Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-6-17

Search This Log:

BeerSerc: Hi there. I finally managed to get the radeon driver working, fglrx works. But I cannot enable KDE 4's desktop effects. Which options do I have to set for that to work?
soreau: BeerSerc: fglrx is not the radeon driver
BeerSerc: oh it's too early. s/fglrx/glxgears/
soreau: Please see the topic
soreau: Oh ok
BeerSerc: just got up, sorry...
soreau: In that case, what is not working with kde effect? Are you trying to use kwin or compiz?
BeerSerc: kwin. I cannot enable them, the control center tells me my X isnt configured the right way
soreau: What is the output of 'glxinfo|grep renderer'?
BeerSerc: software rasterizer
MostAwesomeDude: BeerSerc: If GL doesn't work, use the Xrender backend.
soreau: BeerSerc: Software Rasterizer basically means 3D is borken. Pastebin /var/log/Xorg.0.log
soreau: MostAwesomeDude: Xrender works for kompositing?
BeerSerc: MostAwesomeDude: doesnt work here
BeerSerc: soreau: http://nopaste.com/p/a5CqjgnW0
soreau: (WW) RADEON(0): Direct rendering disabled
soreau: (II) RADEON(0): XAA Render acceleration unsupported on Radeon 9500/9700 and newer. Please use EXA instead.
BeerSerc: ah, it says "disabling DRI" ...
MostAwesomeDude: Scroll way up. Your dri module isn't loading.
MostAwesomeDude: Gonna have to reinstall/rebuild your X.
soreau: and maybe mesa too (?)
BeerSerc: MostAwesomeDude: Hm, recompiled that all yesterday ... should all be up to date. Lets see ...
MostAwesomeDude: soreau: libdri just isn't getting loaded, that's the source of all the problems IMO.
zhasha: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed (libdri.a too old)
zhasha: there's y'ar problem
BeerSerc: grrr it's a dead symlink still from fglrx ...
stikonas: radeon KMS is already in main linus tree
airlied: \o/ finally
airlied: does a little dance
hifi: so, 2.6.31 it is?
stikonas: yes
airlied: well its in staging
airlied: but at least we have a good starting point
stikonas: fedora and ubuntu enables staging anyway
hifi: YES for ubuntu o/
airlied: they won't ship radeon kms
airlied: it takes a lot more than just the kernel
hifi: airlied: ppa my friend
airlied: none of which is released yet
airlied: hifi: true it'll get all of it
hifi: the kernel part was the tricky one
airlied: MrCooper: hey, so this compiz crash what hw you on? (still not seeing it here :(
hifi: everything else is quite easy to get from a ppa
MrCooper: airlied: always my RV350 PowerBook
MrCooper: airlied: it can be tricky to reproduce; I tried with all gtk-demo windows open on a naked desktop but failed, but my normal GNOME session reproduces it reliably
MrCooper: airlied: anyway, I think I've seen people report that assertion failure in other circumstances, apparently something is missing to enforce the assertion?
BeerSerc: reboots with a fresh libdri.so ...
airlied: MrCooper: I'm running lots of things but can't make it happen :( and assert definitely seem to work here.
airlied: I'll keep going I'm bound to hit it eventually
BeerSerc: shiny kde effects now. thanks everybody...
airlied: MrCooper: you don't get a gpu lockup after it or anything?
MrCooper: no, just X restart
BeerSerc: I have heard there's some work going on concerning power saving in latest git, is that worth a try?
MrCooper: airlied: does your session have something like gkrellm that updates constantly? Maybe that 'does the trick' for me
airlied: adds gkrellm, can still alt-tab
airlied: will open up a few more things
MrCooper: airlied: I take it you're using the static application switcher plugin?
airlied: MrCooper: I'm using the one that came with compiz :)
airlied: not sure which one that is
MrCooper: probably the other one
airlied: oh lookee there it is
MrCooper: when you alt-tab, do you see several lines of thumbnails, of all windows? Or just a single line with moving thumbnails
airlied: I've switched over.. that thing is slooow
MrCooper: probably EXA offscreen memory thrashing
airlied: woot crashed it
MrCooper: yay
airlied: got the bo count < 32 fail
airlied: thanks thats half the battle
airlied: failure to revalidate BOs - badness
airlied: I like my debugging msgs
MrCooper: hehe
taiu: great! with mem=2048m cp initializes
taiu: without it and 8G mem it fails
airlied: MrCooper: thanks for that EDID stuff btw
airlied: need to give it a run on the G5 soon
MrCooper: airlied: np, turned out to be less hard than I thought when I saw all those bitfields first
MrCooper: heh, just hit the assertion failure with torcs as well
airlied: MrCooper: how much RAM on that card you have
MrCooper: 128M, but only about 35M for textures
taiu: airlied: http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=2ff2e8a3e1898e692b604424c384f134009dea80
taiu: dows this cause x failure with radeon
airlied: osiris_: it was either your draw or fallback chanegs that busted my bo validate patth
airlied: need to go figure it out
airlied: taiu: it shouldn't, the patch after it should have fixed radeon
airlied: taiu: where I remove _DRM_DRIVER
taiu: argh, fails for me
airlied: got a log?
taiu: is that one also in linus tree
taiu: X or dmesg?
airlied: its only in Linus tree + the fix
airlied: what kernel you using?
airlied: and what config?
taiu: linus git
taiu: KMS
airlied: okay KMS is busted I think, glisse mentioned it but never sent me a patch
airlied: just remove _DRM_DRIVER from the kms file
taiu: in radeon_kms.c kmsret = drm_addmap(dev, 0, sareapage, _DRM_SHM, _DRM_CONTAINS_LOCK| _DRM_DRIVER, ?
airlied: yup just remove _DRM_DRIVER
taiu: ok
airlied: just send patch to Linus
taiu: ok, works better
taiu: now fails further on.. [drm:radeon_cp_getparam_kms] *ERROR* invalid ioctl with kms radeon_cp_getparam_kms
BeerSerc: hey, I just updated to latest git version, and now I am getting no DRI again, with the following message in Xorg.0.log: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch.
BeerSerc: [dri] radeon.o kernel module version is 2.0.0 but version 1.17.0 or newer is needed.
airlied: you enabled KMS.
airlied: wonders how people are doing that
airlied: its not like its obvious
BeerSerc: yes I did
airlied: well then you'll break your current userspace
airlied: and you have
BeerSerc: shouldnt it work with recent mesa and libdrm?
