Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-12-23

Search This Log:


spstarr: airlied: GLSL works on radeon now, with SL it does not crash but you get ~ 1 FPS at least
spstarr: airlied: the IRQ issues im having remain with KMS.. hopefully some kernel folks will provide me some patches to test or just build linux-2.6 git master until then
spstarr: well limited GLSL since its still heavy dev
bridgman_: spstarr; any messages about software fallbacks ?
bridgman_: my recollection was that sl needed a higher level of gl and glsl than was implemented right now, but not sure what the specific extensions are
spstarr: didn't check, I will in a bit
spstarr: bridgman_: i had redirected all debug output to /dev/null
bridgman_: ok so look in /dev/null ;) ;)
spstarr: heh
spstarr: SL really uses GLSL to full, but turning on basic shaders showed a marked improvement when it was enabled
spstarr: bridgman_: I think SL is really good for stressing radeon
spstarr: bridgman_: the IRQ issue im having is kernel, that will keep me from using KMS for time being
spstarr: airlied: you dont have that W500 still?
spstarr: airlied: would be nice to see if you can reproduce the irq issues on that laptop with glxgears
spstarr: bridgman_: they have: 'Basic Shaders', 'Atmospheric Shaders', 'Water Reflections'
agd5f: airlied: it may need a more involved algo. I was trying to with a basic xfer style algo, but things fall apart dealing with buffer larger than the hw, e.g., edid
spstarr: bridgman_: my understanding is the latter 2 are just using GLSL
spstarr: since you must have basic shaders to use atmospheric/water
spstarr: I will certainly try GLSL with debug captured from SL to see what it spits out and what the DRI spits out (if anythign)
spstarr: wonders a GLSL program looks like
bridgman_: it looks like wierd C
bridgman_: you need to get an opengl book
bridgman_: there are probably some good downloads
spstarr: ah
spstarr: i found some examples
spstarr: it looks... Cish
spstarr: so GLSL is not compiled? or it's compiled at runtime?
spstarr: or compiled into something the GPU can recognize
spstarr: glCompileShader()
airlied: spstarr: probably next year, its not my laptop
spstarr: airlied: ok
Ghworg: What do I need to do to get KERN_DEBUG level printks to show up on the console?
bridgman_: spstarr; it's compiled at runtime; the app says "here's some source, compile it and use it"
spstarr: very nice
spstarr: this explains how Gallium can use LLVM to compile GLSL
spstarr: and optimize per-GPU
spstarr: very nice indeed
airlied: Ghworg: echo 9 > /proc/sys/kernel/printk
Ghworg: airlied: Thanks
spstarr: bridgman_: because GLSL is an OpenGL standard the program code is generic, so any GPU that can use GLSL just needs the compiler part per-GPU
spstarr: quite nice
bridgman_: yep, that's how software rendering works as well; JIT compile down to CPU code
spstarr: ahh!
spstarr: I wasn't sure how that worked for swrast
bridgman_: not sure if classic mesa and softpipe do that but it's the plan for llvmpipe
bridgman_: guess I should read the code one day
spstarr: so takes GLSL programs JIT compile for CPU (type)
spstarr: neat
spstarr: simpleWaterF.glsl
spstarr: neat to see the sources
bridgman_: yep, once you get it through your head that the program runs on every vertex or pixel, a bazillion or so in parallel, GPUs start to make more sense
spstarr: i am a little confused with atmosAmbient() are these built-in functions in GLSL or app-level?
spstarr: looks at SL code for this routine
spstarr: ah ok
spstarr: they are functions but there's no #include
spstarr: http://ati.amd.com/developer/sdk/RadeonSDK/Html/Samples/OpenGL/GLSLSimpleShader.html
spstarr: bridgman_: GLSL programs are different than fragment programs though?
bridgman_: there are "low level" and "high level" languages for shader programs
spstarr: ah
bridgman_: both support vertex and fragment programs
bridgman_: low level is arb_vertex_program and arb_fragment_program
bridgman_: high level aka glsl is arb_vertex_shader and arb_fragment_shader
bridgman_: the intermediate language inside mesa (between high level mesa and hw driver) is pretty close to arb_fragment_program / arb_vertex_program
bridgman_: for glsl mesa compiles the high level stuff down to low level IL
bridgman_: and passes it to the driver
MostAwesomeDude: The translation from GLSL to ARB_vp/ARB_fp is intended to happen before attempting JIT.
bridgman_: supporting GLSL needed some new IL instructions to be supported
MostAwesomeDude: GLSL, in essence, is just a series of extra opcodes on top of NV_*_program4.
spstarr: so the mesa does the generic GLSL support but and the actual DRI driver deals with the arb*_programs (only)?
MostAwesomeDude: Which, in turn, is on top of ARB_vp/ARB_fp.
bridgman_: yep
spstarr: ahh
spstarr: that makes sense
MostAwesomeDude: Yeah, the IR that DRI drivers deal with is Mesa IR, which is near-identical to ARB_fp/NV_gpu_program2.
bridgman_: IIRC the IL doesn't *exactly* match arb_*p but it's pretty close
bridgman_: what he said
MostAwesomeDude: Er, not gpu_program. Can't remember the exact NV shader extension.
spstarr: since the arb*_programs differ for GPUs extensions
bridgman_: ?
spstarr: so then as soon as a DRI driver has all the opcode handling for what the Mesa IR spits out it should just basically run when each opcode is implemented?
spstarr: ARB assembly instructions from what im reading
spstarr: its very assemblyish looking indeed from a sample ARB Vertex Shader
MostAwesomeDude: Yeah, the devil's in the details.
spstarr: indeed
MostAwesomeDude: Some opcodes are harder to implement than others. Only a small handful of opcodes are fully native to any GPU; the rest have to be emulated through opcode sequences.
bridgman_: for 5xx MostAwesomeDude went through the fragment IL instructions one by one, and each new instruction made some more stuff work
bridgman_: it was pretty cool to watch, I had no idea what was going on at the time ;)
MostAwesomeDude: Oh man, those were crazy times.
spstarr: that would be!
spstarr: kinda nice to see things just click into place like that as you work
MostAwesomeDude: And now we've got radeon_compiler for that.
MostAwesomeDude: And I actually understand some of radeon_compiler now. :3
bridgman_: oh good, so now two people understand it
bridgman_: is allergic to compilers
spstarr: what does 'IL' stand for?
bridgman_: intermediate language
spstarr: so almost like bytecode
spstarr: but not bytecode in this case
bridgman_: or intermediate representation (IR)
bridgman_: I forget which term mesa uses
MostAwesomeDude: Same thing.
bridgman_: yep
bridgman_: in our Stream stack il is the interface between the compute languages and CAL
bridgman_: OpenCL, Brook+ etc get translated into il, il goes through the CAL API into our shader compiler, and shader code comes out
spstarr: having a generic IL/IR does make that easier
bridgman_: for graphics, GLSL, arb_*p, HLSL etc get translated into il, then il goes through the shader compiler, etc
spstarr: I originally thought GLSL was done in the running 3D app itself and passed to OpenGL vs being separate 'programs' that are JIT
spstarr: the latter makes much more sense to me now
bridgman_: sometimes the shader programs are string constants in the program but storing separately makes for easier maintenance
bridgman_: also allows for separation of shader geeks from app geeks
spstarr: ya
spstarr: what I thought was 'complex' isn't, in terms of the organization of GLSL
spstarr: implementing the compiler for it.. well, INACD (I'm Not a Compiler Developer) :)
bridgman_: if you're interested our IL spec can be downloaded from here :
bridgman_: http://developer.amd.com/gpu/ATIStreamSDK/pages/Documentation.aspx
spstarr: looks
spstarr: the closest I've come to any compiler writing is a interpreted scripting language I wrote in VB
spstarr: http://www.sh0n.net/spstarr/visual.html
spstarr: that was a gross
spstarr: -rwxr-xr-x 1 spstarr spstarr 88599 1999-12-02 00:12 Applet.cls
spstarr: is reminded why he never uses VB anymore or Win32 APIs
bridgman_: VB is too pretty
bridgman_: real software should be ugly, hard to read, and monochrome
bridgman_: in a fixed-pitch font
bridgman_: what's that old saying... "real programmers don't use comments... it was hard to write, it should be hard to read"
spstarr: indeed
spstarr: i had to write a RunDLL() function to load Windows DLLs and execute any routine at runtime
spstarr: lost that code though for that :/
spstarr: I never finished advanced IF evaluation, i was a kid still back then
bridgman_: time for zzz
bridgman_: bbl
aszlig: hm...
aszlig: platform radeon_cp.0: firmware: using built-in firmware radeon/RV730_pfp.bin
aszlig: platform radeon_cp.0: firmware: using built-in firmware radeon/RV730_me.bin
aszlig: however in 2.6.33-rc1 it tries to request firmware for RV700 _after_ these two lines
dileX_: aszlig:
aszlig: so i need this on 2.6.33-rc1, right?
aszlig: hm, iirc it tried to load RV700, lemme check again...
dileX_: you need the files for R600/R700 in addition (license issues)
spstarr: you need the new firmware bits for IRQ
Nightwulf|work: hi all
dileX_: hi Nightwulf|work
aszlig: should be enough to put them in firmware/radeon within the kernel tree and compile them in, right?
aszlig: ah, sec, need to adjust the Makefile
agd5f: airlied: ok. I got a radeon i2c algo working with hw i2c and sw i2c
Nightwulf|work: hum...funny effects when trying to run 2.6.31.y + libdrm-2.4.16 + mesa 7.6.1 with radeon-git....just a corrupted screen with randomly coloured, vertical stripes all over the screen and a mouse curser represented by a similar looking box :P
Nightwulf|work: s/2.4.16/2.4.17
aszlig: dileX_: okay, thanks a bunch :-)
dileX_: n.p.
TBBle: Hmm. Any reason HDMI audio isn't enabled in the rv770 kms driver?
GoGi: glxgears says do_wait: drmWaitVBlank returned -1, IRQs don't seem to be working correctly.
GoGi: Try adjusting the vblank_mode configuration parameter.
GoGi: and then the colors of the gears seem wrong
GoGi: r700 with git drivers
Ghworg: IRQ support needs kernel 2.6.33 or one of the drm-* git branches
twnqx: will a stock 2.6.33.rc do?
Ghworg: Should do
GoGi: does not having irq support create the wrong color effect?
Zajec: airlied: did you avoid 5th patch from Alex for some reason?
Zajec: airlied: or just missed it?
