FIReun: I drag a window around the screen and my cpu fan goes crazy
soreau: FIReun: With which wm?
DanaG: argh, "invalid operation" in glsl... doesn't tell me WHAT operation is invalid.
DanaG: hmm, is there a glsl validator somewhere?
DanaG: er, weird... just removing one "uniform" stopped "invalid operation"
DanaG: or maybe it wasn't the error I thought it really was.
DanaG: ARGH
DanaG: so weird... even with a literally EMPTY shader, it still sometimes gives "invalid operation"
DanaG: oh, I see... the "invalid operation" is somewhere else.
MostAwesomeDude: airlied: Weekend ping. Any opinions on "Re: Mesa (master): st/mesa: Always recalculate invalid index bounds." ?
lordraptor: i favour DanaG's code
DanaG: favor?
lordraptor: yup
DanaG: anyway, yeah, it does seem like my code is at fault.
DanaG: "favor" is a weird way to say it.
FIReun: soreau: e17
lordraptor: hows crossfire dual head workin' on fglrx?
lordraptor: or best to stick with a card...
DanaG: ah. no wonder my thing broke:
DanaG: wallTexID_3D=1939463928
DanaG: wallTexID_3D... forgot to call glGenTextures
airlied: MostAwesomeDude: I've been following along but would need to re-read to form an opinion
MostAwesomeDude: airlied: Basically pipe_buffer::max_index and pipe_draw_elements(..., max_index, ...) are both ~0 in a certain case that Wine hits a lot, and it's apparently our problem.
airlied: MostAwesomeDude: can we work it out from the size of things?
MostAwesomeDude: airlied: Yeah, trying to figure out how it could be done.
MostAwesomeDude: It would be nice if the kernel interceded and force-limited the vertex clamps, but I guess that's not feasible.
MostAwesomeDude: Or at least that's not the way it's written. :3
airlied: isn't max index the min index + sizeof vbo / vertex size?
MostAwesomeDude: Ideally, yes.
MostAwesomeDude: But the kernel maths appears slightly different.
DanaG: glTexImage3D(GL_TEXTURE_3D, 0, GL_RGBA, walls[0][0]->width, walls[0][0]->height, 6, 0, GL_RGBA, GL_UNSIGNED_BYTE, texels[set]); -- weird, it's segfaulting on this.
lordraptor: :s
DanaG: _mesa_TexImage3D (teximage.c:2394)
DanaG: that's one thing I love about having open-source drivers: I get that, instead of fglrx_dri.so
DanaG: http://pastebin.com/jzzDF7eS -- gdb.
DanaG: aah, looks like a case of having one too few or too many levels of pointers.
FIReun: hmm, this doesnt look good -- http://www.pastebin.org/112396
FIReun: appearing at the end of my dmesg
FIReun: [TTM] and drm:radeon_object_list_validate errors
DanaG: argh, this page is screwed up: http://www.opengl.org/sdk/docs/man/xhtml/glTexImage3D.xml
DanaG: lines go way off to the right, with no scrollbar.
DanaG: er, nevermind, it's full-zoom on Firefox, combined with a crappy xml.
GNU\colossus: not for me
DanaG: hmm, try zooming in.
GNU\colossus: no problem
GNU\colossus: i'm using konqueror though
DanaG: Weird.
DanaG: ah.
DanaG: weird... and now it's working fine in Firefox, too.
FIReun: and of course the wm maintainer says the driver must be broken
FIReun: but that it also could be that I'm not using whats in svn
FIReun: grrr
FIReun: wonders if gentoo would make living with source built everything easier
chainsawbike: yep :)
FIReun: adjusting the cooking temp by number of compile threads is tedious
FIReun: I'd rather use the stove
chainsawbike: lol
FIReun: I'm waiting for the package management system take on facebook
chainsawbike: ?
FIReun: if I have a rs480 on amd_64 and subscribed to the "newest" level of updates, then I should be getting packages built with that in mind (from other people subscribed to the same set of common hardware)
chainsawbike: hmm would be interesting
FIReun: so I should get a kernel lightened of all the "generic" crap, the latest radeon drivers that have had changes that apply to my hardware along with all the other dependancies, but built / configured for similar systems
FIReun: and of course all this "specificity" would be tested by all the other ppl subscribed to the same setup, but with the "bleeding edge" level of updates
FIReun: kind of what debian started with in the beginning, but its fallen down since then.
FIReun: its a difficult problem, so I'm not bashing debian
FIReun: the diversity of what I propose would make manual oversight nearly impossible
mcencora: Thunderbird: what opengl command is executed during IDirect3DDevice9_SetRenderState(device, D3DRS_CLIPPING, FALSE)?
