, please upgrade to debhelper 4.2.11
(see Debian bug #248624), otherwise
you will need to run update-modules yourself after installing the module.
cd /usr/src/modules/fglrx ./make.sh
Provided that the module was compiled without error, become root and install it:
mkdir /lib/modules/$(uname -r)/misc cp fglrx.ko /lib/modules/$(uname -r)/misc/ depmod -ae
make-kpkg
is just a matter of using the correct parameters.
cd /usr/src/kernel-source-2.x.y make SUBDIRS=/usr/src/modules/fglrx modules(replace /usr/src/kernel-source-2.x.y with the path to your kernel sources)
Provided that the module was compiled without error, become root and install it as shown in the previous section.
At this point, the kernel module is installed and ready for use; to make sure that it's working, you should now try to load it. You'll do this only once to check that the module was compiled and installed correctly: you won't need to load it manually every time, because the X driver will load it as needed.
rmmod radeon
Now try to load the module by entering the following command as root:
modprobe -v fglrx
If everything goes well,
modprobe prints the path of the module it loaded.
If modprobe can't load the module,
make sure you've installed it correctly,
make sure you've read the paragraph about bug #248624 above,
and look at the output of dmesg for any errors.
Please include the full error message from modprobe -v fglrx
and the output from dmesg and lsmod
if you choose to contact me at this point!
Unresolved symbols:
If you get any "unresolved symbol" errors,
look carefully at the unresolved symbol's name and try to understand
which configuration option it might be related to.
For example, unresolved symbols containing "mtrr" are obviously caused
by a missing CONFIG_MTRR=yes.
If you see errors like "unresolved symbol dev_base_R863b01ad",
with a few hexadecimal digits at the end of the symbol names,
it means that the module was compiled with CONFIG_MODVERSIONS=yes
but the running kernel was not.
In other words, the configuration of your kernel source tree
does not reflect that of the running kernel.
Go back to section 2.1,
make sure that you configured your kernel source tree correctly
and double check that CONFIG_MODVERSIONS is not set.
If you still can't load the module, please read Question 4.7: When I try to run "modprobe fglrx" it doesn't work from the Gentoo ATI Radeon FAQ before contacting me.
fglrxconfig.
Just don't use it, OK?
fglrxconfig is useful if you want a dual-head setup,
but before you try that please make sure that a single-head setup works
by editing your X server configuration file as outlined below.
Also note that fglrxconfig will overwrite your
existing X server configuration file!
Edit your /etc/X11/XF86Config-4:
Section "Module"
...
# These modules are required for 3D acceleration
Load "GLcore"
Load "glx"
Load "dri"
...
# Load "extmod" but omit DGA extension
# (the DGA extension is broken in the fglrx driver)
SubSection "extmod"
Option "omit xfree86-dga"
EndSubSection
...
EndSection
Section "Device"
Identifier "ATI"
Driver "fglrx" # this is the important bit
# If X refuses to use the screen resolution you asked for,
# uncomment this; see "Bugs and Workarounds" for details.
#Option "NoDDC"
# === Video Overlay for the Xv extension ===
Option "VideoOverlay" "on"
# === OpenGL Overlay ===
# Note: When OpenGL Overlay is enabled, Video Overlay
# will be disabled automatically
Option "OpenGLOverlay" "off"
# === Use internal AGP GART support? ===
# If OpenGL acceleration doesn't work, try using "yes" here
# and disable the kernel agpgart driver.
Option "UseInternalAGPGART" "no"
EndSection
Section "Screen"
Identifier "your screen"
Device "ATI"
Monitor "your monitor"
DefaultDepth 24
SubSection "Display"
Depth 24
Modes "1280x960" # this is only an example,
# use your preferred resolution here
EndSubSection
EndSection
Section "DRI"
Mode 0666
EndSection
Stop and restart your X server. Figuring out how to do that on your system is left as an exercise, but usually you'll have to switch to a virtual console (text mode) by pressing Ctrl+Alt+F1, then stop and restart your display manager (kdm, gdm, etc.). If you know no better than to reboot your system, you will be in a lot of trouble if things don't work as expected: you might be stuck with just the console, or things could go really wrong and you could have to reboot into single user mode because your display and keyboard are completely locked. So please do learn how to properly stop and restart the X server.
If it breaks your system, you keep all the pieces and I don't care. (But let me know what happened, OK?)
After your X server starts, check if OpenGL acceleration is enabled. Open a shell and typefglrxinfo or glxinfo:
the "OpenGL renderer string" should say something like "RADEON 9600 XT Generic";
if it says "Mesa GLX Indirect" instead, it means that there is a problem.
Full-screen mode is somewhat broken: sometimes you'll end up with a garbled display, and your only chance of getting your desktop back is blindly closing the program that caused this. I didn't find a workaround yet.
[This problem was frequent in older driver versions, but it rarely happens now.]Recent versions of the driver implement the DGA extension, but there are problems:
MPlayer for example will lock your screen if you use -vo dga(*).
