Nightwulf|work: hi all
evil_core: hi all
evil_core: gallium is still being broken by MAD?
MostAwesomeDude: It's less broken ATM.
evil_core: hmm...but quake3 main screen should still be broken?
MostAwesomeDude: Hm, don't think so. Lemme check.
evil_core: because it is for me. I wonder if I should restart xorg
MostAwesomeDude: OA works here.
MostAwesomeDude: Are you using ioquake3 or old quake3?
evil_core: yes
evil_core: ioquake3
evil_core: but I didnt rebooted machine/restarted xorg after rebuildinf libdrm/xorg
evil_core: restarting xorg didnt helped
evil_core: still total mess in menu
MostAwesomeDude: OA, or original pak0?
MostAwesomeDude: None of us have the original pak0, so using OA would be way better.
evil_core: OpenArena looks OK, but hangups kernel after few seconds
evil_core: in gallium mode
MostAwesomeDude: Yeah, I'm told that it causes hangs, but I have no idea why.
evil_core: but airlied claims it doesnt for his identcial T60p :/
evil_core: maybe because he is using 32bit?
evil_core: anyway quake3 menu and xmoto is broken
evil_core: you can always use pak0 from demo, but xmoto menu is nightmare for me
evil_core: its hard to close it from within
evil_core: supertuxkarts also
evil_core: I think that openarena is exception ;)
evil_core: OpenArena is packaged from binaries in most distros
evil_core: ioquake3 doesnt like most other games
evil_core: and maybe its build against some realy old libGL, because its strange thats not afected by most problems like multitexture s3tc(maybe its content doesnt got one), etc
cxo: Nexuiz is still broken for me (r700)
Nightwulf|work: runs well here (rv790)
cxo: its been broken for the last week or so for me, i've noticed this on the console:
cxo: Draw_CachePic: failed to load gfx/complete
cxo: Draw_CachePic: failed to load gfx/inter
cxo: GL_RENDERER: Mesa DRI R600 (RV770 9440) 20090101 TCL DRI2
cxo: GL_VERSION: 2.0 Mesa 7.8-devel
cxo: Is there a way to see gl errors? like an strace for mesa?
MostAwesomeDude: Bugle...
Nightwulf|work: cxo: nothing like that here...runs smooth
MostAwesomeDude: IF you build your Mesa with --enable-debug, it should gripe on GL errors.
Nightwulf|work: cxo: OpenGL renderer string: Mesa DRI R600 (RV635 9598) 20090101 x86/MMX+/3DNow!+/SSE2 TCL
Nightwulf|work: (@work)
Nightwulf|work: but same string @home with rv790
cxo: at least half the textures are missing here
cxo: alien arena and open arena are fine
Nightwulf|work: xorg-server 1.7.3.902, mesa-7.7, kernel 2.6.32.3
cxo: 1.7.4.3+3ubuntu, mesa-git,drm-git,linux-git,xf86-video-ati-git
Nightwulf|work: cxo: last one applies here too...radeon is git from approx. 5 days ago
cxo: everything seems normal apart for Nexuiz
cxo: i have about 20k fps on glxgears without compiz, about 10k with compiz, over 200fps on openarena (1280x1024)
cxo: should install some of those fancy gfx benchmarks from phoronix
cxo: Isnt there like a basic sanity test program for mesa that just calls every function and checks its return values?
Nightwulf|work: only thing not running ok here is blender...shows some severe redraw errors
cxo: i know i'm on the edge, but people honestly need to do more regression testing
zhasha: cxo: try running piglit
cxo: is that a mesa demo?
Nightwulf|work: cxo: to be honest....i never encountered a project where i could use HEAD/MASTER revisions with so less doubts than this one...never had severe problems
MostAwesomeDude: Nightwulf|work: Well, all those other 3D driver projects must just be better off.
MostAwesomeDude: Oh, wait.