Zajec: airlied: ah, Alex still didn't fix that default mode
Zajec: airlied: it's ok then
kdeman: We have apps querying for shading version 1.20 support like SL.
kdeman: Mostly getting apps to understand the driver is OpenGL 2.1 compatible is a key issue. Piglet testing comes in handy but I don't think it was passing tests for all the GLSL 1.20 opcodes yet.
mortal_: there is this radeon patch that fixes the ogl crash in old radeons
mortal_: could someone please commit it
mortal_: http://www.mail-archive.com/dri-devel@lists.sourceforge.net/msg45406.html
kdeman: Most pre-R500 driver(s) having OpenGL 2.1 functionality may be 'antique-only' unless the HW can TRUELY handle it performance-wise at a decent FPS rate...but good to have a patch for older chips.
kdeman: I'm talking performance at a 'normal' 1024x768 type resolution..not 1600x1200 or higher ones.
kdeman: There was someone working on tessellation on OpenGL since April 2001, so rendering with noise(3d/4d) and tessellation features for testing should be easy to do in the driver(s).
kdeman: See: http://www.flipcode.com/archives/Polygon_Tessellation_In_OpenGL.shtml
spreeuw: I saw people posting that furr benchmark, but theres no linux version of it
spreeuw: it was a very suxually laden furry donut
spreeuw: torus
kdeman: heh...I think that is why it seemed easier to test the 3D engines ported to linux like Unigine and SprintRTS.
kdeman: Not all those individual programs spread across the Net.
kdeman: SprintRTS/SpringRTS/
spreeuw: great game
spreeuw: I have it working
spreeuw: only trouble is textures
spreeuw: for 3/4 of the units and the map
kdeman: compressed texs?
spreeuw: spring now has a mod claled complete annihilation which aims for gpl compliance
spreeuw: and the engine already is
spreeuw: its a very cool fps a bit in the vein of Darwinia
spreeuw: and total annihilation
spreeuw: kdesk: yep I think so
spreeuw: I peeled open a map but those textures didnt seem to be compressed
spreeuw: it dumped out bmps
spreeuw: but maybe the decompiler did it on the fly
spreeuw: the map textures are huge btw
spreeuw: 8k x 4k
spreeuw: or somthing
spreeuw: kdeman: have you played complete annihilation?
mjt: hm. I just noticed that etracer does not work anymore in 2.6.32. It works in 2.6.31.
kdeman: no
spreeuw: its pretty cool, you can build force field shields
spreeuw: they are rendered as transparent bubbles
spreeuw: and you can later the terrain more
spreeuw: alter
spreeuw: like in simcity
spreeuw: and the nukes make deeper craters than in XTA
spreeuw: oops I meant RTS where I said fps
kdeman: mts: I think one of the devs said to pull git for DRM-2.6 (2.6.32) and then pull in another branch afterwards.
kdeman: mjt: git clone git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git; cd drm-2.6; git checkout --track -b drm-radeon-testing
kdeman: (<-- still sleepy)
kdeman: ;)
spreeuw: yep that works good here
spreeuw: here are the steps http://phoronix.com/forums/showthread.php?t=20484
kdeman: spreeuw: Somebody was testing rendering with TF2 which I think had a related issue with rendering textures (I think it was a GLSL issue as well).
kdeman: Some old programs just do basic checks and if they don;t like what they see - they'll either crash or give you bad results..even though the backend API or hardware can really do the work (so, it is more the apps fault than the underlying graphics framework).
mjt: kdeman: are you suggesting to do that that to every user of 2.6.32?
kdeman: Recommend for Radeon graphic users.
kdeman: (or FireGL/FirePro ones as well) ;)
mjt: i don't remember for sure but i think that 2.6.31 was the first kernel which worked out of the box without any drm/radeon patches
mjt: for me anyway
kdeman: Suggested by core developer(s)... as of yesterday.
mjt: had to patch 2.6.30, but 2.6.31 worked ok. Now in 2.6.32, it does not work again...
mjt: and for fun, there's no way to "not touch framebuffer" when enabling kms
mjt: or else there's no console
mjt: (as outlined on that thread)
kdeman: You should only update it if there is something not working for you that you want (well, NEED). You'll want to run the cycle of updating kernel/DRM->mesa->drivers in that fashion of sequencing..
mjt: i've been there several times already :(
mjt: ;)
kdeman: heh
mjt: OpenGL renderer string: Mesa DRI R600 (RS780 9610) 20090101 x86/MMX+/3DNow!+/SSE2 TCL DRI2
mjt: OpenGL version string: 1.4 Mesa 7.6.1-devel
kdeman: Hmm....
mjt: and fbcon is sloooooooow.
adamk_: So what still needs to be done to get the blur plugin working properly with the open source drivers?
kdeman: Something is not right: ' 1.4 Mesa 7.6.1-devel'
adamk_: s/blur plugin/compiz blur plugin/
kdeman: mjt: recheck your DRM/kernel and then refresh mesa 7.6.1-final against it.
kdeman: Get either ' 2.x Mesa 7.6.1-devel' in software rendering to work.
kdeman: or DRI.
mjt: kdeman: sure I will. In next spare time window...
kdeman: heh..ok
mjt: not that it's a great priority - i used etracer just as an example 3d app...
kdeman: mjt: just prevents great pain if you use Kwin or compiz type apps.
mjt: i don't
mjt: i needed 2 things: xv and multiple X sessions (on different vts). xv works for long time, but multiple X started working with kms.
mjt: (i'm with radeon[hd] driver since about oct-2008, with variable success)
kdeman: ahh...I think the move was to radeon recently. The code is more updated now.
kdeman: (well HDMI limiting maybe)
kdeman: Try the recent radeon driver snapshot - if not already tried.
kdeman: They'll probably unify the drivers (somewhat) in the future.
mjt: xv appeared in radeonhd first, at around dec-2008. I switched to radeon at about Mar-2009.
kdeman: you have the recent snapshot?
mjt: no. I updated last time for 2.6.30.
mjt: since that only 2 kernel updates without any patches
mjt: everything (incl. 3d) worked in 2.6.30+patches and 2.6.31.
kdeman: so you really just need a nice refresh to the latest kernel >2.6.32 snapshot.
kdeman: like 2.6.33-rc1-git3
mjt: actually i need to refresh all 3 (or 4) components. What I have now is some insane mix of debian packages with some .so files replaced.
mjt: so 2.6.32 is really broken? :)
mjt: in that case -stable is the way to go...
mjt: 2.6.32.3 or so
kdeman: something like that... I'd probably look at 2.6.33-rc1
kdeman: plus the drm-radeon patches.
mjt: well, 2.6.33-rc does not look like a 2.6.31->2.6.32 regression fix.
kdeman: This was the convertstaion a few days ago circling around the two snapshots.... 2.6.32.2 is on the table but then it was mentioned a new release coming by end of week with all the tree trimmings added to it that you may be looking for.
kdeman: I'd think it would end up in 2.6.33-final.
kdeman: --> /convertstaion/conversation/
kdeman: But that is futurespeak...you may want to look at your mesa build in the meanwhile on why it is not detecting >=OpenGL 1.5 with your hardware.
kdeman: OpenGL renderer string: Mesa DRI R600 (RV770 9442) 20090101 TCL OpenGL version string: 2.0 Mesa 7.7-devel OpenGL shading language version string: 1.10
_KAMI_: Heelo!
_KAMI_: Hello!
_KAMI_: How can I setup the hdmi output of my mobility radon 3470 with radeon driver?
kdeman: Talk to the Radeon HDMI developer(s)...but you also use this as a guide: http://www.mythtv.org/wiki/ATI_Radeon_HDMI
kdeman: ;)
kdeman: HDMI audio setup is another thing to setup with ALSA/SDL.
[Enrico]: kdeman: that guide has some totally wrong suggestions
[Enrico]: [and it uses fglrx]
kdeman: The last time I checked was the radeonhd driver did HDMI and I'd reviewed the fglrx for a working example: http://game.amd.com/us-en/drivers_catalyst.aspx?p=xp/hdmiaudio-xp
kdeman: Enrico: I know..but as mentioned..just a working example to reference for some newbies.
[Enrico]: kdeman: ok but be aware that way of installing fglrx will break the system :D
kdeman: :D
kdeman: Basically setup: " Driver "radeonhd"
kdeman: Option "Audio" "on"
kdeman: Option "HDMI" "all""
kdeman: in xorg.conf
kdeman: and use radeonhd >=1.3.0
agd5f: mortal_: there are still other issues. even with that patch, you'll just hit another problem. the real fix is much bigger
Zajec: agd5f: your patch was commited :)
Zajec: agd5f: i checked my GPU and from AtomBIOS tables it seems it doesn't allow reducing pci lanes
Zajec: agd5f: but reducing with xf86-video-ati did work
Zajec: agd5f: then nof sure... maybe we could ignore table in case of DPMS_OFF and should just reduce to 1?
agd5f: committed to what?
Zajec: agd5f: git
Zajec: agd5f: radeon.h patch
Zajec: and my prevent... patch
Zajec: http://git.kernel.org/?p=linux/kernel/git/airlied/drm-2.6.git;a=summary
agd5f: oh the atombios.h
Zajec: sorry. atombios.h not radeon.h
Zajec: yeah
kdekorte: Morning all... is there any known problems with the drm_radeon_testing kernel? I keep getting crashes like this: http://pastebin.com/m5a9eacb0
agd5f: Zajec: feel free to play with it. It seems to generally be problematic for a lot of r6xx/r7xx cards. I'd like to get engine clock stuff solid first.
agd5f: kdekorte: file a bug
Zajec: agd5f: did you update your patch maybe (increasing issue)?
mortal_: agd5f: what is the real fix? is it coming anytime soon?
Zajec: can do that myself, but maybe you've that fixed already
agd5f: Zajec: just about to
kdekorte: agd5f, ok, just wondering since it was fine yesterday
Zajec: ok, will let you ;) waiting then :)
Zajec: agd5f: could you check if my private patch provides what you have meant?
agd5f: Zajec: will do
agd5f: mortal_: textures have to be emitted pair-wise on r2xx IIRC even is you only use 1, IIRC. the current command checker doesn't deal with that yet
kdekorte: agd5f, what does I file it under Mesa?
agd5f: kdekorte: DRI
kdekorte: agd5f, I'll try and bisect it and find the problem patch
agd5f: kdekorte: excellent! thanks
MrCooper: agd5f: FWIW the KMS legacy LVDS backlight patch I posted recently had some detection logic for inverted backlight levels, mostly copied from radeonfb
agd5f: MrCooper: cool
agd5f: Zajec: ok. updated: http://www.botchco.com/alex/xorg/pm/
Zajec: agd5f: first patch is already in-tree...
agd5f: Zajec: not in drm-radeon-testing
agd5f: it's a cherry-pick anyway
Zajec: uh, little messed trees ;)
twnqx: christmas trees?