BeerSerc: and why isnt it obvious?
airlied: no, we aren't magic, we can't add support for something that isn't there yet
BeerSerc: ah OK.
airlied: you add support to the kernel then you add support to the userspace afterwards
airlied: wonders which bit of the CONFIG_STAGING warning people missed reading
BeerSerc: I read your mail on lkml and thought as you were maybe using it before, it could be in the latest git...
airlied: no its all in branches, drm and 2D driver
airlied: mesa is in master, if built against the right drm branch
airlied: hopefully we'll start merging bits this week
airlied: drm modesetting-gem branch and DDX ~glisse/xf86-video-ati radeon-gem-cs3 branch I think
airlied: will pull the DDX into the main DDX repo soon
BeerSerc: ah OK. Great work by the way. Perfect timing as ATI just discontinued the drivers ...
dileX: airlied: cool, drm-linus part #2 merged!
airlied: need to re-write the DDX code, it got a bit messy as I worked it all out
airlied: okay kms-support branch in the main DDX
Zajec: does radeon have it's tool for registers reading and writing?
airlied: radeontool
airlied: or avivotool for later cards
Zajec: ok, thanks
glisse: airlied: patch are underway
glisse: i am doing more testing just to be sure things work
glisse: at least a little
glisse: it will be in my radeon-dave branch
glisse: they should apply on linus tree cleanly
airlied: glisse: I send the _DRM_DRIVER one :)
airlied: okay libdrm is now on master
airlied: --enable-radeon-experimental-api
airlied: libdrm_radeon even
airlied: and kms-support branch of DDX
airlied: and master of mes
airlied: and master of mesa
taiu: ddx of kms-support works :)
dileX: airlied: where is the _DRM_DRIVER one?
stikonas: airlied: do you know when master branch of DDX will be merged into kms-support? I think master has some scaling fixes for XVideo.
airlied: stikonas: more likely we'll recreate kms support on master
airlied: kms support has some ugly bits I'd rather avoid merging
glisse: stikonas: the bug is that on resize video isn't resized ?
stikonas: glisse: no, I can only see part of the video in the whole media player window
stikonas: only top left part of the video is seen
stikonas: agd5f had some patches
stikonas: glisse: http://stikonas.homelinux.org/files/bbb.png
stikonas: glisse: http://www.botchco.com/alex/xorg/kms_fix_xv.diff or http://www.botchco.com/alex/xorg/kms_fix_clipping.diff but I don't know if they apply now
sylware: Is there documented the access to the r7xx fan control/temp sensor?
sylware: atombios?
Zajec: if i would like to write new kernel module accesing radeon gpu registers...
Zajec: what way should I choose?
vehemens: airlied: kms is glisse cs3 branch renamed?
Zajec: memory mapping? os something other?
stikonas: vehemens: yes
MrCooper: Zajec: why a new one rather than extending the existing one?
sylware: Can we provide Linux GPL only code for the driver?
Zajec: MrCooper: ok, i just found radeon_backlight.c... but it's not compiled by default for me...
Zajec: er, don't understand that module
Zajec: ah, it's part of radeonfb.ko
MrCooper: Zajec: backlight functionality should probably be added to radeon KMS
Zajec: ah, good to know
suokko: http://img230.imageshack.us/img230/8400/screenshot1vlw.png <- interesting screen corruption with radeon KMS and mesa 7.6. I have forced my card to pci moe becaue AGP mode doesn't recover after same corruption.
Zajec: er, just forget there is not KMS for r6xx :D
suokko: metacity is my window manager in that screenhot but compiz did cause similar problems in agp mode too
suokko: card is mobility 9200 (rv280)
airlied: sylware: depends on how useful the code is, but I'd rather keep it useful for BSD if they wanted to take it all
sylware: I do not endorse the BSD philosophy
sylware: Then I'll have to code on the side then
airlied: well whats the code?
sylware: For the moment, nil
sylware: Just getting info
airlied: well we'll worry when you produce the killer feature
sylware: Ok
sylware: BTW, do you know if the fan control on r7xx is documented?
airlied: not yet I don't think, bridgman might appear and tell us
airlied: I think the r7xx has integrated sensor/control
sylware: He told me, it's not documented yet
sylware: airlied: yes
sylware: That what he told me
airlied: its on their list, but Alex is on holidays
airlied: so not sure if we'll get any info until he gets back
sylware: Right... because that would just to calm donw my fan during coding...
sylware: Because it runs to the max, it's *very* noisy
airlied: you could use the atom table to drop memory/engine clock
airlied: the fan might go down automagically
airlied: I'm not sure how automagic the on-board sensor control is
airlied: the bios might configure a useful temp profile in it
sylware: that's a bit what I'm looking for
airlied: well the code to drop memory/engine clocks is already out there
sylware: Ok
sylware: I'll try that
vehemens: BSD want it all!!
airlied: you can probably play with userspace -ati driver to do it
airlied: I think alex added some options to set a low power mode
sylware: airlied: cool
sylware: I'll try all that
airlied: Option "ForceLowPowerMode" "true"
airlied: should do it with the normal ddx
sylware: yeah!
sylware: you saved my day
sylware: I'll try it tonight
sylware: And start reading the specs.
sylware: (without a noisy fan)
suokko: Playing neverputt without lightning is a bit challenging :) http://yfrog.com/7gscreenshot3cwep
max_r: any plans for r6xx-r7xx-3d drm to 2.6.30? Or will be any 2.6.30 kernel with r6xx/r7xx 3d bits?
airlied: need to make sure the bits are useful first, but I think we might start upstreaming some r600 CS code at smoe point with kms
max_r: and about making it useful: are we back to friday weekly drop of code to r6xx-rewrite?
airlied: not sure how it was working, alex is on holidays
airlied: so it might be a bit slower.
max_r: 14:12 < airlied> not sure how it was working, alex is on holidays
max_r: 14:12 < airlied> so it might be a bit slower.
max_r: oops
max_r: sorry
max_r: and how long alex's holidays lasts?
airlied: I think they let him out for two weeks, maybe three
max_r: bridgman said that richard is working on 3d support now. Alex used to commit richard's code every friday, interesting to see who will do it this week if it will ever happen
suokko: Can read what everybody is saying in irc? http://yfrog.com/5jscreenshot4tnjp
slytherin: Can anyone suggest me some workarounds for a problem in dual monitor setup?
adamk_: Only if you tell us what the problem is.
slytherin: adamk_: I have an ibook G4 (powerpc) with radeon 9200 card. I tried connecting my new LCD TV to the VGA output. I get white screen on TV while the machine is booting. TV says the resolution of white screen is 640x480. As soon as GDM starts the TV screen shows 'No Signal'.
slytherin: I have verified that the VGA cable is fine since it works with my friend's laptop. Also my olc PC monitor (Daewoo 14" CRT) with it's own cable works fine with my laptop.
airlied: slytherin: does running xrandr show anythinh?
slytherin: airlied: xrandr shows second display (TV) 'disconnected'. Where as in case of my CRT monitor it shows connected and the resolutions supported.
slytherin: I was wondering if this was anyway related to EDID so I tried IgnoreEDID option. But that did not seem to help.
airlied: try turning on dac detection I forget how though.
airlied: xrandr --output --set load_detection true
slytherin: airlied: Ok. Will try when I go home.
airlied: might be 1 instead of true
airlied: but something like that
slytherin: airlied: anything else I could try? like AGPMode, IgnoreEDID etc?
airlied: the only other thing I can think is to add a mode explicirtyly
airlied: but xorg options won't help
airlied: if load detect fails though it really means the card can't see anything
airlied: we don't do auto loaddetect on some DACs as it returns false positives
slytherin: thanks for help. I will probably login in my night time to give feedback.
slytherin: airlied: by the way, the card seems to see at least something since the TV gets some signal (even though it is complete white screen) at the bootup. It is only after loading GDM the problem occurs.
airlied: yeah the bios probably initis it
stuckey_: Hello
stuckey_: Does anyone know which ATI cards are supported by the open-source driver, and how well they are supported? I just need something that can play a few games and be used to watch x264 movies.
slytherin: damn, I don't know how to deal with openfirmware configuration.
chithead: stuckey_: http://wiki.x.org/wiki/RadeonFeature
slytherin: stuckey_: I believe cards till 9500 are supported well.