Nightwulf|work: MostAwesomeDude: really? i doubt that...btw we talked about code quality in cutting edge code and i found it impressing to have so less trouble with this one
MostAwesomeDude: Nightwulf|work: I'm being sarcastic.
Nightwulf|work: MostAwesomeDude: yeah...i know...i tried to answer that way too...but i'm no native english speaking guy, so please forgive me if i didn't hit that point :-)
cxo: cant believe how ridiculous it is to run phoronix on ubuntu, forgets idea
cxo: zhasha, what are the deps for piglit?
zhasha: X11 and GLX I presume
zhasha: oh and cmake
cxo: it looks like its all in python?
zhasha: I think all the tests are in C
cxo: hates Make variants
zhasha: ccmake .
cxo: oh so i got to compile this first
zhasha: hammer random keys
zhasha: *pray*
zhasha: then make
cxo: did cmake .
zhasha: it's ccmake
cxo: cmake failed
cxo: oh ok
zhasha: yeah, I thought it was a type too the first time
zhasha: typo*
cxo: thats just a curses gui kinda of thing for cmake
cxo: my glew library cant be found or something
cxo: gives up
zhasha: seems to depend on GLEW then
cxo: 'make' that tries to do configure = fail
zhasha: it's the only test suite we got for conformance
zhasha: install glew, try again, hope
cxo: has no patience for cmake
zhasha: you wanted a conformance test suite, I gave you one. Stop complaining
cxo: you guys go to try mesa/progs/demos/tunnel
cxo: so cool
zhasha: how is that cool?
cxo: cos its so 3D
cxo: make it fullscreen, turn off the help message
evil_core: (EE) AIGLX error: Calling driver entry point failed
evil_core: (EE) AIGLX: reverting to software rendering
evil_core: what I can do about that?
evil_core: it causes tearing for Xv
JohnDoe_71Rus: who can speak in Russian?
glisse: airlied: m9 didn't suspend for you ?
glisse: i am pretty sure here rv2xx suspend fine
amarsh04: anyone running radeon driver with DRI enabled, can you try running GNU paint (gpaint) and attempt to draw an ellipse? It crashes the xserver for me using either radeon or radeonhd with DRI enabled
evil_core: can I turn off backlight on T60p r500?
glisse: evil_core: acpi might be exposing something for that
evil_core: xrandr and radeontool doesnt work for me
koolfy: Hey, I just bought a new laptop (sony) with a radeon HD45xx, and compiled the kernel with the sonypi component required to control (for example) brightness of the screen
koolfy: but someone I know who has a sony laptop too says that the brightness has to be controlled by the video card
koolfy: he uses a nvidia-related command to do so
koolfy: is there anything for that with radeon driver ?
evil_core: can I get off tearing while moving windows w/o compositing?
evil_core: vsync doesnt work for me with KMS :/
koolfy: it seems that the module loaded by the binary driver of nvidia launches nvclock to change the brighness when the keys are pressed
evil_core: se noai
evil_core: a(EE) AIGLX error: Calling driver entry point failed
evil_core: (EE) AIGLX: reverting to software renderingse ai
evil_core: sorry, I didnt wanted to paste it
adamk_: Alright, so I've updated my kernel to the latest from git + drm next, and I still get the washed out colors on one port of the x850.
adamk_: Not sure if I care enough to report it, or to just put the HD4350 back into the box. :-)
Tommeh: Sounds like a shame, adamk_
Tommeh: Do you have the same behaviour from fglrx?
adamk_: fglrx doesn't support the x850 any more
adamk_: The problem only happens with that video card, and only with KMS.
mokoloko: btw does radeon driver support any form of antialiasing or anisotropic filtering?
gsedej: Hello! I have RV350 and Ubuntu and I am haveing problems with compositing managers (compiz and KDE). It just randomly completly freezes (REISUB does not work). I tried with KMS and without. I am trying to solve this for quite some time. What can I (as user, not developer) try to locate problem?
