MarcoPau: adamk: hello. you know what, if I reboot the machine xorg will find ati and radeon. however that doesn't happen with radeonhd, which I wanted to try also for the compiz issue. xorg.conf, instead, is never read
dileX_: Linux 2.6.34-rc2 released
suokko: MarcoPau: radeonhd has nothing to do with 3D
MarcoPau: suokko: so there's still nothing to do with compiz and kde decorator crashing
suokko: MarcoPau: Have you reported a bug with backtrace and steps toreproduce the problem?
MarcoPau: suokko: that bug has been reported a whole bunch of times
suokko: MarcoPau: What is the bug number?
MarcoPau: https://bugs.kde.org/show_bug.cgi?id=143419
MarcoPau: suokko: have you ever been thru that?
suokko: MarcoPau: Looks like wrong place for bug report and nothing to do with drivers
MarcoPau: suokko: yeah they say it's a compiz bug but still I can't find any help around
suokko: kde-window-decoration bug
suokko: That is different package
MarcoPau: suokko: so what should I do? do you have any hitn?
MarcoPau: hint*
suokko: MarcoPau: Your best hope is to locate the package that is buggy for you and report it to correct developers
suokko: MarcoPau: Try running the kde-decorator in valgrind and see from the log where the memory errors are coming
MarcoPau: suokko: what's valgrind?
suokko: MarcoPau: debugging tool that is good at checking errors in memory access
MarcoPau: suokko: alright, thanks for the hint
yoshi314: hi there, anybody around for a quick question?
dileX: yoshi314: just ask
yoshi314: i think dri doesn't work on mesa 7.8rc1 with rv515
yoshi314: is it normal that glxgears take 70% cpu time when running ?
Thunderbird: yes
Thunderbird: it is a useless test
yoshi314: hmm i though that cpu burden would be lower
Thunderbird: it attempts to submit as much polygons to the gpu
Thunderbird: it is a very basic GL app
yoshi314: darkplaces also gotten very sluggish, and it stops rendering every few seconds
yoshi314: i'm not sure what's going on
yoshi314: maybe darkplaces also changed something
yoshi314: i guess i'll try disabling BFS then, maybe that's the cause
dileX: yoshi314: whats BFS?
yoshi314: the cpu scheduler
dileX: brainf***
dileX: I once tried a real-time kernel - no good with 3D :-)
yoshi314: well i didn't quite set it for realtime
yoshi314: i played with it before and i didn't have that kind of problems
yoshi314: well, time to see without bfs then. i'm really lost what to try otherwise. disable kms ?
[Enrico]: yoshi314: UMS performance are likely better then the KMS one for now
dileX: yoshi314: try UMS, radeon.modeset=0 or nomodeset (e.g. in grub-line)
yoshi314: ok, rebooting with CFS for now ;)
mjt: is there any reason to prefer amd785g vs amd760g with current radeon driver abilities? :)
Pallokala: afaik no
GoGi: can some form of anti aliasing be enabled?
GoGi: http://www.x.org/wiki/RadeonFeature says "UNKNOWN"
Thunderbird: I would say there are more urgent things to fix; further anti aliasing takes a nice performance hit and it consumes a lot more video memory
mcencora_: Thunderbird: is there an other way to lock a texture besides glGetTexImage (it's slooow if PBOs aren't available)
Thunderbird: I don't think so
Thunderbird: we either use that or glReadPixels (yuck)
mattst88|laptop: glReadPixels is accelerated now, isn't it? I think that's been a rather recent development
evil_core: GoGi: its not, because multisampling is nit implemented
evil_core: GoGi: currewntly you cannot even get smooth vertex line
mcencora_: mattst88|laptop: it's but since we don't support PBOs there's almost no speedup
mcencora_: mattst88|laptop: VRAM->GTT transfer is the most expensive part
evil_core: KMS breaks some things in HW
evil_core: dethkartz doesnt run after using kms(and getting back to ums)
evil_core: Can I find what causes that?
