mattfury: hi i have radeon m9 (mobility) and i cant play many games for very long
mattfury: (on windows)
wirry: try cleaning the fan
mattfury: it glitches after a while: temperature issue you think?
mekius: sounds like a question for someone else :P
mattfury: i just want to play WoW on my laptop...
wirry: yup, i guess temperature
mattfury: ran well for like 5 minutes then gave up...
mattfury: im even running the omega drivers so i hope thats helpful
MostAwesomeDude: mattfury: This channel is for the open-source drivers...
mjt: folks... i need some.. unusual help. 2.6.33, radeon 6.12.99, opengl libs. The issue at hand is that when running some X app from within a chroot which has older X libs and stuff, screen immediately goes blank (in text mode) and machine freezes. It does not happen when i run older X client from remote. It smells like a security issue...
mjt: kms
airlied: and server backtrace?
mjt: i tried the whole night today to get something...
mjt: the machine is dead after tht
mjt: that
mjt: even netconsole is dead
mjt: and i can't reproduce it with 2.6.32
mjt: ..another crash, another reboot
airlied: so everything else the same, 2.6.32 works and 2.6.33 fails?
mjt: . o O { bisect }
mjt: ;)
mjt: yes
mjt: but there's something.. unusual here, which might be of interest. I've a 64bit kernel and 32bit regular userspace which works fine. It's the 64bit chroot that fails.
mjt: in there (in the chroot) the libs are old
airlied: whats the gpu?
mjt: built-in 780g
mjt: hm. Lemme try to collect some strace with O_SYNC
airlied: before going into a bisect might be worth using the 6.12.191 DDX just so we know its up to date
mjt: ..another reboot
mjt: but i just realized... it's what one gets by trying to debug something all night long...
mjt: the issue has exactly _nothing_ to do with radeon or kms
mjt: and my last freeze just confirmed that
mjt: it's kvm, not radeon
mjt: which i used for testing as it was the only X app in the chroot
mjt: almost-fresh mind (i just awake after ~1.5 hours of sleep) and i now finally can see the huge idiocy..
mjt: it's still something to do with kms/radeon
mjt: but not about X
mjt: sorry for the noize.
dileX: mjt: latest ddx is in d/exp (if you wanna test)
mjt: oh, that's something too large ;)
mjt: it pulls tons of other dependencies
mjt: i tried installing debian packages, not once, and each time ended up compiling from sources
dileX: then rebuild in your env
mjt: even that does not work
mjt: due to explicit dependencies
mjt: i should switch to squeeze
mjt: ok
mjt: one more question about this thing I had overnight. Yes it's kvm, which - apparently - tries to create sdl window on console (when no X access is available). But once it tries that, screen goes blank and nothing more happens.
mjt: kvm itself fails, but the thing is in this mode still
mjt: the system is running fine, i can restart gdm for example (when logged in from remote)
mjt: but screen does not show up
mjt: there's _nothing_ in dmesg
mjt: but in Xorg.log i see (retyping, slowly):
edwin: mjt: fwiw I had similar experience with a 32-bit chroot (when giving access to /dev/dri/), everything just locked up
mjt: Fatal server error: xf86OpenConsole: VT_WACTIVATE: failed: Interrupted system call
mjt: edwin: it has nothing to do with chroot really. The "chroot thing" was 100% due to my sleepy brain
edwin: ah ok
mjt: it behaves the same with my regular userspace
mjt: so.. is there any good in trying to debug this version, when an SDL app effectively removes display?
mjt: (rebooting...)
Singularity: Does the radeon driver support 5770 to the extent that I can use 1680x1050 with my monitor (the vesa driver only supports 1280x1024)?
mjt: radeon: You have old & broken userspace please consider updating mesa & xf86-video-ati
mjt: ;)
mjt: dileX: re ddx in exp: but is there gl library?
jcristau: yes, in sid.
mjt: aha, got it. libgl1-mesa-dri
mjt: i were looking at something like libgl1-mesa-radeon-dri
mjt: and this time it worked flawlessly
mjt: 6.12.191
dafox: agd5f, soreau: are you there?
spreeuw: hey, should the doom3 demo work on r600?
spreeuw: it runs but like quake4 it lacks s3tc
alpha_one_x86: Hello, I have strange beviour, without firmware R700_rcl.bin it say it, in kms (console) my screen is down, under kdm it's up, with the firmware both console/kdm my screen is down
alpha_one_x86: I use vga
alpha_one_x86: I have remote controle in ssh on it, if you have question
BioTube: alpha_one_x86: do you have the fbcon module loaded?
alpha_one_x86: In xorg.0.log in have: Timeout trying to update memory controller settings
dileX: edwin: (thinking wiki) now there is 7.8 branch - master got a version bump to 7.9-devel.
alpha_one_x86: no
alpha_one_x86: no fbcon module loaded
glisse: alpha_one_x86: your ddx is likely not kms enabled
alpha_one_x86: How enbale it?
glisse: you need to build ddx from git
alpha_one_x86: I have do that's with overlay gentoo
glisse: something went wrong as you should see the mc message in kms case
alpha_one_x86: what do?
edwin: dileX: should I update to say origin/7.8 instead of origin/mesa_7_5_branch?
dileX: edwin: yeah.
dileX: http://marc.info/?l=mesa3d-dev&m=126783423609212&w=2
alpha_one_x86: ./configure ...... --enable-dri --enable-kms
dileX: "My current plan is to have release candidates for *both* 7.7.1 and 7.8 as follows: * RC1: 3/12 * RC2: 3/19 * final: 3/26"
edwin: dileX: version string stil says 7.8-devel
alpha_one_x86: according to this guide: http://en.gentoo-wiki.com/wiki/Radeon I have setup the git libdrm xf86-video-ati mesa
dileX: alpha_one_x86: see topic
dileX: edwin: that will change with rcX available
edwin: k
alpha_one_x86: dileX: wich topic? have you a url?
dileX: alpha_one_x86: each channel has a topic
dileX: (with general and important notes)
alpha_one_x86: Ah... I don't know well irc, ... how go to this topic?
edwin: dileX: done
BioTube: type in '/topic'
edwin: dileX: http://wiki.x.org/wiki/radeonBuildHowTo?action=diff
edwin: btw any plans on comitting that patch that enables pbo?
edwin: some games won't start without that
alpha_one_x86: fimrware installed and verified, all recompiled...
dileX: edwin: I combine the git clone and checkout parts in mesa section - they belong together
edwin: dileX: ok, then please move the text explaining which branch to choose
edwin: before the clone
dileX: yupp
agd5f: spstarr: you likely have some overclock modes or something that prevents the driver from adding all the power modes
agd5f: the driver only prints the modes that were added, not all the power modes availabl
alpha_one_x86: News: in dvi after reboot kms not work but the ddx drivers yes
dafox: agd5f: hi, I managed to find out differences in 6.12-branch/ums vs master/kms wrt tv-out/PAL
dafox: agd5f: I tracked it to 3 TV_CNTL registers
agd5f: dafox: ok
agd5f: can you send me the list?
dafox: setting those to the values from 6.12 branch fixes it
dafox: http://bugs.freedesktop.org/show_bug.cgi?id=26915
dafox: I posted the results on my bugreport
agd5f: excellent!
da1l6: hi
dafox: thanks. But what happens next? Is there something else I can do too help?
da1l6: With Kernel 2.6.33 KMS i get image distortion on the VGA port, the lines are "wobbling" from left to right and back.
da1l6: This wasn't the case with 2.6.32 KMS. DVI and LVDS work fine. Using UMS also makes VGA work correctly.
da1l6: The card is a M76 [Radeon Mobility HD 2600 Series].
agd5f: da1l6: filed a bug yet?
da1l6: not yet
dileX: edwin: should be more comprehensive now
da1l6: agd5f: any specific info that needs to be attachedworms2?
alpha_one_x86: My card is RV730XT and I use 2.6.33 kernel, kms turn off my screen in dvi, hdmi, vga
da1l6: argh
agd5f: da1l6: xorg log and dmesg
edwin: dileX: looks good
edwin: dileX: except the git URL is wrong
da1l6: wtf? the "forget my password" link requires me to enter my password...
edwin: dileX: hmm, both the old and new URLs work. how comes?
dileX: edwin: I got the one from Vcs-Browser URL
dileX: Clone
dileX: git://anongit.freedesktop.org/mesa/mesa
dileX: ssh://git.freedesktop.org/git/mesa/mesa
edwin: yeah both work
edwin: I'll switch my local copy to use the new url
dileX: so this is the official one - we take that
edwin: yeah
agd5f: da1l6: does 'avivotool regset 0x458 0x00000000' or 'avivotool regset 0x45c 0x00000000' fix the problem?
da1l6: agd5f: where do i get avivotool?
agd5f: da1l6: your distro may provide packages (may be included in radeontool), otherwise: http://cgit.freedesktop.org/~airlied/radeontool
agd5f: dafox: I attached a patch to your bug. can you verify it fixes the issue?
dafox: agd5f: I'll try it right away
dafox: agd5f: yes, that fixes it for me :D
agd5f: dafox: excellent!
dafox: agd5f: thanks for the quick fix. I'll report this to the bug too so you may close it
da1l6: agd5f: thanks, avivotool regset 0x458 0x00000000 immediately fixed it. :)
da1l6: OLD: 0x458 (0458) 0x2a020101 (704774401)
da1l6: NEW: 0x458 (0458) 0x00000000 (0)
da1l6: May i ask what i just did? ;)
Enverex: "direct rendering: Yes" and "OpenGL renderer string: Mesa DRI R600 (RV770 9442) 20090101 TCL" mean that direct rendering and 3D acceleration are enabled don't they?
da1l6: Enverex: yes
dafox: agd5f: Done. Btw I noticed while testing that there appear to be some other issues with the tvout, which I'll look into reproducing later and maybe report some more bugs
Enverex: Hrm, oddly enough Wine complains that it is disabled - "err:winediag:X11DRV_WineGL_InitOpenglInfo Direct rendering is disabled, most likely your OpenGL drivers haven't been installed correctly"
agd5f: dafox: sounds good
da1l6: Enverex: do native 3d apps work?