gsedej: what can I do to locate problem*
glisse: gsedej: sounds like GPU lockup not much you can do
xming: the fan of my GPU is spinning louder and louder :(
xming: but under compiz is better than kwin
maligor: it spins quite fast here even on 2d, albeit i haven't tried very recent versions
xming: fast is not a problem :D Loud is though :p
gsedej: glisse, it didn't freeze in some game (nexuiz) at all... Normally freezes when I shitch viewport
xming: I hope that PM for kms will be merged soon
gsedej: glisse, is this maybe hardware problem? because I didn't see anyone else had similar problem on internet(forums)
gsedej: What options can I try at GRUB? (radeon.modeset=true, etc...)
glisse: gsedej: AGP ?
gsedej: glisse, yes. I think I have AGP2x
glisse: disabling AGP often helps
glisse: but it will be slow
gsedej: it is radeon mobile 9600 on laptop with P4 1,6GHz
glisse: so agp
gsedej: how to do that?
maligor: xming, well, fast here means it sounds like a jet engine
xming: glisse: try disable fastwrite, agpx1, pci mode. in that order
maligor: since it's a stock 3870 cooler where I've modified the bios fan control to keep it under 90C
gsedej: where can I find some "HOWTO" for this options?
glisse: gsedej: depends if you are using kms or not
glisse: if kms disabled man radeon
glisse: otherwise add radeon.agpmode=-1 to your kernel boot parameter
xming: glisse: X --configure will generate a config file with all those options commented out
gsedej: so, how to disenable fastwrite? (in GRUB, right?)
maligor: is it even enabled?
gsedej: how to find out?
maligor: fastwrites and sideband addressing were notoriously unstable features
gsedej: glxinfo?
maligor: I don't really know, I doubt it'd show up in glxinfo
glisse: xming: don't work with kms
gsedej: since I have default, I think i have disenabled fastwright
ossman: agd5f, time to go poke those hw guys with something sharp
gsedej: LOL, console mode does not work... only "_" is blinking (ctrl alt F1-F6)
glisse: ossman: my feeling is that hw bug are hard to track down and understand, and you might have pointed out a particular hard issue and now they curse you every morning and every night ;)
ossman: glisse, as long as I'm memorable ;)
agd5f: ossman: just pinged him again
ossman: agd5f, great
ossman: any of you guys going to fosdem, btw?
jcristau: glisse is supposed to talk there so hopefully he'll go :)
ossman: heh
glisse: agd5f: btw i got a bunch of fixes for r600 irq might help with the kexec issue
agd5f: glisse: excellent
AndrewR: hey all. I have "Forbidden register 0x22D0 in cs at 6708" in dmesg after some wine experiments on rv280/KMS/agp 1.... guess it was R200_VAP_PVS_CNTL_1 ? (form http://www.opensubscriber.com/message/linux-kernel@vger.kernel.org/4305045.html)
amarsh04: anyone like to try to reproduce this bug: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=564592
agd5f: AndrewR: file a bug
AndrewR: agd5f, xorg or kernel?
agd5f: AndrewR: either mesa or drm
agd5f: fdo bugzilla
glisse: AndrewR: yeah it's on my plate might we looking into that today
glisse: also some weird issue with kms & lighting on r200
AndrewR: glisse, in hw tcl mode? I have too dark teapot demo with hw tcl. But it works fine in sw tcl mode. Text still broken (i have kms colortiling ON)
AndrewR: Bug 26061
glisse: yup tcl mode
glisse: AndrewR: is it always broken for you ?
glisse: sometimes for me the it does work, lighting are correct
glisse: i haven't found out yet how it's fix, maybe because i run ums before kms
AndrewR: glisse, not sure. I has it (demo) running only for few seconds. tunnel and some others are fine
AndrewR: glisse, UMS broken for me for some strange reason ... it dies right after Disabling TV message.
