EruditeHermit: airlied, hey
EruditeHermit: airlied, what computer(s) do you use?
EruditeHermit: if you have a laptop which one is it?
Nightwulf|work: hi all
dileX_: xf86-video-ati-6.12.192 released
airlied: EruditeHermit: I have lots :-)
airlied: my normal laptop is T60p + ATI rv530, my travel one is HP2510p which is intel, my office laptops are a HP with an rs480, a T42 with r100
airlied: today I got a T43 with PCIE r300, which does s/r fine
airlied: EruditeHermit: we have tried s/r on your laptop with radeon.agpmode=-1?
EruditeHermit: airlied, yes and that doesn't fix the screen issue
EruditeHermit: the screen not resuming
EruditeHermit: i.e. 23103
EruditeHermit: airlied, what laptop would you recommend?
airlied: really depends on what requirements are, I got a really nice Lenovo X201 tablet today to play with, Intel Ironlake GPU
EruditeHermit: I really need 3 laptops
EruditeHermit: 15.4 inche powerhouse, 13.3 portable light, and a cheap netbook type thing withHDMI if possible again very light
EruditeHermit: recommendations on any of those are welcome
twnqx: for the first i'm quite happy with my lenovo t500
twnqx: except for the GPU lockups with KMS.
EruditeHermit: how are the trackpads?
airlied: twnqx: try latest drm-radeon-testing + latest DDX
EruditeHermit: I remember lenovo trackpads were really small
EruditeHermit: which is annoying
airlied: I didn't like the keyboard one of the T series I plaed with
airlied: compared to the W500 darktama has
EruditeHermit: oh and they place the Fn key on the left of the ctrl key
twnqx: Linux dnnote 2.6.33-gb979b2c #1 SMP PREEMPT Fri Mar 12 10:52:26 CET 2010 x86_64 Intel(R) Core(TM)2 Duo CPU T9550 @ 2.66GHz GenuineIntel GNU/Linux
twnqx: something relevant changed the last four days?
EruditeHermit: preempt I was told is more crashy
twnqx: maybe, but linux IO is fubar enough with forced preemption, unbearable for a desktop otherwise
airlied: try with preempt off, I need to test preempt I tnik
EruditeHermit: airlied, and no I don't have preempt enabled =)
twnqx: well, as-is i only have to try to start thunderbird or filezilla (those two i know) :P
twnqx: for a lockup to occur
twnqx: gonna try with preemption off
airlied: twnqx: d-r-t also has lockup recovery now
twnqx: hm, i thought this is d-r-t
airlied: I only pushed it yesterday
airlied: so if you have 4days old you don't ;-)
twnqx: or wait, this is glisses' drm-next tree
twnqx: err... or is it...
twnqx: how to i make git show me the repo path? :S
twnqx: git://git.kernel.org/pub/scm/linux/kernel/git/glisse/drm-radeon.git <- this is what i'm using
airlied: ah that should have the reset code at least
airlied: there is one fix in my tree for r6xx that he probably doesn't have yet
twnqx: i'll grab d-r-t and try without preemption
twnqx: git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git <- that one?
twnqx: cloning... time to get ready to go to work :)
airlied: twnqx: you could just use git remotes
airlied: saves re-cloning stuff you already have
airlied: or git clone --reference
EruditeHermit: airlied, gnight for today, if you have time tomorrow we can work on it again
twnqx: airlied: too late :]
twnqx: was only 330MB anyway
twnqx: airlied: is there a rebasing needed, or is your tree only based on 2.6.29rc2? :S
airlied: twnqx: its in a branch
airlied: drm-radeon-testing branch
twnqx: git rebase drm-radeon-testing
twnqx: Current branch master is up to date.
airlied: git checkout -b drm-radeon-testing origin/drm-radeon-testing
twnqx: oh, i can't just switch the head?
twnqx: works with glisse's tree :P
Vash63: Is there a firmware needed to get R800 KMS working? And does it work at all in the latest DRM?
airlied: Vash63: not yet, should work in 2.6.34-rc1
airlied: there will be firmware requirements later
Vash63: Ok, cool. KMS works for high res modesetting on DVI?
twnqx: airlied: dunno how to put it, but...
airlied: should do, haven't played much with evergreen yet
twnqx: your current kernel stalls my GPU immediately
twnqx: before even starting X
airlied: twnqx: wow winning, without preempt?