Zajec: ;)
kdekorte: agd5f, I suck at bisecting.. I can never seem to find a good point... I know I had a working kernel last week and I thought I found the place in git that did that... but even going back to that I still get a dead kernel
evil_core: where is kms code stored, in drm dir?
Zajec: drivers/gpu/drm/radeon
Zajec: if you mean radeon's kms
Zajec: KMS infrastructure is just drivers/gpu/drm
Zajec: see drivers/gpu/drm/radeon/Makefile for info about files (you will see these used by KMS)
kdekorte: agd5f, any tips on doing a bisect?
agd5f: kdekorte: set good to the last working commit
kdekorte: agd5f, that is my problem, I think I have a "good" commit but for some reason I can't make it work even at that point in time I think is good
agd5f: kdekorte: maybe some local build issue?
kdekorte: is there a way you can say, roll back the changes to how the git tree was a "3 days ago", I know you can do that in CVS, but don't see how to do that in git cause changes come in with different date stamps
kdekorte: agd5f, I don't think it is a local build issue.. did a make clean on the tree
agd5f: kdekorte: you can if the branch wasn't force updated
kdekorte: agd5f, what is the syntax for that? and does it work when you use git pull --rebase?
agd5f: kdekorte: drm-radeon-testing was force updated
evil_core: Zajec: thx, its enough, I wanted to make sure if diffing gpu and includ/drm was enogh
evil_core: enough*
agd5f: kdekorte: git reset
agd5f: git reset --hard to remove uncommitted changes
kdekorte: ok, I'll poke around with it more, right now git bisect is telling me d6a025d2dcc9ce2300ec36e18ca65ef0e20f6dcd is bad, so I'm going to reset the tree and then revert that one and see if that does it
evil_core: how to build gallium for r500?
evil_core: should I build specifically softpipe, llvmpipe, etc?
kdekorte: evil_core, ./autogen.sh --prefix=/usr --libdir=/usr/lib --with-dri-drivers="" --enable-gallium-radeon --disable-gallium-int
kdekorte: el --with-state-trackers=dri,glx
evil_core: kdekorte: but I want GLSL
kdekorte: unsure about that
evil_core: should I use KMS or non-KMS with drm-radeon-testing?
kdekorte: evil_core, I would use KMS, but right now I can't get it to work on my machine
DanaG: http://www.phoronix.com/scan.php?page=news_item&px=NzgzMA
DanaG: hmm... doesn't mention KMS.
evil_core: kdekorte: r500?
kdekorte: evil_core, no r600
[Enrico]: you need kms to use GLSL afaik
evil_core: yeah, but if gallium will be working like up to now, then I dont need it ;)
Zajec: agd5f: could you check my patch?
Zajec: agd5f: flags patch
agd5f: Zajec: yeah, looks good
Zajec: fine :)
adamk_: FYI, GLSL shows as supported by glxinfo even with KMS disabled, but I have no idea how functional it really is.
taiu: piglit at least does't do anythin badly when you enable GL2_1 and glsl120
taiu: some pbo functions seem to work somehow
taiu: 2_0 and 2_1 together http://people.freedesktop.org/~andrem/piglit/r600-quick-21-12-2009/
evil_core: Zajec: is pm ready?
mikkoc_: i just got a lockup with 2.6.33-rc1 kms enabled
mikkoc_: 2.6.33-rc1 with extra patches from drm-radeon-testing
Zajec: evil_core: of course, not :)
GoGi: I am using svn radeon driver, drm, mesa etc. kernel 2.6.32 with kms and glxgears has clearly the wrong colors
mikkoc_: http://pastebin.com/d2d1b71b1 the first part was repeated many times, I had to cut it
mikkoc_: WARNING: at lib/kref.c:43 kref_get+0x20/0x30() that is
mikkoc_: i was doing nothing at the time of the lockup
GoGi: every other opengl application is also blue
MrCooper: GoGi: what's an 'svn radeon driver', and what kind of machine/card is that?
GoGi: git
GoGi: my package manager is talking about svn
GoGi: 02:00.0 VGA compatible controller: ATI Technologies Inc RV710 [Radeon HD 4350]
ossman: onestone, the bicubic filter used for compiz and radeon xv, have you looked at if it is possible to do something similiar with a catmull rom based function?
onestone: ossman: no
ossman: (or anything else with negative lobes)
ossman: the current function can have a bit too much blurring effect
ossman: but changing it wasn't obvious :/
GNUtoo: hi, I've hands on a r700 with ubuntu karmik...if I want 3d should I recompile all the usual stuff such as xorg,xf86-video-ati,libdrm,kernel module etc...
GNUtoo: s/with/under
GNUtoo: because glxinfo says that there is direct rendering...but...the games are so slow...
GNUtoo: like if there was no 3d
ossman: GNUtoo, direct rendering can still mean software rendering these days
GNUtoo: ah ok thanks
ossman: GNUtoo, look for this line:
ossman: OpenGL renderer string: Software Rasterizer
GNUtoo: yes I've that
ossman: you want it to say something like R600
GNUtoo: I heard of that with the openmoko but I didn't recognize it
GNUtoo: ok is there a quick way to activate 3d?
GNUtoo: without recompiling by hand every trunk thing?
agd5f: GNUtoo: see if your distro provides packages
GNUtoo: which packages? xf86-video-ati?
ossman: GNUtoo, mesa mostly. and an updated libdrm and kernel
ossman: GNUtoo, which distribution are you running?
GNUtoo: karmic
GNUtoo: ubuntu
ossman: that's 9.10?
GNUtoo: yes
ossman: doesn't keep track of the fauna :)
GNUtoo: mmm xorg.log says something about r600_dri.so
GNUtoo: ok
GNUtoo: np
ossman: do you have /usr/lib/dri/r600_dri.so?
GNUtoo: no
ossman: see if "aptitude install /usr/lib/dri/r600_dri.so" does anything
ossman: if not, then I guess they haven't packaged it (or at least not put it in a default repo)
GNUtoo: ok
GNUtoo: thanks a lot I'll try to search what provides that
ossman: np
GNUtoo: I bet radeon_dri.so won't work
agd5f: GNUtoo: that's for r1xx hardware
GNUtoo: ah ok ouch
GNUtoo: thanks a lot so I'll leave it as-is
GNUtoo: and hope to succeed at convincing the person not to install the proprietary ati driver
GNUtoo: or look at what they call ppa
GNUtoo: ah there is one
_KAMI_: only radeonhd driver has hdmi output nor radon?
agd5f: _KAMI_: both support video. rhd supports audio and radeon requires kms for audio
_KAMI_: agd5f: thanks
_KAMI_: I was able to put pic to the hdmi output (on a tv)
_KAMI_: but the picture does not appeared on the tv and laptop lcd in same time
_KAMI_: only one has picture in same time
_KAMI_: I upgraded to 32 kernel is it enough to kms for audio
_KAMI_: agd5f: Can you please review my xorg.conf?
_KAMI_: agd5f: http://www.pastebin.ca/1725091
agd5f: _KAMI_: check the output of xrandr
agd5f: .32 isn't enough. you need airlied's drm tree for audio
ossman: agd5f, does that activate audio permanently? or is it connected to the handling of the display in some way?
ossman: e.g. will turning of the display completely disable the hdmi port so that you can't transfer audio either
ossman: off
agd5f: ossman: turns it on when hdmi monitor is connected. not sure what happens when the monitor blanks
ossman: k
ossman: it's desirable for the music to continue even if the power saving kicks in
ossman: I guess I'll just have to test it :)
spreeuw: nice quake3 looks good in HD
_KAMI_: agd5f: I put the hdmi cable
_KAMI_: and when I wun xrandr
_KAMI_: the lcd and the tv screen is blanked
_KAMI_: I restarted
_KAMI_: with hdmi cable pluged
_KAMI_: and got information
_KAMI_: http://www.pastebin.ca/1725101
_KAMI_: I have hd ready machine
_KAMI_: with resulution of 1360x768
_KAMI_: I See the hdmi output (Toshiba TV)
_KAMI_: disabled
_KAMI_: but when I enabled it
_KAMI_: I have only 3 option for the resulution
soreau: adamk_: Last I asked about blur I was told 'it depends on why it is not working'. afaict they're concentrating on getting piglit tests passed and hopefully it will automatically start working. or at least that's what I gathered
spreeuw: lol 130 fps demo four in 1280x960
spreeuw: max detail trilinear filtering picmip 0
spreeuw: vsyn costs 10%
spreeuw: vsync
ossman: agd5f, is float the only texture format that provides more precision than 32 bits/pixel?
agd5f: ossman: depends what hw
ossman: OpenGL has GL_RGBA16, but it seems to be mapped to plain ol' 8 bits/chan
agd5f: Not sure if GL support 128 or 64 bpp formats
ossman: I figured RGBA16 was 64 bpp. the docs weren't exactly crystal clear
ossman: I need it for a filtering lookup texture. floating point texture support seems to be a bit shaky in mesa
adamk_: soreau, Thanks :-)
rehabdoll: dmesg of my 2.6.32.2 modeset-boot/crash: http://obra.se/2.6.32.2
rehabdoll: 2.6.32.1 works just fine, as does 2.6.33-rc1
rehabdoll: card is 4770
spreeuw: thats not the latest code
spreeuw: try this kernel first
spreeuw: git clone git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git
spreeuw: git checkout -b drm-radeon-testing origin/drm-radeon-testing
lordheavy: drm-linus branch is now merged in the linus kernel (future 2.6.33rc2)
luc_: hello
luc_: how can i check if i have software or hardware acc on ddx driver?
chithead: for exa or xaa, see Xorg.0.log and for xv see xvinfo output
luc_: if i have this?
luc_: OpenGL renderer string: Software Rasterizer
luc_: is software?
chithead: opengl is not handled by the ddx
luc_: ok
adamk_: luc_, 'xvinfo | grep -i textured' should show a textured video adaptor if 2D acceleration is enabled.