Enverex: da1l6, I'll find one or two to install
Enverex: da1l6, Any good examples that aren't huge?
da1l6: glxgears?
dafox: agd5f: what exactly does your patch do? It's a small change but it seems makes a lot of difference in other ways also. For example choosing tvstandard pal60 caused the tv screen to go nuts, but it seems to work now, although I'm not convinced it's really 60hz, the image is good.
agd5f: dafox: this image isn't 60 hz. for tv-out the connector always outputs the native tv timing, pal, ntsc, etc.
agd5f: the mode is just a place holder
agd5f: the timing is native to the tv standard and the image is scaled down to the native tv size
dafox: agd5f: I thought so, but the funny thing is that when I first tried pal60 a few days ago it caused to tv loose picture(turned black) and the OSD to scramble, but now it just keeps the old mode
edwin: [ 5888.176661] [drm:radeon_info_ioctl] *ERROR* copy_to_user
edwin: I have that 3 times in my dmesg
edwin: unfortunately I only noticed now, is that something I should worry about?
paxcoder: hey people. i'm using "rotate" option in "Display" GUI for setting up resolution in GNOME. I rather liked it (I've rotated my LCD screen of course), but it takes a bit of time to draw. Why is that - and can it be remedied?
agd5f: paxcoder: using dualhead?
paxcoder: no
agd5f: paxcoder: what hw?
paxcoder: plain old D-sub attached to an LCD (onboard Radeon)
paxcoder: no 3D support - haven't got installed proprietary "firmware"
agd5f: paxcoder: what chip?
paxcoder: um.. eh Xsomething. how to check?
paxcoder: X1200 or alike
agd5f: paxcoder: you need acceleration for rotation
paxcoder: well that sucks
paxcoder: why?
soreau: paxcoder: Why shouldn't you have acceleration? :)
agd5f: the 3d engine is used to accelerat the rotation
soreau: If you pastebin your X log, maybe we can figure out why you don't have acceleration
agd5f: soreau: he doesn't have the ucode installed
paxcoder: agd5f: well that's pretty inconvenient now isn't it?
soreau: agd5f: Oh sorry, that's a bit over my head
agd5f: paxcoder: install the ucode if you want acceleration
paxcoder: I don't. I want only rotation :-P
paxcoder: ucode ftl
soreau: Well you know you need acceleration for rotation, so..
Enverex: da1l6, "16934 frames in 5.0 seconds = 3386.738 FPS" but to be honest I have no idea what I should be getting
soreau: Enverex: glxgears is not a benchmark
soreau: Enverex: The problem is the wine is heavily geared toward nvidia drivers
Enverex: soreau,
BioTube: Enverex: are you running an AMD64 system?
soreau: glxgears still isnt a benchmark
Enverex: BioTube, Yes
BioTube: make sure you have an IA32 version of the mesa driver installed
paxcoder: 3386.738 FPS <- wth, are you rendering 16x16 icons?
_KAMI_: Hey Developers
_KAMI_: Thank you for the great work
BioTube: paxcoder: glxgears is CPU bound
_KAMI_: I am very happy with Ubuntu 10.04 Lucid/latest PPA DDX/MESA/DRM drivers
_KAMI_: It works well
Enverex: BioTube, Although on my nVidia card I got 26000 FPS on the same CPU...
_KAMI_: I also can USE HDMI with audio
Enverex: BioTube, I'll make sure it's installed
paxcoder: BioTube: it's an easy test I guess, I got 14.949 on my rotated non-3D accelerated display :-P
_KAMI_: on my mobility radeon 3470
_KAMI_: cool
_KAMI_: If I can help with testing I am glad to halp us
paxcoder: there is no "us" without "u"
paxcoder: that's about the corniest thing I thought of whole week.
Enverex: BioTube, Is there a 32bit version of Mesa in the Debian repos by default? If so do you know what it's called? I'm not seeing any 32bit reference to it
BioTube: Enverex: I don't use AMD64
soreau: _KAMI_: Great to hear hdmi audio is working
_KAMI_: soreau: hello
paxcoder: Enverex: what Debian?
_KAMI_: You helped a lot :D
Enverex: paxcoder, Sid
AstralStorm: hello
_KAMI_: thank you so much
_KAMI_: and all of our developers
_KAMI_: If I can provide
_KAMI_: logs
_KAMI_: test
AstralStorm: my card is throwing nice WARNINGS
AstralStorm: Mar 6 16:49:19 destiny kernel: [15028.584621] Querying an unemited fence : ffff880122f50840 !
_KAMI_: or any other resources
_KAMI_: I am glad to help
AstralStorm: rare
_KAMI_: :D
AstralStorm: rv670
_KAMI_: I am using 2.6.33 drmnext kernel from ubuntu kernel ppa
paxcoder: Enverex: that has it. your repos only pull stuff for your architecture.
Enverex: paxcoder, True to an extent. There are lots of lib32/ia32 compatibility packages listed, just nothing mentioning mesa
AstralStorm: oh right
AstralStorm: these are just before the WARNING:
AstralStorm: Mar 6 16:47:39 destiny kernel: [14928.160957] radeon 0000:01:00.0: z/stencil wi
AstralStorm: th no depth buffer
AstralStorm: Mar 6 16:47:39 destiny kernel: [14928.160973] radeon 0000:01:00.0: r600_packet3
AstralStorm: _check:1055 invalid cmd stream 142
AstralStorm: any quick ideas what's wrong? (this is drm-radeon-testing I think, not sure though)
paxcoder: Enverex: this one liner usually does the trick for me: apt-get -t experimental install libgl1-mesa-dri xserver-xorg-video-radeon
soreau: AstralStorm: How do you not know which kernel you're running?
paxcoder: Enverex: try without "-t experimental" first, if it doesn't work, you'll need pinning
Enverex: paxcoder, Already running them
Enverex: (the experimental ones I mean)
paxcoder: Enverex: what's the prob I don't get it.
AstralStorm: soreau: I know which, but I'm not sure which experimental branch is in it
AstralStorm: have to run qgit
AstralStorm: unless you have another quick way to check that :)
Enverex: paxcoder, Wine claims I don't have Direct Rendering enabled...
soreau: AstralStorm: What does uname -r say?
paxcoder: Enverex: you have proprietary blobs/ucode/"firmware" ?
_KAMI_: soreau: where can I read more about
AstralStorm: 2.6.33-zen0, which doesn't tell me much
Enverex: paxcoder, Yup
AstralStorm: it's definitely 2.6.33 or 2.6.33 + something
_KAMI_: the possible xorg configure things
AstralStorm: which "fix drm" commit spells out clearly to me
_KAMI_: is it possible to tune the performance
_KAMI_: and energy saving options?
paxcoder: Enverex: so basically, you have 3D just not in Wine?
soreau: AstralStorm: Perhaps you should know what you're working with before reporting problems with it
agd5f: da1l6: does this patch fix the issue? http://people.freedesktop.org/~agd5f/0001-drm-radeon-kms-atom-spread-spectrum-fix.patch
AstralStorm: soreau: ok, found it
AstralStorm: it's drm-radeon-testing
Enverex: paxcoder, Seems so
soreau: Enverex: Does wine say what gl renderer string it's detecting? or does it just fail? I still think this is a wine problem and not with the driver
AstralStorm: soreau: commit 45fa67f404bb08ef4b04504766753c28e354ecb2
AstralStorm: merged onto 2.6.33
Enverex: soreau, All it says is what I pasted earlier, that it's simply not enabled (and isn't any more specific)
agd5f: AstralStorm: ddx bug already fixed. try xf86-video-ati git amster or 6.12.191
paxcoder: Enverex: well why don't you ask in #winehq ?