Thunderbird: if I remember correctly we execute small helper functions which are in state.c I think
Thunderbird: look for 'state_clipping'
mcencora: Thunderbird: I can't find it. where is the IWineD3DDevice_SetRenderState function implemented? can't find it with grep
Thunderbird: mcencora, it is all COM stuff (vtable magic)
Thunderbird: there is IWineD3DDeviceImpl_SetRenderState
Thunderbird: IWineD3DDevice_SetRenderState is a macro which basically does object->vtable->SetRenderState(object, parameters)
Thunderbird: its our c++ emulation
edwin: if I --enable-debug, then shadowtex crashes here: radeon_common.c:970
edwin: packet0 = state->cmd[i];
edwin: state->cmd is NULL, is that normal?
edwin: hmm, its always NULL
mcencora: Thunderbird: what d3d function causes the rendering output to be displayed? EndScene?
Thunderbird: in general _Present
Thunderbird: d3d works with so called swap chains
Thunderbird: chains of backbuffers and a frontbuffer; _Present is basically 'SwapBuffers'
mcencora: Thunderbird: ok. I'm trying to make the visual/depth_clamp_test to work. but somehow nothing gets rendered
mcencora: Thunderbird: the output color comes from the _Clear function called at the beginning, and that's all
Thunderbird: try to set the OffscreenRenderingMethod to backbuffer first and see if that works
mcencora: Thunderbird: no change
mcencora: Thunderbird: setting Z func to ALWAYS didn't help either
Thunderbird: it might be useful to compile wined3d in a different way
Thunderbird: I would compile it using -DUSE_WIN32_OPENGL
Thunderbird: in that way all opengl calls will be routed through opengl32.dll
Thunderbird: if you use WINEDEBUG=+opengl you will see all opengl calls
Thunderbird: perhaps that can aid your debugging
mcencora: Thunderbird: I think I have an idea what is going on
mcencora: Thunderbird: at the beginning of the test the depth range is set to (0, 7.5)
mcencora: Thunderbird: and these values are passed to glDepthRange
mcencora: Thunderbird: but according to GL spec these values will be clamped to [0,1]
mcencora: Thunderbird: so the wine should rescale the MinZ and MaxZ before passing it to glDepthRange
fabio333: hi there
fabio333: kms is broken on my igp 345m
SnowRaptor: how do I circumvent and error like "Entry 'drivers/gpu/drm/drm_edid.c' would be overwritten by merge" when pulling drm-radeon-testing over 2.6.33?
BioTube: try 'git clean -df'
SnowRaptor: BioTube: same thing
fabio333: is it true that fglrx got opengl video acceleration (youtube)?
fabio333: flash is very slow with radeon...
fabio333: what about kms on my igp345m (rs200)?
adamk: fabio333: Assuming anyone is around to help you, you still haven't given them much to go on.. "broken" is not particularly specific.
fabio333: adamk: ok
fabio333: i would know whether kms will work on old ati cards
fabio333: like rs200, because as for now it doesn't
fabio333: kennel mode setting is only meant for recent hardware?
fabio333: i need it because of plymouth in ubuntu lucid.
adamk: I believe that KMS is supported on all ATI video cards.
adamk: You need to explain how it doesn't work.
fabio333: wait a sec
fabio333: i can't find it, i cant start my ubuntu with kms on
fabio333: i got black screen so i have to use radeon.modeset=0
fabio333: "
fabio333: plymouth is a a graphical boot system for Linux which takes advantage of the kernel-based mode set‐
fabio333: ting (KMS) available for modern graphic cards
fabio333: ---- from man plymouth
SnowRaptor: has anyone tried to merge drm-radeon-testing into 2.6.33's tree?
adamk: fabio333: I know what plymouth is... I would wager everyone in this channel knows :-)
adamk: fabio333: Is this a kernel you compiled yourself?
fabio333: adamk: read moderna graphic cards
fabio333: modern*
fabio333: rs200 is modern?
fabio333: kernel is from lucid ubuntu: 2.6.32-14-generic
adamk: Yes, again I believe all ATI video cards are supported by KMS.
fabio333: https://bugs.launchpad.net/ubuntu/+source/plymouth/+bug/527083
adamk: So it looks like this is a known bug.
adamk: If that's the case, I guess all you can really do is wait.
fabio333: is there a blacklist?
adamk: Not in KMS, I don't believe. I guess there might be one that Ubuntu or Plymouth keeps track of.
fabio333: http://osdir.com/ml/fedora-devel-list/2009-09/msg00771.html ----> mine is igp345
fabio333: what about opengl video acceleration?
fabio333: is it true that flgrx is faster there?
adamk: I've never compared them, so I don't know.
evil_core: fabio333: fglrx is definitely faster in opengl
adamk: Yeah, but he was specifically asking about opengl video acceleration.
evil_core: fabio333: and not fater by some percent, but it got multiplied performance in opnegl
evil_core: hmm...eee?
adamk: But, frankly, it doesn't matter in his case anyway as fglrx is not an option.
evil_core: opengl is faster, but wtf is opengl video acceleration?
evil_core: opengl doesnt have video support
adamk: You can use opengl for video rendering. Not sure if that's what he's talking about.
evil_core: probably about flash problems
SnowRaptor: Any idea how to circumvent an error like "Entry 'drivers/gpu/drm/drm_edid.c' would be overwritten by merge" when pulling drm-radeon-testing over 2.6.33?
evil_core: r300g: add high quality anisotropic filtering for R5xx (disabled by default)
evil_core: how to enable it?