luc_: i have all from git
luc_: except kernel with drm-next
luc_: i have vanilla kernel 2.3.32 final
luc_: and i have no hardware acc on opengl
luc_: is because of drm module from kernel?
adamk: Correct, you do not. But you specifically asked about hardware acceleration from the DDX, which is the 2D driver.
adamk: Let's start from the top the. Please pastebin your /var/log/Xorg.0.log file.
luc_: if i use curent git from drm-testing i will have opengl?
adamk: Since we don't know why you don't have it now, we can't tell you if you will have it then.
luc_: ok
adamk: But if you pastebin that log file, we might be able to tell you what's gone wrong.
luc_: ok here it is
luc_: http://pastebin.com/d4bdad6a6
adamk: Hmmm... There's something wrong with your r600_dri.so .
adamk: Can you pastebin the *full* output of 'LIBGL_DEBUG=verbose glxinfo' ?
adamk_: If I had to guess right now, I'd say that your libdrm and Mesa are out-of-sync with each other.
luc_: i will tray
luc_: ok here
luc_: http://pastebin.com/d2762423f
adamk: Yeah, I'm pretty sure you need to update libdrm from git, and then rebuild Mesa from git. This has nothing to do with the kernel.
luc_: ok
adamk: I had this problem yesterday (or the day before) and solved it by updating both.
luc_: ok
adamk: There was an update to one in git that broke things till the other one was updated.
luc_: doing now
adamk: And, on that note, I'm out of here.
luc_: thank you
adamk: No problem.
luc_: working now
luc_: :)
luc_: mesa rebuild
luc_: beafore this i updated the lib drm
luc_: and mesa but beafore libdrm
luc_: :)
spreeuw: is anybody working on hw accelerated h264 decoding? if so what api has focus for ati
spreeuw: gallium?
Zajec3: AFAIK no
spreeuw: vaapi looks like the coolest
Zajec3: (no one)
spreeuw: it can leech on top of all others
spreeuw: by the time its completed it will be less necessary anyway
spreeuw: only for low power devices its handy
spreeuw: it would open the gate for real free mediacentres
spreeuw: on low electricity
spreeuw: did you see the WD Live player
spreeuw: awesome box
spreeuw: ethernet/HD decoding
luc_: working now
airlied: Zajec3: I only merged the atombios header for completeness I'd already merged the mutex for atom
Zajec3: airlied: yup, saw that later :)
Zajec: airlied: that + pull request
Zajec: airlied: for some reason drm-radeon-testing is outdated
Zajec: doesn't include all that will go to Linus
agd5f: Zajec: new patches: http://www.botchco.com/alex/xorg/pm/
agd5f: I haven't had a chance to test yet
dileX: agd5f: seems like the patchser is not applicable on top of latest drm-2.6 GIT branches
dileX: patchset*
dileX: (anyway, shopping)
agd5f: dileX: they are on top of drm-radeon-testing
bjaglin: Zajec: i am experiencing random lockups with the pm patch V8 here on my RV250, not the same as the bug 25718 AFAIK. Is there some relevant log/debug flags I can set for next time it happens?
Zajec: + switch (type) {
Zajec: + case POWER_STATE_TYPE_DEFAULT:
Zajec: + default:
Zajec: agd5f: is that on purpose?
agd5f: Zajec: ?
Zajec: agd5f: sec
Zajec: agd5f: ah, you used return, ok
Zajec: no... moment
Zajec: agd5f: track code flow for calling radeon_pick_power_state(POWER_STATE_TYPE_POWERSAVE)
Zajec: agd5f: it will go into default:
agd5f: Zajec: ?
Zajec: huh, am I sleepy? ;)
Zajec: default in switch eats all values, doesn't it?
agd5f: default is for non-matching items
Zajec: ups, for my whole life I got wrong understanding of default...
Zajec: i though switch takes first matching and default matches all :|
evil_core: hardhang while playing openarena on gallium
evil_core: ut2k4 stopped working at all, and neverball doesnt work with gallium
Zajec: agd5f: patches looks fine
Zajec: *look
agd5f: Zajec: cool
Zajec: that's what I wanted to implement :)
Zajec: agd5f: not sure what we should be when there are for example two states for battery
agd5f: Zajec: yeah, we fine tune thigns as we go
agd5f: for now using the clocks from the tabels should be safe as they are known good combinations
agd5f: *tables even
Zajec: how can I download all from http://www.botchco.com/alex/xorg/pm/ with wget?
agd5f: maybe wget http://www.botchco.com/alex/xorg/pm/*.patch
jcristau: or wget -r -np http://www.botchco.com/alex/xorg/pm/
dileX: agd5f: thx for info
dileX: hmm, drm-linus in upstream GIT, now
dileX: (cries for reverts)
airlied: reverts of what?
dileX: IIRC seems only to exclude 0001-drm-radeon-kms-add-definitions-for-v4-power-tables.patch. I'll try
airlied: dileX: that patch only just adds the defines, no reason to revert it
dileX: airlied: hehe. no I wont revert, I exclude 0001* from agd5f pm patchset
Zajec: agd5f: great
Zajec: agd5f: it took nicely lowest mode :)
Zajec: agd5f: it just spams a little with Requested/Setting
Zajec: agd5f: i'd make it debug probably
dileX: yeah :-)
agd5f: Zajec: yeah, that's the plan
tesuki: Does this driver work equaly well on FreeBSD as linux and using the same codebase (>90% of the code)?
dileX: agd5f: did your pm-related fixes went into radeon/avivotool?
dileX: s/went/go
dileX: OK, they did. thx
adamk_: tesuki, For the most part, yes. The only major differences are in the kernel DRM, and the fact that the linux DRM now supports KMS/GEM.
Zajec: agd5f: now when using lower engine clock, reclocking happens more often
Zajec: agd5f: because engine works slower and even changing tabs in Opera can make if upclock for a moment
Zajec: agd5f: that's absolutely understandandable and normal
Zajec: agd5f: but less fine thing is I notice corruptions now when reclocking :/
Zajec: now always
Zajec: but sometimes
tesuki: adamk, Thank you.
Zajec: do we have any docs containing any reclocking info?
_KAMI_1: Hi!
Zajec: hey
_KAMI_1: Hali Zajec
_KAMI_1: agd5f: Hello!
_KAMI_1: So I still have HDMI related question
_KAMI_1: I can get picture on TV with HDMI but this is rare case
_KAMI_1: when there is picture in TV no picture in Laptop LCD
Zajec: define "rare "case"
Zajec: "rare case"
_KAMI_1: So when I start laptop
_KAMI_1: HDMI connected
_KAMI_1: I see
_KAMI_1: LCD and TV (HDMI) in the display manager
_KAMI_1: when I enable HDMI
_KAMI_1: the screen blanks
mikkoc: can someone take a look at this? http://pastebin.com/m43c4e845
_KAMI_1: on LCD and HDMI too
mikkoc: 2.6.33-rc1 + patches from drm-radeon-testing
_KAMI_1: when I restart laptop only HDMI show picture during login
dileX_: agd5f: dmesg_radeon-pm.txt:
_KAMI_1: then blanks all of two display
_KAMI_1: so I cant enable LCD and HDMI in same time
_KAMI_1: and when HDMI connected no way to predict what display will be used
dileX_: (last one enforced via 'xset dpms force off')
_KAMI_1: Zajec: do you have idea?
_KAMI_1: or agd5f?
_KAMI_1: http://www.pastebin.ca/1725101
_KAMI_1: xrandr output
_KAMI_1: another question: IRQ's not enabled, falling back to busy waits: 2 0 what is this error?
chithead: this is not an error
chithead: irq support was only recently added to the drm
dileX_: Zajec: when did you get screen corruption? what caused it?
_KAMI_1: chithead: so only need new drm?
_KAMI_1: or what is the way?
AStorm: hello
AStorm: did anyone else report mesa from git segfaults recently?
AStorm: esp. vs 2.6.33-rc drm
_KAMI_1: chithead: What is the state of HDMI for 3470 Mobility?
Zajec: dileX_: your dmesg looks fine
Zajec: dileX_: when I switch tabs in Opera fast, my engine goes upclock
Zajec: dileX_: and shortly after that downclock
Zajec: dileX_: when switching tabs I see corruptions
AStorm: Zajec: your card is probably being overclocked by accident
AStorm: thus the corruption
dileX_: Zajec: unreadable ones or wrong colors?
Zajec: _KAMI_1: i own 34xx mobility and it work great
AStorm: bad form
Zajec: dileX_: for a 1 frame i see about 5 black horizontal lines
AStorm: _KAMI_1: it should work now even with KMS if you use drm-radeon-testing
Zajec: dileX_: after that everything is fine
_KAMI_1: Zajec: Can you share your xorg.conf with me?
_KAMI_1: or any idea?
Zajec: _KAMI_1: i don't own
Zajec: xorg.conf
Zajec: _KAMI_1: do you use KMS ?
_KAMI_1: [KMS] drm report modesetting isn't supported.
Zajec: _KAMI_1: please pastebin your /var/log/Xorg.0.log
Zajec: maybe something interesting in it...
_KAMI_1: ok
_KAMI_1: here it is:
_KAMI_1: http://www.pastebin.ca/1725392
_KAMI_1: thank you in advance
Zajec: AStorm: i don't think it's overclocking
Zajec: _KAMI_1: it looks fine for me :|
Zajec: _KAMI_1: you can create bug report
Zajec: _KAMI_1: Alex could have some idea (after xmas probably)
_KAMI_1: Okay
_KAMI_1: I saw
_KAMI_1: there is never version in xorg edgers ppa
_KAMI_1: I try it the report
AStorm: _KAMI_1: did you boot the kernel with radeon.modeset=1?
AStorm: you'll need this
_KAMI_1: no I did not
AStorm: that is required for KMS. either that, or flipping the STAGING option
_KAMI_1: so I have to add for kernel parameter radeon.modest=1 ?
AStorm: yes
_KAMI_1: I try it
_KAMI_1: and reboot
AStorm: modeset
AStorm: not modest ;>
_KAMI_1: okay?
AStorm: yup
_KAMI_1: with HDMI connected
_KAMI_1: during boot?
AStorm: not sure
AStorm: try both
AStorm: i'll test hdmi w/ r6xx later
AStorm: from drm-radeon-testing
Zajec: _KAMI_1: this way (modeset=1) you will test KMS
Zajec: _KAMI_1: you need updated libdrm and updated radeon with KMS build
Zajec: _KAMI_1: HDMI should work even without KMS!
AStorm: yup, it should
AStorm: 3470 is r500 or r600?
