Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-10-29

Search This Log:


ajami: airlied: what was going wrong with the kernel? Why wasn't it "letting me boot"
airlied: ajami: not sure, lots of fixes for different things, probably a crash in new spread spectrum code
mart: Ghworg, http://www.two-kings.de/tutorials/dxgraphics/dxgraphics16.html similar to fbo's sure lot different at same time
happycube: mostawesomedude: here's a version of r300_emit_query_finish shortened by replacing the switch w/a for loop: http://pastebin.com/m1f3bc799
Ghworg: mart: From what I can tell from the wine source the textures I'm interested in (DXT1-5 or s3tc) just get passed straight through to mesa
Ghworg: Ah well, will just have to wait for someone who actually understands this stuff to fix the real bug
mart: s3tc was somekind of compression alghorithm, what about dxt what this stands for, render to texture?
agd5f: mart: dxt is s3tc, just the ms licensed version
agd5f: directx texture compression
MostAwesomeDude: happycube: I'm not totally sure if that's more readable, but it's quite clever. :3
happycube: thx :) (as for readability... i think the best hope's really good commenting)
MostAwesomeDude: Yeah, there's always a tradeoff between verbosity and complexity in code regardless of how many wall-o-text comments I put in there.
MostAwesomeDude: I mean, I laid the comments on *THICK* when I first wrote it all up, and it's still a jumble in some places.
soreau: dont you hate it when you find yourself back in the same place after a few months going (what the hell was I thinking)
happycube: this code's just intrinsicly complex... there're only so many ways to write a CS register sequence
soreau: and your comments dont even make sense to you ;)
happycube: yeah ;)
soreau: then it clicks, and you crack a crooked smile (clever dog you..)
MostAwesomeDude: I love some of my comments sometimes. :3
MostAwesomeDude: But I don't have very many good ones in r300g.
spstarr: tests today's mesa git
ajami: airlied: I think there may be an issue with r600_tex...http://pastebin.com/d645a8de1
ajami: airlied: or it could just be an issue with the game I am attempting to play. Not sure.
airlied: ajami: looks like r600 mesa bug
airlied: not really looking at those yet since they are experimental
mart: agd5f, need to take a look to which memory pool/framebuffer are they rendered adventually with opengl, compiz had that s3tc option too
mart: there is an open source toolkit for amd also, was something like close to metal, which may be interesting
MostAwesomeDude: I don't know why Compiz would use S3TC.
MostAwesomeDude: Also nobody does CTM.
airlied: MostAwesomeDude: it only uses it when you string random words together from reading 20 firefox links composited on top of each other
airlied: with like a 10% alpha
MostAwesomeDude: airlied: Ooooooh. I see.
happycube: lol
happycube: the CTM documentation is suprisingly similar to the r500 programming manual
happycube: ^ hopefully it got used to help r300 pre-amd-opening-up
MostAwesomeDude: Yeah, CTM was pretty much r500 fragment shaders IIUC.
MostAwesomeDude: But it wasn't really easily accessible until the r5xx manual got released.
happycube: and by then they'd moved on
happycube: also nv40_draw_elements ends with a flush. :P there's gotta be a better way tho.
happycube: (theirs is rather different, though... looks like it optionally flushes the texture cache and optionally sets up a fence)
mart: happycube, so it's good stuff then, cause r500 human readable context is needed to try programming/talking about stuff
happycube: not really - one could do something similar when gallium's finished around TGSI... and then you get something that works on more than one chip type ;)
mart: without screaming on the streets like an idiot trying to figure out , what which loop does, happens also very randomly to lot's of people who can't stick their penis to former girflfirends target buffer
happycube: "this is rocket science, not GPU coding."
spstarr: touches the bottom of his laptop case.... .. mmmm hot
spstarr: might be hot to begin cooking an egg
mart: constantly even so to speak, "..." years and years 10 and counting, so to start negotiating how to kill one of them off or slice throat with their own knives, and forward the kgb agents names to police departements "whith a pistol i carry that dude is a threat to john Kerrys whife"
Ghworg: My laptop runs ~10C hotter than it did under fglrx, it'll be nice when the powersaving stuff goes in and my fan can occasionally switch off
mart: airlied, how randmos is that? how long should i suffer , going into court with such absurds, until it hits you, or you land to your feets or someone here would understand, a faget like matthew garret? every fucking day, every single day same problems
DragonionS: hi all! I have a problem during trying to start Radeon driver with Gallium. I see cursor and screen is black.
DragonionS: What could it be?
MostAwesomeDude: All kinds of things.
MostAwesomeDude: Radeon Gallium drivers are still really really experimental and you shouldn't be surprised if they don't work.
DragonionS: thanks
DragonionS: for the answer
MostAwesomeDude: Are you trying xorg st or dri st?
DragonionS: what does it mean "st"? I have xorg-server 1.7.1, mesa, radeon, drm from git
MostAwesomeDude: Which part of Gallium are you trying out?
MostAwesomeDude: How are you running it?
soreau: MostAwesomeDude: While youre here, can you explain what xorg st is or the difference between it and dri st?