airlied: slytherin: no need if the load detect works
chithead: stuckey_: if you want 3d, the open source drivers support up to radeon x1950 atm
MrCooper: airlied, glisse: looks like someone should bribe you guys to build the DDX kms branches with -Werror :}
slytherin: airlied: got to go. thanks for all help.
glisse: airlied: did you pushed Alex tv dac patch to Linus ?
airlied: glisse: yes (missed the two undefined var warnings)
stuckey_: chithead: Yeah I need 3d
glisse: airlied: ok in drm-radeon-dave on fdo there is one patch for cmd checking do you want me to send it also to lkml ?
stuckey: chithead: the x1950 isn't on newegg... Do you know where one might buy it?
chithead: I think you can get only x1300 and x1600 cards new, x1900 and x1950 can be had for cheap from ebay though
stuckey: chithead: but x1950 are better than the x1600 and x1300??
chithead: yes, x1950 is the fastest you can get with any of the open source drivers
stuckey: chithead: but it isn't sold anymore? Strange...
stuckey: chithead: wow, this thing is a beast
suokko: Is it possible that exa is somhow using same memory location in video memory for different windows?
suokko: Or possible that opengl/exa are using same memory locations for different purporse?
suokko: (KM/DRI2)
suokko: ouh. WEll can't read the irc now :(
glisse: airlied: also my checker needs change to ddx & mesa i am pushing those
suokko: Even the gdi in wine can cause font corruption :(
glisse: suokko: agp ?
suokko: agp that is forced to pci mode
suokko: In agp mode corruption is worse
glisse: should work
suokko: In agp mode whole screen is mess while in pci mode only parts of screen (video memory) is corrupted randomly
suokko: Sometimes some textures are corrupted in 3D and sometiem soe others images are corrupted
suokko: It seems like rendering buffer/frame buffer overlaps with some texture/pixmap data
glisse: suokko: lastest linus kernel ? drm ? mesa ?
glisse: compiz enabled ?
suokko: metacity&compiz both causing the problem
suokko: I'm using a week old build from your git repo
glisse: which kernel ? drm ?
suokko: https://launchpad.net/~xorg-edgers/+archive/radeon-kms :)
glisse: i will see latter if i saw this too with lastest everythings
suokko: I did upload some screenhots showing the corruption: http://img230.imageshack.us/img230/8400/screenshot1vlw.png
suokko: http://yfrog.com/5jscreenshot4tnjp
glisse: looks like wrong pitch
suokko: Ans also some applications have lighting problems: http://yfrog.com/7gscreenshot3cwep (not related to corruptions)
glisse: suokko: does corruption happens right away or only after some times ?
suokko: Randomly
suokko: But instantly with wine
suokko: So seems like more complex opengl apps do cause it faster
glisse: suokko: how much vram do you have ?
suokko: (complex=larger memory footprint)
suokko: 64M if I remember right
glisse: lspci -v should tell you
suokko: Memory at d8000000 (32-bit, prefetchable) [size=128M]
suokko: But wikipedia claims that card should have only 64M
glisse: which card ?
suokko: Mobility 9200
ajax: bar size is not necessarily vram size.
glisse: there is no such things on radeon as fixed amount of memory according to gpu
glisse: it all depends on constructor
suokko: yes, But this laptop was the cheapest one with radeon card so not likely to have extra memory
MostAwesomeDude: Shouldn't the BAR only be 64M on a 64M card?
glisse: MostAwesomeDude: well you should never presume that people doing the gpu bios are clever
glisse: or use logic
MostAwesomeDude: glisse: I used "should," not "is." :3
glisse: oh i missed the subtilities :)
MostAwesomeDude: I guess weird things would happen if addresses > 64M were used, then?
suokko: So it might be possible that address would be truncated so same memory are could be used twice
airlied: MostAwesomeDude: generally you get NMIs
MostAwesomeDude: airlied: Ah.
airlied: at least I always did on my dell.
suokko: vramlimit parameter for radeon module should do the trick?
airlied: it shouldn't ever get that stuff wrong
airlied: it should be in dmesg how much VRAM you have
suokko: drm does detect 128M memory
suokko: But I still think it i wrong
nanonyme: Don't some integrated cards also use system ram in addition to the vram on the card?
airlied: glisse: looks like glyph cache corruption
airlied: the imagse all look fine
airlied: just txt broken
nanonyme: (and cheap discrete cards too)
glisse: airlied: xmoto doens't look fine :)
suokko: xmoto did render outside the window
airlied: but from a 2d breakage pov I think you are aiming at wrong thing :)
airlied: zzzz &
suokko: In agp mode that rendering ot outside area would stay permanent but in PCI mode everything is rendered correctly after I close xmoto
suokko: except sometimes font cache stayes corrupted after closing xmoto (or any other opengl program)
suokko: I did check BIOS and it claimed only 64M vram
suokko: So memory size detection is somehow going wrong
suokko: Semms lie vram liit didn't help
glisse: suokko: pastebin your dmesg
suokko: And I can't read irc now :(
suokko: again font corruption
suokko: Now when I had corrupted system font switching to new font in pidgin helped only for a few seconds
TCW: what was my quitemessage? Should have been around... in the last 30 Minutes
TCW: and at what time was it? :)
adamk_: [08:37] <-- TCW has left this server (Read error: 110 (Connection timed out)).
TCW: adamk_, ah, good morning then to you :)
MostAwesomeDude: adamk_: East Coast?
adamk_: Yeppers.
suokko: glisse: http://pastebin.com/m27920bf0
suokko: hmm. Video playback is also broken :(
kyku: hello
kyku: I have a problem with dual head setup on M690T
suokko: There is green shading in mplayer when using xv
kyku: i connect the dvi monitor but xranrd says DVI-0 unconnected
kyku: can someone give me some hints?
TCW: kyku, M690T is a laptop / mobile chip? Which one exactly?
kyku: the board is kontron kt690 which is supposed to be an embeded design
TCW: O_o
suokko: Seems like some some videos are turned to "greenscale" but I already found a video with worse corruption
TCW: kyku, may I see a lspci output and /proc/cpuinfo of it? Just curious :)
kyku: sure, one moment
kyku: TCW: http://pastebin.com/m5d2829c5
TCW: kyku, cat /proc/cpuinfo?
kyku: TCW sempron 2100+
kyku: you can find the specs on kontron site: http://us.kontron.com/kt690-mitx
kyku: it looks like the DVI doesn't sense DDC data... is it possible that they didn't connect it?
kyku: in Xorg.0.log it says something like: RADEON(0): Port2: / XRANRD name: DVI-0 / Connector :DVI-D / DFP2 INTERNAL_DDI / DDC reg: 0x0
kyku: both LVDS and VGA-0 have DDC regs different from zero
TCW: kyku, interesting one... how much power does it use?
kyku: TCW, i don't know but it heats like hell (w/o fan)
TCW: kyku, how warm does the cpu get?
kyku: i'm testing it on ad embeded busybox system... don't have sensors utils on it
TCW: ok :)
stuckey: Could anyone recommend me a card if I just wanted to play the occasional game and watch x264 movies?
MostAwesomeDude: X1300, X1600, X1900.
MostAwesomeDude: All fully supported. (And decently cheap.)
stuckey: MostAwesomeDude: What's the difference between those and the x1950?
kyku: is there a way to force the output on dvi without connected monitor?