AndrewR: glisse, but i don't care too much bout UMS here
AndrewR: *about
glisse: neither us ;)
RiotingPacifist: [drm] failed to load kernel module "radeon" /n (EE) RADEON(0): [dri] RADEONDRIGetVersion failed to open the DRM /n [dri] Disabling DRI.
RiotingPacifist: I don't have a kernel module i compiled KMS into the kernel without recompiling my kernel is it possible to tell xorg to just check sysfs
RiotingPacifist: or just go for it anyway and crash if im wrong
glisse: RiotingPacifist: for kms you need new userspace
RiotingPacifist: i have libdrm (4.17+git20100114) xserver driver radeon (6.12.99+git20100114)
glisse: compiled with kms support enabled ?
RiotingPacifist: i think so, i got them from a ppa though so it's possible they messed up the compile, can i check without recompiling myself?
glisse: pastebin will xorg log
glisse: full
RiotingPacifist: http://pastebin.com/d1938e8f6
RiotingPacifist: I have no xorg.conf, and the only "error" in dmesg is ath5k related
glisse: RiotingPacifist: according to log kms is disabled on your system
RiotingPacifist: sudo cat /sys/module/radeon/parameters/modeset = 1, suggests it works in the kernel does, the error suggests that xorg is failing because it's not a modules.
RiotingPacifist: Do you think its an error with libdrm or xorg itself, rather than the kernel?
glisse: RiotingPacifist: pastebin full dmesg
adamk: Perhaps a firmware issue...
RiotingPacifist: http://pastebin.com/d4061e664, firmware issue is possible im on 2.6.33 i seam to recall firmware issues in earlier release but did not use them myself
BioTube: i'm pretty sure you've got to disable radeonfb for KMS to work
adamk: Oh, get rid of radeonfb.
RiotingPacifist: thx
RiotingPacifist: wait, erm KMS is already working in the kernel (well according to sysfs), are you sure thats causing the problem?
glisse: what in sysfs tells you that kms works ?
BioTube: RiotingPacifist: it says the option's enabled, not that it's working
glisse: according to dmesg it's definitly off
RiotingPacifist: dmesg has "[ 1.360552] [drm] radeon kernel modesetting enabled. ", i don't see the line that tells me it's off (your probably right i just can't see the line)
adamk: KMS will not work with radeonfb compiled in. Period.
adamk: So that's where you should start. If it doesn't solve the problem, then come back and let us know :-)
RiotingPacifist: ok, thx
RiotingPacifist: everything works now sorry for the stupid question i hadn't checked for radeonfb. I assumed if sysfs said that KMS was working it was right, is there any easy way to tell?
adamk: RiotingPacifist: 'cat /proc/fb' should tell you that you are useing radeondrmfb
adamk: And, again the parameters you see via sysfs are just the parameters passed to the kernel.
RiotingPacifist: i mean is there an easy way to tell if kms is working in the kernel?
adamk: I could pass one to the kernel that says "adamrules" but that doesn't mean it will do anything :-)
adamk: 'cat /proc/fb' :-)
adamk: I'm not sure it's always 100% accurate. Heck, even /var/log/Xorg.0.log should tell you if KMS is enabled in the kernel or not.
RiotingPacifist: just after you said Xorg.0.log should be able to tell, my screen shifted itself to the left and locked up, will see if i can reproduce it, quite a strange glitch not sure where it came from kde probably as xorg and dmesg logs said nothing of interest
gsedej: hi! how can I check if I successfully used radeon.agpmode=-1 (or others) in desktop (Ubuntu)
xming: dmesg
gsedej: ok... so if I put radeon.modeset=1, it will enable KMS and DRI2?
xming: it ssould
gsedej: dmesg says: Uknown boot optiom "radeon.agpmode=-1": ignoring
gsedej: what did I make wrong?
jcristau: nothing
gsedej: jcristau, so, why is it ignoring?