mcencora_: airlied, glisse: the state atoms emission is broken for bigger atoms. not sure about others but vpp, vpi, r500fp are certainly affected
mcencora_: the count field in almost every union of drm_r300_cmd_header_t is unsigned char (max 255)
mcencora_: that allows us to write max 64 constants for VPU
mcencora_: IIRC wine uses 256 constants always when relative addressing is used
mcencora_: so such apps are broken
suokko: mcencora_: UMS only right?
mcencora_: also programs that are longer than 64 insts will be broken too
mcencora_: unfortunately not
mcencora_: KMS is affected too
suokko: hmm. I think kernel side is not using drm_r300_cmd_header_t in kms
suokko: code
mcencora_: the drm_r300_cmd_header_t holds the count fields, which are used in atom emission (see emit_vpu function in r300_cmdbuf.c for example)
mcencora_: s/emit_vpu/check_vpu/
mcencora_: the cnt var is read from the drm_r300_cmd_header_t.vpu.count which is uchar
suokko: mcencora_: That structure shouldn't be used in KMS
mcencora_: suokko: but it is
suokko: So mesa should be fixed
mcencora_: suokko: any ideas how to fix it without rewriting major pieces of code in all radeon drivers?
suokko: mcencora_: That should only affect r300
codingrobot: glxgears shows ~550fps on radeonhd 4200 with 6.12.5 and radeonhd from git, is it supposed to be that slow?! i'm running debian squeeze with 2.6.34-rc2
suokko: codingrobot: glxgears is not benchmark
codingrobot: xbmc is extremely slow too
mcencora_: suokko: r600 doesn't use the state atoms?
codingrobot: that's why i trieg glxgears
suokko: mcencora_: It uses them differently
suokko: And r200 does it differently too
Jonimus: codingrobot: what is the output of glxinfo|grep Open
mcencora_: suokko: good, less code to change
suokko: To obad that probably needs separate emit functions for KMS and UMS to keep code cleaner
codingrobot: renderer: Software Rasterizer; opengl: 2.1 Mesa 7.7.1-DEVEL; shading language: 1.20
codingrobot: i'm running w/o KMS
Jonimus: codingrobot: thats your issue, DRI isn't working,please pastbin your dmesg and xorg.0.log
mcencora_: suokko: do you know where the extra value for KMS case in check_vpu comes from?
codingrobot: Jonimus: i booted with radeon.modeset=0
codingrobot: drm is working and the firmware is there, r600_dri.so i loaded too
suokko: mcencora_: I don't know
Jonimus: codingrobot: like I said please pastbin your dmesg and xorg.0.log so we can take a look to see the issue
MarcoPau: suokko: I'm running valgrind kde4-window-decorator. is that ok?
codingrobot: http://pastie.textmate.org/private/leukje41pcpefmvycgq3dg
Jonimus: codingrobot: hmm thats odd, I don't see why you would be using the swrast, everything there looks good
Jonimus: codingrobot: can you post a dmesg as well
codingrobot: and thats the kernel output when staring X: http://pastie.textmate.org/878643
codingrobot: before that it's just [drm] Initialized radeon 1.32.0 20080528 for 0000:01:05.0 on minor 0
Jonimus: codingrobot: what Distro do you use?
codingrobot: debian testing
MarcoPau: suokko: http://pastebin.com/N9Lc7sn2 this is the output, if you're so kind to take a look I'd be grateful
Jonimus: hmm, I think the userspace is just maybe to old then, see the link in the wiki on how to build a newer libdrm, mesa, and ddx
codingrobot: yesterday i tried installing fglrx but it failed, so i just removed the crap and reinstalled the xorg packages
adamk: codingrobot: Psatebin the output of 'LIBGL_DEBUG=verbose glxinfo'
codingrobot: oh.. thats what i got on stderr: http://pastie.textmate.org/878650
adamk: Heh...