Enverex: paxcoder, What card are you using?
AstralStorm: that's ddx? thanks, upgrading now
soreau: Enverex: Have you tried asking in the wine channel?
AstralStorm: :)
Enverex: paxcoder, Because I'm the only one insane enough to bother using Wine with the XOrg Radeon driver
soreau: Enverex: Since it is likely a wine issue, we can't do much to help you
Enverex: soreau, Didn't think so but thought it was worth asking anyway
da1l6: agd5f: here is a 2 in 1 bug report which seems to include the issue i have https://bugs.freedesktop.org/show_bug.cgi?id=25741
paxcoder: Enverex: nah, I ran stuff on Wine too, like CS. Before I found out that I was using firmware blobs (later were removed from main). But my card is X1200 or something like that. So I can't compare to your game selection.
soreau: Enverex: There are many people who like to run 3D apps in wine so if you can get them to fix their program, you'd be doing everyone a favor
da1l6: agd5f: will try to test the patch, its against the kernel, i guess.
agd5f: da1l6: yes
Enverex: paxcoder, Actually I'm not running anything high end, just basics like Deus Ex, Morrowind and some of those 64KB demos (for testing)
paxcoder: oh
Enverex: soreau, I've done a lot of work in Wine before, problem was it was always with the binary blob drivers (nvidia or fglrx). I'll have to poll the devs again and see who's running what these days. I've been out of the project for ages due to lack of any free time
Thunderbird: hi Enverex ;)
Enverex: Thunderbird, Talk of the devil. What are you running these days? (and yes, I'm still going out of my mind trying to find a new laptop)
paxcoder: unrotates
Thunderbird: Enverex, in my laptop I have a radeon
Thunderbird: to all people who don't have 3d in wine if they are on 64-bit, really make sure your 32-bit opengl is correct
Thunderbird: wine should show a warning about it on (>1.1.36) saying that direct rendering isn't enabled
Thunderbird: or install a 32-bit glxinfo and expect what it shows
AstralStorm: yeah, I'm having issues with that too, thanks to Arch being non-multilib
Enverex: Thunderbird, That's what I'm running into
AstralStorm: or rather, not having a multilib compiler
AstralStorm: so that mesa is older
Thunderbird: further WINEDEBUG=+wgl wine yourapp.exe should also some some basic glxinfo-like output
Thunderbird: if the app you are running is a 3d app
AstralStorm: if it's not, also
AstralStorm: because wine uses opengl for displaying 2D surfaces too
Thunderbird: only in some cases if you set ddraw to use opengl
AstralStorm: which is the default
AstralStorm: :)
Thunderbird: no it isn't
AstralStorm: ? oh really? so what is?
AstralStorm: (and why I couldn't run anything with broken mesa there)
Thunderbird: by default we use gdi
Thunderbird: wined3d barely works without opengl
da1l6: agd5f: can't test the patch right now, will report back as soon as i have the time (propably tomorrow). Many thanks for your help.
Thunderbird: in theory it should work without but it tends to break quickly
AstralStorm: the app in question is directx 8 2D
AstralStorm: no d3d
Thunderbird: there is no directdraw 8
AstralStorm: yeah, that's directdraw 7
Thunderbird: so if it is 2d and uses directx8, it is a 3d app
AstralStorm: no, it needs new directsound ;p
AstralStorm: funny how devs can't work with a step older one. it's all convenience
dileX: edwin: by following above discussion - which pkgs did you need to compile libdrm/mesa 32-bit on 64-bit-system? {binutils,gcc}-multilib? can you add reqs to wiki? thx in advance.
AstralStorm: dileX: 32-bit glibc, 32-bit gcc
AstralStorm: 32-bit binutils
AstralStorm: alternatively, multilib enabled all of the above
AstralStorm: oh, and obviously X libs
AstralStorm: all the deps.
AstralStorm: it's usually easier just to have a 32-bit chroot
AstralStorm: although less convenient to run apps
AstralStorm: (have to add a bunch of bind mounts for instance)
AstralStorm: the most important bind mount being your home dir
AstralStorm: second, /tmp
AstralStorm: (and the stuff like /proc /sys, blah blah read a guide to chroots)
dileX: AstralStorm: background was to run gl-apps in 32bit wine
dileX: (see wiki)
dileX: http://wiki.x.org/wiki/radeonBuildHowTo#Troubleshooting32-bitapplicationson64-bitsystems
AstralStorm: dileX: ... obviously, you need 32-bit mesa and stuff
AstralStorm: dileX: build "foo from git" won't work if you don't have 32-bit prerequisites
AstralStorm: which they conveniently omitted
AstralStorm: even if you happen to have multilib toolchain (that's what required for -m32 to work)
dileX: AstralStorm: so its edwin who successfully compiled, tested and updated the wiki. I am only asking for clarification. no 64-bit-system here
AstralStorm: he can't use ./configure
AstralStorm: for reasons unknown to man
AstralStorm: I mean for mesa, what's the deal with the variables
AstralStorm: doesn't look like convenience to me
AstralStorm: I don't see the part where he adds /opt/xorg/lib32/dri to ld.so.conf
dileX: might missing
AstralStorm: it just happens on debian that it's there
dileX: edwin: /etc/ld.so.conf.d/mesa32.conf?
evil_core: hi all
Nille02: Thunderbird: if you enter linux32 glxinfo you get the 32bit glxinfo output but unless this wine print the warning that direct rendering is off every time on 64bit for me ;)
Thunderbird: if wine prints the warning, something like is still wrong though
Thunderbird: it means we really detect indirect rendering
Thunderbird: the check also makes sure that the X11 server is 'local', so it won't appear if you are working remote
Nille02: ah ok
Thunderbird: this remote check works by checking whether the display socket is a unix socket or not
PinkFreud: I appear to have borked my Xorg setup when attempting to get KMS to work. I have a dual displays, configured to be one large desktop via Xrandr. I now see part of one desktop overlaid on the other. The output of xrandr shows 'Screen 0: minimum 320 x 200, current 2560 x 1024, maximum 1680 x 1680'
PinkFreud: ... in other words - help!
PinkFreud: I suspect the desktop overlay issue is related to the maximum being smaller than the current desktop. I've no clue why such a ridiculous resolution is now the maximum, though. 2560x1024 has been working just fine up until I started screwing with KMS.
evil_core: vsync is still broken in mesa?
edwin: dileX, AstralStorm: indeed I added /opt/xorg/lib32 to my /etc/ld.so.conf./a-local-xorg.conf, similar to how I added /opt/xorg/lib as suggested earlier by wiki.
evil_core: edwin: it sometimes doesnt work, for me i.e. now
evil_core: edwin: change LD_LIBRARY_PATH in /etc/environment
evil_core: master is still 2.6.32 based?
evil_core: or 2.6.33 now?
edwin: dileX, AstralStorm: as for required packages, I have ia32-libs (but thats only runtime packages), the usual for building 32-bit stuff (gcc,binutils,libc)
edwin: evil_core: I don't need LD_LIBRARY_PATH, if its in /etc/ld.so.conf.d
edwin: AND I properly compiled mesa
edwin: i.e. with --enable-glx-tls on Debian
evil_core: I too
evil_core: but it stopped working, dyunno why
edwin: dileX, AstralStorm: and I have the usual Xorg includes, the 32-bit X libs are in ia32-libs already
edwin: evil_core: what does ldconfig -p |grep libGL.so show?
evil_core: show us ldd /usr/bin/glxgears ;)
edwin: ldd /usr/bin/glxgears
edwin: linux-vdso.so.1 => (0x00007fff9f53c000)
edwin: libGL.so.1 => /opt/xorg/lib/libGL.so.1 (0x00007f96c22d0000)
edwin: ....
evil_core: http://pld.pastebin.com
edwin: http://pld.pastebin.com/K7ggVZgJ
evil_core: and LD_LIBRARY_PATH is not set?
edwin: nope
dileX: edwin: pls name the pkg(-reqs ): apt-get install pkg1 pkg2... Put a code-snippet for your ld-conf-file.