evil_core: aniso affects performance at all? I remember that AA were problem for me, but aniso up to 8x on gf4mx never
wulfen: test
evil_core: Floating point exception - I got this error when running one of the GrubbyGames games(fizzwizzle),
evil_core: se noai
evil_core: a DisplaySize 423 318se ai
evil_core: I added such line to /etc/X11/xorg.conf according to one ubuntu topic, but didnt helped(probably bad section)
evil_core: It was some formula to multiply it by something and divide res by 75(font size?)
evil_core: and I would prefer to avoid changing xorg.conf for such workarunds(xorg restart requried)
Thunderbird: mcencora, I don't know much about that area but if the gl spec says the data will be clamped to 0-1.0, it might mean that it is done by the driver
Thunderbird: if the values are too large
Thunderbird: I think there are more APIs with this behavior
evil_core: se noai
evil_core: a--with-dri-drivers=r300,swrast se ai
evil_core: aits enouh to build gallium?
Thunderbird: mcencora, btw also add a test for this to piglit ;) (so that other drivers don't get the issue; I think the driver is supposed to do the clamping)
mcencora: Thunderbird: hmm, this sentence "Thus, the values accepted by glDepthRange are both clamped to this range before they are accepted" would mean that it's clamped before it gets to hw
Thunderbird: let me look something up
Thunderbird: this could be related to nvidia drivers
Thunderbird: they have (or had) some clamping settings you could override
Thunderbird: it could be that some nvidia specific behavior slipped into wine
Thunderbird: but also realize that a lot of drivers allow things which are outside of the spec
Thunderbird: and that way because apps rely on them become 'part of the spec'
Thunderbird: it seems this clamping override was related to texture coordinates
edwin: how can glDepthRange reject values? the only error I see there is an invalid operation if called after glBegin
edwin: so isn't it supposed to silently clamp to 0-1 always?
Thunderbird: yes, that's what I mean as well; don't the color calls behave the same way?
evil_core: - 640x480 59.9 59.4
evil_core: + 848x480 59.7
evil_core: + 720x480 59.7
evil_core: + 640x480 60.0 59.4
evil_core: why it differs?
evil_core: its diff between kms and ums
evil_core: it probably causes that some games in wine doesnt run under wine
evil_core: probably problem is 59.9 vs 60.0 refresh rate for 640x480
mcencora: Thunderbird: hmm, I think wine should use glFrustrum for setting minZ, maxZ (not glDepthRange) if ARB_depth_clamp isn't available
mcencora: Thunderbird: if I understand the docs correctly glDepthRange can be used to increase the Z resolution
Thunderbird: I don't know much about that area yet (I have never done much in that area but recently resumed a lot of 3d hacking) but I do remember that the depth code is very fragile
Thunderbird: something to do with opengl using -1..+1 and d3d using 0..+1 or the way around
evil_core: anybody would help me with finding splution that Dethkarz runs under UMS, but not under KMS?
Thunderbird: but a lot of code still needs to be moved to shaders but the last fixed function part doesn't have a high priority
Thunderbird: unless someone volunteers to hack on that :)
Thunderbird: the patches exist but they have to be cleaned up and some leftover performance issues have to be fixed
edwin: the shadowtex corruptions are visible on piglit depth tests too
edwin: https://bugs.freedesktop.org/show_bug.cgi?id=24818#c6
edwin: each time I run those tests I get some random colors (between pink and black)
soreau: MostAwesomeDude: FWIW, today alpha blur and mags fisheye mode effects for compiz are working with classic mesa
soreau: alpha blur causes a crash in the gallium driver though
soreau: and doom3 runs really fast with st3 or whatever textures, but they're kinda rainbow and messed up :)
soreau: without them, it renders all correctly for the most part but it's a bit slower
soreau: and gallium's consistently slower than classic
Wizzup: soreau: what driver @ alpha blur? r300 or r600
soreau: Wizzleby: 300
Wizzup: ok
soreau: I don't think r600g works at all yet
Wizzup: ah,silly me
[Enrico]: mhm i have some problem playing nexuiz on my r200 (fedora 12 up to date). with kms it works, but very slow (not playable) with settings at minimum. in UMS it starts but seems freezed (but if i push f10 it stops so it is not, just the graphic is frozen)
[Enrico]: i wonder why it doesn't work in UMS
[Enrico]: mhm no ok in UMS it works, but less then 1 FPS in the first screen
Jonimus: Gah, xorg-edgers just broked my fresh Ubuntu install I did for my Mom :/
FIReun: xorg-edgers != mom
soreau: lol
amarsh04: agrees with FIReun
Jonimus: All I did was enable it, upgrade and reboot, now I have a broken system
Jonimus: this is why I switched to Arch a few years back
Jonimus: but I figured Ubuntu is supposed to be easy and all
Jonimus: and what happens, my X brakes faster than I thought possible
Jonimus: and of course switching to a TTY failed as well :/
soreau: Jonimus: Which version of ubuntu and which kernel version?