twnqx: white screen, color lines, fading to black
twnqx: Linux dnnote 2.6.33-50659-g7fc3c45 #1 SMP Tue Mar 16 09:41:07 CET 2010 x86_64 Intel(R) Core(TM)2 Duo CPU T9550 @ 2.66GHz GenuineIntel GNU/Linux
airlied: wierd that should be quite close to glisse's tree
twnqx: as per glisse's recommendations i use modeset=1 dynclocks=0 dynpm=0
airlied: yup sounds sane
airlied: what gpu?
twnqx: Chipset: "ATI Mobility Radeon HD 3650" (ChipID = 0x9591)
airlied: okay wierd, thats pretty close to what I booted on earlier
airlied: maybe try git revert 324abcaa8df637bc5ba01fe119390b48b52e202b
twnqx: maybe i should see what happens withpout the "Vritual 3600 1200" in my xorg conf
twnqx: wait no
twnqx: xorg.conf is irrelevant before X start >_>
twnqx: is there a doc on git bisect procedures?
airlied: yeah if it dies before X thats wierd
twnqx: hm.... i just remembered
twnqx: my acpi reboot script leaves dmesg
airlied: has some instructions
airlied: mainly git bisect start and good/bad until you find one
airlied: might be worth trying d424b925f7092b9d95e0a8556872349abe79d9b6
airlied: as a good point to start with
airlied: assuming it works
twnqx: just wondering about one thing
twnqx: glisse said that card would be an rv630
twnqx: but it always loads rv635 firmware
airlied: yeah smae firmware
twnqx: ah ok
twnqx: what is GTT memory?
airlied: its RAM that the GPU can access via a translation table
twnqx: wow, so you reserve 512MB system ram?
airlied: its dynamic
twnqx: d424b925f7092b9d95e0a8556872349abe79d9b6 as good for bisect?
airlied: yeah probably a good start point, its whats in gliise and my tree
airlied: needs to do a bisect now a swell
twnqx: interesting, changes in iwl wifi Oo
glisse: yeah i am about to plug a r5xx :)
airlied: glisse: I'll try and confirm its the fence changes
glisse: yeah it either that or the reset is not reseting enough after suspend
dileX: someone tried out new power-mngt patches from
Terman: hmm, looks like they are not fox 5xx
twnqx: 2.6.33-50612-gb792210 == good, vmlinuz-2.6.33-50613-g86cb2bb == bad
twnqx: drm/radeon/kms: use lcd pll limits when available <- breaks on my laptop
twnqx: i think...
dileX: hmm, pm2 patchset does not apply on yesterdays drm-radeon-testing
twnqx: let's see, reverting to drm-radeon-testing head & reverting the patch
twnqx: make install; make modules_install
twnqx: airlied: for sure 86cb2bbfda2cf402aee46779ee90bbb7d915482b breaks. drm-radeon-testing with that reverted boots fine
twnqx: also, still fade to white & no gpu reset with filezilla start :P
twnqx: given what the patch does... my bios is broken \o/
twnqx: well, or the patch.
airlied: twnqx: did we ever add a quirk for you?
twnqx: are there any quirks? :>
twnqx: let me add some printk to see what values by bios provides
airlied: agd5f: ^^
airlied: twnqx: hopefully Alex can spot it
twnqx: yeah, just found the quirks section
airlied: twnqx: also try booting with radeon.new_pll=0
twnqx: it doesn't really make a difference that i use radeon as a module, right?
glisse: airlied: btw maybe we should push the reset stuff for 2.6.34 as it seems helping people with lockup on r6xx/r7xx
twnqx: doesn't :(
glisse: twnqx: you are the only person reporting it didn't help so far
twnqx: yey, i'm special :X
dileX: (pm2: 0003-i2c-algo-bit-Add-pre-and-post-xfer-hooks.patch is in upstream - commenting)
airlied: glisse: yeah I'm hoping to push it once we fix the regressions ;-)
twnqx: radeon_atom_apply_lvds_quirks <- should i add my laptop to that?
airlied: twnqx: file a bug for Alex, he knows that stuff best
airlied: twnqx: okay with that reverted does no preempt help?
twnqx: err... ehere are my dmesgs....
twnqx: ah. no, no messages whatsoever from the drm after Initialized radeon 2.1.0 20080528 for 0000:01:00.0 on minor 0
twnqx: however my ddx is still 4 days old, should i update that too?
dileX: why not :-)
twnqx: good. i can test my quirk then
TuomasT: Where is the status page?
twnqx: *sigh* these kernel version names are getting worse...