Curan: stuckey: http://en.wikipedia.org/wiki/Comparison_of_ATI_Graphics_Processing_Units#Radeon_R500_.28X1xxx.29_series
zhasha: MostAwesomeDude: NOT fully featured though
zhasha: stuckey: that goes for you too. There's no GL2.0/1 on those cards yet, or any card for that matter
Curan: stuckey: oh, i just saw, you seem to be German, then read the German WP article, it's better I'd say
zhasha: unless you use the proprietary driver which now only supports cards after HD2000
Curan: stuckey: here http://de.wikipedia.org/wiki/ATI-Radeon-X1-Serie
stuckey: zhasha: What's GL2.0/1?
zhasha: MostAwesomeDude: mind if I remove the ioctl and rename the files?
zhasha: stuckey: OpenGL 2.0 / 2.1
stuckey: What's this do?
zhasha: it's the 3D library
zhasha: only GL 1.3 is supported at this point
zhasha: maybe GL 1.4
stuckey: What does it do/why would I need it?
zhasha: but all the shader stuff didn't enter till version 2.0
zhasha: now the r500 series (X1000+) are all quite capable of running shaders, which every new game uses, but the open driver doesn't support it
stuckey: ehh... I'll probably just be using this thing to watch x264 movies...
zhasha: SO, if you want to play newer games, and you're going to buy an ATI card, get a HD2000 card or better since all previous cards don't have shader support in the drivers available
zhasha: H.264/AVC will be decoded in software no matter what you use from AMD
stuckey: Do nvidia cards have hardware x264 decoding?
zhasha: through VDPAU, yes, but it's all proprietary AFAIK
[Enrico]: afaik GLSL is a WIP so it might works
zhasha: [Enrico]: in the gallium driver, which is not usable at all
stuckey: What's the difference between Fast video overlay and faster video overlay?
stuckey: and fastest...
stuckey: What's GLSL?
suokko: http://en.wikipedia.org/wiki/GLSL http://en.wikipedia.org/wiki/Xv http://en.wikipedia.org/wiki/Xvmc
suokko: fast video is xv, fastest video is xvmc
stikonas: http://en.wikipedia.org/wiki/X_video_extension
bridgman: sylware; VBIOS should set up a default profile that adjusts fan speed as temperature changes, so reducing power consumption on the chip should slow down the fan
sylware: bridgman: can I change the temps theshold through atombios?
bridgman: I don't know, we haven't looked at that much yet... trying to get the stinkin' 3D working first ;)
sylware: right.
sylware: airlied pinpointed the DDX force low power mode
bridgman: we know that the chips started including a fan controller around rv670 but don't know yet how many boards actually used the new controller
bridgman: I imagine all that info is in the bios but not sure yet
sylware: bridgman: have a r7xx, hope sapphire used the on GPU controler for its vapor stuff boards.
bridgman: no you don't ;)
bridgman: I'm kinda joking, but it seems like the third party temp/fan chips might actually be a bit easier to program than the ones on the GPU dies
sylware: REally?
bridgman: and I don't have to worry about IP review for the third party chips (yay ;))
bridgman: depends on what you mean by hard, I guess... the third party chips have to be programmed through i2c but I think our chips have more registers and more complex stuff
sylware: Indeed, that would requiere to ask IP stuff for all fan controllers in the proprietary driver
bridgman: again, not sure yet though
sylware: Anyway, the futur is on the GPU
bridgman: it's like voltage regulators though, there are only about 3 chips that actually get used AFAIK
sylware: I don't know the structure of atombios yet... but can I dynamically discover the services it provides?
bridgman: anyways, we'll get there... but top priority was getting code out to give some control over power consumption, so we don't have to worry about overheating
sylware: Indeed...
bridgman: for better or for worse we're putting fan noise after 3d in priority ;)
[Enrico]: agrees
sylware: Well, now I just can pray that the fan controller is well adjusted regarding its temp thresholds.
sylware: I'll tell you when I force the low power stuff on the DDX.
bridgman: I haven't seen any that are really broken, it's just that some of them are kinda crude
sylware: Hope sapphire did it right...
bridgman: getting the programming right is pretty complicated, it's a feedback loop with some wierd delays from the thermal mass of the chip
bridgman: so you want the fan speed to be a *bit* ahead of the temperature but not too much
sylware: err... don't know if I will remember my thermic lessons.
sylware: Well... till we have the registers to monitor and set (or even program), it should be fine
Wizzup: Ok... so one might or might not remember that problem I had with radeon the last few days, I just tried fglrx with kernel 2.6.30, and after applying all the patches, it gives the same blank screen as radeon. :o
bridgman: I think so... usually the two "problems" are "too loud even when the chip isn't that hot" and "not loud enough when the chip is really hot"
bridgman: I haven't seen both in the same BIOS yet ;)
bridgman: Wizzup, did fglrx work with an older kernel ?
bridgman: trying to figure out if this is a clue or not ;)
[Enrico]: Wizzup: still a black screen? there is somethin weird here
[Enrico]: Wizzup: btw better to tru fglrx with .28. but imho you have to write a good xorg.conf
Wizzup: bridgman: I am going to try that now
Wizzup: [Enrico]: Hmm.. I don't think there's much wrong with it, though. The same sort of config worked on my older laptop with the radeon driver
[Enrico]: Wizzup: last time you were missing the modelines, and may be the autodetect is failing
Zajec2: what needs to be done for r6xx/r7xx KMS on kernel part? just extend r600.c/rs780.c/rv770.c files with proper code? or something more? do we need touch memory management or is this ready for r6xx/t7xx?
glisse: Zajec2: i have a r6xx branch WIP that shows what is needed
glisse: it's a start
glisse: i will get back to r6xx soon
Zajec2: glisse: http://git.kernel.org/ and ctrl+f and "glisse" gives no result :)
Zajec2: could you point me to your repo?
glisse: cgit.freedestkop.org drm-next
Wizzup: [Enrico]: Hmm... Well, switching back to radeon isn't that hard. I could give it one last try soon
Wizzup: I'll try fglrx first
Wizzup: I want at least some sort of life in my laptop's display. ;)
Zajec: glisse: thanks
[Enrico]: Wizzup: if i remeber well you card is not supported anymore by fglrx
Wizzup: [Enrico]: But it's very new
Wizzup: Well, quite new at least
[Enrico]: Wizzup: mhm refresh my memory pls, show me the model
Wizzup: ATI 3650 Radeon HD Mobility
[Enrico]: Wizzup: oh my bad it is supported by fglrx
Wizzup: Let's hope so, haha
Zajec: err, what does mean 0xFFFFFFFFUL;?
Zajec: it's probably more C than radeon question... anyway, can sb tell/
phoenix64: all bits set in an unsigned long.
Zajec: thanks
Zajec: did't know about such suffixes
suokko: glisse: Seems like firefox&some gnome apps&sdl games don't cause in corruption so it is only opengl/wine that uses features where the bug is present
Zajec: slowly I started to understand how DDX works... and now this KMS it's one big mess in my head again :P
Wizzup: Heh, my 2.6.28 panics - so much for trying fglrx.