jcristau: gsedej: if radeon is a module, as opposed to builtin, the kernel doesn't know at boot time that radeon.agpmode is a valid option
jcristau: so it whines
jcristau: but then when it loads the radeon module it should dtrt
gsedej: so, where it says which agpmode is?
ifvoid: hi
hnsr: hi
ifvoid: does anyone know if I shoudl do something special to get audio on hdmi to work with my HD3200 card?
ifvoid: running 2.6.33-rc3, which should have the patches that support it
ifvoid: is there anything I need to do for example to fiorce use of KMS?
jcristau: make sure your X driver supports kms, and boot with radeon.modeset=1
ifvoid: hm, demsg indeed reports that kms is off
ifvoid: which driver version do I need?
jcristau: master. built against libdrm_radeon.
ifvoid: oh, no release yet?
jcristau: no
ifvoid: ok, let's first see if the radeon.modeset=1 kernel parameter works
jcristau: aiui it will happen once radeon kms is out of staging
ifvoid: well, I can compile it myself for now
ifvoid: hmm
ifvoid: with radeon.modeset=1, booting hangs with "firmware: requesting radeon/R600_rcl.bin"
jcristau: ifvoid: http://people.freedesktop.org/~agd5f/radeon_ucode/
jcristau: put it in /lib/firmware/radeon/
ifvoid: thanks
ifvoid: let me see
ifvoid: (a kernel oops/unbootable system seems like a bit of a heavy penalty for a missing firmware though ;) )
BioTube: it should fall back to unacellerated mode
ifvoid: well, it doesn't ;)
jcristau: this is supposed to be fixed in dave's tree
ifvoid: filing a bug about the missing firmware in debian first
jcristau: according to http://lkml.indiana.edu/hypermail/linux/kernel/1001.1/02878.html
jcristau: it's not missing firmware, debian doesn't have 2.6.33
ifvoid: jcristau: well, the firmware-nonfree package doesn't have it
BioTube: true, but it probably will once 2.6.33 makes it into debian
ifvoid: hmm, still hangs when the firmware is requiested
BioTube: did you build radeon in?
ifvoid: yes
BioTube: then you need to rebuild it with the extra firmware
ifvoid: ah
ifvoid: will the kernel find the firmware in /lib/radeon?
ifvoid: uh
ifvoid: /lib/firmware/radeon
BioTube: only if you set CONFIG_EXTRA_FIRMWARE_DIR to /lib/firmware
chithead: ifvoid: see the wiki link in the channel topic, it contains a section on firmware
ifvoid: ah
ifvoid: overlooked that
jcristau: or just build radeon as a module :)
MostAwesomeDude: Argfl. Hate Ubuntu so much. How do I make radeon early-load?
ifvoid: put it in /etc/modules ?
MostAwesomeDude: Yeah, let's see if that works.
MostAwesomeDude: Yeah, that did it.
ifvoid: jcristau: do I need any specific branch of libdrm? or is the LMS support in master?
ifvoid: *KMS
BioTube: it's in master
jcristau: it's in master, but you need to configure with --enable-experimental-radeon-api
jcristau: or whatever it's called
BioTube: --enable-radeon-experimental-api
ifvoid: is it also in 2.4.17?
BioTube: yes
ifvoid: ah great
rehabdoll: http://www.youtube.com/watch?v=ALEly8dKVCo
rehabdoll: bag in box + microwave ftw
AndrewR: mmm.... in like this LMS typo .. Linux Mode Setting (because implemented only in Linux kernel so far)
xming: ifvoid: I find building radeon as modules is much easier and more flexible
ifvoid: yeah well
ifvoid: I had some trouble with misixng 64 bit kernel and 32 bit userland
xming: well you need git for libdrm, mesa and radeon driver
xming: and kernel .33-rcX or radeon testing
xming: and xorg server >= 1.7
xming: Obscene_CNN: ding dong
Obscene_CNN: yes?