Xavier says:
I just gave it a try and the result was a reboot of my computer (I tried launching xdtv). I think DGA implementation is still a bit buggy :)
Therefore, I suggest to disable the DGA extension in your X server configuration (see above).
It appears that the ATI driver does not support the DVI port on my Radeon 8500 LE: the monitor goes into standby mode as soon as X starts up. It's not a hardware problem, because I can see Linux boot in text mode on the DVI port; also, the Windows XP driver works fine (not that I would want to run XP for anything but a few games).
This is what I see in the XFree86 log with my card:
... (--) Chipset ATI R200 QL (R8500) found ... (II) fglrx(0): Primary head: Monitor -- NONE Connector -- None DAC Type -- Unknown TMDS Type -- NONE DDC Type -- NONE (II) fglrx(0): Secondary head: Monitor -- NONE Connector -- VGA DAC Type -- Primary TMDS Type -- NONE DDC Type -- VGA_DDC
Several people told me that they are happily using DVI on different cards, so I borrowed a 9600 XT from a friend and DVI worked out of the box, without even having to adjust the XFree86 configuration. With the borrowed 9600 XT, the primary head says something along the lines of "Connector: DVI-I", "TMDS Type: Internal".
Someone also suggested that I try with
Option "MonitorLayout" "TMDS, AUTO"in the XFree86 config file, but this doesn't make any difference.
Currently, I don't know of anybody who can use the DVI port on an 8500 with the fglrx driver. If you do, please share your experience.
Update: As of April 2005, I've installed Ubuntu's X.Org 6.8.2-10 packages, and I was curious to see if my DVI port works with the X.Org “radeon” driver. At first I got a blank screen again, but with the MonitorLayout option, the DVI port does finally work.
Carsten wrote about his experience on a IBM T41p laptop:
I've installed the latest package from you (8.10.19-1) and it works fine apart from the DVI as you mentioned on your website. I'm using a IBM T41p with a ATI Technologies Inc M10 NT [FireGL Mobility T2] (rev 80). Apart from that the driver even works together with radeonfb used for the terminals. Also suspend to RAM seems to work fine. After wakeup X and terminals come up cleanly again. Thus, I would use this driver always if there weren't this DVI problem ... until then I'll work with the standard radeon driver from X.
Jack reports that he's seeing lots of static at 1600x1200 on his digital display. He says that "this problem at 1600x1200 is a known issue due to approaching the video bandwidth limit of the hardware", because "the TMDS standard for DVI is fixed at 165 MHz which means that you are pushing things to get 1600x1200 at 60 Hz". He researched the issue and he found this article from ATI: Horizontal lines or screen flicker within Windows when using a DVI flat panel.
He further says: "The only fixes are to 1) reduce the refresh rate at higher resolutions to keep the pclk away from the upper limit of 165 MHz or 2) replace the DVI connection with an analog one."
It appears that the ATI driver will refuse any screen size that your monitor doesn't list in its VESA DDC info; for example, I can't get my preferred resolution of 1280x960(*) with my Samsung CRT monitor. The symptoms are the following: if multiple resolutions are listed in XF86Config (such as "1280x960" "1152x864" "1024x768" "800x600"), the display will open at 1024x768 with a noticeable delay (a couple of seconds).
There are two workarounds:Option "NoDDC" in the "Device" section,
as shown above:
this is harmless as long as you got your monitor timings right.
Kernel 2.6.5's "make clean" removes include/linx/version.h
and several other files that are required to build external modules.
A workaround is the following:
$ make-kpkg ...options... kernel_image $ make-kpkg clean $ make-kpkg ...options... configure $ make-kpkg ...options... modules_image(i.e do an extra "make-kpkg configure" before "make-kpkg modules_image")
Another workaround is to apply a patch that restores a sane behavior (from the Linux Kernel mailing list).
Do not install kernel-package 8.086 that "fixes" this problem by not doing a make clean: it's worse because the kernel makefile will now relink all kernel modules during the build of external modules.
CONFIG_REGPARMCONFIG_REGPARM
because it changes the kernel ABI.
Starting with version 3.7.6-5, I have applied
this patch
for the fglrx kernel module that should fix this problem.
Upstream version 3.9.0 should have fixed this problem,
according to the release notes,
but users report that it is still broken.
Starting with version 3.9.0-3, I have applied yet another patch,
kindly supplied by Alexandru Fomin, that hopefully fixes it once and for all.
Upstream versions 3.11.1 and later have this problem fixed properly.
Kevin Schlichter reports: "I'm running an SiS chipset, and found that the AGP support doesn't work in the 2.4 series, and not in the 2.6 series until 2.6.5. I think there are some patches available, but I didn't test any of them."
If your card is not supported, you can force the driver to recognize it
as a different card by overriding the ChipID.
Look up the PCI ID of your card with lspci -n, then search for
a similar model and ID in the list of supported chipsets.