Jonimus: 9.10 and I think there is a 2.6.31 kernel on there
Jonimus: I just did a fresh install
soreau: If you use xorg-edgers, you need to update the kernel as well
Jonimus: soreau: ok from where?
soreau: especially because there was a libdrm API change around xmas last year IIRC
Jonimus: ahh KK
soreau: I guess you could just use the lucid kernel
Jonimus: so how do I force Ubuntu to not boot into X?
Jonimus: since as soon as gdm starts up Its all down hill
soreau: Boot into recovery mode?
jcristau: add single to the kernel command line
jcristau: or what soreau said, which is probably the same
Jonimus: Ok how do I force grub to show up, I can't remember what key it is, I'm not used to it being hidden
soreau: Esc
soreau: Then E
Jonimus: ahh KK thanks
Jonimus: I'll try that, though how do I get the newer kernel then?
Jonimus: I'm fine with here being on a lucid system since iirc lucid is the next LTS release
soreau: I guess you could add the lucid repo, install the kernel, then remove the repo before your system notices
soreau: But it might be a good idea to check in a distro-specific channel
Jonimus: kk
MostAwesomeDude: soreau: Alright, thanks. Can't do anything about it right now though.
soreau: MostAwesomeDude: No worries, just giving a babbling update ;)
SnowRaptor: Any idea why DSVisualizer's 3d window is completely black and the cursor disappears when goes over it?
BeteNoire: hi, any chance that newer radeon driver fixes lockups of x 200 m?
SnowRaptor: BeteNoire: not sure, I still get some random lockups
SnowRaptor: BeteNoire: when do you use to get yours?
BeteNoire: switching to tty
SnowRaptor: BeteNoire: are you using KMS?
BeteNoire: no, but when i tried it didn't change anything
BeteNoire: last version i use was 6.12.4, and the screen went crappy when i tried to change to tty
SnowRaptor: I'm using xpress 1150 (200m renamed) and I'm having little problem with KMS and airlied patches
SnowRaptor: but I'm getting xf86-video-ati , mesa and their buddies from GIT
chithead: if you have problems with vtswitch, try 6.12.191 with kms
BeteNoire: trying this requires recompiling bunch of packages on my gentoo, and failure requires downgrading them back, i need to be sure it works
SnowRaptor: BeteNoire: you just have to unmask the ~9999 versions of some packages
SnowRaptor: and probably forget about gentoo-sources
SnowRaptor: although it would be nice to have an ebuild of airlied's patches!
BeteNoire: i am not used to breaking something that works, i'm on 2.6.28 tuxonice with 1.5.3-r6 xorg-server and 8.593 ati-drivers
mfedyk: airlied: do you know if the tv-out patches made it into rhel5? I have an older card I want to use with myth-tv...
mattst88|laptop: mfedyk, you want to run rhel5 on your myth-tv box?
airlied: mfedyk: in theory the r500 driver on RHEL5 might work
airlied: mfedyk: you'll have to pick it in xorg.conf
mentor: Is there an available reference for the 3D registers on r6xx?
mentor: is confused
xig: i am running linux 2.6.33 (gentoo)/Xorg 1.7.5 trying to get rid of the "IRQ's not enabled, falling back to busy waits: 2 0" message. i followed the instructions here: http://www.phoronix.com/forums/showpost.php?p=86849&postcount=251 and the modules (drm+radeon) seem to work, but i still get that message :O
Jonimus: xig: you need kms enabled
xig: Jonimus: what does kms stand for?
Jonimus: xig: Kernel Mode Setting
xig: is this a kernel configuration option or where is this to be enabled?
Jonimus: try adding radeon.modeset=1 to your kernel cmd line and rebooting
Jonimus: you should then get a hires console(tty)
xig: thanks, i try
xig: Jonimus: i love you
Jonimus: hugs xig
agd5f: mentor: http://www.x.org/wiki/DataSheets
xig: Jonimus: well, there is still one issue: the message is gone, but i see glxgears having just 1/3 of the framerate it had when the message did occur
xig: Jonimus: is this normal?
mentor: yay!
mentor: licks agd5f
Jonimus: xig: depends on your video card
xig: Jonimus: its mobility radeon 4200 (RV620) in my notebook
Jonimus: heh then I'd say thats possible, what does glxinfo | grep Open tell you?
xig: Jonimus: five lines
xig: Vendor: Mesa Project; Renderer: Software Rasterizer
xig: version: 2.1 Mesa 7.7; shanding lang version 1.2
xig: (fifth is empty extensions line)
soreau: MostAwesomeDude: That reminds me.. I wanted to ask why gallium reports gl1.5 on rv350 now? I didn't really look into why it might be..
xig: Jonimus: that software rasterizer think looks suspicious, doesnt it
xig: thing*
MostAwesomeDude: soreau: Not sure.
soreau: xig: Yes. your X log might give a hint as to why it's using swrast
soreau: MostAwesomeDude: How could it be though? http://pastebin.com/fgv5zCpM
soreau: MostAwesomeDude: It's on master branch at commit de4ee20578a79e024b0de83c40648112f42c994e
MostAwesomeDude: soreau: http://pastebin.com/efWWf8AA
MostAwesomeDude: NPOT and two-sided stencil are disabled.
soreau: MostAwesomeDude: Not sure how to enable/disable them in the config.