adamk: What group owns /dev/dri/card0?
codingrobot: chowned it and glx shows AMD as vendor string
adamk: You'll want to add yourself to whatever group was owning that device.
adamk: Becuse when you reboot, it will revert back.
suokko: MarcoPau: I would need source code to understand what goes wrong. But it sounds like call invalid function pointer
MarcoPau: this one is more eloquent: valgrind -v http://pastebin.com/fxCizpCH
MarcoPau: suokko: is there anything I should do with that? maybe reporting the whole valgrind -v kde4-window-decorator to bug kde?
Jonimus: codingrobot: add yourself to the video group, that should hopefully fix it then
codingrobot: adamk: at least we know where the problem was. fps is now 1700 but xmbc show artifacts and i get a msg "IRQ's not enabled, falling back to busy waits: 2 0"
Nerihsa: hi
suokko: MarcoPau: kde4-window-decorator is not part of kde
suokko: mcencora_: It looks like there is already special emit function for KMS but that extra 5 is wrong
MarcoPau: suokko: compiz bugs?
suokko: MarcoPau: yes.
Garrappachc: hi
Garrappachc: just one question
Garrappachc: my graphics is Radeon HD5450
Garrappachc: I use Gentoo
Garrappachc: with drivers from x11 overlay
Garrappachc: and almost everything works
Garrappachc: but there's no 3d accelerations
Garrappachc: i mean there is
Garrappachc: as glxinfo says
Garrappachc: but compiz doesn't work
soreau: There is no 3D support for HD5xxx yet
Garrappachc: aaah
Garrappachc: ok ;)
Garrappachc: when can I expect?
soreau: http://www.x.org/wiki/RadeonFeature
Garrappachc: yap
Garrappachc: i know this site
soreau: HD5xxx is r8xx or Evergreen
Garrappachc: i visit it everyday
Garrappachc: ;)
Garrappachc: also radeonhd
Garrappachc: there's kms available for my Radeon
Garrappachc: in 2.6.34
Garrappachc: but xorg works not very well on it
soreau: it?
Garrappachc: on 2.6.34
Garrappachc: with kms
Garrappachc: i know i may be seen as a noob
Garrappachc: but i used to use intel
Garrappachc: so i am not updated ;)
soreau: You probably dont want to use radeonhd
Garrappachc: why?
mattst88|laptop: it's not developed anymore.
Garrappachc: aaaah
Garrappachc: i didn't know ;)
soreau: most all of the work is happening in the radeon driver
Garrappachc: yes
Garrappachc: i have up-to-date drivers
Garrappachc: form x11 overlay
Garrappachc: *from
yangman: I'm surprised it works with a 5450 at all
Garrappachc: so am I ^^
Garrappachc: but for example Alt+Ctrl+F1 gets me blank screen
Garrappachc: everything i can do is to came back to x
Garrappachc: Ctrl+Alt+F7
Garrappachc: so not everything works
Garrappachc: but i dont need xorg.conf
Garrappachc: and i have composition enabled
Garrappachc: there's one thing to make me happy ;)
mcencora_: suokko: yay! the Hitman Blood Money game works when fixing this big state atoms
yangman: Garrappachc: if you're running configfile-less, then you're not running radeonhd ;)
mcencora_: suokko: there're some major rendering artefacts (Z order problem, missing textures) but it works
suokko: mcencora_: btw, Yo ucould extend the state atom to hold extra members if you define check and emit function
Garrappachc: yangman: yap, i know
Garrappachc: but i used to think that radeonhd is newest
Garrappachc: testing
Garrappachc: unstable
Garrappachc: but newest
Garrappachc: as arch-wiki says
suokko: mcencora_: then yo ucan generate the header in emit from etra varaibles
soreau: Garrappachc: Do you have a particular question about the radeon driver?