MostAwesomeDude: soreau: Sure!
hoo-hah: hi guys, any of you on gentoo with x11 overlay? seems I can't build mesa-9999.
hoo-hah: I'm trying to wgetpaste the portage log, but my net is shaped
hoo-hah: and it struggles to get uploaded
MostAwesomeDude: dri-st is the radeon_dri.so driver. It can be used to do GL.
soreau: For instance, I have gallium installed to nonstandard prefix and loading only certain gl progs by setting LIBGL_DRIVERS_DIR. Would it be any different if I were using xorg st too?
DragonionS: I've installed mesa with Gallium USE-flag that compile mesa with "enable-gallium-radeon" option
hoo-hah: I'll try uploading through ompload
MostAwesomeDude: xorg-st is the modesetting_drv.so driver. It can be used with Xorg to do regular 2D.
DragonionS: than I have the latest radeon driver from git
amarks: hoo-hah: what is the build failure?
hoo-hah: http://omploader.org/vMm5kOA
DragonionS: I am loading ttm driver from kernel
MostAwesomeDude: DragonionS: What programs are you trying to run? Just X?
DragonionS: no, gnome
DragonionS: :)
MostAwesomeDude: Hm.
hoo-hah: from the log, gmake[5]: *** No rule to make target `compiler/libr300compiler.a', needed by `r300_dri.so'. Stop.
MostAwesomeDude: Could you pastebin your Xorg.0.log?
amarks: hoo-hah: yes i am on gentoo, i have hit this issue since brian paul changed the makefile
amarks: hoo-hah: do you need r300?
hoo-hah: nope
DragonionS: wait a bit - I'll switch to gallium...
amarks: hoo-hah: then remove r600 from the ebuild and rebuild the manifest
amarks: sorry
amarks: r300
hoo-hah: wait. I think i need r600
hoo-hah: ah, kay
amarks: hoo-hah: it is something gentoo specific as mesa git compiles file manually
amarks: s/file/fine
soreau: MostAwesomeDude: So.. Would it make any difference if I were using xorg st too? (for progs using r300g)
hoo-hah: amarks: ah cool. I've never taken a look inside the ebuild
hoo-hah: amarks: do you reckon that I can remove other chipset support while i'm there?
hoo-hah: all I'm after is r600 support
amarks: hoo-hah: you can, i did not
hoo-hah: or does this mean I'll need to modify the other 2 ebuilds
MostAwesomeDude: soreau: Yeah, xorg-st is a lot less tested and possibly broken w/dri-st.
soreau: MostAwesomeDude: That is to say, it would make a difference. It would be worse. (?)
amarks: hoo-hah: just edit the driver_enable video_cards line
hoo-hah: yeah
MostAwesomeDude: soreau: Yeah, probably.
soreau: ok thanks for clarifying
soreau: I will probably try it anyway ;)
MostAwesomeDude: DragonionS: Hm. Looks like your X server's not using Gallium directly, so it should work fine.
ajami: airlied: so what exactly is this new spread spectrum code?
DragonionS: thanks for the answer. I wish you good luck in your work and as ATI user. ;)
blueldr137: hi
blueldr137: does anyone here use a X1600 card?
blueldr137: glxinfo can't identify my chip id, and therefore puts me in indirect rendering :(
blueldr137: would installing the latest 'xserver-xorg-video-chips' package allow my chip to be identified?
MostAwesomeDude: That's for a Chips video card, actually. :3
MostAwesomeDude: xserver-xorg-video-ati is what you want.
blueldr137: oh lol :P
MostAwesomeDude: Also, for 3D, you're going to need to update Mesa; IIRC it's called "libgl-mesa-dri" on Debian.
blueldr137: ok
MostAwesomeDude: You *are* on Debian, right?
blueldr137: yes
MostAwesomeDude: Stable, or testing?
blueldr137: stable
MostAwesomeDude: Ah.
MostAwesomeDude: Could you pastebin the output of "LIBGL_DEBUG=verbose glxinfo" so I can be sure I'm not misleading you? :3
blueldr137: sure... my deb version is lenny too
blueldr137: http://paste.debian.net/50241/
blueldr137: thanks
MostAwesomeDude: Yeah, Mesa's just too old, that's all.
blueldr137: so I just need to update 'libgl-mesa-dri' and not 'xserver-xorg-video-ati'?
MostAwesomeDude: Oh, wow. Lenny's libgl1-mesa-dri is version 7.0, that's ancient.
MostAwesomeDude: Yes, with the caveat that you're going to have to pull it from squeeze. :3
MostAwesomeDude: $ aptitude -t squeeze update libgl1-mesa-dri
blueldr137: ok... squeeze is the upcoming deb release right?
MostAwesomeDude: Or something along those lines; haven't done Debian in a bit.
MostAwesomeDude: Yeah, squeeze is current testing.
blueldr137: ok I'm sure I'll be able to figure it out now lol
blueldr137: thanks
blueldr137: I'll try to update it now :)
hoo-hah: urgh
hoo-hah: who was I chatting with earlier regarding gentoo ebuild?
hoo-hah: (lost my irc session log)
hoo-hah: are the latest mesa xf86-driver-ati and libdrm dependent on xorg-server 1.7+?
hoo-hah: I remember getting dri2 for my r600 card working with xorg 1.6.x
hoo-hah: xorg log at http://dpaste.com/113608/
airlied: ajami: spread spectrum is used on laptops to make the LVDS not interfere with other things
hoo-hah: ah, nmv
hoo-hah: *nvm
hoo-hah: turns out it was due to evdev module having to be rebuilt
hoo-hah: with 1.7.1 xorg upgrade
amarks: hoo-hah: yes that is necessary with any major upgrade
hoo-hah: amarks: I always thought it'd be taken care of with xorg-server upgrade
hoo-hah: does this mean I'll need to explicitly upgrade any input modules I'm using (mouse,kbd)?