MostAwesomeDude: soreau: r300_screen, get_param.
soreau: MostAwesomeDude: ?
xig: soreau: i cant find a real "error" in my X log, but there are some lines at least containing the word SWRAST, it would be nice if you could take a look: http://codepad.org/KKaRk5jn
MostAwesomeDude: soreau: You have to edit the code. src/gallium/drivers/r300/r300_screen.c:r300_get_param has several if (r500) statements.
soreau: xig: Does 'dmesg|grep drm' show anything interesting?
xig: no, nothing
soreau: MostAwesomeDude: when did these get disabled or when did the version go from 2.1 to 1.5?
soreau: xig: Can you just pastebin the complete output of dmesg?
soreau: oh here's the problem
xig: soreau: dmesg here: http://codepad.org/gpKJYdaw
soreau: [drm] failed to load kernel module "radeon" (EE) RADEON(0): [dri] RADEONDRIGetVersion failed to open the DRM
xig: soreau: wtf
soreau: xig: You need to build your kernel with radeon DRM
xig: soreau: i am so sorry, i did completely overlook that error
xig: soreau: thanks for your help
xig: although i built with drm, but i gonna check it
xig: (i even build a differen drm module from some git repository (r6xx-r7xx-3d)
soreau: Well there's userspace libdrm, then there's the kernel side drm
xig: drm.ko was the product of what i built (from git)
xig: and radeon.ko
xig: thats not userspace, is it?
soreau: Doesn't sound like it
soreau: it couldn't find or load them whatever it was
soreau: why did you compile those out-of-tree modules?
xig: i did that while trying to solve that now solved problem with the IRQ message, i found that post:
xig: http://www.phoronix.com/forums/showpost.php?p=86849&postcount=251
xig: and followed the instructions
soreau: In any event, you would need to make sure loading the modules with modprobe does not show any errors
soreau: or, use the code that comes included with the kernel
xig: drm loads without problems, and (i just discovered) radeon does not load because it does not know modeset parameter
Jonimus: xig: you said you were running a 2.6.33 kernel so that post does not apply to you
xig: Jonimus: oh, damn. okay i try then falling back to the normal DRM and radeon and try with modeset again
soreau: Yea I don't see any reason to build out-of-tree modules
soreau: 2.6.33 should be new enough
xig: this will take a minute, thank you in advance
Jonimus: xig: yeah the out of tree one you built is older than the ones in 2.6.33
xig: ohh :D
xig: damn
xig: hey there is that kernel option
xig: "Enable modesetting on radeon by default - NEW DRIVER"
Jonimus: yep :P
xig: sounds pretty much for what i want, but
xig: help text says
xig: R[12345]xx is supported
xig: and "Works is underway to provide support for R6XX"
Jonimus: xig: .33 supports r[67]xx
xig: aight
xig: hires console appeared - looks well so far
xig: but still software rasterizer
DanaG: r600: Number of extensions: 109 GL 1.3 status: 9/9 present GL 1.4 status: 16/16 present GL 1.5 status: 3/3 present GL 2.0 status: 7/8 present - Missing these extensions: --- GL_EXT_texture_non_power_of_two GL 2.1 status: 2/3 present - Missing these extensions: --- GL_ARB_shading_language_120
xig: xorg log: [dri] radeon kernel module version is 2.0.0 but version 1.17.0 or newer is needed.
xig: this message is weird
soreau: xig: Rebuild libdrm, mesa and xf86-video-ati
xig: aight
soreau: It wouldn't hurt to use the -9999 packages from the x11 overlay
xig: how do u mean this?
soreau: aren't you using gentoo?
xig: i am, but i didnt know a x11 portage overlay
xig: what exactly is it for?
soreau: 9999 packages will provide latest developments from git
soreau: you can ask more about overlays in your distros support channel
xig: in general i know about overlays, thanks:) sounds pretty cool
xig: soreau: i rebuilt the three but still get the same message
xig: did u notice it complains about having 2.x , and _wanting_ _newer_ than 1.x so the message is basically nonsense?
Jonimus: xig: yeah
soreau: That has since been fixed in git afaik
xig: which package actually causes the problem?
soreau: xig: You have to have new enough versions of these three and you must build libdrm then mesa and ddx against that libdrm
BioTube: you've got a 6.12.x ddx
soreau: of course the package manager should take care of that
soreau: I think it is old ddx too
xig: ddx?
Jonimus: xf86-video-ati
Shuren: DanaG, what program give you that output?
DanaG: http://github.com/MostAwesomeDude/madsnippets/blob/master/glxinfo.py
DanaG: pipe glxinfo into it.
DanaG: it also needs a hashbang added.
Shuren: thank's
xig: is desperate
DanaG: I do think the wording of that message could use a tweak:
DanaG: IT should say "need 1.17.x", not "1.17.0 or newer"
xig: soreau: now updated xf86-video-ati from x11 overlay, and rebuilt libdrm and mesa. still that version conflict. i am now recompiling xorg-server for undefined reasons but dont think this will help
DanaG: where 'x' is literally 'x', to mean 'anything here'
DanaG: did you build libdrm with --enable-radeon-experimental-api ?