ifvoid: xming: oh, bah I need a new xorg, too?
ifvoid: not according to the wiki link in topic
xming: I think you need xorg 1.7
ifvoid: 1.6.2 suffices, according to the wiki
xming: Obscene_CNN: my glxgears perfromance regression was not caused by your patches
Obscene_CNN: I saw that
xming: ifvoid: packages system of my distro pulled xorg 1.7.x for git radeon driver as requirement
BioTube: it shouldn't be
Obscene_CNN: I may have a new set of patches this weekend. They should perform even better. I'm doing coding changes from things I find in the assembly output.
xming: Obscene_CNN: I just built a kernel w/o your patches and tried ut2004, it was so horrible slow
Obscene_CNN: cool, They definitely have an effect
xming: Obscene_CNN: why don't you have a git repo?
xming: Obscene_CNN: about 5~10% in glxgears, but huge in ut2004 in some areas (pumps up the min fps quite a bit)
Obscene_CNN: to be truthful, I don't do much software developement
xming: Obscene_CNN: but doesn't really increase the mac fps too much (about 5%)
xming: getting patches from pastebin then past in my vi then run patch is a bit painful :D
xming: BioTube: so >= 1.6.2 should be fine?
BioTube: xming: AFAIK, the interface between DDX and X itself hasn't changed
michup86: hi i have some difficulties to run modeset=1 (which im passing via grub) i use radeon hd 3200 and ati 6.12.4-r1 (built in kernel) and kernel 2.6.31-r6
xming: does .31 has kms for radeon?
BioTube: not for r600
michup86: i thought it has
michup86: oh
soreau: Yes, for
xming: not for 3200
soreau: michup86: You will need a newer kernel, probably .33 for that card
soreau: Or if you'll be building your own kernel, use drm-radeon-testing
BioTube: 2.6.32 add r600 support
michup86: damn... okay btw i thought 3200 is based on r600 or r670
soreau: It is an r6xx
michup86: okay i lost you somewhere in the middle but i get the point... update kernel that is
Obscene_CNN: grabs lunch
michup86: i know its not relate to this place but i thought you might help me a bit more... if i upload new kernel and run make oldconfig
michup86: -sorryjustsmashedmykeyboardneedtofix
michup86: now its better :) so if i run make oldconfig it will use settings from previous kernel right? and will just ask for new options... if i will hit just enter it will set all news as default?
michup86: or do i have to specify yeach of it y or n etc
xming_: it depends, but normally hitting enter will do
ifvoid: hah!
ifvoid: unbelievable!
ifvoid: it works!
michup86: tks
ifvoid: now to see if soudn works
xming_: via hdmi?
ifvoid: yep
ifvoid: that why I needed all this experimental stuff
ifvoid: hmm, maybe first get X working again
michup86: and default answer for this kernel setting is highlighted by upper case of a letter?
soreau: Yes
xming_: Y, M or N :D
michup86: ok
michup86: the reason im asking is because i tried to jump from 31 to 32 kernel version and i did it this way with such assumptions and it failed to stand up after reboot with kernel panic
soreau: Well you have to copy the old .config file to the new kernel directory first..
michup86: yes i did
michup86: from 31 to 32 directory then run make oldconfig answer by hitting enter to everything
soreau: and after running make oldconfig, it might be a good idea to double check the most important settings with make menuconfig
michup86: run make -j4 make modules_install make install
michup86: checked /boot
michup86: update grub
michup86: reboot
michup86: suddenly panic
BioTube: did you remember to make the initrd?