BeteNoire: lots of xf86-video-ati updates on gentoo lately
BeteNoire: any chance that some of these fix my problems?
BeteNoire: i have x200m which lock ups on switching to tty
airlied: BeteNoire: yes I tihnk that was fixed
BeteNoire: have you tested it, airlied? :]
airlied: no but we did get a commit from someone to fix that
BeteNoire: what's crtc?
airlied: piece of gpu that scansout memory
airlied: for displaying
BeteNoire: ok, so cross your fingers for me, airlied, as i'm going to recompile half of x deps to use new radeon driver
twnqx: http://pastebin.org/114225 and it works.
twnqx: it's really my bios that had bad pll clocks...
twnqx: airlied: it's not the new pll clock divider calculation algorithm, though :P
airlied: twnqx: what values does the bios havE?
twnqx: 45000/60000 for the lcd, 90000/120000 otherwise
twnqx: exactly half the values...
twnqx: well, 450/600 (i implied the multiplication with 100 as per the code)
airlied: I wonder are they meant to be used for some lower power mode
airlied: or did they just them wrong
airlied: can you open a bug, attach your bios and the patch
twnqx: either that, or it's like that other case where clocks need to be multiplied by 2
airlied: hopefully alex can get the AMD bios guys to take a look
twnqx: how do i get the bios extracted?
airlied: cd /sys/bus/pci/devices/
airlied: echo 1 > rom ; cat rom > /tmp/rom ; echo 0 > rom
twnqx: -r-------- 1 root root 131072 Mar 16 12:24 rom <- readonly? :P
airlied: won't matter
airlied: zz &
twnqx: good night
twnqx: guess i'll crash my laptop now...
twnqx: yepp, thunderbird main window kills KMS :P
TuomasT: Can I get vertical sync from 60Hz to 75Hz for KMS?
TuomasT: In kernel
dileX: agd5f: your pm2 patches on top of linus-tree ends up here w/ KMS enabled in VT with stripes.
Vash63: Mmm, latest git update in the last few days seems to have sped up 2d a lot.
Vash63: Firefox is much faster now.
Vash63: Still a bit of a delay when I grab a tab and try to move it around that isn't there on Windows, I'll have to try it on my friend's nvidia system to see if it's graphics related.
TuomasT: OpenGL version string: 1.5 Mesa 7.7 <- Is this correct for Radeon 3850?
TuomasT: 2.0 would be nice
adamk: Update Mesa then.
adamk: OpenGL version string: 2.0 Mesa 7.8-devel
TuomasT: As soon as I find ebuild for mesa-git ..
dileX: agd5f: http://paste.debian.net/64461
twnqx: agd5f: https://bugs.freedesktop.org/show_bug.cgi?id=27106
Wizzup: TuomasT: x11 overlay
TuomasT: Wizzup: Yes, I just added it, but 'emerge =media-libs/mesa-9999' gives me not found
Wizzup: Did you unmask it?
TuomasT: Wizzup: I did put media-libs/mesa at /etc/portage/package.unmask
TuomasT: Wizzup: Solved, I had bad layman info at make.conf
TuomasT: Receiving objects: 11% (39322/329081), 22.74 MiB | 229 KiB/s <- Woah, how big is this. 10Gb ?
TuomasT: Well, more like 300 MiB based on that
twnqx: git is like "all versions ever written, straight on your harddisk"
TuomasT: I guess should be able to fetch only a specific version
Wizzup: TuomasT: you'll only have to do that once iirc
twnqx: TuomasT: at least gentoo caches it.
TuomasT: Any idea about this question: Can I change default vertical sync from 60Hz to 75Hz for KMS in kernel?
TuomasT: I'd be nice to have the same vertical sync in consoles and X
TuomasT: My old OpenGL 1.0 program is smooth otherwise, but has periodic stuttering, and some of my new OpenGL 2.0 applications fail at Error: Incompatible type returned by call to 'max' when compiling GLSL programs
TuomasT: I guess give it some time
che: actually i found an easier to reproduce testcase for the problem with iris: https://bugzilla.redhat.com/show_bug.cgi?id=574021
che: ember is also an ogre based game that shows the same error message when it is beeing started.
lordheavy: che: it can be reproduce with the samplers demo provide with ogre3d
che: lordheavy, really?
che: lordheavy, which demo?
lordheavy: yes, if you run a second demo, the bug is hit
che: lordheavy, is it a driver bug or an ogre bug?