[Enrico]: Wizzup: don't tell me you have used the same .config used for the .30
Wizzup: I tried oldconfig, indeed. But I've dropped that, hehe
nha: airlied: thanks for writing down those instructions on your blog; it's always good to have these pointers in our forest of branches
stikonas: the forest of branches has almost come to an end
stikonas: only DDX needs checking out a branch
nanonyme: kicks Google for giving him "Bad Girls Blog" when he googled airlied blog
glisse: nha: i started deleting various dead branch in my tree, i think we should do that on drm repo too
osiris_: airlied: about the split patch. this (max_index - min_index >= 65536) condition doesn't seem correct to me. we could have 1,2,3,3,4,4,5,5,6 ... indices and even without reaching index 65536 we would have exhausted the count limit
nha: glisse: I'm always in favour of killing dead branches
osiris_: airlied: also docs doesn't say that VAP_VF_CNTL.NUM_VERTICES is actually n-1, so the max is 65535 not 65536
glisse: osiris_: iirc doc are wrong on that point
osiris_: airlied: the limits.max_indices should be 65535 too
kyku: hello, is it possible to use a monitor without ddc ??
osiris_: airlied: also the limits.max_vb_size is little bit too low I think. let's say we have 8 4-float attributes, that would give us only 2048 vertices
osiris_: airlied: I'd suggest doubling this value
mcgreg: airlied: the drm from 2.6.31 (that just have been accepted by linus) .. is it the very current one that will work with 3d for 6xx/7xx?
glisse: mcgreg: no
glisse: it will work only with 2d
glisse: i think for 3d you need Alex's drm branch
[Enrico]: btw congrats for the merge in .31 :D
mcgreg: well.. I hoped so. the 6xx/7xx drm branch doesnt compile on .30 yet.
kyku: are you the authors of the xf86-video-ati driver?
mcgreg: the authors ... airlied , agd5f (not here it seems) and a few other I dont know all of them ;)
chithead: if you look at the git history you will find the current contributors to xf86-video-ati
kyku: I got a board with DVI-D but no DDC... I have to get it working
kyku: i'm reading through the sources and gathering info from the net
adamk_: Have you tried adding a modeline for it?
kyku: yes with xranrd... no luck
adamk_: xrandr --newmode and xrandr --addmode to do it dynamically.
adamk_: What happens when you try?
kyku: nothing
kyku: the monitor stays in idle mode
chithead: is a crtc assigned to the output according to xrandr --verbose ?
kyku: chithead, by default: no
kyku: I turned off LVDS and assigned a CRTC to DVI-0
kyku: the VGA-0 blinked but DVI-0 stayed idle
MostAwesomeDude: zhasha: If it's really, really bugging you, then go ahead. I won't be at a place where I can modify it until tonight; I got delayed for yesterday, and now I'm at a conference.
zhasha: meh.
kyku: btw. I've seen some references to load_detection on the net, it seemed to turn on S-VIDEO for some guys
MostAwesomeDude: I mean, it's not a problem, is it? It needs to be changed, but AFAIK there's not exactly any pending patches.
kyku: xranrd --output S-VIDEO --set load_detection 1
kyku: however I get an X error if i try to set it on DVI-0
kyku: does anyone know what this flag is responsible for?
adamk_: It has to do with DAC load detection. Apparently it's disabled for some GPUs since it doesn't work properly for every one.
adamk_: But that's only related to tv-out.
adamk_: I did have to enable coherent_mode on my x1900 to get anything to display at 1280x1024 and 1024x768 on one DVI port (but not on the other).
kyku: adamk, you sure that it is only related tv-tou? i'm seeing in sources that it is being set for all types of connector
adamk_: No, I'm not really sure at all. That's what I picked up from a conversation in here earlier this morning. Someone else was having a tv-out problem and that was offered as a solution.
adamk_: a possible solution, rather.
MostAwesomeDude: kyku: Load detection only makes sense for analog outputs, like VGA and TV. LVDS and DVI are digital.
adamk_: Ahhh... That makes sense.
kyku: btw. how does it detect if a monitor is plugged in?
MostAwesomeDude: For analog outs, it uses a voltage check which requires flickering the monitor.
MostAwesomeDude: Digital outs just kind of work.
nanonyme: MostAwesomeDude: I kinda wonder why analog outs haven't been obsoleted already. ^^
MostAwesomeDude: nanonyme: VGA is prevalent, and analog TVs are always around.
MostAwesomeDude: Additionally, nobody likes DisplayPort.
nanonyme: MostAwesomeDude: Well, I kinda hoped DVI would kill off VGA and flatscreen TV's would kill off old analog TV's.
nanonyme: Seems it's not happening.
MostAwesomeDude: nanonyme: My parents' flatscreen has both analog and digital inputs.
MostAwesomeDude: Deprecated but not obsoleted.
nanonyme: MostAwesomeDude: I know. That still makes some sense. But they do sell FullHD displays with *only* analog connectors.
nanonyme: It's as if some vendors wouldn't want to change at all.
chithead: displayport is great, with 1.2 you can connect up to four monitors to one port
chithead: I think this will be a killer feature for anyone who uses many monitors (eg. multiseat) provided that the drivers properly support it by then
kyku: MostAwesomeDude, do you think it is possible to drive an DVI output without a connected monitor?
MostAwesomeDude: kyku: Sure, you can use xrandr to force outputs on.
kyku: MostAwesomeDude, so I just add mode in xranrd, set crtc, and do eg--left-of VGA-0 ?
MostAwesomeDude: kyku: You could. Not sure why you'd want to do that on a disconnected monitor though.
kyku: i'm thinking abot a monitor w/o ddc
kyku: or a board with dvi but w/o ddc
MostAwesomeDude: Oh, you mean, a busted monitor?
MostAwesomeDude: Yeah, that's part of the reason that the feature exists.
MostAwesomeDude: TVs often have busted/non-functional load detection.
kyku: and if that doesn't work, then what can it mean?
MostAwesomeDude: Bad cable, bad output, bad driver (sometimes...)
MostAwesomeDude: Bad TV input...
kyku: but the monitor works in clone mode until I startx
MostAwesomeDude: That's not a guaranteed behavior, that's just the way that it comes up.
kyku: then no matter how much xrandr wizardry I do it stays disconnected
MostAwesomeDude: Hm.
MostAwesomeDude: Can't you use --force --on or whatever it is?
kyku: i'll try tomorrow
kyku: MostAwesomeDude, but wait the's no --force option :)
MostAwesomeDude: Huh, so there isn't.
kyku: How does it always say Output LVDS connected in Xorg.0.log even if I don't have an LFP?
Zajec: glisse: is KMS on r6xx testable?
ajax: kyku: because connection sense on lvds isn't really a done thing.
ajax: if we see an lvds connector it's sort of hard to know whether there's actually a panel on the other end of it
stikonas: Zajec: I think that you will not have any acceleration at all
stikonas: but you will be able to use fbcon
nanonyme: Does it start X atm, btw? Iirc it worked at one point but was broken at another, not sure what's status quo.
glisse: Zajec: i only tried to compile it
glisse: haven't even tested it
glisse: and wont do much more than allow fbcon to load
nanonyme: Oh, right. glisse's branch. *shrug*
glisse: even that might not work
nanonyme: Does it benefit you if you get testing results? :)
glisse: nanonyme: well i dont expect it to work so just dont bother
nanonyme: Right.
ajax: hm, no bridgman.
nanonyme: MostAwesomeDude: Btw, you had a post about UVD and shaders on Phoronix a while ago. Wouldn't happen to have an idea of what exactly it is that you can do and what you cannot do without the UVD specifications? :)
Zajec: could someone explain please, how does KMS work?