PinkFreud: uhgh. anyone? why would a modern radeon driver, running on modern radeon hd hw, declare the maximum screen size to be 1680x1680?
evil_core: http://wklej.org/id/291458/
edwin: evil_core: that looks good, is that with LD_LIBRARY_PATH set?
evil_core: # ldd /usr/bin/glxgears | grep libGL
evil_core: libGL.so.1 => /usr/lib64/libGL.so.1 (0x00007f4c00b98000)
evil_core: no, but it doesnt work
dileX: edwin: why did you put libdrm to /usr/lib32 - is also /opt/xorg/lib32 possible with adding $LIBDRM32_DIR to the configure-line of mesa32?
edwin: dileX: because /usr/lib32 already has the libdrm from ia32-libs...
edwin: which is ancient
dileX: edwin: so you are overwriting an existing lib?
edwin: yeah, I haven't tried another solution
edwin: I'll have to try if /opt/xorg/lib32 works
edwin: after I put back the old libdrm
evil_core: but wshy its not working for me?
evil_core: # cat /etc/ld.so.conf.d/a-local-xorg.conf
evil_core: /opt/xorg/lib64
evil_core: /opt/xorg/lib
edwin: evil_core: you've run ldconfig, right?
evil_core: yes
edwin: can you paste eu-readelf -a /usr/bin/glxgears?
edwin: (or readelf -a /usr/bin/glxgears)
evil_core: http://wklej.org/id/291464/
dileX: edwin: http://wiki.x.org/wiki/radeonBuildHowTo?action=diff&rev2=107&rev1=106
evil_core: dileX: it didnt helped me ;)
PinkFreud: gives up, declares radeon to be broken, goes back to the (working!) radeonhd.
evil_core: PinkFreud: what is broken?
PinkFreud: yep, no more ridiculous 1680x1680 limitation in radeonhd. When did radeon break?
evil_core: Automatic merge failed; fix conflicts and then commit the result.
evil_core: 320x240 should be enough for all!
PinkFreud: evil_core: Xrandr, in the radeon driver, has suddenly decreed that my maximum resolution will be no larger than 1680x1680
edwin: evil_core: did you reboot since adding the new path into /etc/ld.so.conf.d/*?
PinkFreud: this, of course, leads to a rather useless desktop. I'm using a pair of monitors configured to use 1280x1024 each.
PinkFreud: so, somewhere along the line, the radeon driver has broken. I was previously using 6.12.2. Apparently, in 6.12.5, Xrandr no longer works.
PinkFreud: I would have thought that to be a large enough bug for someone to notice, but I guess I'm wrong. :)
evil_core: yes, on 64bit system I got many hangups due to broken Mesa/drm
evil_core: how to update to master mesa?
PinkFreud: and, of course, the 6.12.2 package is no longer available. screw it. I'm sticking with radeonhd for the moment, it appears, until the radeon devs decide to fix xrandr.
evil_core: git branch
evil_core: * drm-radeon-testing
evil_core: master
evil_core: I have broken something
AstralStorm: no, that's probably not mesa
AstralStorm: there's no drm-radeon-testing branch there
AstralStorm: :)
evil_core: its drm
bridgman: PinkFreud; have you set a Virtual line in xorg.conf ?
edwin: evil_core: ldd -v /usr/bin/glxgears ?
PinkFreud: bridgman: I have not.
bridgman: I think you're just seeing different default behaviour in the two drivers
bridgman: figure out what maximum you want and add a Virtual line to xorg.conf in the device section
evil_core: http://wklej.org/id/291480/
PinkFreud: er, why would that be? The old driver was able to figure it out on it's own.
bridgman: are you sure ? I thought radeon always based the defaults on the first screen
edwin: evil_core: and ldd -v /opt/xorg/lib/libGL.so.1?
PinkFreud: bridgman: sure, and the first screen is 1280x1024.
evil_core: http://wklej.org/id/291484/
bridgman: goes back and looks at the logs again
PinkFreud: the driver appeared able to figure out that, if it used the same settings for the second display, it'd come up with a desktop of 2560x1024, which was correct.
bridgman: what desktop size do you want ? just add a virtual line and see if that works
edwin: evil_core: prelinking!
edwin: evil_core: try re-prelinking
evil_core: glxgears?
edwin: yeah
edwin: if it got prelinked (by a cronjob)
evil_core: stupid idea
edwin: then it'll keep using the old lib
evil_core: its from system package
edwin: well do you use prelinking on the system?
evil_core: but Icannto relink tc-elite and many other games
edwin: nono, not relink
evil_core: probably not
edwin: prelink is somthing that runs from a cronjob (if you installed the package)
evil_core: I dodnt added any prelinking to cron
evil_core: not sure if cron is running at all
edwin: and it speeds up startup time of programs
edwin: do you have a /usr/bin/prelink?
edwin: or /usr/sbin/prelink
evil_core: crontab -l
evil_core: no crontab for root
edwin: /etc/cron.daily, and others
evil_core: # whereis prelink
evil_core: prelink:
evil_core: so I should prelink all>
evil_core: all*
edwin: no, I thought you had alraedy prelinked the current glxgears
edwin: if not, then thats not the problem
evil_core: $ ums_exec ldd /usr/bin/glxgears | grep libGL
evil_core: libGL.so.1 => /opt/xorg/lib64/libGL.so.1 (0x00007f1ac4cb8000)
evil_core: I am starting games usign simple script
edwin: what distribution?
PinkFreud: bridgman: no go.
PinkFreud: same thing. max 1680x1680
bridgman: can you pastebin your xorg.conf and xorg log ?
evil_core: cat ~/bin/ums_exec | wklej
evil_core: http://wklej.org/id/291493/
evil_core: PLD
evil_core: its PLD
evil_core: pld-linux.org
edwin: mine is Debian
evil_core: theres no installer
edwin: I don't know why for me it works w/o setting LD_LIBRARY_PATH
bridgman: that doesn't sound right, but then again most of us have 1680x1050 monitors so... ;)
evil_core: maybe prelink or something ;)
edwin: no, that binary is not prelinked now
edwin: since it can't prelink the new libGL
edwin: prelink: Could not prelink /usr/bin/glxgears because its dependency /opt/xorg/lib/libGL.so.1 could not be prelinked
evil_core: anyway i used KMS up to few days ago, so I were usign thos scripts to run gallium/non-gallium
PinkFreud: bridgman: nevermind. I see what I did wrong there.
PinkFreud: *now* Virtual works. I'm still not sure why it's required now, though.
evil_core: bridgman: I got only 1600x1200 thinkpad ;)
bridgman: evil_core; so you wouldn't notice either ;)
bridgman: I haven't really paid attention to that part of the code so not really sure if it changed or not, maybe someone else here will remember
evil_core: could I add hdmi to mine T60p?
twnqx: "add"?
evil_core: it got DVI on docking station
twnqx: my t60p only had a vga out iirc
twnqx: ah
evil_core: and only VGA on the edge
evil_core: so its internally DVI somehow also
twnqx: you can use a DVI->HDMI cable on the docking station
twnqx: i wouldn't recommend you try to mod your laptop :P
bridgman: agreed, unless you have teeny tiny fingers
evil_core: I am brave enough ;)
bridgman: I think laptops are built somewhere really small
evil_core: I am not using external monitor, but it would be good to got dvi and hdmi sometimes ;)
evil_core: I were modding SNES consoles
bridgman: DVI and HDMI use the same basic signalling, so converting from one to t'other is usually just a cable/dongle thing
evil_core: it would be good to put dvi instead of vga on the edge
evil_core: to got both VGA and DVI
evil_core: or maybe something more advanced
bridgman: I imagine laptop mfgs are moving to DVI-I instead of the VGA 15-pin but for a long time all the projectors in meeting rooms only had vga plugs
evil_core: I got spdif on dock also
evil_core: so maybe full hdmi + vga
bridgman: the DVI adapter having been lost in the first two weeks after getting the projector
twnqx: lenovo is moving straight to vga+display-port
bridgman: do you have the docking station ? If not, there are usually "mini docks" that have just a couple of connectors and cost much less
twnqx: my t 500 has that.
agd5f: PinkFreud: it's required with ums since memory management is static. with kms, it's all dynamic and you can resize at will
evil_core: I bought T60 dock for less than 50$
bridgman: good deal
twnqx: :(
twnqx: i paid 100€ for mine
evil_core: with serial/parrarell ports
twnqx: does your laptop crash if you undock without notifying, too? :ÜP
evil_core: no
twnqx: meh
twnqx: mine does
twnqx: i have to use the button on the side
evil_core: its T60p
twnqx: or linux freezes
evil_core: I too
evil_core: its full dock?
evil_core: mine isnt
twnqx: yeah
evil_core: only port replicator
twnqx: ah
twnqx: no, it's the full dock
evil_core: w/o PCMCIa/SATA
evil_core: i didnt wanted full dock, to obig for me
twnqx: uh...
evil_core: I put 500GB HDD inside, but not enought for h264 torrentz ;)
twnqx: would have to lookup, but it's at work
twnqx: i only use it for convenience anyway
evil_core: I too, still got only one AC adapter ;)
evil_core: but I dont have to attach ethernet everytime
twnqx: i have a keyboard, too
evil_core: some time ago I got connected amplifier to it, but its crap - not sound
evil_core: I got trackball, but I am not using it
evil_core: mine trackpint got broken a bit
evil_core: its moving simetimes in some direction
evil_core: not sure if its fixable by cleaning or not
twnqx: by the way, displayport.. anyone working on audio? :P
evil_core: but whoo nned keyboard for thinkpad(best keyborads ever)
evil_core: on hdmi it worked I heard
evil_core: with radeonhd probably
PinkFreud: agd5f: interesting. still - earlier (non-kms) drivers didn't require it.