_KAMI_1: ok I updated to the latest version of radeon driver
_KAMI_1: (091222)
bridgman_: 3470 is rv620
_KAMI_1: OpenGL renderer string: Mesa DRI R600 (RV620 95C4) 20090101 x86/MMX/SSE2 TCL
AStorm: oh
AStorm: I see KMS HDMI has just been added in drm-radeon-testing
AStorm: so no point in testing that w/o the branch pulled into the kernel
AStorm: first, I need that mesa problem sorted out
_KAMI_1: So what is my task?
AStorm: _KAMI_1: which version of xf86-video-ati is it?
AStorm: it looks like your HDMI hasn't been detected properly
_KAMI_1: xserver-xorg-video-radeon 1:6.12.99+git20091222.9d0f3af7-0ubuntu0sarvat
_KAMI_1: I am using this one
_KAMI_1: others
_KAMI_1: xserver-xorg-video-radeonhd 1.3.0+git20091208.edeb1e7a-0ubuntu0tormod~kar
AStorm: looks fresh
AStorm: hmm
AStorm: might be a bug or just that missing r6xx HDMI support
_KAMI_1: server-xorg-video-ati 1:6.12.99+git20091222.9d0f3af7-0ubuntu0sarvat
_KAMI_1: should I try modeset=1?
AStorm: nah
_KAMI_1: AStorm: http://www.pastebin.ca/1725101
_KAMI_1: this is how detected
AStorm: oh right, which version of the kernel are you running
_KAMI_1: 2.6.32
AStorm: _KAMI_1: so it works
AStorm: try xrandr --output HDMI-0 --auto
_KAMI_1: Linux AgentM 2.6.32-02063202-generic #02063202 SMP Sat Dec 19 11:00:49 UTC 2009 i686 GNU/Linux
AStorm: that should enable the screen
AStorm: there are other xrandr options to try as well :)
_KAMI_1: it enabled the HDMI output
_KAMI_1: picture appeard in the tv
_KAMI_1: but laptop display is blanked
_KAMI_1: so I tipe it as blind
_KAMI_1: the resulotion of tv is not so high
cutchyacokov: Hello everyone! I just got KMS and dri2 working with libdrm, mesa and radeon driver from git as well as drm modules built in to the kernel 2.6.32. Everything is working fine on one monitor (xrandr) but the other is tearing and looks like it is still using dri1. Any ideas?
AStorm: _KAMI_1: it shouldn't blank the laptop
AStorm: try xrandr --output LVDS-0 --primary
bridgman_: cutchyacokov; tearing during video playback ?
cutchyacokov: bridgman_: Yes video playback and normal desktop compositing
cutchyacokov: bridgman_: Specifically DVI-0 works perfectly but it seems like DVI-1 isn't using KMS at all
bridgman_: 2.6.32 wouldn't have the new sync-to-vblank logic so you're probably just getting the initial tear-free code in the X driver
bridgman_: there were some limitations there in dual-head, don't remember the details
cutchyacokov: Ah, ok. Will this be fixed in 2.6.33?
Zajec: bridgman_: do you hardware details like reclocking?
Zajec: bridgman_: any idea why relocking can cause 1 corrupted frame?
cutchyacokov: It also isn't just the vsync switching between Framebuffer console and desktop is instant on DVI-0, as expected but slow on DVI-1 as if KMS wasn't enabled
bridgman_: making reclocking reliable is why our power management people have grey hair and smoke a lot
bridgman_: it's probably the nastiest part of the driver code, different tricks for every GPU
Zajec: :P
Zajec: bridgman_: can you make them share some tricks? :)
bridgman_: yeah, the problem is that they have to be dug out of the source code, which is a long & painful process
_KAMI_: reee
_KAMI_: updated driver
Zajec: heh ;)
_KAMI_: and added radeon.modeset=1 to kernel boot
_KAMI_: did you see my messages that I wrote
AStorm: yes
_KAMI_: without enabled lcd screen?
AStorm: _KAMI_1: it shouldn't blank the laptop
AStorm: try xrandr --output LVDS-0 --primary
AStorm: my replies to them
AStorm: now, KMS + HDMI is only done in drm-linus (maybe) and drm-radeon-testing (certainly)
Zajec: AStorm: i don't think you understand HDMI ;)
AStorm: uhm, on r6xx that is
Zajec: AStorm: i believe _KAMI_ just wants picture by HDMI
AStorm: I've read the commit messages
AStorm: Zajec: he got one
Zajec: AStorm: latest HDMI patches gives sound (audio) over HDMI
AStorm: but his LVDS got blanked
AStorm: do they? ;p
Zajec: standard (picture) HDMI is in KMS since first days :)
AStorm: http://git.kernel.org/?p=linux/kernel/git/airlied/drm-2.6.git;a=commit;h=dafc3bd515d6079406986f7748b48adb3c7170f2
AStorm: are you sure? ;>
AStorm: the patch suggests otherwise
Zajec: yeah
Zajec: this patch is for audio
AStorm: mmmh
AStorm: now I have audio too? how nice
Zajec: :P
AStorm: btw, why did it blank the LVDS for him? same encoder?
AStorm: quite a bit of manufacturer's fail there
Zajec: i don't think so... _KAMI_ attached Xorg.0.log and it shown one encoder for LVDS and one for HDMI
AStorm: hmm
Zajec: which was totally fine for me...
AStorm: so why did xrandr --output HDMI-0 --auto blank the LVDS ;>
AStorm: it shouldn't have
Zajec: have no idea :/
AStorm: _KAMI_: next time, try xrandr --output HDMI-0 --right-of LVDS-0
AStorm: just to be sure
AStorm: (or --left-of)
AStorm: _KAMI_1: did you get the xrandr tip?
_KAMI_1: AStorm: yes I got it
_KAMI_1: But I have only sw opengl this time
_KAMI_1: (checking the log)
rnoland: airlied: agd5f : is the pci gart mapped into kva?
_KAMI_1: When I blanked the LCD with enabling HDMI
_KAMI_1: the randr --output LVDS-0 --primary did not helped :(
Zajec: _KAMI_1: did you try enable HDMI with KMS (with KMS == with modeset=1)
_KAMI_1: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch. -- in xorg.log
_KAMI_1: what is this-?
rnoland: airlied: agd5f : nm, i see that it is....
_KAMI_1: Zajec: I put radeon.modeset=1 to kernel boot parameters
_KAMI_1: (II) [KMS] drm report modesetting isn't supported.
Zajec: _KAMI_1: you don't have correct libdrm
_KAMI_1: dri] This chipset requires a kernel module version of 1.17.0,
_KAMI_1: [dri] but the kernel reports a version of 2.0.0.[dri] If using legacy modesetting, upgrade your kernel.
_KAMI_1: [dri] If using kernel modesetting, make sure your module is
_KAMI_1: [dri] loaded prior to starting X, and that this driver was built
_KAMI_1: [dri] with support for KMS.
_KAMI_1: [dri] Disabling DRI.
_KAMI_1: what I should do?
AStorm: _KAMI_1: it means your libdrm is outdated, as Zajec mentioned
_KAMI_1: Okay
_KAMI_1: how to update it?
AStorm: it's ubuntu? hmmh
_KAMI_1: I used all module from xorg edgers ppa
AStorm: no idea really
_KAMI_1: yes it is
Zajec: _KAMI_1: checkout, compile with experimental radeon api, and install
Zajec: _KAMI_1: or just don't use KMS ;)
AStorm: ok, which is the best subpixel rendering patch I can get for freetype?
AStorm: and the setting
MostAwesomeDude: That message is my doing; it means that the DDX isn't built for kernel modesetting.
_KAMI_1: Zajec: It is not easy way
Zajec: and to build DDX with KMS you need to have libdrm with experimental radeon API :)
_KAMI_1: what I have to check out and build it
_KAMI_1: only libdrm?
Zajec: _KAMI_1: libdrm, and then (after installing drm) also xf86-video-ati
bjaglin: _KAMI_1, it might also be that X is loaded before radeon, I had this problem in Karmic. See https://wiki.ubuntu.com/X/RadeonKMS
AStorm: MostAwesomeDude: even better ;p
AStorm: _KAMI_1: it means your xf86-video-ati was built against old libdrm
_KAMI_1: bjaglin: so you did it without build?
_KAMI_1: I should try
bjaglin: _KAMI_1, yes i used the workaround suggested there in gdm.conf. Also, if you pick up the 2.6.32 from the lucid repositories, the timing is different, and no workaround is necessary. You should try that first if you are not confident with building
_KAMI_1: I have own builds from openoffice :)
_KAMI_1: but X.org is a sightly new area for me :)
_KAMI_1: bjaglin: I did the modification what you mention before
_KAMI_1: I retart an come back soon
_KAMI_1: ok?
bjaglin: good luck :)
_KAMI_1: gooooooooo
_KAMI_: hi
_KAMI_: bjaglin: No succes
bjaglin: _KAMI_, with the workaround in gdm.conf you mean?
_KAMI_: radeon.modeset=1 the not possible to boot however I fixed gdm.conf
bjaglin: did you try to enable kms after boot?
_KAMI_: bjaglin: I added the boot parameter
_KAMI_: or other?
bjaglin: if that does not work, then you really have a version mismatch, but i doubt it if you are using xorg-edgers
bjaglin: no, i mean booting without KMS, and then unloading radeon and reloading it with KMS as explained in the first paragraph there https://wiki.ubuntu.com/X/RadeonKMS
AStorm: _KAMI_: you must build the firmware into the kernel
_KAMI_: without kms I have DRI again
bjaglin: _KAMI_: https://wiki.ubuntu.com/X/RadeonKMS "You can also try out KMS without rebooting..."
_KAMI_1: reee
_KAMI_1: when i enable modeset I am not able boot
_KAMI_1: the sreen blanked (switched to black)
_KAMI_1: as soon as I command : sudo modprobe radeon modeset=1
chithead: possibly framebuffer console is not enabled in your kernel (or the module not loaded)
_KAMI_1: what is te name of module?
chithead: fbcon
AStorm: btw
AStorm: I had a similar problem with 2.6.32.2
AStorm: funny thing is that 2.6.32 clean worked fine
AStorm: 2.6.33-rc1 is fine as well
_KAMI_: reee
_KAMI_: I have no success with KMS :S
_KAMI_: DO I need it for HDMI?
_KAMI_: or what is your next idea?
_KAMI_: Did I miss something?
_KAMI_: from :38:53?