amarks: hoo-hah: yes, typically you rebuild all x11-drivers
amarks: hoo-hah: or update to the latest
amarks: hoo-hah: mouse is 1.5.0, keyboard is 1.4.0 and evdev is 2.3.0
hoo-hah: yeah, that's available with ~amd64
hoo-hah: (what I'm running)
amarks: sometimes this happens as part of the dependencies, sometimes not
hoo-hah: I just didn't expect to have to manually rebuild. evdev now appears in world
amarks: when u emerge xorg-server it will warn you
hoo-hah: amarks: does your world file contain those input drivers you've installed?
amarks: absolutely
hoo-hah: odd. I thought they'd just get puleed in as dependencies based on what's in make.conf
amarks: they are separate packages, why wouldn't they show in world?
hoo-hah: dunno. I've had problems with dependencies ending up in world, that should have not been there
hoo-hah: I'm sking in #gentoo to be sure
amarks: hoo-hah: why do you care if it's in world or not?
hoo-hah: amarks: if the driver rebuilds ideally should be done as oneshot
amarks: hoo-hah: i've never used oneshot really, i'm not across why it is necessary
amarks: hoo-hah: i.e. for most stuff, you want it recorded so it can be updated
ajami: airlied: that error I showed you earlier the r600Delete? that issue is causing a lot of problems actually. These stemed from the latest master.
ajami: Everything requiring OpenGL as far as I can tell if getting a segment fault. I try opening comiz, segment fault. Glxgears, segfault.
ajami: if->is
ajami: agd5f, glisse, airlied, anyone working on r600 classic ^^^
ajami: and now I am getting either gpu hardlocks or kernel deaths. I just had to hard restart my comp. Anything I can do to help narrow this/these issues down, let me know.
ajami: VLC using Xv hasn't had any issues. Heck 720p playback actually works flawlessly on my computer now. Never, not with propriotary drivers, fglrx/windows, has my comp been able to do so.
honk: that's pretty poor :}
twoerner: glisse: 4G does not fix my poblem
twoerner: glisse: to have only 4G does not fix my problem
DJJeff: ATI T200 Unified AVStream Driver has a yellow sign
DJJeff: whats the best fix?
Nightwulf: hi all
glisse: airlied: your patch post pone corruption
glisse: got them now after 2h or desktop usage
glisse: this thing is painfull to debug
agd5f: ajami: make realclean and re-install
agd5f: glisse, airlied: regarding the font corruption, have you tried adding an HDP cache flush when mapping/unmapping a bo in vram?
glisse: agd5f: good idea :)
glisse: i need to stop working on 3 differents things a the same time
agd5f: glisse: yeah, me too
spreeuw: hmm any idea why my 3d fullscreen windows are not fullscreen? they are centered 1:1
spreeuw: and slightly darker too than they should be
agd5f: spreeuw: laptop?
spreeuw: no an eizo tft
spreeuw: on a 3450 pcie
spreeuw: it started after I updated my xserver to the current stable
spreeuw: no errors in the log
agd5f: spreeuw: xrandr --verbose and check what scaler is set to
spreeuw: scaler: off
spreeuw: in 2d desktop
agd5f: spreeuw: what res is the 3d app trying to switch to?
spreeuw: 1280x960 which is available in xrandr
agd5f: and what's your monitor's native res?
spreeuw: native is 1920x1200
spreeuw: current
spreeuw: does tha texplain the sclaer off atm?
agd5f: spreeuw: does xrandr --output --mode 1280x960 do the same thing?
agd5f: replace with whatever output you are using, DVI-0, etc.
spreeuw: nope that works
agd5f: spreeuw: I suspect it's an issue with xvidmode trying to change the mode on the wrong crtc or ouptut
spreeuw: ok thanks I'll see if its something I forgot to build
agd5f: spreeuw: if it was an xserver change, you might try bisecting that
agd5f: spreeuw: I doubt you missed anything. xvidmode just isn't multi-head aware
spreeuw: with the release 7.5 dde I get a black screen
spreeuw: the git version works
spreeuw: so it may indeed be head related
_moep_: ATI Mobility Radeon HD3470 <- should be work with radeon?
adamk_: _moep_: Yes, it should work.
_moep_: ok :)
_moep_: btw: I'm searching a new radeon card. atm I've a Radeon HD 4350 it doesnt work (I think a radeonbug, bug report is open) so I want to buy a new card but dont know which one
adamk_: Well if you are having problems with one radeon, it could be something to do with your computer itself, so there's no guarantee that another one would have more luck.
mcgreg: _moep_: well, I have a radeon 4670. it is working very nicely
taiu1: seems i can also use INDX_OFFSET on r600
taiu1: agd5f: what you think of this version? http://people.freedesktop.org/~andrem/0001-r600-use-AUTO_INDEX-for-draw-saves-cmd-buffer-space.patch
_moep_: mcgreg: like this one http://geizhals.at/deutschland/a430965.html ?
agd5f: taiu1: looks good. are the indices always sequential?
agd5f: taiu1: nevermind
mcgreg: ATI Technologies Inc RV730XT [Radeon HD 4670] , mine is from sapphire, pcie2.0, 512MB gddr3
mcgreg: very similiar to my yes
agd5f: taiu1: looks good to me
taiu1: o
taiu1: k
_moep_: mcgreg: and do you use DRI?