bridgman: PinkFreud; I always required it on my system
agd5f: PinkFreud: they do, some earlier versions set the default size differently though due to bugs int eh early versions of xrandr 1.2
bridgman: at least when I borrowed a second display ;)
agd5f: twnqx: I've looked into it. similar IP issues to hdmi audio
twnqx: Oo
evil_core: and al lwe know that T60p is best thinkpad ever ;)
evil_core: so it would be nice to mod it
twnqx: it's more then flip the bit for "mux the digital bitstream in"?
twnqx: that.. sux
PinkFreud: agd5f: so a bug in xrandr was setting my default virtual screen size appropriately? sheesh. bring that bug back. :P
bridgman: PinkFreud; we'd need a different bug for each user though...
agd5f: PinkFreud: no, the driver used to set larger default virtual sizes several releases back. problem was, it tended to break systems with limited vram
PinkFreud: ahh hah. Now I'm getting it.
agd5f: PinkFreud: you still needed a virtual depending on the size of the screens attached. it just happened to be larger than the current size the core randr code sets up
PinkFreud: agd5f: so, with the larger default size in earlier versions, my dual 1280x1024 configuration was coming in at, or under, that, and Randr was able to set the size appropriately?
agd5f: PinkFreud: yes
soreau: agd5ftw(TM)
PinkFreud: heh, yeah, I get it now.
twnqx: hm, 81°C on the gpu and 86 on the CPU ain't good, right?
PinkFreud: twnqx: that seems a mite toasty
twnqx: effing firefox running at 100% for no reeason again
Jonimus: PinkFreud: yeah on my laptop before KMS the defualt Virtual size was just big enough for my one monitor
Jonimus: is there still no way to be checking GPU temps yet?
twnqx: my ibm laptop gives them via ACPI :D
PinkFreud: I'm still having issues convincing KMS to work, but that'll have to wait until I have more time to play with it
PinkFreud: as least I'm back to a working desktop.
twnqx: has issues convincing KMS to keep working :P
PinkFreud: twnqx: you're further along than I am :)
Jonimus: twnqx: yeah but thats an external sensor, I thought most cards had one on chip or at least a dedicated one
bridgman: all of the GPUs have an on-chip sensor (a diode junction) - it's the translation from diode current to something readable that is implemented all over the place
bridgman: until recently that translation was done by an external temp/fan controller hooked up by i2c to the GPU
mcencora: agd5f, airlied: I've just checked the speed on UMS without ColorTiling. guess what, the FPS in sauerbraten and glxgears went down to values for KMS
bridgman: that's why agd5f has been working on the i2c stuff in drm
bridgman: the more recent GPUs have some of that control logic on the GPU die
Jonimus: bridgman: ahh ok, but the currently exposed i2c devices are not the one with the chip at least not according to LM_sensors.
bridgman: "not the one with the chip" ?
bridgman: I'm not sure exactly what has to be done in order to expose the new i2c code to lm_sensors
bridgman: I'm kinda hoping agd5f knows ;)
twnqx: to lm_sensors? nothing, just create a kernel level i2c bus
bridgman: I'm sure in theory it's all automatic but in practice there usually seems to be some kind of hint required
twnqx: lm_sensors will scan it, but if the temp driver is embedded, you'll need a chip driver
twnqx: i saw it scan my nvidia's busses without being asked :P
bridgman: I know we still need drivers for the newer on-chip stuff but I'm thinking more about the external fan/temp chips
twnqx: just if the chip is non-standard you'll need
bridgman: most of the 6xx and everything earlier was third party fan-temp chips off the cpu die
bridgman: most of the 7xx use on-chip, and there's a transition between late 6xx and early 7xx
bridgman: I think
mjg59: The chip type is, IIRC, in the thermal tables
bridgman: presumably something needs to read the thermal tables and explain the contents to lm_sensors
mjg59: The main problem is that hwmon is utterly useless for in-kernel use
bridgman: I just don't know what that is
mjg59: bridgman: Not even that
mjg59: bridgman: The driver parses the thermal tables and identifies the chip type, then performs an hwmon register request with the chip address. That'll trigger the module load.
bridgman: so how does, for example, lm_sensors know what kind of chip is on each i2c bus ?
twnqx: if the chip is iedntifiable on standard addresses
twnqx: lm_sensors will just find it if you expose the bus
mjg59: lm_sensors is, at this point, just the userspace component
mjg59: Which can optionally scan for chips by probing addresses and then attempting to identify the chip types that are known to ship with those addresses
twnqx: bridgman: probing
twnqx: (see sensors_detect)
mjg59: But it's utterly irrelevant here
bridgman: is not a big fan of probing
mjg59: Once the chip is loaded, it'll be exposed via hwmon
mjg59: lm-sensors is then one potential consumer of that, but there are others
twnqx: never looked into the hwmon framework
bridgman: do we still need to do something in the driver so that hwmon knows about the controllers & i2c buses ?
mjg59: bridgman: Yes
mjg59: bridgman: If the driver knows what kind of chip there is, it needs to trigger driver loading and binding for that
mjg59: bridgman: I wrote code to do this for nouveau
bridgman: ok, tx
mjg59: http://lists.freedesktop.org/archives/nouveau/2009-November/004087.html
bridgman: you mentioned hwmon vs kernel issues ?
mjg59: Yeah, hwmon only exposes parameters via sysfs
mjg59: So the kernel has no way of using those values itself
mjg59: I modified the thermal layer to make it generic enough that it can be used for gpu fan control, but that requires that the driver be able to read the temperature itsef
mjg59: Which hwmon doesn't currently allow
agd5f: twnqx: there's an entry in the powertables that specifies the external chip and the i2c bus and address
agd5f: twnqx: avivotool romtables
agd5f: and you can see the info
twnqx: i don't really need it
twnqx: mind elaborating on the IP issues for audio over hdmi?
agd5f: you'll then need to add the i2c bus based on the entry in the i2c gpio table and then you can attach the hwmon driver
bridgman: yeah, it's not our IP, it came from multiple sources, and tracing each bit of information back to the original supplier and original IP transfer is a big honkin' job
bridgman: which takes big honkin' amounts of time
bridgman: ^^audio
agd5f: twnqx: I've been planning to add all the i2c buses by default and then add pointers back from various components that need them rather than adding buses on the fly like we do now
twnqx: hm, that would mean it's not in the specs amd opened so far?
bridgman: right - the specs & code we released so far is for graphics
bridgman: when we started the project audio wasn't even on the radar
bridgman: it was over in the chipset area; different part of the company
twnqx: and along came hdmi and displayport, happily throwing things around :)
bridgman: yeah, my top priority right now is making sure things don't get too much worse
bridgman: DRM is creeping into everything
twnqx: yey for digital restriction management
bridgman: yeah, obviously the DRM mob and the free software mob need to talk more, rather than leaving us to sort out the damn conflicts
twnqx: you can talk as much as you want... opening those parts want work as it'd allow people to just turn it off regardless
bridgman: yeah, that's always the challenge -- finding ways to carefully cut between the functionality people want and the drm stuff so that we can let X/DRI drivers do things without putting DRM at risk on other platforms
bridgman: and obviously all of the DRM stuff is closely guarded even inside the company, so a lot of times the initial response is "no you can't release that and I can't tell you why not" ;(
twnqx: hehe
bridgman: welcome to my world
twnqx: sad, but understandable until that crap is finally ruled illegal by court
bridgman: yeah, and before that happens someone is going to have to rule that media playback on PCs is an "essential service" or fundamental right
twnqx: nah
twnqx: just that full, unrestricted documentation has to be available for a product you want to sell
edwin: but is the audio only a problem if it comes from a DRM guarded device, or is it a problem if its just an audio file from my HDD?
bridgman: just an audio file
twnqx: i can see on my laptop the audio just comes from spdif
bridgman: the problem is that the audio HW and drm HW kinda have to be tightly integrated, just like video decode and drm
twnqx: or would the radeon chip be an audio device on its own?
bridgman: the GPU exposes a few devices, one of them is an HD audio controller
twnqx: oh.
agd5f: it shows up as a separate pci device
agd5f: and is driven by an alsa driver
bridgman: the problem is that there is a base HD Audio spec which is publicly available, but everyone expects the HW to do more than that
twnqx: i only have this 00:1b.0 Audio device: Intel Corporation 82801I (ICH9 Family) HD Audio Controller (rev 03)
edwin: 01:00.1 Audio device: ATI Technologies Inc RV710/730
bridgman: I think that's the one on your chipset - BIOS probably turned off the one on the GPU (assuming you have 6xx+ GPU)
twnqx: indeed
bridgman: edwin's, on the other hand, is enabled
twnqx: hm
bridgman: sometimes there's a BIOS option
edwin: and there is a HDA ATI HDMI device, which only has a mute/unmute button
twnqx: but i remember initially graphics adaptors had an spdif in
twnqx: and no own audio device
twnqx: so... is that setup possible with the r600 as well?
agd5f: twnqx: the on chip audio is purely for hdmi/dp
bridgman: this is what makes things exciting - depending on which audio output connectors a system includes, they may use the audio controller on the chipset or on the GPU or both
twnqx: that'd be pretty annoying
twnqx: having to switch the audio device based on where you want to use the audio :(
twnqx: i'd expect the mobile parts to optionally just provide a HDA codec that can go on the bus
twnqx: (like nvidia does e.g. for the ion)
agd5f: er, I guess except for IGP chips. in that case, the on chip device can be used for both direct connect and hdmi. I suppose discrete could as well in laptops and such
twnqx: well, this is just a radeon mobility :P
twnqx: but gonna look in the bios on next reboot
bridgman: twnqx; you can reboot any time, you know ;)
twnqx: but i don't want to :X
twnqx: and i don't have a reason to update the kernel until glisse/spstarr report r600 to no longer crash
twnqx: (with kms)
PinkFreud: r600 still crashes with kms, eh?
twnqx: well, there's a test build that features chip reset within 2s :P
mmp: hello; sorry for repeating question another person asked on bugzilla, but what is the cause of little or no progress on 25408 (RS600 and modesetting/DRI2) for few months? is there anything I can help with?