lordheavy: don't know :-) should look at radeon source
lordheavy: to know if the assert is here
lordheavy: but it really looklike a radeon bug
TuomasT: Should "OpenGL version string: 2.0 Mesa 7.9-devel" mean all of OpenGL 2.0 should be supported, but there may be bugs?
che: lordheavy, do you know anything more about the bug?
che: lordheavy, sorry to keep onbugging hehe... i just want to gather as much info as possible.
lordheavy: mesa/src/mesa/drivers/dri/radeon/radeon_texture.c here is the file
che: just checking out :D
che: basically everything else i have tried worked
che: yofrankie is missing ground textures (or is it shaders?!)
che: maybe related
Wizzup: I thought blenders game engine doesn't use shaders if they aren't available
xrmx: hello, is possible to have multiple xserver driving the same card, like one xserver per head? sorry if it is a silly question :)
jcristau: xrmx: no
adamk: Depending on the version of the X server and driver, you can have one X server manage each head as a separate screen, but that's as close as you can get to what you're asking, I think.
xrmx: jcristau, adamk: thanks for the answer
adamk: Wow, gordboy is quite the ass.
adamk: Pardon my language.
BeteNoire: well well, it works on me x200m crap :)
BeteNoire: but i get in xorg log:
BeteNoire: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed to open the DRM
BeteNoire: [dri] Disabling DRI.
adamk: Pastebin the full log file.
BeteNoire: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed to open the DRM
adamk: Just do not paste it here.
BeteNoire: that difference between clipboards :/
adamk: Sounds like you don't have the radeon DRM support compiled into the kernel or as a module.
adamk: Pastebin the output of 'dmesg | grep drm'?
adamk: Odd, I have no idea why it's failing to open the DRM.
BeteNoire: btw. unfortunatelly i cannot switch to tty - still :(
g-zu: I have a 512M video card but with 2.6.33 and KMS it only recognized 256M, is that normal at the moment or did I do smth wrong/forgot some step?
agd5f: g-zu: glisse only recently added support for vram amounts larger than the pci bar window
adamk: That's normal at the moment.
g-zu: ok, thanks
g-zu: good to know
BeteNoire: adamk: could you help me how to diagnose problem of lockups when switching to tty and file some good bug report?
suokko: BeteNoire: 1. check if ssh connection stays up 2. If yes collect dmesg and Xorg.log after lockup. If no try to setup netconsole to capture kernel log with another machine in lan
BeteNoire: suokko: i think its all welll with ssh as i can switch back to x
suokko: If defaul logs don't show anything just before lockup setting drm.debug=15 may help providing some usefull info
BeteNoire: how do you set up drm.debug?
suokko: BeteNoire: So it isn't lockup. Do you have fbcon?
BeteNoire: you mean framebuffer?
BeteNoire: dmesg doesn't show anything after switching
suokko: BeteNoire: your configuration is somehow wrong. kernel module doesn't load correctly. Are you sure that you _don't_ use radeonfb or uvesafb? Also you need fbcon for KMS console
BeteNoire: well, i've compiled radeonfb as i would like to use it
BeteNoire: i guess i should disable it completely
adamk: Using radeonfb and radeon at the same time can definitely lead to problems.
BeteNoire: ok, i'll disable it, despite i'd like to use it
mjg59: BeteNoire: drmfb will give you a high-res console
adamk: Yeah, KMS is much better.
BeteNoire: sorry, i don't know much about that new things
BeteNoire: should i compile agpgart in?
mjg59: BeteNoire: There's no need to use radeonfb with KMS - KMS gives you the same advantages
edwin: shouldn't the kernel config reject radeonfb+radeon?
mjg59: edwin: No, you could build both but blacklist one
BeteNoire: ok, so why kms is in some "staging drivers"?
edwin: it isn't
chithead: kms is not in staging anymore since kernel 2.6.33
BeteNoire: oh, i do have 2.6.32 as tuxonice patchset in gentoo
adamk: It's still quite stable on 2.6.32.