DanaG: And is radeon (and its dependencies) in your initramfs image?
Jonimus: xig: your xserver should be fine, you just need libdrm, then xf85-video-ati, then mesa
soreau: xig: Do you have VIDEO_CARDS="radeon" in /etc/make.conf?
xig: y
Wizzup: xig: what libdrm and mesa?
Wizzup: they should also be from overlay
Wizzup: http://dpaste.com/171971/ uhm, that's mine, may be outdated
Wizzup: (keywords)
xig: Wizzup: overlay doesnt seem to provide never versions of those
xig: x11-libs/libdrm-2.4.19 media-libs/mesa-7.7-r1
soreau: xig: I would install all -9999 versions
xig: soreau: :/
Wizzup: [I] media-libs/mesa
Wizzup: Available versions: 7.5.2 ~7.6.1 ~7.7-r1 {M}(~)9999[1]
Wizzup: package.unmask: http://dpaste.com/171972/
Wizzup: (don't add the portage)
xig: thanks
Wizzup: it may be a bit outdated
soreau: xig: build libdrm first, then mesa and ddx against that libdrm (all -9999 versions)
xig: ddx == xf86-video-ati?
soreau: yes
Jonimus: xig: yeah
xig: okay, i do that and then report back, ty all
Wizzup: looks forward to power management in kms
Jonimus: Wizzup: there is wsome but not much yeah
Wizzup: read the phoronix article about more power manage stuff emerging in 2.6.34, looked quite promising
Jonimus: Wizzup: I already have some of agd5f's dynPM patches applied, its not much but way better than nothing
Wizzup: :-)
Wizzup: I'm on UMS mainly because my charger gets dangerously hot with kms
Jonimus: ahh
shadowmaster: hm. Power management. I'm not sure I even have that with UMS.
shadowmaster: at least I never use such things like turning off the screen. I tend to suspend the laptop to RAM/disk instead.
soreau: there's a force low power mode or some option you can set iirc
Wizzup: driver "radeon"
Wizzup: Option "ForceLowPowerMode" "true"
shadowmaster: and DynamicPM
shadowmaster: so I'd not lose or win anything (well, the framebuffer consoles are pretty) with KMS :D
soreau: Also, gl apps play nicely with compiz when using dri2
lordraptor: are there any driver options i can use yo speed up radeon?
Wizzup: shadowmaster: well,kms has no corruption for me, while ums does
lordraptor: mind you KWin doesnt animate with dual monitors
shadowmaster: argh, yeah, that's something I'm missing
odinsbane: I've installed everything from xorg-edgers and I still cannot use compiz. The system boots up fine into 2.6.32, but doesn't et past the loading in 2.6.33.
BioTube: odinsbane: did you install the rlc firmware?
odinsbane: BioTube, I believe so.
odinsbane: BioTube: I installed the karmic package but I am running jaunty.
odinsbane: BioTube: When I enable compiz with jaunty it kicks me out of gnome back to the login screen.
odinsbane: also glxgears fails
BioTube: see if 'find /lib -name R600_rlc.bin' turns up anything
odinsbane: BioTube: it does.
BioTube: strange then
odinsbane: BioTube: There was something about the boot parameter modset=1, I am not sure if I applied that.
odinsbane: Also I don't know how to check if the firmware was installed correctly.
BioTube: if KMS isn't enabeld, then it wouldn't ask for the firmware
DanaG: speaking of power management... my gpu has two battery states, and the PM stuff in KMS uses the wrong one.
xig: does anyone know whats going wrong when modprobe -v -i pcspkr fails silently, with nothing appended to dmesg and the module just not getting loaded?
xig: just doesnt get loaded*
xig: sorry
odinsbane: BioTube: do you know how I enable KMS ?
BioTube: odinsbane: passing radeon.modeset=1
odinsbane: in grup I can modify the boot options, so I would just append that to one of the lines?
BioTube: to the kernel line
odinsbane: (I only used modset=1
odinsbane: Ill give it a go
FIReun: tourists
Jonimus: lol
DanaG: hmm, how do I turn off that pm debug spew?
DanaG: it floods my dmesg and renders it useless.
odinsbane: Well glxgears works now.
Jonimus: DanaG: disable dynPM or remove the printf's from the patch
DanaG: hmm, a future idea: dynpm=1 or dynpm=2, where 2 is debug.
odinsbane: I still cannot enable compiz though, but it doesn't kick my out of gnome when I try.
odinsbane: I still have a DRM error message in my Xorg.0.log, where should I look for the firmware errors?