PinkFreud: heh
PinkFreud: picked the wrong time to try to get kms working on my rv610
twnqx: hm
twnqx: i should figure out when to go to the airport :S
twnqx: so plane departing in 5h...
bridgman: you're already late ;)
bridgman: mmp, my guess is that part of the problem is too few people having rs600s
mmp: bridgman: well... from what point of view? too few developers, or too few interested people, or pure economic reasons? ;-)
dileX: hmm, openarena has red-ish color corruptions (linus-tree, mesa latest)
agd5f: mmp: too few developer, ie, none
agd5f: and not too many users either
agd5f: there weren't too many made
Jonimus: agd5f: so how hard would it be to write a patch that exposes the i2c line specified in the table for the thermal table?
agd5f: Jonimus: not too hard
bridgman: a couple of developers have tried to find rs600 on ebay etc but not much luck yet
mmp: agd5f: yup, that's reasonable, but still there are people (like me:) who own it and would like actually get something from that card :)
mmp: bridgman: it's integrated, they could be only looking for motherboards, right?
agd5f: mmp: I understand :) The problem is I don't see anything obviously wrong, so It probably requires hw to debug
Jonimus: agd5f: thank you for your work on the ddc i2c ports, I was finally able to shut up the annoying red light on my LG monitor :P
bridgman: mmp; AFAIK yes
agd5f: mmp: yeah. I'm not sure if any motherboards are available or not. Might have been a few laptops only
mmp: agd5f: as long as remote access is good enough, I think at least I (and maybe also Dariem; the original reporter) would give access to our machines...
mmp: not too great, because you still don't have visual feedback
mmp: but still better than nothing
agd5f: mmp: that's an option. ping me or glisse next week. maybe we can take a look
mmp: agd5f: ok, will luckily do
bridgman: somewhere there's a picture of a system on ajax's desk with 2 30" displays and a webcam
mmp: agd5f: I can set up a webcam, so you can even see what's happening on the sccreen :-P
mmp: :-))
bridgman: then airlied worked on the other side of the planet developing and testing drivers on that system
bridgman: don't think he would ever do it again though ;)
mjg59: bridgman: You say that...
mjg59: Actually, in the end I think we got that laptop shipped to ajax last time
ajax: i think i lost that picture in the great annarchy disk crash of 2009
bridgman: ahh, maybe I'll stop looking then
bridgman: it was a great picture
mmp: well, remote debugging of hardware drivers is one of the purest form of the evil (I already gave access to my PC for debugging purposes, and I guess it's a hell of work)
ajax: it would have been http://people.freedesktop.org/~ajax/30-love.jpg
mjg59: 404
ajax: and apparently the wayback machine does not have a copy of it. sad.
mjg59: Sorry, I see what you mean
agd5f: is ~ actually backed up now, or is that still just a good idea?
eviljames: Figured I would report back, xorg-server 1.7.5 libdrm 2.4.18 mesa 7.7 and ati 6.12.191 on slackware64-current... kms + dri2 work like a charm.
ajax: agd5f: that is an excellent question and one i do not know the answer to
ajax: so i would assume "still just a good idea"
agd5f: mcencora: I guess that explains some of the kms slowdown :)
Jonimus: notices no difference between KMS and UMS on his two r7xx cards :/
mcencora: agd5f: yeah, would it be possible to you the Composite hooks to do the blit in dri2?
agd5f: mcencora: you could probably create a composite wrapper for the dri2 copy hook and use composite rather than the copy hook
agd5f: exa copy hook that is
da1l6: agd5f: tried your 0001-drm-radeon-kms-atom-spread-spectrum-fix.patch against kernel 2.6.33, unfortunately it does not apply: http://pastebin.com/Am9hfbu3
agd5f: da1l6: it's against drm-next or drm-linus
da1l6: agd5f: hmm... need to figure out how the compile those. for vanilla .33 i just need to add the patch line to the official arch PKGBUILD
DanaG: Jonimus: red light?
DanaG: sorry, reading backlog.
Enverex: Can someone running 32bit Debian throw me a copy of glxinfo?
Jonimus: DanaG: the power light on my monitor is annoyingly bright but I can turn it off through DDC controls
DanaG: ah. ddc/ci?
Thunderbird: Enverex, just grab it from packages.debian.org
Enverex: Thunderbird, Already have
DanaG: http://hpfansite.com/hp-elitebook/dream-color-2-rumors/
DanaG: OOh, 10-bit color in a laptop? spiffy.
DanaG: er, 30-bit.
Axtroz: hello everyone
Axtroz: i cant seem to get opengl 2.1 working with a X1600
adamk: Axtroz: Using the gallium3d driver?
Axtroz: No
adamk: Well that would be why.
Axtroz: using the radeon driver
Axtroz: Any hints on how to enable the gallium driver?
adamk: radeon is the name of the 2D driver and kernel modleOuD
Enverex: wonders if adamk just fell on his keyboard
Axtroz: I see. Coz I get really neat OpenGL 1.5 with it :)
adamk: D'oh.
adamk: Well, it's Mesa that provides the 3D functionality.
adamk: Within Mesa, there is the regular r300 driver, and there is also the r300 gallium driver.
adamk: Usually called r300g.
adamk: No one ships with r300g enabled, so you'd have to build it yourself.
Enverex: Just out of interest, what distro are you all using?
Axtroz: Slack-13
Axtroz: Ok, thanks for the info
Axtroz: Will try it out :)
adamk: Axtroz: You'll need to build a kernel with KMS, update libdrm and Mesa.
Axtroz: In short... Upgrade to -current
Axtroz: :)
adamk: That's an option. You'd still need to recompile mesa.
adamk: Perhaps with a newer version. I don't remember what -current uses.
Axtroz: uses 7.7 now
adamk: r300g sees a lot of development so I would normally recommend Mesa from git anyway.
Axtroz: Oh, I see. Well thanks I'll hit the git in this case :)
Axtroz: Btw, excuse my ignorance... Can I compile on another computer that doesn't have a radeon card? This won't affect the build, right?
Axtroz: I'm kinda stupid :D
adamk: Yes, it's fine.
Axtroz: Great. Thanks!
adamk: I regularly build drivers for my intel machines on other boxes.
glisse: agd5f: know off hand why mesa r7xx compiler add temporary mov btw mad instructions ?
adamk: Axtroz: BTW, this isn't a laptop GPU, is it?
FIReun: adamk: are you implying laptop GPUs are stupid? d-:
FIReun: they're just graphically challenged
Axtroz: adamk: no, this is a desktop one
Axtroz: which runs really neat, 2700 FPS with glxgears.. compared to 450 with a Toshiba laptop with a X1200 :D
FIReun: glxgears is not a benchmark, supposedly
Axtroz: FIReun, I know its not a benchmark, but still
Axtroz: :)
FIReun: but I get 393 with an rs480 on turion64 and 2week old git
TGEN: anyone know when radeon(4) will start supporting GLSL?
FIReun: 3d games are fairly unplayable (ioquake) and I get weird horizontal line distortion (like tearing but different) with full screen playback (only with the radeon driver)
FIReun: the tearing almost looks like some kind of block alignment error
FIReun: or someone's idea of a watermark
FIReun: but only with -vo xv (-vo gl works fine)
dmb: FIReun: are you using fglrx?