AStorm: _KAMI_: you don't *need* it for HDMI
_KAMI_: Then, whatif?
_KAMI_: any other idea?
AStorm: _KAMI_: it worked
AStorm: it just wasn't supposed to blank your LVDS
AStorm: maybe it's some HDCP stuff
agd5f: Zajec: you might try removing the DRM_INFOs in the reclockign code. might be adding too much latency so the reclock misses the blanking period
agd5f: might also be that the reclock is taking too long in general. maybe we schedule it against a vline irq to add some more time
Zajec: hmm, ok
agd5f: Zajec: might also be too big a step, might work better if we gradually step down the clocks
Zajec: it is not
agd5f: dileX_: looks good. thanks for testing
Zajec: tried 110MHz <-> 130MHz
Zajec: same thing
agd5f: ok
dmz001: I just created a new fedora 12 server. With all the latest updates. I'm using a radeon x1300 and am unable to get X to display on the DVI port.
dmz001: I ran X without an xorg.conf and tried some experiments (vodoo programing here) in trying to force the DVI port.
dmz001: I had some issues with the text mode display. But I got around that by passing nomodeset to the kernel during boot. That allowed the text to be visible.
dmz001: I tried running without a monitor attached to the VGA port, but that didn't help.
_KAMI_: AStorm: okay
_KAMI_: but what i have to do?
Hackus: Hi?
Hackus: Amupme pm
Hackus: Anyone on?
Hackus: :-)
Hackus: Hello.
Hackus: Anyone on Radeon?
Droste: hi
Hackus: Hello
Hackus: hello
BioTube: if you've got a question, ask it
Hackus: I do have a question!
BioTube: then ask it!
AStorm: _KAMI_: #radeon
AStorm: I don't have the time right now
_KAMI_: AStorm: :)
Hackus: I have a 3870x2, and I was wondering if I will at this time be able to get it to do DRI, as it would seem VGA Arb prevents it from doing so.
_KAMI_: Thank you for your help
Agent_Jacobs: _KAMI_ - Any luck yet?
_KAMI_: Agent_Jacobs: no luck with kms
_KAMI_: but in overall I am happy with opensource driver
Hackus: Works great with just ONE video card. :-)
_KAMI_: the video, compiz is much more reponsible than with fglrx
_KAMI_: thank you guys
Agent_Jacobs: Lemme review the chat log and I'll see if I have any suggestions for KMS.
Hackus: My BIOS won't allow me to turn a card off either.
Hackus: Worse, its a LAPTOP.
Agent_Jacobs: Had a real time myself, so...
Hackus: So, pulling a card is a major pain.
Hackus: So, was that an AWESOME question or what? :-)
BioTube: I don't have any experience with vgaarb
BioTube: but I remember reading it shouldn't be needed for cards with DRI support
Hackus: My Xorg log says it will not enable the DRI with a VGA Arb device.
Hackus: Is there a way to turn off a card in Xorg?
BioTube: I'm not sure, but either airlied or agd5f mentioned that KMS opts radeons out of vgaarbs
Hackus: What exactly does that mean? I turn on KMS?
Agent_Jacobs: Reviewed the chat log, KAMI. I'm unclear on one thing.
Agent_Jacobs: Did you do modprobe fbcon, then modprobe radeon modeset=1?
Agent_Jacobs: If so, was it the same problem.
Agent_Jacobs: er, that was a question, so...
Agent_Jacobs: should be problem?
Droste: hackus: try it :-)
Hackus: Ok, so let me get this straight....
_KAMI_: Hi Agent_Jacobs
Hackus: I do a modprobe fbcon; then put modeset=1 in my radeon.conf file in modprobe.d?
_KAMI_: thank you for your help
_KAMI_: so I did not try fbcon
Hackus: reboot?
BioTube: Hackus, no
BioTube: modeprobe fbcon; modprobe radeon modeset=1
Agent_Jacobs: Yup. For both Hackus and KAMI.
Hackus: Ok.
Hackus: Do I have to restart X?
BioTube: yes
Hackus: Should probably try this at Runlevel 3.
Agent_Jacobs: I tried it at console first.
Hackus: I will try that. Then I will be back after it doesn't work. :-( )
Agent_Jacobs: No X running.
Hackus: Merry Xmas!
Agent_Jacobs: 1 day short :)
Hackus: You can't unload the radeon module while X is using it can you?
Droste: no
BioTube: nope
Hackus: I thought so.
Agent_Jacobs: Just kill GDM/KDM/whatever + X.
Hackus: Alright. but if I do that I will lose my IRC session. :-)
Hackus: Going to logout and try. Thanks guys.
Agent_Jacobs: brb as soon as you can.
Agent_Jacobs: We will await news with eager anticipation.
_KAMI_: just a minute I have a dicsussion in the other channel
Droste: :-D
_KAMI_: Agent_Jacobs: :D
Droste: why isn't fbcon a dependency for radeon?
Droste: just out of curiosity :-D
Agent_Jacobs: Because it's not necessary, I presume.
Agent_Jacobs: Your shell is still good after modprobing radeon, you're just blind.
Agent_Jacobs: Try it sometime... modprobe radeon modeset=1, screen goes blank.
Agent_Jacobs: Then modprobe fbcon, screen comes back.
Droste: I know, but thats not really a preffered state :-D
Droste: preferred*
Agent_Jacobs: Yeah, I would personally recommend doing both, but...
Agent_Jacobs: Module deps is for what's required, not for what's really convenient.
_KAMI_1: reee
_KAMI_1: I modprobed fbcon
_KAMI_1: but I have same problem
_KAMI_1: Agent_Jacobs: I have problem with HDMI
_KAMI_1: when I enable it blanks the Laptop's LCD display
Droste: ok :-)
_KAMI_1: and not using native resulotion of TV (HDMI)
dmz001: Help with radeon x1300 DVI port stays black. Only VGA port is usable. Under Fedora 12 with all updates
Droste: _KAMI_: xorg.log?
Droste: dmz001: xorg.log? :-)
dmz001: new to this how do I get it to you?
Agent_Jacobs: pastebin works
dmz001: one moment....
Agent_Jacobs: KAMI, I'm assuming you have to reboot to solve the issue?
_KAMI_1: http://www.pastebin.ca/1725617
_KAMI_1: Agent_Jacobs: Rebot is not enough
_KAMI_1: sadly
_KAMI_1: when I rebot with HDMI connected
_KAMI_1: the login screen apperas on the tv
_KAMI_1: and the lcd (of laptop ) is blank
Hackus: WOW.
Droste: did you try to enable it with xrandr?
Hackus: The colors. :-)
_KAMI_1: Droste: yes I tried
_KAMI_1: but every time I enabled hdmi
_KAMI_1: my lcd ges blank
Hackus: Enabling radeon modeset=1 and starting up the X server results in a big time crash.
Agent_Jacobs: I like explosions!
BioTube: Hackus: sounds like lack of KMS supporrt
_KAMI_1: (I have a no way to switch back
dmz001: Droste: http://pastebin.ca/1725623
Hackus: I wonder if I should build a 2.6.32 kernel with KMS built in and try it.
_KAMI_1: and the tv resolution is lover than the monitor can display
_KAMI_1: (1360x768)
Hackus: What do you think?
BioTube: Hackus: I meant lack of ddx KMS support
Hackus: Ok, what is ddx?
BioTube: 2d driver
Hackus: Really?
BioTube: yes
Hackus: I thought all of that stuff was already done for the RV600 chip family.
_KAMI_1: Agent_Jacobs: http://www.pastebin.ca/1725101 detected res
Agent_Jacobs: Hackus, did fbcon take over after loding your modules?
BioTube: Hackus: the ddx can still be compiled without support for KMS
Agent_Jacobs: loading*
Hackus: Oh I see.
_KAMI_1: Agent_Jacobs: my conf file: http://www.pastebin.ca/1725091
Hackus: Your saying it has support, nut just in combination with this new fangled KMS stuff.
Hackus: Is that right?
BioTube: right
Hackus: So its a ..... GASP regression!!!
BioTube: if the ddx doesn't support KMS, it'll try to take control of the video card, resulting in corruption
BioTube: Hackus: it needs to be compiled against a libdrm that supports KMS
Droste: dmz001: did you try to boot with kms?
Hackus: heavens to mergatroid.....
dmz001: What is kms and how do I boot with it?
Hackus: yes, I did put modeset=1 in my grub startup line.
Hackus: I think that turns it on right?
BioTube: yes
Agent_Jacobs: The kernel part, anyway.
Droste: dmz001: boot with radeon.modeset=1
Hackus: Well, would getting a libdrm version be as simple as getting the source packages and setting a few autconf rules?
Agent_Jacobs: What distro, Hackus?
Hackus: Rebuilding my Fedora RPMs?
Hackus: Fedora 12
dmz001: Just add that exact text to the boot command line?
Droste: yes
Hackus: Or would you guys suggest I go and get the latest build someplace?
dmz001: Droste: can you tell me what that will be doing?
Hackus: (i.e. libdrm)
BioTube: Fedora 12 should have support
dmz001: Droste: One point. I had to add nomodeset to the boot line to allow the text mode to be visible on the dvi port.
Droste: oh ok
Hackus: really?!
Hackus: So, it has support that just doesn't work at the moment? :-)
Hackus: LOL
BioTube: I don't use fedora, but that's what I heard
_KAMI_1: Agent_Jacobs: any idea?
Hackus: Ok, so let me get this straight.
Hackus: Supposedly, Fedora 12 already has a ddx that has KMS support correct?
Droste: dmz001: are both outputs blank without nomodeset?
Agent_Jacobs: KAMI, HDMI doesn't work with Radeon or Radeon KMS?
dmz001: Droste: Only the DVI. VGA always seems to work.
Agent_Jacobs: Does LVDS (laptop screen) work with KMS?
dmz001: Droste: without nomodeset lose DVI text. with nomodeset only lose DVI under X
_KAMI_1: Agent_Jacobs: kms not working at all
_KAMI_1: and hdmi notworking well with radeon
dmz001: Droste: I attempted to see if not connecting monitor to VGA did anything. Didn't help
_KAMI_1: I can get picture on hdmi
Hackus: Well, if its any consolation, all of my Intel GMA machines work great with KMS. :-)
_KAMI_1: but my main display is blank when i enable hdmi
Agent_Jacobs: KAMI, you enable with Xrandr?
_KAMI_1: LVDS is not working with kms
_KAMI_1: yes
Agent_Jacobs: Ok, for KMS it kills when you modprobe radeon modeset=1?