Garrappachc: not really
Garrappachc: thanks for everything
pzanoni: hello! I just got a radeon hd 4770 (rv740) with very recent software (mesa 7.8, kernel 33, kms) and I can't even watch videos with XV. Is that expected or am I missing something?
suokko: pzanoni: see build how to in topi
soreau: pzanoni: What happens when you try to watch a video with xv?
pzanoni: soreau: mplayer -vo xv tells me that there is no xv available
pzanoni: soreau: and xvinfo tells me there are no adapters present
soreau: pzanoni: Can you pastebin your x log?
pzanoni: soreau: http://pastebin.mandriva.com/17560
soreau: pzanoni: You should get rid of the vga= line for your kernel
pzanoni: soreau: ooops... missed that grr
soreau: but that probably doesnt have anything to do with xv
pzanoni: soreau: thanks
pzanoni: let's try
pzanoni: soreau: maybe the xv support was not added for this chip yet:
pzanoni: ?
soreau: no, it should work
Jonimus: hmm I guess that means I should update the ATI page on the arch-wiki about radeonhd
suokko: pzanoni: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch.
soreau: pzanoni: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch.
soreau: heh
svenstaro: Good morning arch people
soreau: pzanoni: You need to rebuild libdrm, then mesa and ddx against that libdrm
soreau: If you use latest libdrm, you shouldnt need --enable-experimental-radeon-api
Jonimus: hello svenstaro
soreau: Does fbcon have hi-res console on evergreen yet?
syncer: Hi all!
syncer: can sombody tell me which chip fully supported by radeonhd with 3d?
syncer: sorry for my poor english
svenstaro: syncer: This channel is not really about the radeonhd driver though.
svenstaro: Mostly the free -ati driver.
syncer: ah.. ok
Curan: syncer: the support of the radeon driver can be seen at http://www.x.org/wiki/RadeonFeature
syncer: thanks!
Curan: (btw: AFAIK radeonhd isn't developed anymore)
syncer: Curan: it's merged with -ati?
Curan: syncer: what is merged with -ati? the -ati driver package contains the ati wrapper, which loads either the radeon driver or the r128 driver, depending on your hardware
syncer: Curan: sorry for stupid question
Curan: syncer: but I'm not aware of any code merged between radeonhd and radeon, though I might be wrong, I'm not a dev for X stuff
svenstaro: In most cases, the radeon driver is preferred to radeonhd.
Curan: syncer: just idle around here to stay informed what I'm able to see soon ;)
Curan: syncer: no need to apologize, at least not to me *g*
syncer: Curan: can you advice best to use with -ati driver chip?
Curan: syncer: hm, hard to say: older but well supported are the chips up to R400, I think the latest driver and kernel bring R500 and R600 up to speed too (didn't test that extensivly so far... no time)
Curan: syncer: R700 (evergreen) is very much work in progress
soreau: syncer: Did you look at the radeon feature matrix wiki page that Curan gave you a link to?
Curan: syncer: so probably a R600 chip would be best atm (which is btw more or less what's shown on the wiki page I linked to earlier)
syncer: soreau: yes
soreau: syncer: I think the r5xx chips probably have a good bang to buck ratio
adamk: Uhhh... r800 is evergreen.
syncer: soreau: Curan svenstaro thanks!
adamk: r700 has the same support as r600.
soreau: indeed
Curan: adamk: yeah.. typo
syncer: adamk: thanks!
Curan: sry
syncer: and one more question, can be compiled -ati under MIPS for example?
Curan: syncer: yes, Debian does it, see e.g. https://buildd.debian.org/status/package.php?p=xserver-xorg-video-ati
syncer: many thanks guys!