FIReun: dmb: I wouldnt be here if I was (:
dmb: oh, was going to say i'v seen watermark corruption bug with fglrx
adamk: You'd be surprised how many people show up here with fglrx :-)
FIReun: I wish ioquake looked as much improved on the radeon driver as movies do (excluding the watermark/tearing issue)
FIReun: adamk: well its still a radeon card, right? d-:
Moobyfr: Hi
Moobyfr: I have working 3d with r525 card
Moobyfr: but there are some rendering issues with the game springrts
Moobyfr: seems that some mapping doesn't work
Moobyfr: any developer aware of that ?
mcencora: Moobyfr: probably not if there's no bug report
Moobyfr: I'm not sure it is a radeon bug, springlobby has some particularities with videos cards as they use a lot of opengl
DanaG: heh, instead of having glxgears keep being "not a benchmark"... what would it take to make it BE a benchmark?
DanaG: Textures, for one, of course.
DanaG: Maybe more geometry?
FIReun: DanaG: its not comprehensive at all
dafox: is there any real benchmark then
dafox: for windows you have 3dmark, is there anything similar for linux?
TGEN: FIReun: teapot works?
FIReun: TGEN: what? I drink coffee
TGEN: no, teapot(1)
TGEN: one of the Mesa demos
FIReun: I dont see that on this system
FIReun: digs a bit
FIReun: 30fps with teapot
TGEN: hmm, that's not a lot
FIReun: no doubt (rs480 remember?)
TGEN: I don't remember, I wasn't in this channel before ;)
FIReun: ah, no worries
TGEN: I get ~220 fps on an X1950XT 256
FIReun: must be nice (:
TGEN: ~110 on an HD4670
TGEN: well, the thing I need it for right now (GLSL) doesn't work
TGEN: Mesa just returns GL_FALSE when I try to compile shaders
FIReun: hmm, 44 without the help overlay
TGEN: yah, the help overlay slows things down a bit
maligor: dafox, well, Unigine has some fancy proprietary benchmarks
FIReun: I'd be happy with 44fps in ioquake (:
maligor: never tried those on the open drivers tho
TGEN: well, I imagine ioquake (being based on Quake 3?) is a heavier load
TGEN: but OpenGL support is rather lacking, not just my aforementioned GLSL issue, but it seems the best it can do is OpenGL 1.4
FIReun: TGEN: halflife in windows (vsync enabled) beats everything, followed by ioquake with fglrx, and radeon+kms coming in just over radeon+ums
TGEN: for my HD4670, Mesa reports 1.2 even (r700 is 3.0 hardware)
FIReun: TGEN: 2d support in the radeon+kms is outstanding
TGEN: I dunno about KMS yet, it appears to be a Linux-only thing
FIReun: TGEN: which is what this channel is for
TGEN: Linux?
TGEN: I thought this channel was about radeon(4)
maligor: plenty of other os's using radeon
FIReun: TGEN: you guessed wrong (:
TGEN: I use it on both Free- and NetBSD
maligor: I think OpenSolaris too?
TGEN: maligor: wherever X.org works, I'd imagine
FIReun: yes
TGEN: I'd hope so at least :)
FIReun: i assumed you meant windows or something
maligor: TGEN, as for opengl versions, git head should report OpenGL 2.0 support
TGEN: no, on Windows I tend to use whatever drivers are supplied officially
FIReun: it happens
maligor: dunno about stable releases
TGEN: as long as they don't break things, like nVidia's latest ;)
TGEN: maligor: including at least GLSL 1.3?
maligor: I think it's 1.10
FIReun: I only still have windows around for Solidworks (not doing much with it lately) and occasionally a fit of halflife gaming
TGEN: should be enough, for the moment I only require 1.3
TGEN: (wasn't even aware there was a GLSL newer than 1.5?)
maligor: I'm not running on the radeon driver right now tho
TGEN: FIReun: most games I play are Windows-only, but as a gaming platform, I can manage with it
maligor: catalyst gives GLSL 1.50
TGEN: yah, but there's no Catalyst for FreeBSD
TGEN: never mind NetBSD :)
maligor: do the intel drivers work in fbsd?
TGEN: yes
TGEN: NetBSD as well
TGEN: I use it on my laptop, but it has a rather old Mesa (7.0.4), which also can't compile my shaders
maligor: just wondering since I think they dropped UMS support
TGEN: yeah, I read about that
TGEN: bunch of idiots :P
TGEN: killed EXA and XAA too while they were at it
maligor: yeah, X wants to be root
TGEN: Packard wants it, I guess
TGEN: I don't think GEM is supported on anything other than Linux either for the time being
maligor: KMS is fantastic however, it improves the nethack immersion
TGEN: heh
TGEN: I've forgotten what the main bullet point was for KMS
dafox: I have to use windows for gaming nowaday too, but I can still remember when 3d drivers on linux were good and I could play counterstrike in linux with better fps than windows :p
TGEN: I don't really mind it, I have a dedicated PC for games :)
TGEN: keeping work and games strictly separated also helps you from getting distracted when you have deadlines to meet ;)
dafox: but xf86-video-ati is steadily improving, so I have good hopes that it won't be too long before I can play again
dafox: and it's far better for other stuff like compiz (actually works as opposed to the fglrx which supported my card) and tvout (strange vertical-stretch bug in nearly every version of fglrx)
TGEN: ah, I use neither of those
FIReun: I would gladly see compiz evaporate in favor of excellent 3d performance (:
dafox: maligor: unigine looks a bit too hardcore/advanced for older graphics cards. Those screenshots look like they would take a week to render on my hw :p I was think more like 3dmark 2001
TGEN: I like the eye candy of those Unigine demos
dafox: I like the eye candy of compiz :p
TGEN: compiz actually annoys me to no end :P
maligor: I like the eye candy of @
dafox: well, mostly the wobbly and things like window preview, expo, e.g. the usefull features. Not the over-the-top fire painting nonsense, although it is nice to showoff what is possible
dafox: maligor: @?
FIReun: TGEN: I've recently installed a system with e17 -- they do "fake" compiz, plenty of eye candy for me -- fastest wm I've had in a long time too.
dafox: how do you 'fake' compiz?
FIReun: dafox: lots of animated window effects
FIReun: no transparencies tho
dafox: xfce does dropshadows and transparency too, but that to me doesn't really say 'compiz', when I think of compiz I think of wobbly windows
dafox: that's just compositing
TGEN: FIReun: XFCE4's compositor is adequate for me
dafox: TGEN: I also switch to that quite often when I need some extra performance, but don't you miss the extra bling?
FIReun: TGEN: e17 is faster, especially at startup
FIReun: TGEN: I was using xfce before switching
Wizzup: thinks transparent terminals are distracting
dafox: Wizzup: what about fake transarency on your terminal than (e.g. just a background)
Wizzup: That's ok
Wizzup: As long as I can't see irssi through them :)
agd5f: glisse: not sure. richard might know.
TGEN: dafox: like I said, the extra bling only serves to annoy me :P
TGEN: FIReun: I doubt I'd notice
TGEN: XFCE4 is nigh-on instantaneous here (i7 920 @ 3.6GHz, X25-M G2)
dafox: I think you're mostly going to notice on older systems, but even on my 5+ year old laptop xfce doesn't take more than a few seconds to start
wirry: TGEN: with this hardware even gnome doesnt take more then one second...
FIReun: see, I have trouble thinking of a 4-5yo system as "old"
FIReun: has a p133 pictureframe over his desk... its **old**
TGEN: well, I still have a somewhat operational MicroVAX 2000
TGEN: but I wouldn't use it for anything serious anymore
TGEN: that's what the new toys are for :P
wirry: hrhr, i have an amd 3800+ x2 with 4gb ram for an old desktop...and an laptop with a 2ghz c2d, 2gb ram and a x1600m...^^
FIReun: I still have my titanium powerbook, but I dont even turn it on...
FIReun: still quite usable for most things tho
TGEN: my laptop's similar, C2D T7300, 4Gibyte RAM, though Intel graphics
FIReun: be even snappier if I put linux back on it
TGEN: FIReun: the thing with laptops is that their batteries degrade, making them lose their purpose over time
FIReun: just think, all that hardware will be "old" in 3-4 years (:
wirry: hmm well, its not old, but i rather use my thinkpad (1,6ghz, intel graphics) now...
TGEN: <- X61
FIReun: TGEN: I just make em ento desktops w/external mouse/keyboard and/or display
FIReun: take up little footprint, often much quieter than "desktops"
wirry: x61t :P
wirry: (it was cheaper Oo)
TGEN: FIReun: I don't have space for that on my desk (even if it's 5m wide) :P
TGEN: wirry: oh, I got it for relatively cheap as well, including the 3G HSDPA modem option
FIReun: well my car is faster than yours! d-:
wirry: nice
wirry: i use my cellphone as modem
FIReun: notes how far the convo has gone off topic
TGEN: my Renault Twingo has a whopping 75 bhp
TGEN: you sure it is? ;)
wirry: hey, i just came home, its after midnight...how can you expect i can stay on-topic?