michup86: heh, soreau, which might that be ;)
michup86: initrd i have never used that before
michup86: btw im running gentoo
soreau: michup86: Your file system and device drivers for your pc
soreau: You don't need initrd
soreau: especially not on gentoo ;)
michup86: soreau... i had somewhere list of my device drivers but i lost it some time ago... i just was hoping not to hear that ;) i dont like to investigate this... its kind of time consuming
soreau: A kernel panic can be caused by a variety of things though, and usually it's just a misconfiguration issue
soreau: michup86: lspci is your friend ;)
xming_: paste the panic
soreau: michup86: As long as you're using a known working .config file you shouldn't have trouble but if you do, I'd ask in #gentoo first and show them whatever message you had. But first, let's assume this will justwork :)
michup86: yeah but i tend to have problems to find everything in there so im checking by vendors ID and matching them with source code which contain them
michup86: yeah... wish me luck, thanks for help
BioTube: 32 has the localyesconfig and localmodconfig targets, if that helps
michup86: 64 im afraid
michup86: amd x2
BioTube: i meant 2.6.32
michup86: oh
airlied: glisse: yeah I think it depends no how the BIOS sets it up
airlied: snice we see lots of M9 resume failure in bugs
airlied: my M7 doesn't even go into D3 check dmesg at suspend time
airlied: for your M9 it might just refuse to go into D3
MichaelLong: M9, rv250?
gsedej: Where in "dmesg" can you check your KMS, DRI, DRI2, AGP settings?
roysjosh: gsedej, you can see some of them or their results by `dmesg | grep drm'
airlied: MichaelLong: yeah
MichaelLong: yep I'm suffering too :>
gsedej: roysjosh, thx. Do you have ideha where to check AGP settings? (there is non in xorg.conf)
ifvoid: hah!
ifvoid: X also works!
roysjosh: gsedej, there should be an agp line in the output from above (dmesg | grep drm | grep -i agp) or maybe just (dmesg | grep -i agp)
ifvoid: jcristau: thanks a lot!
gsedej: roysjosh, ok, there is line some about agpgart-intel, but one line is "... putting AGP V2 device into 4X mode"
ifvoid: hmm
gsedej: so it means I am running in AGP 4x mode (which is HW max)
ifvoid: crash when I try to play video
roysjosh: gsedej, ok, so it's in 4x mode - yes
gsedej: roysjosh, so if I put in boot "... radeon.agpmode=-1" is should disenable AGP?
roysjosh: gsedej, yes
ifvoid: hmm
ifvoid: but still no sound, it seems...
gsedej: Hi! puting "radeon.agpmode=-1" does not work in ubuntu 9.10. How can I alternatively change agpmode? Should I change xorg.conf? (there is now almost nothing in there)
BioTube: it only work if you're using KMS
soreau: gsedej: What 'does not work' about it? What are you trying to do ultimately?
soreau: In that case he'd need radeon.modeset=1 radeon.agpmode=-1
gsedej: soreau, computer just freezes when using copositing manager (viewport switch). disenableing agpmode might solve problem (right?)
jcristau: so when you say "does not work" you mean it still freezes? or something else?
gsedej: I haven't tried that combination yet...
soreau: gsedej: First try simply radeon.modeset=1
jcristau: then why do you say it doesn't work?
gsedej: modeset works
soreau: This will enable kms and may alleviate the problem
gsedej: if I enable it
gsedej: but I think it freezes even faster
gsedej: sorry. I am trying to solve this problem for half year :S
gsedej: but 3D games are working OK (nexuiz, neverball, even some .NET (mono) opengl frameworks work ok)
gsedej: ok, now I have pcimode :)
ifvoid: yeah, audio also works \o/
gsedej: YEEEESSSSS COMPOSITING IS NOT FREEZING ANYMORE :D:D:D:D I solved problem that took me half an year :D
gsedej: so now... much less performance :(
ifvoid: thaks for your help!
gsedej: thanks from my side too :)
maligor: gsedej, much less performance in what? moving windows?
gsedej: nexuiz works like... very slow :D
gsedej: and wobbly windows are slower too
maligor: try without compositing
gsedej: bit I didn't freeze yet :P
maligor: yeah, I think I've seen mentions that kms lowers performance
xming: pci mode kills perfromance
maligor: oh.. he's in pci mode?