_moep_: with your card?
_moep_: a crap offline :D
ajami: agd5f: my bad. Will do that from now on before asking.
ajami: airlied, glisse, agd5f all seems to work fine now.
agd5f: ajami: make clean is probably enough
ajami: agd5f: on the contrary of the patch breaking everything, the game I have been trying to get to work in wine happens to work now. (without changing anything with wine).
agd5f: cool
Pallokala: am I using somehow from libdrm-git when there hasn't been any changes for a long time in it (gentoo git-ebuilds here)
Pallokala: s/from/wrong/ + ?
soreau: Pallokala: What makes you think you're using wrong libdrm?
Pallokala: the fact that I havent seen any changes for a 1-2 weeks
Pallokala: but I might just be paranoid because things are working quite well here
Ghworg: Only changes since 2.4.15 are to intel and nouveau anyway
UnNamed: osiris_: hey, i figured more or less what was making some apps slow, no need of sysprof: set migration to greedy, and speed went to old XAA level ("smart" is like default, or maybe a bit faster, but still in the >1sec range)
spreeuw: hmm what are these firmwares the kernel installs?
spreeuw: MKDIR /lib/firmware/radeon
spreeuw: INSTALL /lib/firmware/radeon/R100_cp.bin
spreeuw: etc
UnNamed: spreeuw: microcode for the video cards, that before were in the .c files
mjg59: spreeuw: It's the firmware for the command processor on the card
UnNamed: spreeuw: no idea what you lose without the command processor... 3d for sure, maybe more
spreeuw: kay going to boot into the 32 rc5
spreeuw: nice , seems to work
spreeuw: without special drm trees
airlied: glisse: yeah its painful, I just thought it might be useful for debugging help that patch
airlied: but Michel is probably right about it being coherency issue
glisse: yeah i think it's a coherency issue too
glisse: i fail since monday to find a standalone test case which exhibit the issue
glisse: and debugging it with X is painfull as it takes times for my laptop to go corrupted
glisse: tomorrow i will clflush like crazy in all place i can think of
glisse: airlied: do you have access to ich9 or or 10 motherboard ?
airlied: glisse: no but darktama's laptop hangs pretty badly, need to check what it has
glisse: r600 ?
glisse: hard lockup no ssh ?
airlied: glisse: yeah its a dual-gpu laptop
airlied: even with aspm off it hangs on X login, however it also dies when acpi vidoe module loads
airlied: so it needs a bit of debugging
glisse: i need to cleanup my fence rework patch and my lockup detection patch
airlied: if i load acpi video before radeon I can at least boot to X where it hangs
glisse: interesting
glisse: i need to go pickup older r600 gpu tomorrow, i forget some gpu at friend's place
glisse: hopefully one will lockup on my ich8
glisse: finds strange to hope for a lockup
airlied: I think its ich9 specifc
airlied: its like pcie is the new agp ;-)
glisse: there is also bugs report with ich10
airlied: yeah I believe they are the first time aspm is used
glisse: well i think benh would say that hw engineer just create des saloperies de bugs ;)
airlied: might have to find someone with vista installed ;-(
airlied: or maybe fglrx is enough
glisse: i installed a windowsxp for the x800 suspend/resume issue
glisse: took me time to find a windowxp cd
fcami: waves.
orzel: Hello. I have this very strange error in my Xorg.0.log since i've enabled KMS in the kernel : [dri] radeon kernel module version is 2.0.0 but version 1.17.0 or newer is needed.
orzel: isn't 2.0.0 'newer' than 1.17.0 ?
chithead: orzel: you are probably using non-kms capable ddx
chithead: the error message is indeed misleading though
orzel: chithead: yes, i'm recompiling stuff to have it
orzel: (following the howto in the topic)
orzel: i'm using the vanilla kernel as in 'from git few hours ago'
agd5f: orzel: you need to recompile your ddx with kms support
orzel: yeps, doing so right now
orzel: "First of all check that you don't load radeonfb, uvesafb or vesafb module.".... -> those are kernel modules, not X modues, right ? like "Remove CONFIG_FB_VESA=y/m , CONFIG_FB_RADEON=y/m from your kernel config" ??
orzel: (it's from the howto)
agd5f: yes
orzel: ok
chithead: vesafb is not a module
chithead: but vesafb remains inactive unless you pass vga= kernel parameter
orzel: anyway, this is not very clear.. i have to guess the change to do in the kernel config, no ?
chithead: if you have radeonfb or uvesafb modules, you need to prevent them from being loaded
orzel: I have never seen uvesafb, i guess fbcon is CONFIG_FRAMEBUFFER_CONSOLE
orzel: and i dont know what "radeondrmfb" is
adamk_: That's the one you want.
adamk_: It's just part of the radeondrm.
orzel: so, CONFIG_DRM_RADEON=y is enough
orzel: is there a way to know if kms is enabled before x starts ? i've grepped my dmesg but can't find anything
fcami: orzel: your screen should be setup correctly i.e. native resolution
adamk_: Well the second the radeon drm module loads, your console screen should switch to the monitors native mode.
adamk_: You may want to enable radeondrmfb support by default in the staging section.