FIReun: my car is a figment of my imagination, so yes, it is both faster and slower than yours
wirry: at least my "old" laptop got amd graphics :P
TGEN: that's hard to compete with, FIReun :P
FIReun: wirry: amd64-1.6Ghz w/1.3Gram, xpress200radeon
wirry: i got a bike :x
Araneidae: Hi: I'm running the radeon driver with an ATI Radeon X1650, and having some funny issues with dual monitors
FIReun: wirry: I dont think my bike has air in the tires
wirry: mine has...at least it had last time i checked
FIReun: wirry: doesnt everyone live on american time?
Araneidae: Had a *lot* of grief with radeonhd, but the radeon driver seems better -- but still managed to choose a stupid resolution for one monitor
FIReun: wirry: I thought that was the *rules*
wirry: err...no!
Araneidae: Anybody interested at this time?!
TGEN: Araneidae: are you using preferred modes?
Araneidae: Don't know. Can pop up my xorg.conf and xrandr output if interested
TGEN: Araneidae: also, observe your xrandr output
Araneidae: This is xrandr: http://paste.debian.net/62803/
wirry: rules are made to be ignored, arnt they?
Araneidae: This is my xorg.conf: http://paste.debian.net/62847/
TGEN: Araneidae: looks like it's picked the biggest resolution possible on both monitors
Araneidae: Ah: yes, I'm setting a preferred mode
TGEN: Araneidae: err, scratch that, the second monitor has a smaller one
Araneidae: Well the problem was it chose 1600x1024 on DVI-0
evil_core: X61X == fail, not flexView!
Araneidae: The problem was that DVI-0 can't actually show 1600x1024!
wirry: flexview?
evil_core: so ebverything newer than X60X sucks!
TGEN: Araneidae: in your Xorg.0.log, what does it report about that monitor?
evil_core: IPS screen
wirry: hmm..
evil_core: new LCD displays are crap
Araneidae: TGEN, how do I find that? Grep for DVI-0?
evil_core: you cannot buy laptop now with decent screen
wirry: dunno what my screen is, but it shouldnt be tn
TGEN: Araneidae: it's usually small enough, you can look around with less
wirry: the viewing angle is just too big
evil_core: http://www.thinkwiki.org/wiki/Flexview
evil_core: wirry: its newer TN :P
Araneidae: Well, my Xorg.0.log is a couple of thousand lines, not sure what to give you. I can paste a grep for DVI-0...
wirry: !ping
wirry: hmm...
wirry: well, i kinda like it...
TGEN: Araneidae: look for the mode lines a little bit after a line like:
TGEN: (II) RADEON(0): I2C device "DVI-0:E-EDID segment register" registered at address 0x60.
wirry: better then my other laptop / samsung 24" screen
TGEN: you can ignore the VESA modes
TGEN: (II) RADEON(0): Printing DDC gathered Modelines:
Araneidae: This sort of thing: (II) RADEON(0): Output: DVI-0, Detected Monitor Type: 3
TGEN: stuff after that is interesting
Araneidae: (II) RADEON(0): Supported VESA Video Modes: and it then goes on to list stuff *not* including 1600x1024!
TGEN: VESA modes are pretty much legacy
TGEN: scroll down further
Araneidae: From here: (II) RADEON(0): Printing DDC gathered Modelines: ?
Araneidae: (II) RADEON(0): Modeline "1600x1200"x59.9 161.00 1600 1712 1880 2160 1200 1203 1207 1245 -hsync +vsync (74.5 kHz)
Araneidae: Lies
TGEN: are you looking at the modelines for the right monitor?
Araneidae: Oh: That;s from the other monitor
Araneidae: Here's around the region of interest, I guess: http://paste.debian.net/62849/
Araneidae: Afraid I'm not sure what you're looking for here
FIReun: wishes someone here played ioquake on a rs480, so settings could be shared
Araneidae: Dammit, I missed: that's the DVI-1 lines!
TGEN: well, if there's a modeline reported as 1600x1024 for the other monitor, the DDC data from the monitor is faulty
TGEN: Araneidae: looks like that's all data from the correct monitor
Araneidae: Well, DVI-0 certainly can't do 1600x1024, and I did set a preferred mode ...
wirry: i never heard of 1600x1024...
wirry: only 1600x1200
Araneidae: Here's a snippet around DVI-0: http://paste.debian.net/62851/
Araneidae: wirry, you're right
Araneidae: Either way, I needed to `xrandr --output DVI-0 --mode 1280x1024` after an X11 restart
TGEN: Araneidae: those are only the VESA modes for DVI-0
Araneidae: Oh
TGEN: there should be Modelines reported for it as well
Araneidae: Well, I'm lost, so here's the last 1500 lines of Xorg.0.log: http://paste.debian.net/62852/ !
Araneidae: It's a little wordy
TGEN: Araneidae: weird, it prints both EDID vendors at the same time
TGEN: (II) RADEON(0): EDID vendor "HWP", prod id 9745
TGEN: (II) RADEON(0): EDID vendor "INL", prod id 9925
TGEN: and lists the data for the correct monitor twice
Araneidae: Which lines?
TGEN: or wait
Araneidae: I mean line numbers, then I can follow along
TGEN: the file hasn't been cleared
TGEN: it contains old data as well
Araneidae: um
Araneidae: I was running radeon-hd (which was awful), changed the Driver line in xorg.conf, and did an X restart. Could that have anything to do with it?
Araneidae: Think I'll remove the radeon-hd .dpk altogether and try a machine restart, if that will help
TGEN: is that paste clean?
Araneidae: um. Think so, but I'll do it again
Araneidae: Clipboard set with: cat /var/log/Xorg.0.log | tail -n 1500 | xsel -i
Araneidae: Here we are: http://paste.debian.net/62853/
TGEN: hmm, still looks horribly broken
TGEN: it does EDID/DDC probes over a dozen times
Araneidae: Hmm. Removing xserver-xorg-video-radeonhd also wants to remove xserver-xorg-video-all ... sound sensible? (Debian)
Araneidae: Yes, very odd
TGEN: yes, it probably means that xserver-xorg-video-all is a meta package clustering a bunch of X.org drivers
Araneidae: Oh, that makes sense.
Araneidae: Oh my. It pulls in *everything* including the kitchen sink and more
TGEN: you never know when you might need one
TGEN: it won't likely interfere with radeon(4) though, you can leave it be
TGEN: anyway, I've got to go sleep
Araneidae: Me too, to be honest
Araneidae: Ah well, I'll put this machine to bed, see what it has to say for itself tomorrow.
Araneidae: Thanks for looking, anyhow!
Araneidae: My cursor shimmers. Is this a known bug?
Araneidae: I've certainly seen this before
Araneidae: Haha: the first google result for radeon shimmering cursor is: http://www.radeonhd.org/?page=archive_display&c=radeon&m=5&y=2009&d=2009-5-05
FIReun:
FIReun:
FIReun:
FIReun: ok well, Unigine locks my X up
FIReun: oh well.
spstarr: hmm
spstarr: well now that's some new errors
spstarr: when GPU softlocked
spstarr: [26782.902409] Uhhuh. NMI received for unknown reason a1 on CPU 0.
spstarr: [26782.902432] You have some hardware problem, likely on the PCI bus.
spstarr: and then the GPU did a badness DMAR fault
FIReun: hmm, so after installing all the crap I need to get the 32bit mplayer to run on my 64bit system -- there are no "watermark" vsync like horizontal lines with fullscreen -vo xv playback
FIReun: anyone have an idea what might be responsible for the buggy xv playback?
FIReun: buggy 64bit libxv / libxvidcore ?
dafox: FIReun: what exactly do you mean 'watermark'?
FIReun: dafox: horizontal lines that dont move
soreau: Hi. My drivers cannot run compiz though X log and glxgears look ok and I can run other 3D apps fine. Compiz gives "Fatal: Root visual is not a GL visual"
soreau: This happened after an upgrade of user space components though I have not downgraded to confirm this is the problem directly
laumonier: someone knows when a new version of radeon will be ready?
FIReun: heh
FIReun: you might have to be a little more specific than that
BioTube: laumonier: the first 6.13 RC is out, IIRC
justlooking_: i was just wondering, given Cooper Yuan doesnt seem to have made any more changes to http://cgit.freedesktop.org/mesa/mesa/?h=pipe-video since 2010-02-01 has he given up and moved onto other non xvmc video stuff things ?
III: hello, got a few ???--> i am trying to add cario dock but I am seeing a black image behind my dock. Went searching around that net and it seems this is due to compiz not being enabled. I tried to enable the Extras on my system but now I am seeing --> Xlib: extension "RANDR" missing on display ":0.0". RandR extension missing --> http://pastebin.com/qqEHfVM5 | setup info
III: can any point me in the right direction....?
III: btw: currently running 3 displays as single monitor...
cxo: ... so putting 400watts worth of GPU and 125watts worth of CPU into a half-depth 3U box with a 480watt PSU may sound like it might work, but dont try it
spstarr: man, im causing the GPU to softlock so much with firefox
spstarr: hope glisse has the GPU recording stuff soon, then i can give him the foul info