Wizzup: what is the error?
odinsbane: (II) [KMS] drm report modesetting isn't supported.
odinsbane: well maybe not an error
Wizzup: hasn't seen it before but guesses you either have old kernel or libdrm
odinsbane: libdrm is from xorg-edgers and the kernel is 2.6.32
Wizzup: doesn't know and wanders off to bed. :-)
odinsbane: I would try the 2.6.33 but I don't even get to a login screen with that.
odinsbane: night
Jonimus: odinsbane: I'd grab the 2.6.33 kernel
Jonimus: but thats just me
FIReun: 2.6.32-2 works here
odinsbane: Jonimus: I have the 2.6.33 kernel, but I cannot boot into it with radeon enabled
odinsbane: Although I haven't tried including the option radeon.modset=1
FIReun: thats a must for KMS
Jonimus: hmm my mom's desktop worked with 2.6.33, it was totally broke on 2.6.31, thought I'm thinking I still should have gone arch and not Ubuntu on it
odinsbane: http://www.pastie.org/869805
odinsbane: That is a snippet of my xorg log concerning drm
FIReun: Jonimus: I'm curious about arch
FIReun: havnt tried it yet tho
Jonimus: the radeon kernel module didn't load hmm, try pastbining a dmesg
Jonimus: FIReun: What do you use now?
FIReun: Jonimus: it started life as a debootstrapped sid install
Jonimus: lol, wow
odinsbane: http://www.pastie.org/869808
Jonimus: well Arch is def a different beast from Debain
FIReun: deboostrap is the king tho
FIReun: I'm tired of a dvd sized iso full of crap I dont want
FIReun: but apt needs channels
FIReun: exim4 is installed by default too
odinsbane: Line 782 seems to imply that I don't have my radeon.modset=1 in the right place.
Jonimus: odinsbane: its modeset not modset
Jonimus: thats your issue
Jonimus: you spelt it wrong :P
odinsbane: That was not a nice reboot. modeset=1 I couldn't get into gnome
odinsbane: I also couldn't get to a terminal with alt+ctrl+fX
odinsbane: Is there a way to access a sys.log from my previous boot?
xig: hey soreau and Jonimus, remember me? after long time of git downloading the -9999 versions (of libdrm, mesa, xorg-server and ddx), i finally solved that version conflict
xig: but:
xig: its still using software rasterizer -_-
xig: although i have no (EE) line any more in my xorg log
xig: it just says:
xig: 53.316] (II) AIGLX: Screen 0 is not DRI capable
xig: and then enables swrast
soreau: xig: Can you pastebin your X log?
xig: yes, mom
soreau: o.o
xig: moment*
xig: soreau: http://codepad.org/O6iKN8Ip (this is xorg log and dmesg)
soreau: xig: I don't see anywhere in the dmesg output that talks about drm or firmware which is odd
xig: not? i get 6 lines when grepping for drm
xig: sry
xig: wrong machine
xig: i get plenty of lines but maybe i pasted the output from the wrong machine, wait a second
xig: soreau: the dmesg output seems being truncated at that paste service
xig: i paste it elsewhere
xig: soreau: here we go: http://pastebin.com/CPYpApz9
chithead: [ 5.100301] r600_cp: Failed to load firmware "radeon/R600_rlc.bin"
soreau: xig: See line 1202-1203 of that paste
xig: i just discovered:/ where do i get that blob from?
soreau: There is a link in the topic here..
soreau: It has a firmware section I do believe
xig: i take a look thanks
odinsbane: Okay I boot with radeon.modeset=1 and I couldn't get to gnome, and the other ttys would show a login screen, but I could still login.
odinsbane: So I got the Xorg.0.log http://www.pastie.org/869830
odinsbane: and the dmesg out put. http://www.pastie.org/869834
xig: i have R600_me.bin and R600_pfp.bin, but X wiki is talking about R600_rlc.bin
xig: which one shall i choose?
Jonimus: xig: you need r600_rlc.bin as well
Jonimus: you need them all
Jonimus: odinsbane: you need the firmware see the topic
odinsbane: Jonimus: I tried to insall the firmware.
Jonimus: well you must have put them in the wrong place
odinsbane: Jonimus: I put them in the right place but I couldn't do the CONFIG_FIRMWARE_IN_KERNEL=y stuff
odinsbane: Jonimus: then I went and got the karmic linux-firmware-nonfree for karmic and tried that too. Which appears to have not worked.
BioTube: did you put the firmware in /lib/firmware/radeon?
odinsbane: BioTube: yes
Jonimus: odinsbane: see the link in the topic about the extra firmware that will not be included till Lucid
odinsbane: I'm looking at the trouble shooting extra firmware, and I don't know where to set the 'kernel-config' parameters.
BioTube: odinsbane: those are for building a kernel
odinsbane: I got my kernel from the ubuntu ppa
xig: Jonimus, soreau: it finally works! tons of thanks for all your help
Jonimus: odinsbane: then all you need is to dl the firmware and put it in the right spot
odinsbane: ls /lib/firmware/radeon/*
odinsbane: /lib/firmware/radeon/R600_rlc.bin /lib/firmware/radeon/R700_rlc.bin
odinsbane: Jonimus: also the dmesg appears to have used the firmware.
Jonimus: hmm ok
odinsbane: [ 16.751300] platform radeon_cp.0: firmware: requesting radeon/RV635_me.bin
odinsbane: It doesn't have an error following that.