orzel: i booted with radeon and vga fb, but without vga=* in grub. And i had a graphical boot. Somethind i never had before trying this kernel. So i guess at least drm is working
adamk_: orzel, What does 'cat /proc/fb' show?
orzel: hey :)
orzel: it says '0 radeondrmfb'
orzel: sounds good
adamk_: Yep.
adamk_: You need libdrm compiled with KMS support (there's some flag to ./configure that you need) and then xf86-video-ati compiled *after* libdrm is installed.
orzel: yes, i think i got this, but i still have to check how i'm supposed to do that with gentoo
orzel: (gentoo provides 'live' package from git)
orzel: (btw, i have a somehow recent ati 'card', which is the hd4200 embedded in some recent amd chipset, not sure how supported this is, we'll see)
wilco_: orzel: I'm running a hd4570 which is also a embedded card, works suprisingly well
orzel: wilco_: wait&see :-)
orzel: i have mesa-7.6 without any special configure option... i needn't recompile it, need I ?
orzel: libdrm ok, compiling *-ati now. It says "Kernel modesetting: yes". Sounds good too.
orzel: see ya, i dont know if i'll get back soon :) testing
orzelf: wel.. it's not a big success :/
orzelf: when starting X the screen get all ugly, and after few seconds, both displays said "no signal". Keyboard is not working. I'm now talking from my laptop. The computer stlll pings, there's hope
adamk_: Log in remotely if you can and grab /var/log/Xorg.0.log
orzelf: i have it, no error in it (EE).. everything seems fine. Weird..
adamk_: Well I'm still of the opinion that KMS is not ready for general consumption.
orzelf: (II) RADEON(0): Setting screen physical size to 952 x 317
orzelf: wait, this is strange :)
adamk_: And considering that you have to jump through hoops to get it working, I guess that's a common belief :-)
adamk_: But it can't hurt to open up a bug report.
orzelf: adamk_: i'm just trying, not yet complaining. i'm aware it's bleeding edge
orzelf: (II) [KMS] Kernel modesetting enabled. <---- was starting ok..
orzelf: if i stop and restart xdm, i have dark screens (lot better than no signal), i can move the mouse (a small clock) all around. Somehow it almost manage to do it
orzelf: ah, though it segfaults with "[mi] EQ overflowing. The server is probably stuck in an infinite loop."
orzelf: well, i'll keep both Xorg.0.log and go back to my usual settings
chithead: eq overflowing is probably an issue with some input driver
orzelf: the bt mentions evdev
orzelf: chithead: i did not recompile evdev after the switch to kms... i should have M
orzelf: ?
chithead: if you upgrade the x server, you need to recompile all drivers afterwards
orzelf: i did not update xserver. the howto says xserver>=1.6.2
orzelf: i have 1.7.1
agd5f: orzelf: should be fine
agd5f: orzelf: check dmesg
orzelf: agd5f: i've rebooted since then. But i have the /var/log/messages... nothing strange in it
mishehu: hi folks. I just installed slackware 13 on my machine, and I have a Radeon HD 4890 card. When I launch X it finds the first DVI connector without any trouble, but the second display is identified as HDMI-0 though I have nothing connected to it. The second DVI also has a monitor connected to it. I get video out DVI-0 but do not get anything out of DVI-1 when running at 1680x1050 (both monitors support this resolution)
mishehu: if I switch the resolution on HDMI-0 to 1360x755, then I get video out HDMI-0 but nothing out DVI-0
agd5f: mishehu: what version of the radeon driver are you using?
chithead: check "xrandr --verbose" output that they are using different crtcs
agd5f: mishehu: make sure you have 6.12.4 or git master
mishehu: agd5f: 6.12.2
mishehu: aa
mishehu: ok, I'll grab newer
mishehu: goes to build from source
mishehu: ugh this is going to be a pain it seems
orzelf: if i have radeon DRM built-in the kernel, but without kms, i wont have the drmradeonfb or whatever it's called, right ?
chithead: orzelf: correct
orzelf: chithead: ok, thx.
Tanktalus: this screen-update problem with xorg 1.7.1 is annoying :-(
twnqx: any idea what could cause "drmRadeonCmdBuffer: -22. Kernel failed to parse or rejected command stream. See dmesg for more info." - with nothing in dmesg?
soreau: Can you pastebin your syslog?
chithead: probably you are using drm kernel module which does not support r600 3d
twnqx: nothing in syslog either
twnqx: http://pastebin.com/d3125a9bf is actually the only thing besides [drm] Initialized radeon 1.31.0 20080528 for 0000:01:00.0 - and i guess that's too old, indeed
twnqx: the problem is actually not that opengl isn't working (glxgears gave that message), the problem is that aegisub is crashing, taking X with it in an unkillable state
twnqx: and i have to reboot to get rid of it :X
twnqx: sooo is there already any vanilla kernel with a more current radeon drm module?
twnqx: also... OpenGL renderer string: Mesa DRI R600 (RV635 9591) 20090101 TCL <- wold indicate that the kernel module is ok, wouldn't it?
chithead: no
twnqx: :\
chithead: you need the radeon drm from kernel 2.6.32-rc1 or later
twnqx: uh
twnqx: well....
twnqx: i guess i can't just drop that into a 2.6.31 tree?
honk: huggles drm-next
twnqx: shoots honk
twnqx: ebuild?