BeteNoire: ok, so i go further testing it on 2.6.32
BeteNoire: what AccelMethod should i use?
chithead: stick to the default
Obscene_CNN: if you have a r600 + exa is the only option
BeteNoire: so XAA
soreau: exa with any recent version of X
soreau: also with any recent version of the drivers, exa should be the default
Obscene_CNN: notes that XAA is a better option for some apps if your card supports it
BeteNoire: http://pastebin.com/xiPykPCp ← could you look at my xorg.conf if it looks reasonable?
chithead: for xserver 1.6, xaa might still be better in some cases and without compositing. but exa has improved in 1.7
soreau: and you dont need an xorg.conf with recent versions of X in a lot of cases
BeteNoire: so i try to change to exa
Obscene_CNN: EXA doesn't have acceleration for diagonal lines, XAA does
BeteNoire: whatever it means
BeteNoire: soreau: but i want my xorg.conf :]
soreau: well what you have is way overkill imho
BeteNoire: soreau: xorg.conf is a way i go since couple of years using linux, so i'm used to it and like it
soreau: yea, looks like that conf file is more than just a couple years old :P
soreau: doesnt matter really if it works
Obscene_CNN: notes that unfortunately there is a push to eliminate xorg.conf
twnqx: wishes he could use his system without but is hit by radeon not allowing an external 1920x1200 display without increasing virtual
adamk: No one is trying to eliminate xorg.conf.
adamk: Developers are trying to make Xorg as intelligent as possible so that xorg.conf files will not be needed for the vast majority of users.
adamk: There is a huge difference between those two concepts.
twnqx: so reserve enough space for two connnected 2560x1600 displays by default, or make memory allocation dynamic >_>
twnqx: stupid 2048 limit
chithead: with kms the limit is 8192
twnqx: just kms freezes al the time
twnqx: and yes, i'm likely the only one where even glisse's reinits don't trigger
twnqx: anway, wanted to test a patch i found on the bug tracker... brb
fabio333: hi there
fabio333: is it true that opengl video rendering (like adobe flash) is slow with radeon and a lot faster with fglrx due to hardware acceleration?
suokko: fabio333: flash != opengl video rendering
fabio333: suokko: so what xvideo?
maligor: I think flash actually had some opengl rendering mode
fabio333: but there is no acceleration
fabio333: maybe it uses directx on windows and opengl for mac and linux
suokko: There is exa cceleration but flash doesn't even use that correctly
fabio333: my last question is about kms
fabio333: i have to start with radeon.modeset=0 ... will kms be supported on old harware?
fabio333: like rs200...
fredrikh: i think flash renders to an xshm image
Tommeh: fabio333: flash uses DXVA on Windows, but they've publicly complained about the lack of a good video acceleration API under Linux (which is rubbish IMO) that they can use.
chithead: fabio333: kms is supported on all radeons
fabio333: so it's like using mplayer -vo x11 file.flv
Tommeh: Their reasoning was that the API's available on Linux don't allow for data to be rendered, passed back to the application, and then displayed.. Or something. :)
fabio333: but cpu usage it even higher .... like mplayer -vo gl file.flv
fabio333: chithead: not for me
Tommeh: fabio333: http://blogs.adobe.com/penguin.swf/2010/01/solving_different_problems.html
Tommeh: It's interesting, albeit upsetting.
chithead: maybe your kernel is misconfigured
adamk: Or maybe you have encountered a bug.
fabio333: it's a old card
adamk: Yes, we know that.
suokko: Tommeh: And full of excuses
adamk: I told you before that KMS should work on all radeons.
fabio333: from plymouth man i read kms is a modern cards features
suokko: fabio333: Any ati card made after 1999 is modern then ;)
maligor: yeah, just tried it on r100 I think
maligor: or it might be a r200, some laptop chip
suokko: rs200 is r100 card
spstarr: suokko: nice
edwin: hmm yuv, why isn't video stored as rgb then?
Tommeh: suokko: I know. I wish I knew more to form a counter argument, but I don't know quite that much about the APIs available.
suokko: edwin: because yuv takes half of te space of rgb
Tommeh: For instance, can they actually do what they want with VDPAU/VA-API?
suokko: Tommeh: xrender based higer level rendering apis should work for them
suokko: But they just want to do software rendering
fabio333: suokko: rs200 ia r100 card right
edwin: why don't they just use OpenGL to compose the elements on the video?
adamk: And it is supported by KMS.
fabio333: when i start with kms on my screen starts to blink
fabio333: using ubuntu lucid right now
maligor: edwin, I wouldn't be surprised if their codebase sucked totally
maligor: just have to look at the time it took to get a 64bit version out
edwin: oh they do opengl, but only if glx vendor is not SGI
edwin: they say they use that to detect if glx is software rendering or not
Jonimus: edwin: you talking about flash, yeah I remember reading about that
maligor: I'm not so sure, the fullscreen rendering is so dreadful even if your glx vendor isn't SGI
edwin: yeah just read that blogpost about flash and accelerated rendering
fabio333: i will make a few test with my card mith kms...
fabio333: see later
Tommeh: edwin: I do wonder what the difference between accelerated video + adding interactive overlays is, against video with interactive buttons. :)
suokko: fabio333: try booting without plymouth and xserver starting
suokko: Then see if there is errors in dmesg
suokko: fabio333: also then checking dmesg and Xorg.log when you have problems may help
twnqx: is there any distribution with life dvd/cd that supports kms?