Zajec: i tried to understand source of kms-support but can't get it
Zajec: what is done in kernel, what is done in DDX and how DDX communicates with kernel?
nanonyme: would suspect over libdrm
Zajec: oh, great... and libdrm to make it harder :D
nanonyme: I attempt not to guess since I think it's important you don't get false information but since KMS is afaik kinda together with DRM and libdrm is the userspace part of DRM, ddx would have to talk with libdrm to do the modesetting.
Zajec: what is "bo" in radeon_bo_*?
ajax: buffer object
Zajec: nanonyme: makes sense :)
MostAwesomeDude: nanonyme: UVD can do MPEG-2 and MPEG-4 IIRC. I have a feeling that it doesn't expose single stages, but I really don't know much about it.
MostAwesomeDude: is just going off bridgman's comments
chithead: uvd can also do vc-1 iirc
nanonyme: It can, yes.
nanonyme: At least according to the marketing papers I've seen about it. It seems to be better in that respect than the original Purevideo by nVidia.
MostAwesomeDude: Well,
MostAwesomeDude: IIRC Purevideo may not even exist.
MostAwesomeDude: It might all be done with shaders and mirrors.
nanonyme: *shrug*
nanonyme: I wouldn't know if it was. I just read that it leaves VC-1 for CPU.
nanonyme: I've just gotten the impression UVD really *is* separate from GPU.
nanonyme: Guess we'll see exactly by how much when the docco comes out.
MostAwesomeDude: nanonyme: Oh, it is. I think it's probably just another chip sitting on the board, addressable through the MC just like the 2D and 3D engines.
nanonyme: Yeah.
Zajec: err... what the hell is kernel mode setting? ;) i was sure it's that kernel sets resolution, crtc, plls, outputs, connectors... but it doesn't seem so
nanonyme: That's what I've been told it would do. *shrug*
MostAwesomeDude: Zajec: Currently, all the modesetting code is in the DDX (radeon or radeonhd).
MostAwesomeDude: And KMS moves it into the kernel.
nanonyme: MostAwesomeDude: Oh, it's slowly migrating?
MostAwesomeDude: nanonyme: Well, it's an all-at-once thing, and it's done, and it's in 2.6.31.
nanonyme: Right.
MostAwesomeDude: I just keep using the wrong tenses because I'm still in a bit of "I can't believe it happened" shock.
nanonyme: ;)
Zajec: but where in new KMS code can you see (for example) CRTC magaing?
MostAwesomeDude: Zajec:
Zajec: i also don't see in DXX anything like "hey, libdrm, enable CRTC2"
MostAwesomeDude: Do you mean, in the actual kernel tree, or what utilities are used?
Zajec: err, sry, don't understand question
MostAwesomeDude: You need to look in radeon-gem-cs3 or some other branch that uses KMS.
Zajec: MostAwesomeDude: radeon-gem-cs3 is the same as kms-support, right?
MostAwesomeDude: Zajec: I'm not familiar with the kms-support branch, give me a second.
nanonyme: Wonder if devs bother porting over KMS support to radeonhd.
Zajec: yeah, i can even see commit: "Merge branch 'radeon-gem-cs3' of git://anongit.freedesktop.org/~airlied/xf86-..."
MostAwesomeDude: nanonyme: AFAIK glisse, airlied, and I are not going to do it for rhd.
Zajec: nanonyme: i'm talkin about radeon :)
nanonyme: Zajec: Just a general thought.
MostAwesomeDude: Zajec: Yeah, looks like it's the same as radeon-gem-cs3. Guess that's the branch I'll be talking about from now on.
Zajec: ok, so could you explaing what is flow of starting X, for example?
nanonyme: Zajec: Since it seems KMS might nail most of the differences between the drivers so wondering if devs will take advantage of the situation and shift to a single driver for KMS time. :)
Zajec: which side checks for attached output, what calculates plls, what sets crts?
MostAwesomeDude: Zajec: Kernel does all the modesetting.
MostAwesomeDude: Userspace just asks for kernel to do the modesetting.
MostAwesomeDude: nanonyme: xf86-video-modesetting
MostAwesomeDude: Additionally, check out the Gallium xorg state tracker.
rektide: i'd really like a blog post on how state trackers equate to a functional working implementation
Zajec: rektide++
MostAwesomeDude: rektide, Zajec : Maybe at some point. If I get bored enough during the conference.
Zajec: rektide <<= 8 ;)
nanonyme: MostAwesomeDude: Not a good immediately solution. There's still lots of systems without KMS+mm and will be for years so you'd need radeon/radeonhd for a long time still.
MostAwesomeDude: nanonyme: That's okay; it's backwards-compatible.
MostAwesomeDude: We won't remove the old ioctls for a long time.
nanonyme: Right.
nanonyme: MostAwesomeDude: Though I was just talking about Linux. Other *nix's without KMS+mm might be pretty much screwed by where things are heading still. :)
MostAwesomeDude: nanonyme: BSDs are always way behind in terms of DRM compat, but they are also usually not behind on their updates.
MostAwesomeDude: Some BSDs, like Dragonfly, probably wouldn't have much of a problem.
nanonyme: Hmm, right.
nanonyme: And since there seems to be developer interest in ATi chips on Solaris, they'll probably get their own stuff set up in time too.
MostAwesomeDude: I don't think there's any arguments against libdrm on other *nix.
MostAwesomeDude: I'm personally in favor of it.
nanonyme: Sure, it's neat.
Wizzup: [Enrico]: Apart from the ModeLines, any other suggestions?
[Enrico]: Wizzup: no...... here it "just works"
Wizzup: k
nanonyme: Oh, you've still probs?
nanonyme: Wizzup: Btw, you could always try doing genkernel on 2.6.28.
nanonyme: In case your kernel panicked on boot.
Wizzup: nanonyme: I don't feel comfortable with it, I've never had issues with compiling a kernel really. I guess I'll try it in a few days with 2.6.28 again, but for now I'd really prefer radeon to fglrx.
nanonyme: Wizzup: Could try downloading radeon's sources, running configure and seeing if it complains anything.
nanonyme: In case there's a problem with the ebuilds.
Wizzup: from the git repos?
nanonyme: Well, the web interface for the git repo has the versions available too.
nanonyme: But yeah, you could also use a real git repo. :)
nanonyme: "Real" and real. *shrug* Anyway.
Wizzup: run the configure.ac ?
Wizzup: nvm
nanonyme: autogen.sh. :)
nanonyme: If you don't have configure.
Wizzup: Oh, darn. brb
nanonyme: autogen.sh will generate configure as one of the first things it does.
Wizzup: nanonyme: It seems to run just fine
Wizzup: the autogen.sh, that is
nanonyme: No warnings even?
Wizzup: nNope
Wizzup: looks like an ordinary gnu automake
nanonyme: Are there any special USE flags available for the ebuilds that might be wrong?
Wizzup: I don't think there are any use flags except for 'debug'
nanonyme: Hmm.
Wizzup: I could enable the debug flag... Perhaps it will uncover some problems
Wizzup: What also bothers me is that it doesn't detect a mouse/keyboard with radeon, but does with vesa :-/
Wizzup: brb
Wizzup: :-)
Wizzup: nanonyme: Is there anything else I could try to generate some useful debug output?
osiris_: airlied: could you also remove the TODO comment right before r300DrawPrims, and the not needed anymore check and comment in r300RenderPrimitive?
airlied: Zajec: you had kms qs?
nanonyme: Wizzup: Sorry, no idea.
Wizzup: Ok
Wizzup: Ok - I sort of get my framebuffer back, it seems. (After killing X with kill -9.)