Curan: (but I'm pretty sure the other distros do that too, but I'm not familiar with them)
yokto_: hi up to now i have always used env variables and symlinks to test gallium from git without actually installing - now with mesa 7.8rc i have a radeong driver in the mesa package but i don't know the proper way to use it - I cant put it in xorg.conf since there is no radeong ddx driver
soreau: yokto_: There is actually a radeong state tracker that you can test..
soreau: There's the mesa gallium driver radeong_dri.so and the x st radeong_drv.so that mesa can be configured for iirc
yokto_: so in where is radeong_drv.so in xf86-video-ati?
yokto_: i see radeong_drv.so is in mesa but how can it be usualy all the *_drv are in xf86-video-* packages
stikonas: yokto_: gallium3d is not yet ready, don't use it
yokto_: ok
Anpu: Hi, I have a problem with ATI Radeon 9250 and opensource drivers. I m getting following errors whenever I move mouse while running one specific program (which doesnt work anymore)
Anpu: Failed to allocate :
Anpu: size : -27392 bytes
Anpu: alignment : 4 bytes
Anpu: domains : 2
Anpu: program name is "Regnum Online" (native Linux MMO)
Anpu: it worked before in 9.04, but not anymore after Kubuntu upgrade
Anpu: (I m currently running Lucid beta)
Anpu: anyone has any idea whats the message i keep getting on ever mouse move?
Anpu: every*
GNUtoo: hi, that is missing in ubuntu karmic and derivated: /usr/lib/dri/r600_dri.so how do I add it? is it missing because 3d can hang the computer?
soreau: GNUtoo: No, it's missing because karmic released before mesa supported 3D for r6-7xx
GNUtoo: ok I saw that on the net
GNUtoo: which ppa do I need
soreau: If you want 3D for these cards on karmic, you will need a newer kernel
soreau: and packages from xorg-edgers repo
GNUtoo: I have 2.6.33.1 running right now
GNUtoo: so xorg-edgers
GNUtoo: can it lock my computer?
soreau: sure
GNUtoo: ok
GNUtoo: so as before with x700
GNUtoo: will depend on the games
soreau: It can blow up your machine and catch your house on fire
soreau: we're still not responsible ;)
soreau: What card model do you have as reported by lspci|grep VGA ?
GNUtoo: 01:05.0 VGA compatible controller: ATI Technologies Inc Device 9710
Ghworg: radeon turned me into a newt
soreau: Did you say that's x700?
soreau: if so I think it only needs r300 mesa driver which is already installed in karmic by default
GNUtoo: Solefald, x700 was my previous card in my laptop
GNUtoo: now I've a new desktop
GNUtoo: I used gentoo with x700 in the past
GNUtoo: was not very stable
GNUtoo: I restart xorg
soreau: what is a Device 9710 then? :P
soreau: I don't see 9710 in the wiki gpu comparison chart
maligor: sounds like a device id
maligor: err.. pci id I mean
maligor: 9710 RS880 [Radeon HD 4200]
soreau: RS880 maybe?
soreau: oh yea
GNUtoo: thanks a lot it seem to work fine
soreau: cool
rehabdoll: when can one expect a 6.13 release? days? weeks?
evil_core: http://cgit.freedesktop.org/~airlied/mesa/log/?h=r300g-buffer-rework
evil_core: shoukld I use this branch for r500?
airlied: evil_core: half of its merged now
airlied: I need to rebase the other bit to finish it off
evil_core: and it really will give so much fps?
airlied: evil_core: thats the theory, though it got slower with some fixes, need to sysprof again
evil_core: hmm..so I will stay with master for now maybe
evil_core: btw, I got problem with Dethkarz in Wine(Glide), it stops running after loading radeon in KMS
evil_core: even after reverting to UMS on the fly
evil_core: dunno what causes that, I love KMS, but I love Deathkarz more ;)
evil_core: the worst is I am forced to reboot, because KMS is messing something in hw
evil_core: anybody can give me links to patches imporving performance on r500?