BioTube: twnqx: Fedora ought to
soreau: lucid does/will
twnqx: maybe i should try something else, as i seem to be pretty alone with my issue
twnqx: but gentoo with enlightenment dr17 allows some reproduction that's... unique
suokko: twnqx: Can you compile a debug kernel with lockdep compiled in and load radeon module with netconsole enabled?
suokko: Then try to collect the kernel deadlock
twnqx: oh, i don't have kernel deadlocks
twnqx: i just cleanly reboot from an acpi event
suokko: twnqx: Does ssh still work?
suokko: So you jsut get gpu hang that prevents the reset
twnqx: yes, with the typical "fade to white"
twnqx: wait, prevent what reset?
twnqx: gpu reset?
spstarr: remains in intel GPU for now
twnqx: i'd even do that
twnqx: if the intel GPU was connected to the displayport >_>
twnqx: i wonder if it could just be some huge bitblt operation, since the two apps that cause it are fullscreen on startup
twnqx: and today thunderbird first gave a small window for plugin update that worked perfectly... then, when the mainwindow came up... *boom*
suokko: twnqx: You would need glisse's work in progress cs capture patches :)
twnqx: ACTUALLY that was - at least in the ddx - for a long time
twnqx: but i was too lazy to finish my command decoder
suokko: twnqx: If you have the cs captured you could upload it to the bug report so someone else can analyze it too
twnqx: i mean.. it's not more than 4 lines to patch the ddx, since all commands pass through one function
suokko: twnqx: yes. If you don't have any dri clients running
twnqx: no, i don't...
suokko: twnqx: But you don't know if the latest submited is the one that caused the hang because there is queue in kernel
twnqx: there's the problem indeed
twnqx: i just wanted to validate the commands
twnqx: but then i realized that i'd to track register states and whatnot
suokko: twnqx: Also one solutions to know what was going on just before freeze is to add debug message to start of every acceleration function
suokko: Then yo ucan just look the log file that was fenerated to know what was the last few acceleration operations happening
twnqx: Type 3, Opcode 6f, 2 bytes data
twnqx: ERROR: Unknown command
twnqx: ok, that was where i stopped writing the decoder :P
twnqx: Type 3, Opcode 10, 1 byte data
twnqx: at least that was decoded!
fabio333: if i do sudo modprobe -r radeon drm...
fabio333: then sudo modprobe radeon mosedet=1
fabio333: that should work: i get a black screen
BioTube: fabio333: you need to load fbcon to get a terminal with KMS
fabio333: something to do with framebuffer?
BioTube: framebuffer console
dileX: fabio333: see topic/wiki - there is a section with kernel parameters to set
fabio333: i'm going to add fbcon to modules... kms will work?
dileX: fabio333: I would built-in fbcon
fabio333: you mean i have to recompile?
BioTube: fabio333: not if you've already got fbcon built
BioTube: try modprobe fbcon and see if anything happens
fabio333: i will try it later because i fear something wrong
dileX: fabio333: next kernel build - thats for sure if you wanna follow radeob OSS WIP development ;-)
fabio333: radeonfb is not required i suppose
BioTube: fabio333: KMS doesn't work with it
fabio333: so it's blacklisted
fabio333: do i need the rhgb option?
fabio333: for plymouth i mean
BioTube: I don't know
fabio333: last question
fabio333: i read that 3d i out when kms is on with r100/200 : do u know something about it?
dileX: fabio333: simply unset CONFIG_FB_RADEON
fabio333: 3d acceleratiobn*
fabio333: thank u see later: i will try with fbcon
Leonidas: stupid question, but how do I build mesa from git?