Zajec: airlied: more than one :/
Zajec: airlied: i totally don't get it... don't understan what is done in kernel, what in DDX
Zajec: maybe you could find a moment to write down some explaination on blog?
Zajec: i belive it's not only me interestied in this
Zajec: you know... which part detects monitors, which attachs outputs to CRTC, which one wakes up CRTC? which receive xrandr requests now?
Zajec: or actually... when KMS starts it's work? as boot time? or at X starting ?
chithead: if you have framebuffer console, right after drm is initialized
Zajec: which part is responsible for registers storing and restoring on VT switching?
airlied: Zajec: all registers are done in-kernel
airlied: userspace just calls the KMS API to do everything
Zajec: chithead: does this console framebuffer works on top of KMS?
airlied: KMS API is in xf86drmMode.h
airlied: Zajec: fbcon works on top of fbev which works on kms
airlied: the kernel exposes a randr like interface
chithead: kms implements the framebuffer api, for the most part at least
airlied: xrandr requests go to the DDX and get translated into kernel requests
airlied: drmmode_display.c in the kms-support branch
airlied: contains most of the userspace code
airlied: we don't use any of the register read/write or mode setting code in the DDX when kms is enabled
airlied: just use that file
Zajec: hm, ok, start getting this
Zajec: wow, reading drmmode_display.c gives much light on this
Zajec: err, why drmmode_output_mode_valid returns TRUE always?
airlied: the kernel deals later it could be made a bit smarter
Zajec: hm, everything that gives us quite messed driver
Zajec: wouldn't it be eaiser to have to separated drivers?
Zajec: one for KMS with dropper registers operations?
Zajec: just asking, didn't think much about it :)
nanonyme: 21:45 < MostAwesomeDude> nanonyme: xf86-video-modesetting
Wizzup: Is the more than 2gb ram "bug" still present? Because that might solve my problems
Zajec: Wizzup: there is kernel option that fixes it
Zajec: don't know if it's on by default
Wizzup: Would you happen to know what option that would be?
nanonyme: Zajec: As in, there's this vision of having a generic ddx that uses KMS for modesetting and Gallium for all accel.
nanonyme: All the card-dependent code would be elsewhere.
airlied: Zajec: its messy but otherwise you can't have it just work
suokko: Is there any easy way to force x to reload all fonts?
Wizzup: Trying the MTRR cleanup now
chithead: suokko: for X fonts, try xset fp rehash, for fontconfig fonts fc-cache
suokko: chithead: too bad not helping :(
chithead: then maybe reloading fonts is not what you want
suokko: heh. It helps but some text aren't redrawn even if I resize windows
suokko: or some applications still have the corrupted font data
Wizzup: Is it possible that this bug (FAQ 10.7) http://wiki.x.org/wiki/radeonhd also applies to radeon? That would definitely explain my problems
nanonyme: Maybe?
nanonyme: Wizzup: Do you have MTRR satitizer enabled? ^^
Wizzup: Yes
Wizzup: (Sorry for the long paste)
[Enrico]: Wizzup: you can try to search in the bugzilla if you want http://bugs.freedesktop.org/buglist.cgi?query_format=specific&order=relevance+desc&bug_status=__open__&product=xorg&component=Driver/Radeon
[Enrico]: and btw PAT rocks
Wizzup: [Enrico], I've found the bug reports. They state that it's a bug in the kernel
Wizzup: What is PAT?
nanonyme: [Enrico]: BIOS bugs don't. :/
[Enrico]: Wizzup: a kernel option for x86 machines
[Enrico]: Wizzup: it is mtrr related
Wizzup: I've tried that
[Enrico]: i guess it is suggested
nanonyme: That is, in some cases the BIOS gives wrong MTRR information.
[Enrico]: Wizzup: which brand/model is you pc?
nanonyme: That's afaik what the sanitizer is for.
Wizzup: [Enrico]: Toshiba A350 12D
[Enrico]: nanonyme: well PAT itself is a sanitazer isn't it?
soreau: Where is agd5f when I need him, damn
soreau: Do we have AMD spec releases for tv-out on r3xx?
soreau: I want to get 1024x768 tv-out working and I'm half way there but I need the timing specs
soreau: Or, maybe I need the timing specs for ntsc in this mode?
soreau: Idk what I need but I want to get it working
nanonyme: Hmm...
nanonyme: soreau: Does that imply driver devs would usually need to implement separate paths for NTSC and PAL? >.<
soreau: The only way I got this far was by regdumping fglrx with radeontool under agd5f's instruction
soreau: nanonyme: No, pal's in radeon_tv.c too
nanonyme: Right.
soreau: But they are different timing afaict, so yes (?)
nanonyme: Well, as long as the timings are available... :3
soreau: nanonyme: I don't understand exactly what I need but since there's only one hard coded available mode it's kinda hard to tell
Wizzup: sleeps
Vash63: I'm having some issues with XDM...
moeSizlak: is kernel -191 gonna offer me any radeon fixes?
Vash63: (EE) AIGLX error: dlopen of /usr/lib64/dri/(null)_dri.so failed (/usr/lib64/dri/(null)_dri.so: cannot open shared object file: No such file or directory)
moeSizlak: yeah i wouldnt expect that file to exist ;)
Vash63: Right.
Vash63: Why is it trying to load it?
Vash63: startx is working right now with EXA and everything.
Vash63: According to the Xorg.log
MostAwesomeDude: Vash63: Which video card?
Vash63: R700.
Vash63: http://vash63.googlepages.com/Xorg.0.log.old
Vash63: That's the full log.
MostAwesomeDude: Hm. Well, there's no r600_dri.so anyway, so I wouldn't worry about that.
Vash63: Well, it's crashing.
Vash63: KDM loads up.
Vash63: I put in my password and it crashes the second I hit enter.
MostAwesomeDude: Hm.
Vash63: That's the only error I see in my Xorg.0.log.
Vash63: I dunno if it's related but I can't get wgetpaste working either >_>, 'cat /var/log/Xorg.0.log | wgetpaste' is trying to upload /dev/null to them.
MostAwesomeDude: Vash63: $ wgetpaste /var/log/Xorg.0.log
chithead: alternatively, curl -F file=@/var/log/Xorg.0.log nopaste.com/a
Vash63: http://vash63.googlepages.com/Xorg.0.log.old
Vash63: Er
Vash63: Apparently nothing was received. Perhaps the connection failed.
Vash63: If I add the verbose I get 2009-06-17 16:27:34 (88.6 MB/s) - `/dev/null' saved [873]
airlied: moeSizlak: it has the radeon AGP speedups + one stability fix
Vash63: Hmm, curl worked. Oh well for now.
Vash63: http://nopaste.com/p/aSEWwujCV
Vash63: Easier to read log.
moeSizlak: airlied: thx
Vash63: Er, wrong paste.
Vash63: http://nopaste.com/p/aee6tg45v
Vash63: There's the broken one. One is from startx, the other from XDM.
Wizzup: Vash63: May I ask how many memory you have in the computer?
Vash63: 4GB.
Wizzup: And kernel/distro?
Vash63: Gentoo, 2.6.30
Vash63: amd64
Wizzup: Right.
Wizzup: Do you have x11-dri emerged? (I'm not sure but you might need it)
chithead: 2.6.30 has drm support for r700 cards already
Vash63: startx is working and seems to be using EXA, scrolling and resizing is fast.
Vash63: But that's twm.