gsedej: maybe agp 1x will work ok
gsedej: now I understand, I have to test :P
xming: isn't that as quick/slow as pcimode?
maligor: I think agp 1x is 66mhz
gsedej: maybe there is hope I will buy ATI in next laptop :P
xming: gsedej: have you disabled fastwrite?
maligor: I'd just skip on laptops ;P
gsedej: xming, I haven't find way to change/check it, but I thing it is disenabled by default
xming: in the xorg.conf
gsedej: maligor, I have nvidia on desktop, and I have almost none problems
xming: and bios might have that too
maligor: laptops and desktops aren't made equal
maligor: the crap they pull in laptops is sometimes unbelievable
gsedej: xming, my xorg.conf is almost clean (some 10 lines only)
xming: you can always put the desktop on your lap :p
xming: Option "AGPFastWrite" "0"
maligor: I think their entire testing procedure is: change variables until windows boots
xming: under Section "Device"
gsedej: BTW: BLUR effect in compiz will be slow?
xming: slow as hell
gsedej: xming, I should try fastwrite with AGP 4x?
xming: froze my computer, took me several minutes to turn it off again
xming: gsedej: don't try fastwrite, disable it
xming: gsedej: then agp 1x, then 2x than 4x
gsedej: xming, blur does not work on my sister HD 47xx card with proprietary
gsedej: ok :)
gsedej: btw, where is the problem of "blur"? no OpenGL 2.x support?
soreau: gsedej: Depending on which blur method used, it needs various features
soreau: Compiz alpha blur already works with the open driver on r3-5xx cards with 4x Bilinear and Gaussian methods
gsedej: soreau, all 3 (gaus, blinear and minimap) are very slow at me...
gsedej: I have RV350 (9600m) and works slooow
soreau: gsedej: You are probably using an older version of mesa if you're not using xorg-edgers repo
soreau: It just recently started working
gsedej: yes :)
gsedej: super :) I will check, when I make compositing stable (and fast)
soreau: In your case it's probably falling back to software routines since you're using old mesa
gsedej: how to check if AGPFastWrite is disenabled?
glisse: airlied: well fixing acpi/pm field might do the trick on enough hw that we shouldn't bother that much on the remaining one unless we are provided with the hw
glisse: wonders since how many years r2xx hav been around
glisse: not that old, last model was in 2004
gsedej: agpmode=1 is not stable.... So I need to use PCI mode?
gsedej: or is there other way?
glisse: weird r300 & rv350 are 2002 & 2003
agd5f: m6/m7 was like 2000 IIRC
gsedej: I have m10 (or 11)
gsedej: how to check if Option "AGPFastWrite" "0" really works?
soreau: Your X log should say if it didn't use the option
gsedej: soreau, sorry, can't find x log. where is it normally?
BioTube: /var/log/Xorg.X.log
gsedej: Xorg.0.log? or 1,2,3,4,5,failsafe?
Wizzup: 0 is the latest afaik
Wizzup: most recent*
BioTube: depends on the screen number the log's from
gsedej: ok
gsedej: (WW) RADEON (0) Option "AGPFastWrite" is not used
gsedej: (--) RandR disenabled
FootLocker: with last commits suspend on rv635 started working, thx for great work
xming: that 0,1,2,3,4,5,... is not due to logrotate, but it's display no.
gsedej: anyway, ubuntu says AGPFastWrite is by default off
gsedej: with AGP off Compiz is stable
gsedej: no freeze with 30 windows opened (normaly it freezes with 10 :D)
soreau: gsedej: You should probably try xorg-edgers repo packages
gsedej: I can remeve xorg-edgers package if it won't work?
soreau: I'm pretty sure that's why 'packages' were designed ;)
gsedej: soreau, ok :D I think I am really stupid now :D
gsedej: there is no way to enable stable agp now... so let's try edgers:P