honk: Linux honk 2.6.31-rc9-27183-g273fad2
chithead: alternatively, you can use the fedora 2.6.31 kernel, it contains the necessary patches. or you can extract the patches from srpm and apply to your own kernel
honk: nah
wilco_: twnqx: check http://wiki.x.org/wiki/radeonBuildHowTo
twnqx: i read that
honk: http://git.kernel.org/?p=linux/kernel/git/airlied/drm-2.6.git;a=summary
twnqx: i just hate git.
honk: even kms works now ;)
twnqx: there's no need to have all obsolete code pieces around
honk: though it's missing some powermanagement ^^
twnqx: steals honks kernel
twnqx: well, off to install 2.6.32_rc5
spstarr: kms no work for r6xx models... yet
honk: huh?
chithead: kms works very well on my rs780
twnqx: umm the 2.6.32 should have it....
honk: works just fine on my (two) r700s, too
spstarr: locks up GPU for me at moment
spstarr: though i have not tried -rc5 + drm-next yet to confirm stil
mishehu: ugh. Ok, I've got ati radeon 6.12.4 drivers installed on my system, and I get both of my screens to output video. undortunately though, one of the displays only allows me a resolution up to 1360x765 instead of 1680x1050.
mishehu: the monitor does in fact support 1680x1050.
twnqx: for it with xrandr
mishehu: twnqx: is that the --fbmm parameter that I need to set or a new modeline to add to correct the problem?
twnqx: i had to override the detection on my displayport with a modeline
agd5f: mishehu: use xrandr to change the mode
twnqx: if the right resolution is already there (check xrandr -q) use that
UnNamed: mishehu: you define the mode with newmode if not listed, then you use addmode
UnNamed: mishehu: and finally use --output with --mode to make it the one to be used
twnqx: 2.6.31.5-00228-g0ac9a1d... wish me luck :X
twnqx: hm ok
UnNamed: mishehu: also, pay attention to maximum value reported by xrandr, you will need 3360x1050 or more
mishehu: grr I don't know why this getting wrong information for this monitor
mishehu: weird, I'm getting this in Xorg.log:
mishehu: (II) RADEON(0): Modeline "1680x1050"x0.0 146.25 1680 1784 1960 2240 1050 1053 1059 1089 -hsync -vsync (65.3 kHz)
mishehu: ?? at 0Hz refresh??
twnqx: OpenGL renderer string: Mesa DRI R600 (RV635 9591) 20090101 TCL DRI2 is where i'm now...
agd5f: mishehu: ignore that
agd5f: it's correct
twnqx: k, glxgears works now... let's see if aegisub still crashes X
UnNamed: mishehu: really, try xrandr and see what it says
mishehu: UnNamed: it certainly doesn't show a modeline for 1680x1050.
UnNamed: mishehu: and what is the max size?
mishehu: I need to find out how to make the modeline
mishehu: hmm sec
soreau: mishehu: cvt 1680 1050
mishehu: soreau: hmm never knew about that cvt program heh
soreau: xrandr --newmode
twnqx: i always google them :P
mishehu: alright, let me try it out
mishehu: twnqx: I was about to do that
mishehu: crap I got called down to dinner
mishehu: will have to work on this after dinner
mishehu: thanks for the help thus far
twnqx: aegisub doesn't take down X any more, and glxgears works...
twnqx: however i have segfaults in _mesa_texstore_argb8888 :\
agd5f: twnqx: shoudl be fixed now in git
twnqx: k, my mesa is from 27/09/09 :P
twnqx: configure: error: C compiler cannot create executables
twnqx: <3
chithead: twnqx: probably you messed up your CFLAGS or worse
twnqx: no...
twnqx: /usr/lib/gcc/x86_64-pc-linux-gnu/4.4.2/crtfastmath.o:(.ctors+0x0): undefined reference to `no symbol'
dmb: interesting....
twnqx: yes :P
dmb: your compiler is messed up dude
twnqx: it just compiled boost, xulrunner and firefox just fine.
twnqx: removes -ffast-math
twnqx: grml where does that actually come from
chithead: probably all software you built with -ffast-math will need rebuilding
twnqx: i luckily didn't build anything, it's only in the mesa-git ebuild
twnqx: /tmp/portage/portage/media-libs/mesa-9999/work/Mesa-9999/conftest.c:42: undefined reference to `dlopen' is more likely the real reason now
twnqx: ah. well...
twnqx: need newer libdrm, too
twnqx: git breaks all automatic version checking :(
Ralesk: never know what is newer and what is older? :)
twnqx: never know the actual version
UnNamed: look at tags?
twnqx: i bet -mfpmath=sse collides with -ffast-math...
dmb: twnqx, yeah, dlopen is in core glibc i believe...
dmb: shouldn't be undefined
twnqx: no, in -l dl
dmb: ok, even more corer
twnqx: but i'm stuck at expat now
dmb: get a core i7 :)
dmb: things compile faster
twnqx: i have one... in my other place
twnqx: also, all my desktops have nvidia cards...