Leonidas: I probably need to generate a ./configure file, but I have no idea how to do that
BioTube: pass the configure options to autogen.sh
dileX: Leonidas: http://wiki.x.org/wiki/radeonBuildHowTo#Mesa3D
chithead: see the wiki link in the channel topic
Leonidas: thank you all, looks great
Leonidas: punches himself for not reading the topic
Leonidas: hmm, so it built, but it didn't produce a r300_dri.so
Leonidas: although I did request the r300 driver in autogen
dileX: Leonidas: see in your $PREFIX dir
Leonidas: dileX: nope, it's not there either. Actually, the $PREFIX/lib only contains libEGL, libGL, libGLU, libglut and libGLw
Leonidas: dileX: no dri folder
Leonidas: oh, wait
Leonidas: make linux-dri
dileX: Leonidas: see wiki - configure-option --with-dri-driverdir=$DRI_DRIVER_DIR
Leonidas: ok, rebuilding with --with-dri-driverdir
BeteNoire: what you think of this http://pastebin.com/zDRT41gk ?
adamk: BeteNoire: That looks normal.
BeteNoire: and it finally works
BeteNoire: amazing how fast is switching to tty
agd5f: glisse, airlied: I just ran into a crash in the gart code when unloading the radeon module: http://pastebin.ca/1842799
glisse: agd5f: with radeon-testing ?
agd5f: glisse: yes
glisse: likely double call
glisse: bah maybe i should try doing the stack init stuff
glisse: will look into it tomorrow if no one beat me to it :)
edwin: hmm multisampling is not implemented?
edwin: hmm line/poly smoothing either
edwin: is this just a missing implementation, or a bug somewhere?
agd5f: edwin: in r600? not implemented
edwin: r600 driver, yes
edwin: ok, I'll wait :D
edwin: any progress on fixing shadowtex btw? (the corrupted depth textures)
edwin: are my comments here of any use in identifying where the bug is? https://bugs.freedesktop.org/show_bug.cgi?id=24818#c5
spreeuw: does the inteldriver have an irc channel?
[SilverFox]: wheres the windows ati channel?
spreeuw: I'd like to know if it supports s3tc
spreeuw: getting sick of the lack of it
wirry: windows user only use mirc on qnet :>
adamk: spreeuw: #intel-gfx
chithead: spreeuw: I don't think it supports s3tc
adamk: I believe it does via the external dxtn library.
chithead: yes, with external libdxtn, s3tc should work
DanaG: argh, Wine fail: 32-bit Wine on 64-bit host tries to load DRI drivers from /usr/lib/dri -- and then fails with "wrong ELFCLASS elf64".
ajax: use a real multilib OS
Thunderbird: it is not a wine issue; wine just uses libGL
Thunderbird: on some distributions LIBGL_DRIVERS_PATH is not correct (this is at least the case on ubuntu)
DanaG: hmm, but every other GL app has no issues.
DanaG: ah, I guess every other GL app is 64-bit.
DanaG: I see, it IS just the lack of libgl_drivers_path var.
chb: evening - I've got a problem with kms and power management -> drm] not in vbl for pm change 00020002 00000000 at entry/exit - only happens when using 3D stuff - card is 4850hd and drivers are latest xorg-edgers ppa
chb: is that bug report worthy?
spreeuw: popped in my old 1900
spreeuw: holds up pretty well
spreeuw: it has some text artefacts that were gone on r600
spreeuw: in spring
DanaG: interesting... running wine with hl2dm gives "you have old and broken userspace please consider updating mesa and xf86-video-ati
Thunderbird: libGL is telling you that?
DanaG: dmesg is telling me that, actually.
adamk: I had that in dmesg, too. Updated both and the message went away.
adamk: I had never seen that before.
DanaG: Version: 7.9.0~git20100313.2615bba1-0ubuntu0sarvatt
DanaG: xserver-xorg-video-radeon: Version: 1:6.12.191+git20100314.67e81c8f-0ubuntu0sarvatt
adamk: Yeah, I was running from git, too.
DanaG: hmm, and it's spewing "radeon 0000:01:00.0: vbo resource seems too big for the bo"
fisxoj: so... which branch of mesa is best for radeon right now (most features, not most stable)? I used to get a 7.8 version string in glxinfo when I compiled it myself, now it's back to 7.7.1
chithead: most features you will find in git master
fisxoj: that's what I was using
Wizzup: is also getting the old and broker userspace error, even though he updated half a day ago
airlied: thingks I'll just remove the warning
DanaG: weird... wine-ing "gpu caps viewer" reports a DIFFERENT version of MESA than native glxinfo reports!