airlied: a 1.8 X server cah elp dri2 in theory
airlied: not sure if all patche sare in yet
soreau: airlied: With this new change that causes default mode to be 800x600 and max of 1024x768 until you add 1280x1024 mode on rv350, adding a mode line in xorg.conf 'fixes' it, but the screen is moved over to the left about 1.5 cm
airlied: soreau: I don't remember mkaing a change like that, we should be picking 1024x768 by default I thought
soreau: airlied: Well I'm using 2.6.34_rc1 I believe, not drm-radeon-testing
evil_core: soreau: it sounds like higher refresh rate on some LCDs
soreau: if it is working in d-r-t, I could try latest of that
soreau: evil_core: It's crt
soreau: afaik
soreau: but it's not lcd
otaylor: soreau: lcd vs. crt is probably irrelevant, sounds like DDC problems
airlied: soreau: ah its the broken hw 2ic
evil_core: so its shouldnt be problem than as long as timings are correct and supported by display
soreau: airlied: Yes, that is when the problem started
airlied: soreau: there is a patch to disable that
soreau: airlied: at the introduction of i2c changes
airlied: just have to merge it
soreau: airlied: I'd rather it be fixed than disable it :P
airlied: soreau: yeah we are having some trouble figuring out hw i2c clocks on some chips
soreau: airlied: Ok
airlied: its not really a useful feature if its broken
soreau: right
airlied: we mainly only really need it for accessing on-board i2c chips
soreau: airlied: but if it goes into .34...
mcencora: airlied: have you seen my messages about problems with big state atoms?
airlied: mcencora: nothing comes to mind
airlied: soreau: hence why we are merging a ptach to turn it off by default
soreau: airlied: I see
soreau: That makes sense
soreau: airlied: I assume before .34 release, it will be disabled by default and have some option to enable it?
evil_core: drm-radeon-testing is based on 2.6.33?
evil_core: LD fs/built-in.o
evil_core: error: Bad exit status from /home/users/evil/tmp/rpm-tmp.63358 (%build)
evil_core: I got such error when trying to build 2.6.33.1 with diffied DRM dirs from drm-radeon-testing :/
evil_core: looks strange, because no errors displayed, only bad exit
AstralStorm: hello there
AstralStorm: seems aspect scaling mode is broken *again*
AstralStorm: (KMS, r6xx, start of march, 2.6.33 + drm-radeon-testing likewise)
AstralStorm: actually all scaling modes end up with the same result, wth?
AstralStorm: maybe option values have changed or something?
AstralStorm: xrandr --output LVDS --set "scaling mode" aspect does zilch
AstralStorm: no error, but no effect
AstralStorm: same with center
AstralStorm: hmm, Center works
AstralStorm: but what's the name of the aspect-correct scaling?
AstralStorm: not Aspect
DanaG: xrandr --verbose
DanaG: may tell you what the choices are.
AstralStorm: heh it does
AstralStorm: "Full aspect" blah
evil_core: /usr/bin/install -c -d //opt/xorg/lib/xorg/modules/drivers
evil_core: /bin/sh ../../../../../../bin/minstall -m 755 radeong_drv.so //opt/xorg/lib/xorg/modules/drivers
evil_core: Unknown type of argument: radeong_drv.so
evil_core: gmake[5]: *** [install] Error 1
evil_core: can I do something with that?
DanaG: typo.
evil_core: wheres typo?
evil_core: its mesa from master
evil_core: 64bit version build, 32bit not
evil_core: radeong is gallium state tracker
DanaG: oh.
DanaG: I didn't know that.
DanaG: =รพ
soreau: now ya know
k-man: hi, i am trying to get tv-out working, when i boot it works but once it switches to graphics mode for xorg, the tv can't cope with the signal it gets from the card i think. also the monitor that i have connected also says "out of range"
k-man: how can i work out where the problem is?
damentz: hey guys, is it necessary to define the extra firmware for R600_rlc.bin and R700_rlc.bin?
damentz: or will those be found anyway?