UnNamed: dmb: what credit card number? so i can charge one to your card :]
dmb: :)
twnqx: strings /usr/lib64/libexpat.so.1.5.2 | grep XML_ParserCreate *whine* it is there....
twnqx: tries with gcc 4.3.3
twnqx: so my gcc 4.4.2 is slightly damaged...
twnqx: nooooo
twnqx: that breaks all the programs that were - successfully - compiled with 4.4.2
twnqx: head -> desk
chithead: your system is seriously messed up
twnqx: no
twnqx: mesa sneaks the -ffast-math in
twnqx: that's the only broken part of it
soreau: I think it is safe to say if it's working for everyone else except you, the problem is somewhere on your end
twnqx: well indeed
twnqx: but my compiler only has problems with -ffast-math :P
twnqx: and ater removing that from mesa's configure it compiles fine now.
chithead: many applications use -ffast-math, including mesa, mplayer, ffmpeg, etc.
twnqx: it's 1:40 in the morning
twnqx: not going to recompile gcc _now_
twnqx: k, the segfault moved... to radeonTexSubImage2D -> memcpy
twnqx: *sigh*
twnqx: and you're right, mplayer doesn't build, and after an update erarlier it is missinf some libs
twnqx: i don't want to go to bed, feeling like i lost against a computer :X
meoblast001: configure.ac:33: error: must install xorg-macros 1.2 or later before running autoconf/autogen
meoblast001: where might those be found?
rehabdoll: http://xorg.freedesktop.org/releases/individual/util/
rehabdoll: but your package-manager should provide it
meoblast001: rehabdoll: what's the package name on say, ubuntu?
rehabdoll: no idea
biotube_: should be xserver-xorg-dev
meoblast001: biotube_: do you know where it would put the macros?
meoblast001: i would guess /usr/share/xorg but i'm probably wrong
biotube_: meoblast001: should be whereever they belong
biotube_: on debian I have no problems building
biotube_: apt-get build-dep xserver-xorg-video-radeon
meoblast001: biotube_: but i need this macro in the top level directory of my build
rehabdoll: its just one file: /usr/share/aclocal/xorg-macros.m4
rehabdoll: on my system
meoblast001: thanks
meoblast001: time to build this :)
meoblast001: i think my drivers are old
meoblast001: everything is really buggy
meoblast001: and any complex 3D application makes the whole xserver crash
mishehu: soreau: ok, I do I did the xrandr --newmode with the output of cvt. I see that it added it to the wrong screen when I just run xrandr by itself. and whenever I try to do xrandr --output HDMI-0 --mode 1680x1050_60.00, it complains that mode isn't found.
meoblast001: so after i make install, i should have the latest version installed?
UnNamed: mishehu: add to the right output
meoblast001: i'll try
soreau: mishehu: You need to do xrandr --addmode
UnNamed: mishehu: xrandr -h gives you a brief list of options, including how they relate to each other (--auto goes after --output)
meoblast001: :(
meoblast001: drivers crashed again
meoblast001: i think the card itself was having issues because my monitor went into sleep mode
twnqx: yey, it's the result of building gcc with installsources -_-
meoblast001: twnqx: ?
twnqx: i was getting rid of an annoying udev bug this afternoon... and a change i did to my system to get usable debug information broke building of my new gcc >_>
meoblast001: ooh
meoblast001: i think i'm starting to get the picture of why there are no free software 3D applications
airlied: wonders what blender is
meoblast001: well, why the amount is so low
UnNamed: airlied: i was going to say that... but then i wonder why driver writers never test those
twnqx: and that in turn broke compilation of mesa
UnNamed: runs
meoblast001: half the free software community doesn't even have 3D acceleration :/
airlied: UnNamed: driver writers generally don't get to test much
twnqx: airlied: try to get radeon to work aegisub -_- i guess i tripped a doeznd bugs from between "app segfaults" and "app breaks X, turning it into unkillable state"
twnqx: to work with*
meoblast001: any who, does anyone know why an.. uh... X1300 (i think that's the name of the card) would crash on any complex 3D application
UnNamed: meoblast001: most i know just use propietary drivers (so one would say the apps side is about the zero cost)
meoblast001: Blender, Extreme Tux Racer, and 3D applications i write are the only things that actually work
airlied: meoblast001: how does it crash? hard hang machine?
airlied: or segfault?
meoblast001: well, it depends
mishehu: BAM! kick it up a notch
meoblast001: when i tested Cube2 the other day, the screen just hung still, and the system became unresponsive
mishehu: squeezes a spice weasel into his display
mishehu: soreau: ok, got it.
mishehu: I suppose I'll have to go through this routine everytime I launch X then. :-/
soreau: mishehu: Not necessarily
meoblast001: i just updated my drivers (i think) and Super Tux Racer causes the screen to go into sleep mode (meaning the card isn't putting out a signal after that)
soreau: mishehu: You could use the modeline in xorg.conf or automate the process by putting those commands in a script to run after you start X
soreau: also, you could ask the devs here if it would be worth filing a bug
twnqx: wonders why 1920x1080 devices on displayport, while it worked, where given 1920x540 modelines
soreau: Or if your monitor just sucks ;)
mishehu: soreau: nah, it's a good monitor.
soreau: so file a bug about it
twnqx: http://nopaste.info/410e2da40f.html :(
mishehu: soreau: I might just do that
UnNamed: twnqx: 2*540=1080... concidence? interlaced?
twnqx: no, they didn't work
twnqx: i had to manually override it with a 1920x1080 modeline
Jonimus: I keep getting this error "AIGLX: Screen 0 is not DRI2 capable" when trying to enable the exprimental 3d on my HD4670, I have the updated drm module installed and the mesa pkg's from git as well
twnqx: and a drm-next kernel?
mishehu: what's a drm-next kernel?