DanaG: 1.5 Mesa 7.7.1-DEVEL
DanaG: Mesa DRI R600 (RV635 9591) 20090101 x86/MMX/SSE2 TCL DRI2
DanaG: glxinfo gives:
DanaG: OpenGL renderer string: Mesa DRI R600 (RV635 9591) 20090101 TCL DRI2
DanaG: OpenGL version string: 2.0 Mesa 7.9-devel
BioTube: sounds like Wine's linking against another library
chithead: or you are running a 64 bit system with different 32-bit mesa
DanaG: No wonder it's giving me "you have old and broken userspace"
Thunderbird: use a 32-bit glxinfo
DanaG: ah, "getlibs" with the 32-bit package from xorg-edgers fixed it.
DanaG: ooh, it's even good enough to run FurMark.
spstarr: airlied: I thought glisse needed to commit something to mesa to stop the 'broken userspace' error
DanaG: I figured out my issue: it turns out ubuntu's xorg-edgers still leaves the 32-bit with the default Mesa.
DanaG: default being 7.7.
airlied: spstarr: he did I thought, I don't think we really need the warning
airlied: its not like the old -ati or mesa can do bad things
spstarr: oh, im not sure if he did or not
agd5f: spstarr: he did. it's likely something else he didn't catch
agd5f: or a corner case
spstarr: airlied: do we have glean for Fedora?
spstarr: or do you just build it from source
spstarr: agd5f: ok
airlied: spstarr: source, I was hoing someone would package piglit but it hasn't happened yet
agd5f: airlied, glisse: r7xx hangs the box solid on module unload. might be related to the oops I posted earlier
spstarr: agd5f: it doesn't do anything anyway the notice
spstarr: airlied: ok, i'll build now
airlied: spstarr: we normally just use piglit
airlied: which runs glean
spstarr: compiling piglit says i need glean headers though
agd5f: with radoen-testing
spstarr: oh wait
airlied: glew is in fedora
fisxoj: how do I get the r600 opengl 2.0? I only get 1.5
chithead: fisxoj: kms and mesa 7.8
fisxoj: I thought I was using 7.8...
spstarr: and freeglut we have also
spstarr: Linking C executable ../../bin/fp-rfl
spstarr: /usr/bin/ld: CMakeFiles/fp-rfl.dir/fp-rfl.o: undefined reference to symbol 'sqrt@@GLIBC_2.2.
spstarr: that is odd.. we're not linking in -lm
spstarr: same for glx-multithread no -lX11 linked in
spstarr: strange, but i fixed it in the cmake files
DanaG: hmm, is that RS780NM really an R600-level 3D core?
DanaG: Or is it one of those weird hackups that's not really R600?
airlied: DanaG: its r600 level alright
airlied: it might be a bit cut-down vs a discrete r600
airlied: but it supports the same things as an r600
DanaG: Now if only I could get one without shipping from Japan. =þ
DanaG1: [13242.497551] radeon 0000:01:00.0: ffff880018828c00 reserve failed for wait
DanaG1: [13272.174373] radeon 0000:01:00.0: ffff880018828c00 reserve failed for wait
DanaG1: [13428.140386] [drm:radeon_fence_wait] *ERROR* fence(ffff88002f318040:0x00426109) 510ms timeout going to reset GPU
DanaG1: got that.... and then no more. Kernel is dead -- heartbeat LED off.
DanaG1: hmm, that keeps happening randomly.
DanaG1: Screen goes blank, and kernel dies.
FIReun: sounds more like you need a doctor
DanaG: ooh, new radeon KMS PM patches.
DanaG: Just need them in drm-next. =þ
dileX: DanaG: if you get 0001-drm-radeon-kms-pm-fix-typo-in-power-table-parsing.patch and 0002-drm-radeon-kms-gfx-init-fixes-for-r6xx-r7xx.patch from d-r-t GIT branch. furthermore disable 0003-i2c-algo-bit-Add-pre-and-post-xfer-hooks.patch (its in linus-tree).
dileX: then new pm2 patches should work fine w/ linus-tree
dileX: DanaG: http://files.iniza.org/drm-radeon-kms/drm-radeon-testing/
agd5f: dileX: I have newer ones up
dileX: agd5f: trying new patchset
dileX: DanaG: http://paste.pocoo.org/show/190556/ (compiling now...)
DanaG: hmm, what I mean is, I don't want to have to compile my own; I'll wait until I can get it here: http://kernel.ubuntu.com/~kernel-ppa/mainline/
dileX: agd5f: with new pm2 patchset (23) I get a black screen here on rv515 - logs give no big hints
dileX: any suggestions on how to dig deeper?
dileX: thats what I tried: modprobe -r -v radeon drm ; modprobe -v drm debug=1 ; modprobe -v radeon modeset=1