Wizzup: x11-drm*
Wizzup: Ok, then you def don't need it :)
Vash63: I think the problem is that AIGLX is trying to load (null)_dri.so instead of swrast_dri.so or whatever.
Vash63: I just don't know why.
Vash63: Hmm. Actually both open swrast_dri.so....
Vash63: XDM just tries to load the other one later.
MostAwesomeDude: Actually, XDM shouldn't have anything to do with it. :T
Vash63: Well, startx works and xdm doesn't >_>
Vash63: So I don't know what's causing it.
Wizzup: Vash63: Would you mind pasting your xorg.conf? I'm having issues that are alike to yours.
Vash63: I'm just on a generic one right now.
Vash63: I'm using evdev and ati which Xorg -configure used by default anyway.
Vash63: So I haven't changed anything.
Wizzup: Thanks. That didn't work for me... :-) I'll better get some sleep - I wish you good luck. (at least X starts)
chithead: Vash63: does startx still work if you start something else instead of twm? on gentoo it is defined in XSESSION in /etc/rc.conf
suokko: http://pastebin.com/m6234d65a null pointer deference in R200 mesa master
Vash63: Should I try kwin? I don't know if I have anything else installed.
chithead: Vash63: the comment above the XSESSION line tells you where to look
Vash63: There's no XSESSION line at all.
chithead: Vash63: at the bottom. or are you running an unstable system?
Vash63: Unstable.
Vash63: Ah, you'd have to make it.
Vash63: Says the openrc guide.
chithead: yes, and /etc/X11/Sessions/ contains the list of sessions
Vash63: Xsession and kde-4.2 are available.
Vash63: brb restarting X...
Vash63: Uh. That worked.
Vash63: I'm in KDE now.
Vash63: This definitelys ounds more like a KDE problem than an ATi one at this point...
Vash63: How do I make sure EXA is working?
chithead: if Xorg.0.log contains messages about exa then it is enabled
Vash63: Yeah, and window resizing is blazing fast.
Vash63: That's usually slow without it, right?
chithead: window moving and scrolling is jerky without exa
moeSizlak: yeah i cant even watch an avi fullscreen anymore
moeSizlak: not sure when that happened
Vash63: Ok, well thanks for the help. Guess it's either an XDM or KDM problem...
Vash63: tries #kde
chithead: Vash63: there is no shortage of login managers, try slim/wdm/gdm/...
Vash63: Actually, i'm going to see if plain xdm works.
Vash63: Seems like the easiest trial.
Vash63: i've noticed when I have KDM I seem to get more stuff in KDE like multimediakeys and stuff, is that just an XDM thing or do I have to be in KDM for it to pass that stuff to KDE?
airlied: glisse: btw I think we may call radeon_viewport too much
airlied: I think it may cause excecssiv flushing :)
moeSizlak: any idea why i can no longer watch a simple avi in fullscreen w/o excessive skipping?
airlied: moeSizlak: got that new kernel?
moeSizlak: yes
airlied: hmm then not really sure, maybe we need to optimise something more
airlied: a profile would probably be needed from sysprof or oprofile
airlied: try doing it without audio just in case its not gpu :)
moeSizlak: ok i wanna disable pulseaudio first to see if its the culprit
moeSizlak: oh
moeSizlak: good call
airlied: i normally just use mplayer, and -ao null to test
MostAwesomeDude: Goddammit PulseAudio. :C
airlied: I need to stop blaming PA for everything :)
airlied: I forgot to plug my headphones in a few weeks ago and blamed PA for my sounds not working
moeSizlak: no i used -ao null and get same problem
airlied: okay then its likely something that needs profiling
moeSizlak: i get glitches too, when i use 'e' in mplayer fullscreen to pan/scan
airlied: does nomodeset work?
suokko: When mesa should set pointer to driDrawablePriv?
moeSizlak: dunno, kms was working well for me so i havent been using it
moeSizlak: by it i mean nomodeset
Vash63: Hmm, ok, xdm works fine.
airlied: suokko: make current time I think
airlied: its a bit early for questions like that for me .
suokko: :) xmoto is crashing in intialization because r200 dri driver has null pointer there
suokko: http://pastebin.com/m6234d65a
moeSizlak: is profiling mplayer something easy to do or no?
airlied: suokko: you got latset mesa?
suokko: yes. master
suokko: demos in mesa seems like working. At least none crashed that I tryed
moeSizlak: airlied heres a sysprof file from me running mplayer: http://www.mediafire.com/?sharekey=eb9edca6ccd2a54ed8f14848abf485dde04e75f6e8ebb871
airlied: moeSizlak: my popup blockers seems to annoy that site
moeSizlak: airlied how about this one http://www.teradepot.com/d3igufu6wxc5/mplayer_profile.dat.html
airlied: probably need some debuginfo packages installed
airlied: but I see quite a lot of time in the kenrel
airlied: which suggests we still need to optimise some things
moeSizlak: yeah when i zoom in mplayer fullscreen with 'w' it introduces maroj glitchage on left side of screen that never goes away
moeSizlak: er, with 'e' not 'w'
suokko1: http://pastebin.com/m4da9aae4 Seems like wrong order of calls to makecurrent and update viewport
Vash63: Hmm. Anyone know the best way to lower my fan speed? R700 at high fan speeds is kinda scary.
Vash63: Do I have to edit the BIOS through Windows or is there a better way yet?
[Enrico]: afaik the default powemode is powersave with radeon O.o
Vash63: I dunno but it's way louder in Linux than Windows.
yangman: Vash63: no fan control yet
Vash63: Any idea how far off it is? I'm ok with rc kernels and stuff.
airlied: you can try Option "ForceLowPowerMode" "true" in xorg.conf
yangman: no one's attempted it afaik
yangman: although my guess is the controller is either on the same chip as the temperature sensor or controlled via the similar methods, so maybe once that's in place
[Enrico]: someone today has talked about that, and the result was "just after the 3d stuff :D"
[Enrico]: is waiting patiently the 3d stuff for r6xx :D
[Enrico]: i've also tried the .30 kernel to be able to use EXA an xv, but kde4 composite performance with xrender are horrible :D
[Enrico]: i know i don't need composite..... but it is so nice :D
[Enrico]: and btw xrender performance with radeon are better than with fglrx, so congrats guys :D
[Enrico]: and thanks :D
Vash63: Heh, tbh I care more about fan controls than 3d.
Vash63: Though I want both.
Vash63: Loud fan makes it hard to use even with 2d.
[Enrico]: Vash63: eheheh i see..... i don't have a fan control here. hd 3470
Vash63: Dual GPU cards are a lot louder than something like that.
[Enrico]: indeed
Vash63: So you can see my problem >_>
Vash63: It's about as loud as my AC.
[Enrico]: of course
airlied: anyone seeing that max bo validate assert please try mesm aster
spstarr: airlied: CONFIG_STAGING enabled in rawhide or is this .for .31-rc?
spstarr: is this newttm or original ttm?
airlied: spstarr: rawhide hasn't got .31 in ityet
spstarr: ok, so i have to wait, I moved to rawhide too soon i guess
airlied: rawhide is still working on .30 kernels
spstarr: oh 2.6.30-git12
spstarr: there isn't even a .31 git snapshot yet
spstarr: i'll just wait til rawhide has the bits, - but - I can't test since r6xx is WIP and not testable yet
spstarr: gotta love state of flux + hot code :)
spstarr: airlied: and I thought r3xx was flux =)
spstarr: boggles