Hackus: Big time crash for me anyway.
brro: I know this channel is for the free software driver but does anyone know offhand if the radeon boards can handle something known as edge overlap? It's when you use the display outputs for say a set of display projectors and you want some of the screens to have overlapped image data.
Hackus: LOL
_KAMI_1: Agent_Jacobs: KMS enabling the the screen switch black
Droste: dmz001: ok.... and it used to work with older version of your distro?
_KAMI_1: and no text consol no gfx at al
_KAMI_1: l
Hackus: Dude, at the moment two cards are a problem. :-)
Hackus: My laptop is new, so it has only had the pleasure of using fedora 12.
dmz001: Don't know. This is a "new machine". Redirected a windows box to a linux server.
Agent_Jacobs: KAMI, assuming this is through LVDS, it sounds like KMS is loading and either dying or fbcon isn't there.
Droste: maybe try a live cd of an older version, to check if it's a problem of the new drivers
brro: hmm, racer.nl looks interesting. it talks about some of the edge blending I was curious about.
_KAMI_1: Agent_Jacobs: what is your advice?
dmz001: Droste: What's the risk of trying the modeset=1?
Hackus: Explosions. :-)
Droste: no :-) maybe blackscreen until x loads
BioTube: I definitely recommend loading fbcon before x for optimal results
Agent_Jacobs: KAMI, I don't want to ask you to repeat anything.
dmz001: Droste: Hmm, I'm not running level 5. I have inittab at 3
_KAMI_1: Agent_Jacobs: I can do it
Agent_Jacobs: I also don't want to troubleshoot two issues at once, so I'll have a look at what you've done with xrandr right now.
_KAMI_1: is it possible to collect log for this problem?
dmz001: Droste: I'll try out some of my other live CDs.
dmz001: Thanks....
Droste: you could try it this way: kill x, unload radeon, modprobe fbcon, modprobe radeon modeset=1, start x
Agent_Jacobs: Possibly. Lemme look at the chat log. I'll see if I have any ideas wrt xrandr.
Hackus: I hope you have better luck than I did when I manually loaded radeon like that amd started X.
Hackus: :-)
Hackus: Not sure what to do next, any suggestions?
Droste: hackus: with newer libdrm this will work for you too :-)
Hackus: Ok, so where do I get this newer libdrm?
dmz001: Droste: What if the screen stays black. Other than reboot what's the (blind) reversal?
Hackus: Oh I hate it when that happens.
Droste: don't really know... I think reboot would be the best thing to do :-)
Droste: hackus: fedora?
Hackus: I am not sure what you mean Droste, I have all of the latest posted updates.
Hackus: libdrm is what package do you know?
Hackus: Or can I get it someplace buiild from source code.
Hackus: ?
Droste: yes you can :-)
dmz001: Thanks.
Hackus: Well, the latest package I have is 2.4.15
Hackus: libdrm I mean.
Agent_Jacobs: Btw, Hackus, when you loaded fbcon, then radeon with modeset=1, did you have console?
Hackus: yes I did have the console.
Hackus: It only blew up X, I didn't lose control of the console or anything.
Agent_Jacobs: Ok, so the kernel stuff is there. Sounds like you need a new X. Either source or updated RPMs.
Droste: I'm using opensuse... maybe someone else knows where to get fedora packages
Agent_Jacobs: Well, not X, just the radeon driver.
Droste: and libdrm
Agent_Jacobs: That too.
Droste: hackus: http://dri.freedesktop.org/wiki/Building <-- wiki for building them from git
Agent_Jacobs: At that point, if you have picture you can check glxinfo.
Hackus: Oh, OK.
Hackus: So, I have to rebuild the KERNEL modules and the X module, correct?
Hackus: for radeon?
Droste: th kernel module should be ok
Hackus: Ok, so just the X module?
Droste: just the library for x to access it (libdrm) and the radeon userspace driver
Agent_Jacobs: Yeah, the X driver and associated stuff.
Hackus: Ok. I will try that. Hopefully it will build OK.
Hackus: otherwise I think I will just wait.
Hackus: And of course complain a lot till its fixed. LOL
Droste: :-)
Hackus: Thanks guys!
Agent_Jacobs: It might be easier if you can find a good 3rd party repo to use.
Hackus: I will try it and if it doesn't take too long report back.
Hackus: I will look for one....
_KAMI_1: So the current driver is not perfect
Hackus: C U L8r.
Agent_Jacobs: Good luck.
_KAMI_1: but it is good enough for me :)
_KAMI_1: VGA output is okay
_KAMI_1: only hdmi has problem :S
Agent_Jacobs: KAMI, looking back in the logs, you can get picture when you go xrandr --output HDMI-0 --auto ?
_KAMI_1: Yes I get picture in hdmi
_KAMI_1: but it turns off LCD
dmz001: Droste: modprobe fbcon : not found
Agent_Jacobs: dmz, custom kernel?
Droste: ok, custom kernel?
dmz001: Droste: modprobe radeon modeset=1 : Blanked the DVI port in text mode
Agent_Jacobs: No, is yours custom or prebuilt?
Droste: yeah without fbcon you can't see the console with kms
dmz001: Where do I find fbcon. yum search didn't show it up.
Droste: did you build your own kernel?
dmz001: No. Straight out of the box.
Agent_Jacobs: Version?
Agent_Jacobs: KAMI, when HSMI loaded, that was when you grabbed that xrandr output?
Agent_Jacobs: HDMI*
_KAMI_1: no
_KAMI_1: i just connected hdmi
Droste: dmz001: you'r using fedora?
_KAMI_1: and this is output of xrandr
Agent_Jacobs: Still on LVDS, then?
dmz001: Droste: Yes FC12
Droste: what's the output of "uname -r"?
dmz001: Droset: 2.6.31.6-166.fc12.i686.PAE
Droste: ok... this is quite old
Agent_Jacobs: KAMI, can you turn on HDMI with something like this:
dmz001: Droste: Hmm? Old? I just installed.dec-12
Agent_Jacobs: xrandr --output LVDS --auto --output HDMI-0 --right-of LVDS ?
Agent_Jacobs: That would be a dualhead example.
dmz001: Droste: I just poked a bit in /var/log/messages. A host of kernel: [drm] messages
Droste: no errors? I think kms is working, but you need fbcon to see the console
dmz001: Droste: Any chance that fbcon is static? I see one of the messages shows: kernel: fbcon: radeondrmfb (fb0) is primary device
_KAMI_1: Agent_Jacobs: how can I go back if primary is blanked?
Droste: could be :-) but you didn't see the console after loading radeon with modeset=1?
Agent_Jacobs: KAMI, you don't have a console or a menu, then?
dmz001: It blanked the DVI and switched the VGA to a tiny font
Droste: ah ok, tried starting x and enable dvi over xrandr?
dmz001: Droste: Expecting a bit much... Can you specify the command lines.
Agent_Jacobs: Crib off what I put up here for KAMI.
Droste: xrandr --output DVI-0 --auto
crib: off!
dmz001: Any chance I lose both screens? What is recovery?
Agent_Jacobs: Use the full line:
Droste: don't think so
dmz001: At the moment I am only running nomodeset on the command line. After a reboot
Droste: yes, try it the same way as before, unload radeon, load radeon with modeset=1, start x and enable dvi-0 over xrandr
dmz001: Tried it... Both screens went dark for a while
dmz001: Oh. Sorry.
dmz001: Okay off we go...
Agent_Jacobs: Mirrored display, now?
dmz001: Err no, only the VGA is live.
Droste: even after the xrandr command?
dmz001: Yup.
Agent_Jacobs: Pastebin an xrandr -q result, pls.
Agent_Jacobs: Welcome back, KAMI.
Droste: dmz001: do what agent_jacobs told you to do :-D
_KAMI_: hi Agent_Jacobs :)
_KAMI_: I had to restart
_KAMI_: no picture on lcd
_KAMI_: no picture on hdmi
Agent_Jacobs: I figured. Can you open a command line now?
dmz001: http://www.pastebin.ca/1725666
_KAMI_: however both of the has signal
_KAMI_: but only black screen
_KAMI_: sure
Droste: strange
Agent_Jacobs: Yeah, both displays show active according to xrandr.
Droste: dows anybody know where to get a newer kernel for fedora? :-)
Droste: try xrandr --output DVI-0 --mode=1280x800 --rate=75
Droste: maybe just a problem with the resolution
dmz001: Doesn't like the command line. Something wrong...
Agent_Jacobs: Ok, try this command line please KAMI.
_KAMI_: okay :)
Agent_Jacobs: xrandr --output LVDS --auto --output HDMI-0 --right-of LVDS
Agent_Jacobs: hold that.
Agent_Jacobs: xrandr --output LVDS --auto --output HDMI-0 --auto --right-of LVDS
_KAMI_: is it same as before
_KAMI_: ok
_KAMI_: but how can I came back without restart?
Agent_Jacobs: No, auto screen res and refresh for HDMI.
Agent_Jacobs: I forgot it.
Droste: dmz001: does it say whats wrong?
dmz001: Droste: No. Just gives me usage: ... etc.
Agent_Jacobs: 1280x800 isn't a define res.
Agent_Jacobs: I'm thinking.
Droste: it was in the xrandr -q output :-D
Droste: try 1280x1024
Agent_Jacobs: Yeah, that's a good mode for VGA and DVI.
Agent_Jacobs: According to the pastebin :)
Agent_Jacobs: *bonk* I see it there on the modelines.
dmz001: Same thing. I cut and pasted your command line. It doesn't like something
Droste: :-)
Droste: hm
Droste: ah..
Droste: xrandr --output DVI-0 --mode 1280x800 --rate 75
Droste: it's late here :-D
_KAMI_: Agent_Jacobs: So I entered what you want
_KAMI_: LCD blanked
dmz001: Bingo!
_KAMI_: TV has displayed picture
dmz001: Is this a clue or the workaround?
_KAMI_: I was able to come back when HDMI disconnected
Droste: both ;-)
_KAMI_: entered xrandr to terminal
Hackus: Hello.
Hackus: Well, I am back!
_KAMI_: switch to consol1 and back to gfx
Hackus: :-)
Hackus: My VGA Arb error is gone! Woot!
Droste: you could make a script to execute the command when gdm/kdm starts
Hackus: I get a different error now. :-)
Droste: but you will never get console output on dvi
Agent_Jacobs: Well, KAMI, that is most weird.