UnNamed: airlied: i was kidding in part, and the other part pointing that most of focus seems to be games
agd5f: twnqx: 1920x540 is 1080i. the xserver edie parser didn't handle thos correctly until recently
agd5f: *edid
airlied: UnNamed: most of my focus is compiz or gnome-shell
UnNamed: airlied: that seems to be recent
soreau: twnqx: See the wiki link in the topic
soreau: It is a branch were most of the kernel side radeon work is stored
twnqx: *i* know that,
Jonimus: anyone have any idea on what is causing my issue?
Jonimus: it was working until a day or two ago :(
twnqx: do_wait: drmWaitVBlank returned -1, IRQs don't seem to be working correctly. <- is that actually normal?
biotube_: twnqx: yes, the code was written for them only recently
Ghworg: I've got more games running now than I did when I used to use fglrx. Of course some of that is due to wine having got better in the meantime I bet
twnqx: radeon gives me a display, fglrx a black screen... i should be thankfull for that :X
mishehu: fglrx was a heaping pile of monkey feces, imo
Jonimus: anyone?
mishehu: I got several gray hairs from trying to us it
Jonimus: od I need to give mroe info, if so tell me what?
mishehu: Jonimus: can't help you man, I just got back into using the opensource driver as it is
twnqx: Jonimus: my displayport was working (except for audio), and next update nothing would be detected any more... shift happens.
Jonimus: mishehu, agreed, and on this laptop I need it to work or I can't finish my school work.
Jonimus: darn
twnqx: next laptop WILL have an nivida graphics chip...
Jonimus: I guess I'mm be building the rc kernel and hoping it fixes it :/
Jonimus: ok my next question is, is there a way to access the i2c ports on the card?
Jonimus: one of my monitors the ddc controls are the only way to access some settings so I need to point ddccontrol at the i2c port of my card but three is no /dev/i2c
twnqx: ok, gcc is fixed -_-
Jonimus: ahh I found a more descriptive error "(EE) AIGLX error: dlopen of /usr/lib/xorg/modules/dri/r600_dri.so failed (/usr/lib/xorg/modules/dri/libdricore.so: undefined symbol: driDispatchRemapTable)
Jonimus: (EE) AIGLX: reverting to software rendering"
chithead: Jonimus: your mesa is not compiled with r600 support
chithead: you will only have 2d
Jonimus: I'm building it from -git so it should :/
twnqx: my mesa is defective.
twnqx: but that's left for tomorrow. good night.
chithead: Jonimus: you probably did not enable r600 in the configure options
Jonimus: k, I'll take a look and try again
chithead: or you installed mesa to another place than /usr/lib
Jonimus: I'll rebuild libdrm-git and try again, I'm using a split mesa
chithead: libdrm-2.4.15 should be enough
Jonimus: chithead, I'm useing everything else from git so I'll just stick with the -git version
Jonimus: but thanks for the help
spstarr: builds mesa git master
spstarr: im doing daily builds if r6xx or i915 changes in trunk
cxo: How come the gallium interface driver for r600 to r800 is the same, but they all have different mesa drivers?
airlied: cxo: they don't have different mesa drivers
airlied: there is only r600 for those
cxo: ok it just looked weird, http://www.x.org/wiki/RadeonFeature
cxo: So has anyone tried an r800 out?
airlied: there is not support for it ye
airlied: yet
airlied: alex has/is getting hw
cxo: How does that work, if it uses the same mesa driver?
airlied: you still have to add support to the driver
cxo: you mean kernel <-> hardware?
airlied: also not only mesa, need to add kernel/ddx support first
airlied: you add support in 3 places
MostAwesomeDude: Arg, that page is irritating me.
cxo: too many yellow blocks, not very pleasing to the eye
airlied: MostAwesomeDude: remove the rhd colums ;-)
cxo: hurry up and make all those yellow blocks, green
MostAwesomeDude: airlied: Stop readin' my mind. :3
MostAwesomeDude: I think I'm going to; it's not like it's in the katamari.
cxo: So how come radeonhd and radeon are still separate?
cxo: I thought that one of them was just going to fall back once certain things were done
airlied: cxo: the first rule of #radeon is don't ask
cxo: way too much politics in open source development
airlied: therer is way too much in all development
airlied: I don't think open source has proportinally more or less, just you can see it
cxo: specifically in foss, because everyone has an opinion, though there still could be a boss telling what to do, its not the same when you're getting paid to do it
airlied: getting paid and open source are in no way mutually exclusive
cxo: Of course not, it just alters a few attitudes
airlied: considering everyone in the raedon vs radeonhd saga was full time employed
airlied: and had management
cxo: then obviously everyone was trying to push their own agenda
airlied: and that happens in both open and closed source world
airlied: as I said you just get to see it
cxo: perhaps you're right, its just more visible with open source
cxo: I just think that if no one is getting paid, there would be more opinion pushing going on, but eventually, due to shear need for collaboration, everyone would be on the same boat. But if someone was pumping money into the project from one end, collaboration with 3rd party devs would still be important, but doesn't become vital anymore
cxo: in other words, people are more likely to put aside their differences, when they are forced to work together
airlied: I don't think pay comes into it near as much as personality
airlied: though money can be useful in personality altering