Put the new ID in the X config file's Device section:
Section "Device" Identifier "ATI" Driver "fglrx" ChipID 0x1234 # replace 1234 with the ID you want to use [more options as needed] EndSection
For example, the unsupported Mobility Radeon 9600 with PCI ID 0x4E52 works fine if you tell the driver to use it as if it was PCI ID 0x4E50. Thomas Wollner wrote: "my card is an Mobility Radeon 9600 M10 with pci-id 4e52 and is not found by the driver. this card is also known under pci-id 4e50. how can i tell the fglrx driver to see the 4e52 card as an 4e50?"
I told him to try with ChipID 0x4E50 in the XFree86 config file's Device section;
he answered that "it worked!!!" and provided the following excerpt from the XFree86 log file:
(**) ChipID override: 0x4E50 (**) Chipset MOBILITY RADEON 9600/9700 (M10/M11 4E50) found
Several people have confirmed that the ChipID override works for them;
the most common unrecognized chips are those with PCI IDs 4e51 and 4e52
(they are integrated chips only found on laptops, as far as I know)
and they work fine with ChipID 0x4E50.
Debian is a “pure” amd64 system, so
64-bit libs go into /.../lib and
32-bit libs go into /emul/ia32-linux/.../lib.
This means that the fglrx driver requires a link
from /usr/X11R6/lib64 to lib
because it is compiled for a “hybrid”(???) system
that puts 64-bit libraries into /.../lib64.
My packages currently don't create such a link automatically because I'm not sure it doesn't break or cause problem to something else. Also, I think this link is currently being created by the Debian installer (or more likely by some base package like base-files), so you might not need to create it yourself. If this link doesn't exist on your system, you can create it with the following command:
ln -s lib /usr/X11R6/lib64
This link is required for 3D acceleration to work.
By the way, several people wrote to ask where the link should be created.
The answer is to read ln's info page and learn about symlinks.
Enter the command exactly as shown and the link will be in the correct place,
no matter which directory you are in.
Trust me.
I maintain a list of setups that are reported to work fine. Please send me a note if your configuration differs significantly from those already listed.
| CONFIG | introduced | status | notes |
|---|---|---|---|
| HIGHMEM4G | ? | works | |
| PREEMPT | ? | works | |
| REGPARM | 2.6.5? | works | requires patch (already included in my packages) |
| 4KSTACKS | 2.6.6 | works | |
| 1GLOWMEM | CK patch | works | Con Kolivas' 1g_lowmem_i386 patch |
Unless otherwise noted, the driver is known to work with any combination of these options enabled and of course with all of them disabled.
I'm currently running with 1GLOWMEM, PREEMPT and 4KSTACKS. I used to use HIGHMEM4G, but 1GLOWMEM is better if you have “only” 1G of RAM.
Contrary to what I had previously written here,
the ATI driver does support the Xv extension,
but you need to have
Option "VideoOverlay" "on"
in your X server configuration (see above).
Roberto JP says: "There are a few more advantages, from a developer's viewpoint. The ATI driver supports many more (important) OpenGL extensions than the open source driver does. If you're using vertex or fragment programs (i.e pixel shaders) in your software, this is very important."
Once upon a time, there was a mailing list for discussion about these packages. Now that official Debian packages are available, these packages have become largely obsolete and the mailing list is no longer useful.
An archive of the mailing list is still available for historical purposes.
Click here to learn how this package was born.
Versions 3.2.5 and later officially support the Radeon products again. These products were initially listed as supported in the documentation for version 2.9.6 of the driver, but later on all references to them were removed, leaving only the FireGL series as officially supported. However, the driver itself was not changed and later versions (up to and including version 2.9.13, which was the latest version available from ATI until 3.2.5 came along) still supported the consumer products too. I still can't understand why it was so, but now support for the Radeons is officially back and everyone should be happy.
So you now have OpenGL working and you want a game that puts it to good use? That would be Doom 3. The Linux demo is freely available.
Several id Software games have a Linux version and/or demo. Quake 2 is still one of the best games ever; unfortunately, the Linux version is not available from id Software's site, but there is a package in Debian.
The Wolfenstein derivatives are also quite good. Enemy Territory is particularly addictive and it also happens to be free, but you'll need to find some friends to play with because it is multiplayer only.
Tux Racer is one of the cutest games around. And now, PlanetPenguin Racer is delivering even more challenging tracks.
If you are interested in the latest, greatest and probably broken package source, you can get it from my CVS repository here:
$ cvs -d:pserver:anonymous@stanchina.net:/home/cvs/fglrx login CVS password: [enter empty password here] $ cvs -d:pserver:anonymous@stanchina.net:/home/cvs/fglrx co fglrx-installer
I moved the source for the official Debian packages from CVS to git. Please see the official Debian packages page for the details.
Several users are translating this page into various languages. Here are a few “official” guidelines:
This is the language translation of Flavio's ATI Linux driver packages for Debian. Please refer to the original page for informations on the latest version of the driver.And the "Working setups" section (in your language, of course):
Flavio maintains a list of working setups on his pages. Please have a look there.
Copyright © Flavio Stanchina
July 22, 2009 at 00:48:28