Droste: but: i think this was fixed in a newer kernel
Hackus: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch.
Hackus: [dri] radeon kernel module version is 2.0.0 but version 1.17.0 or newer is needed.
Hackus: Ok....so that is just doesn't make any sense to me. :-)
_KAMI_: Agent_Jacobs: screen 0: 3360x1080 888x285 mm 96.05dpi / crtc 0: 1440x900 59.9 +0+0 "LVDS" / crtc 1: 1920x1080 50.0 +1440+0 "HDMI-0"
_KAMI_: verbose log of your command
Hackus: Last time I checked my Ven logic stuff, 2.0.0 is newer than 1.17.0. :-)
Droste: dmz001: the problem is that your monitor in dvi-0 reports that he's able to handle 1920x1200 at 60Hz but it doesn't seem so
MostAwesomeDude: Hackus: The log message is disingenuous; I fixed it in a newer version of the driver.
dmz001: Droste: Well in windows it sure can.
MostAwesomeDude: But basically it means that the kernel module's too "new."
Hackus: Ok, so it should be working now right?
MostAwesomeDude: You need to rebuild your DDX with kernel modesetting.
Droste: hackus: did you build the libdrm and drivers on your own? then something went wrong
Agent_Jacobs: KAMI, that's the output after restarting X with your laptop screen working again?
mentor: Hackus: Did you make sure to load the radeon module before starting X?
Droste: dmz001: maybe a bug in the kernel module... mostawesomedue: do you know if there is a fedora 12 repository with a newer kernel?
Droste: MostAwesomeDude*
Hackus: No, I don't think so.
MostAwesomeDude: Droste: What's "newer" mean? F12's kernel is up-to-date enough for most things, but there's no bleeding-edge kernel.
dmz001: We just did this without the radeon.modeset
MostAwesomeDude: There might be one for F13/Rawhide...
Hackus: I rebuilt and installed the ddx stuff, which I think is in this repo: git clone git://anongit.freedesktop.org/xorg/driver/xf86-video-ati
Agent_Jacobs: libdrm as well?
Droste: newer than 2.6.31.6 :-)
_KAMI_: Agent_Jacobs: that is outout of your switching command
Droste: dmz001: you could tr a f13 live cd if there is one :-)
Agent_Jacobs: Oh, ok.
Droste: hackus: did you boot with radeon.modeset=1?
dmz001: I didn't see one offered.
dmz001: I could try Knoppix.
Hackus: Well, I think I might have missed a package. :-)
Hackus: More work to do. Darn it, this is complicated. I want my mommy.
Droste: dmz001: I don't think they have a newer kernel than fedora :-)
Agent_Jacobs: Make sure you build an up to date libdrm before doing the radeon DDX code.
Hackus: Could be a while....
Hackus: I think I am going to call it a night....my finals where a killer.
Hackus: Been up now for about 20 hours.
Droste: hackus: :-D first build and install libdrm, after that build and install xf86-video-ati
Hackus: Ok.
dmz001: What is actually going on here? This card and monitor on the DVI worked fine under windows.
dmz001: So where can the kernel get it wrong?
Hackus: Thats because Microsoft has bought off the industry and pays ATI under the table not to write good drivers for LINUX.
Hackus: :-)
Agent_Jacobs: Could you do 1080p in Windows?
dmz001: We see that with xrandr I can get it at least to 1280x1024.
Hackus: Purely specualtion on my part......OR IS IT!!!!
dmz001: Sorry, what is 1080p under windows?
Droste: dmz001: try switching the mode again. just to check if thats the problem: xrandr --ouput DVI-0 --mode 1920x1200 --rate 60
Agent_Jacobs: 1920x1080 screen resolution. Does Windows use that, or does it use a lesser resolution?
Droste: typo... it's output not ouput
dmz001: 1920x1200 DVI blank. 1600x1200 blank 1680x1050 Success
Agent_Jacobs: It's all in the resolution Windows is defaulting to.
dmz001: Windows did 1920x1200
Hackus: Gad....
_KAMI_: Agent_Jacobs: Thank you for help
Hackus: Huge number of merge conflicts from libdrm HEAD. :-(
_KAMI_: I have to go
Hackus: Ok, I give up....for now.
Hackus: Going to bed.
_KAMI_: Can I come back later to solve this problem?
Hackus: Nighty night.
Hackus: Thanks for the help.
Droste: hackus: merge? ok night
Hackus: No, merge failed....in quite a spectacular manner too.
Droste: dmz001: maybe the refreshrate is the problem
Hackus: I need chocolate after a bad merge. :-)
Droste: does windows also use 60Hz for this modes?
Agent_Jacobs: KAMI, sure. Before you go, I'll give you my thoughts.
Hackus: night!
Agent_Jacobs: xrandr is accepting your commands and thinks it's activating both displays.
_KAMI_: okay
_KAMI_: yes looks okay
_KAMI_: I do not understand why not use smaller resulution
_KAMI_: for hdmi
dmz001: Droste: Use something lower than 60?
Agent_Jacobs: I don't think it will make a difference. HDMI is activating.
Agent_Jacobs: And you get picture.
Droste: yes :-) maybe 50Hz
Agent_Jacobs: It's almost like there's a modesetting quirk with the laptop.
Droste: xrandr don't accept that rate, so we can't find out in linux :-)
_KAMI_: Agent_Jacobs: but why swith of lcd then?
_KAMI_: s/of/off
dmz001: Droste: Tried 50... Only one that worked was 1680x1050
Agent_Jacobs: That's not 100% clear.
Agent_Jacobs: I'm assuming the laptop came with Windows from factory?
dmz001: Agent_Jacobs: Was that for me or _KAMI_
Droste: dmz001: yes, with xrandr we can't find out
Agent_Jacobs: KAMI
_KAMI_: Agent_Jacobs: not installed with windows only dos
dmz001: Droste: We can't find out what?
Droste: dmz001: if 50Hz works, to get 50Hz with xrandr we have to add a new mode and... i never did this before :-D
Droste: if you pass --rate 50 it uses 60 because 50 is currently not available
dmz001: Well I got the modeline from cvt
_KAMI_: Agent_Jacobs: Maybe report it as a bug?
Droste: you can try it :-) the syntax is explained in xrandr --help. you need to create the mode with --newmode and add it to DVI-0 with --addmode
Agent_Jacobs: Perhaps. Definitely you will need X.org.log for before and after turning on HDMI output.
Agent_Jacobs: You might find clues there.
Agent_Jacobs: If you do file a bug, include as much information as possible.
_KAMI_: okay
Agent_Jacobs: Sorry, as much relevant information as possible.
_KAMI_: :)
Agent_Jacobs: ex xrandr -q before and after, X.org.log model information for your laptop and a step by step walkthrough demonstrating the problem.
dmz001: Bingo.
_KAMI_: ok
dmz001: It went into extended mode. The screen is now spread across two screens on two different sides of the room :-)
dmz001: So, until they fix the problem How do I do this?
Agent_Jacobs: Alrighty! Talk to you later, KAMI.
Droste: copy the commands in a text file, as first line insert #!/bin/sh, save it, chmod +x the file and put it into the autostart folder of kdm or gdm
Droste: and always boot with radeon.modeset=1
dmz001: I'm currently NOT running with radeon.modeset=1. That left me with a blank DVI port in the modprobe experiment.
dmz001: And how do I disable the VGA port or at least force it to mirror?
Droste: also with xrandr, for mirror (clone) mode you should use the same resolution on both screens
Agent_Jacobs: Disable the VGA with xrandr --output VGA-0 --off
Agent_Jacobs: Mirror the DVI onto VGA with xrandr --output VGA-0 --same-as DVI-0
Droste: clone: xrandr --output VGA-0 --same-as DVI-0
dmz001: Thanks folks... Let me play.
Agent_Jacobs: Although you can do vice versa and not know the difference.
dmz001: Now is this a bug in fedora / kernel? Or in X where it doesn't get the full set of modes. Or something else.
dmz001: As I said, windows had no problem and it was 60Hz refresh if not higher.
Droste: could be all :-) and there were fixes with edid since 2.6.31.6 :-)
Droste: I'm not that deep into it :-) also just a user ;-)
dmz001: Thanks... Now for some education.
dmz001: How do I find out what the TLAs being used mean? I did a search on KRM and nothing with meaning to x came up.
dmz001: And drm? that definitely is a difficult search
Droste: it's KMS ;-)
dmz001: Sorry. I meant KMS.... Is that my monitor having a period?
Droste: :-D hopefully not, maybe start here: http://www.x.org/wiki/radeon :-)
dmz001: Thanks.
dmz001: Good night.
Droste: night
Agent_Jacobs: If you can get Radeon+KMS working for VGA, you can try doing xrandr to get DVI going.
Agent_Jacobs: But maybe for another day :)
dmz001: What IS KMS? And how do I know if it is working
Agent_Jacobs: Kernel Modesetting
Agent_Jacobs: It's where the kernel handles telling your video card what to do.
dmz001: Modesetting is working on the VGA. Without the nomodeset line in the boot I lose the text on the DVI
dmz001: So that makes it sound like a kernel bug
dmz001: I don't want to come up blind. Too many things to go wrong.
Droste: if I remember correctly mirror/clone was not available for the console in 2.6.31
Droste: so this can be fixed with a newer kernel
Agent_Jacobs: Could be. If you can get it working on both with xrandr, it's just a matter of included features for the framebuffer console.
Droste: (works for me with 2.6.32 (vga and dvi))
dmz001: Okay... So I have to wait it out with fedora. Or build my own?
Agent_Jacobs: For dual head console and KMS together, sounds like it.
Droste1: sorry, got disconnected, did you say something?
Agent_Jacobs: He's wanting to know his options for updating the kernel.
Agent_Jacobs: You missed one other, and that's find an updated kernel RPM.
dmz001: I'm not trying for dual headed. I don't have the room :-) But I need to get the DVI working.... I've appropriated my kids monitor to run the VGA port. And I
dmz001: need to get it back to them. ;-) Not very patient when they can't get at their games.
Droste1: as mad suggested: maybe theres a newer kernel for f13, which can be installed in f12 :-)
Agent_Jacobs: Maybe. Don't temp the wrath of the children, though :)
dmz001: I'll take a look.... Tomorrow...
Agent_Jacobs: Ok, take it easy, then.
dmz001: I really want to find a way to turn off fedora's plymouth... I LIKE seeing the messages flash by.
Droste1: I like that too :-D
dmz001: Good night..........
Droste1: night :-)