Jonimus: yeah I jsut realized that
Jonimus: but your xorg log is more interesting
Jonimus: it appears to be using the fbdevhw driver....
xig: how can one control settings like enable/disable vsync (radeon)?
odinsbane: whoops
Jonimus: xig: driconf
odinsbane: I get some straing error because of version missmatch
Jonimus: hmm
odinsbane: Maybe I should go through and try to build the drivers from git.
Jonimus: ahh yes, you don't have the newer packages from xorg-edgers installed
Jonimus: at least thats how it appears
odinsbane: It seems as though it doen't recognize 2.0.0 > 1.17
Jonimus: odinsbane: yeah thats an old bug that has been fixed if you update to xorg-edgers or build from git
chithead: that error is because your ddx is too old for kms
soreau: odinsbane: That is because you have too old libdrm/mesa/ddx
chithead: need xf86-video-ati-6.12.191 or newer for kms
xig: Jonimus: when are driconf's settings applied? disabling vsync didnt had any effect, even after X restart
Jonimus: xig: that I have no clue on, though I'm not sure if you can disable vsync with KMS enabled
xig: okay
odinsbane: chithead: xorg-edgers jaunty package only includes 1.6.12.99
odinsbane: So I can compile them from the git, or I could enable the repo for all and use the distro independant packages.
odinsbane: Where should I get libdrm from?
soreau: odinsbane: Which distro?
odinsbane: soreau: I found it.
odinsbane: -ubuntu jaunty
odinsbane: It seems I could have saved a lot of trouble by updating to karmic
ToastyJustice: Well, I've tried updating using radeon, radeonhd, and updating my kernel, but I still can't get this thing to use something other than the software rasterizer.
odinsbane: Thank you guys, I finally got compiz to work with radeon drivers, I still cannot use the 2.6.33 kernel, and I haven't tried it yet but I'm not using modeset=1, I had to use modeset=0.
soreau: odinsbane: Did you pastebin an X log from the failed kms X session?
melkor: Yes but not since updating to the new mesa
melkor: I installed mesa in the wrong directory the first time, then I set modeset=0 and fixed my mesa.
melkor: Ill reboot and give it another try.
ToastyJustice: I'm most intrigued by melkor's progress, since I'm still working on getting mine going
DanaG: hmm, one thing I've found: it's best just to put stuff in /usr -- anywhere else makes it a pain to use.
melkor: yeah modeset=1 still will not let me reach the login screen, but now it is bright white.
melkor: I'm getting the Xorg.0.log
ToastyJustice: So guys, I've tried the radeonBuildHowTo, and a few other tutorials elsewhere, but, I can't seem to get this thing working. I have the 2.6.32 kernel, as I found the 2.6.33 to be buggy, but still my 4350 won't enable acceleration
melkor: http://www.pastie.org/869893 <- thats my xorg.0.log with modeset=1
melkor: ToastyJustice: what distro?
ToastyJustice: ubuntu 9.10
melkor: did you get the linux-firmware-nonfree package?
ToastyJustice: Nope
melkor: Did you d/load the firmware files then?
ToastyJustice: I may have, I've gone through so many thigns today. I'm going to try the linux-firmware-nonfree, see if it gets anything going
melkor: ToastyJustice: also what do you get when you do the glxinfo | OpenGL are you using the newest mesa drivers?
ToastyJustice: One sec, I'm just rebooting the system
melkor: Well I keep getting an error from java3d about irq's not enabled, but it redraws better than fglrx, but that could also be the new xorg.
_KAMI_: Driver still cool! Radeon 3470, Ubuntu 10.04 A3, KMS enabled, 2.6.33 drmnext driver
_KAMI_: Thank you for the hard work
ToastyJustice: melkor: Haha. Now it's giving me a segmentation fault when I try to tun glxinfo. But last I remember, it was using the software rasterizer with Mesa 7.7
melkor: ToastyJustice: did you use the modeset=1 boot parameter? I was getting segfaults from glxgears when I didn't
melkor: ToastyJustice: and the radeon.modeset=1 actually made my system not boot, so I am currently using modeset=0 and everything is working fine except for an error here and there.
melkor: _KAMI_: how is ubuntu 10.04?
ToastyJustice: melkor: where can I check the modeset=1? I haven't set that anywhere myself that I recall
melkor: ToastyJustice: I set it in /boot/grub/menus.lst on the kernel line for a kernel I wanted to enable it. (if you are using grub)
melkor: ToastyJustice: you can try it out without changing the menu.lst by hitting 'e' when you select your kernel at bootup.
_KAMI_: melkor: I am using Alpha 3 version of ubuntu with X org edgers PPA repository
melkor: I'm holding out for the release, but its really tempting.
ToastyJustice: melkor: don't see a menus.lst in the grub folder. I'll try it using the bootup options
melkor: ToastyJustice: its /boot/grub/menu.lst ... in jaunty.
melkor: ToastyJustice: I set the parameter by appending radeon.modeset=0 to the kernel line.
_KAMI_: melkor: I seems better than 9.10
_KAMI_: it