MostAwesomeDude: Shaders always operate on four-component vectors of single-precision floats.
cxo: like how does the gpu access its memory, it needs to address it somehow right? what are those sizes?
spstarr: airlied: ok im in Fedora, still installing components...
spstarr: I dont know why Fedora has issues with my brightness kI know Fedora 12 is old now
spstarr: so maybe .33-rc8 fixes since Debian it didnt have issues (maybe userspace)
spstarr: time to install KDE group then start X
spstarr: and test GPU lockup with Fedora 'rawhide/13' kernel
spstarr: brb
MostAwesomeDude: cxo: Ah, that varies. Usually 128-bit or 256-bit bus.
airlied: spstarr: btw I just kicked off a kernel in koji with all fixes in it
airlied: 2.6.32.8-55 is building
cxo: Are radeon big or little endian?
airlied: little
DanaG: wishes somebody would make an ARM device with an ATI GPU of some sort.
cxo: ATi cards would have to wery enewgy efficient
airlied: dmb: imageon was that
airlied: or something close
airlied: maybe it was mipssy
dmb: i think you mean DanaG :P
MostAwesomeDude: Huh. So I just got back from an industry reception.
MostAwesomeDude: Barco was there. Apparently they use ATI/AMD stuff for their hardware.
DanaG: Barco? Sounds familiar... POS? (point of sale)?
MostAwesomeDude: And they "work with AMD to enhance the graphics drivers for features such as enhanced color depth."
DanaG: wishes the mobility FireGL / FirePro cards had stereo connectors.
MostAwesomeDude: So an internship there would, as an educated guess, be an fglrx hackfest. :C
spstarr: airlied: in Fedora now
spstarr: airlied: it's good to be back :)
spstarr: 2.6.33?
spstarr: 2.6.33-0.44.rc8.git0.fc13.x86_64
airlied: spstarr: oh I should do a rawhide build as wel
airlied: I only did F13
airlied: F12
spstarr: oh, ok
spstarr: im gonna build my git kernel on laptop now and reproduce GPU lockup
spstarr: with stock Fedora KDE/mesa/libdrm/ddx
spstarr: rawhide
DanaG: hmm, I wonder if my hdmi port would support 10-bit color.
spstarr: brightness keys work in X, but not in console
spstarr: very odd...
spstarr: i dont know yet if its a kernel option or userspace setup doing that
Animeking: graphics cards, and energy efficiency in the same breath, IMPOSSIBLE
Animeking: seriously though, the only current cards i see energy efficient is the new 5xxx series, and even that takes up a lot of energy ( i think)
spstarr: rawhide anaconda == major fails.. wasted 2 cds :/
spstarr: the iso glisse told me to get weeks ago failed with python exception, today's build blew up formatting disk (LVM said partition was in use)
spstarr: so F12 -> F13 :)
DanaG: hmm, how about getting something on par with, oh, the 9800 series, but massively shrunk?
spstarr: hearts LZMA
Nightwulf|work: hi all
airlied: most of the eneregy is getting to the RAM
airlied: really fast RAM needs lots of power
DanaG: I'm also curious about Marvell's new thingy that can do 1080P and 3D at the same time... but no discussion if it can do 1080p WITHIN 3d.
DanaG: =þ
spstarr: building my existing kernel source
spstarr: im not going to pull any updates with my kernel just use what I had last.
spstarr: hmm
spstarr: rpm --target - unknown option
spstarr: i suspect kernel.org's scripts/package/Makefile is outdated for Fedora 12+
spstarr: rpm used to have -ta
spstarr: (make an RPM from a tarball)
thansen: anyone in here know if ati driver works with.. http://www.notebookcheck.net/ATI-FirePro-M7740.18791.0.html
spstarr: it might help if i had rpm-build installed :)
thansen: from what I can tell it shouldn't be a problem
spstarr: ok, now its compiling kernel
Pallokala: spstarr: it should work AFAIK if you have recent-enough distro
spstarr: it does
spstarr: was missing rpm-build stuff
Pallokala: oh, sry I meant to talk to thansen
thansen: Pallokala: yeah, I run gentoo
thansen: I asked earlier, but was unclear whether or not I'd be able to power an external 2560x1600 screen
spstarr: almost done compiling
agd5f: thansen: the hw is capable, depends on te connectors used on the laptop
thansen: agd5f: I'll have a full display port setup
agd5f: i.e., if the laptop only has a single link dvi port, you'll be limited to single link modes
agd5f: thansen: DP should be fine
thansen: the question that arose was about some 2048 limit in mesa or something silly
agd5f: thansen: passive DP to DVI adapter may be an issue
agd5f: but native DP should be fine
agd5f: thansen: the 2k render limit was on older chips. t6xx+ have 8k render/texture limits
agd5f: r6xx+
thansen: it'll be pure...precision m6500 with the new dell 27" 2711 that has dp connector on it
thansen: k
spstarr: /root/rpmbuild/RPMS/x86_64/kernel-2.6.33rc8-1.x86_64.rpm
spstarr: ok
spstarr: now let's see if i can GPU lockup with it
thansen: agd5f: and I presume it should be able to drive a 3d desktop at that res without any issue whatsoever
agd5f: thansen: yes
spstarr: bbr
thansen: is tired of laggy intel X3100 and old ati X1400
spstarr: hmm
spstarr: airlied: seems to fail to boot kernel
spstarr: either the initramfs is fscked or something else
DanaG: wait, m6500... already have it?
DanaG: what's the GPU in that?
spstarr: the firmware is in the initramfs
twnqx: morning
twnqx: i'm still to incompetent to use radeon-drm-next :(
rah: who is Andrew Randrianasulu?
AndrewR: me
spstarr: hehe
spstarr: gah
spstarr: darcut is using WRONG kernel
spstarr: lol
spstarr: no wonder this is fscking up
spstarr: 2.6.33-0.44.rc8.git0.fc13.x86_64 = 2.6.33-rc8 (mine)
spstarr: !=
rah: AndrewR: I've responded to your question in bug 26582; I expect furtherance of the discussion into the lands of debugging
AndrewR: rah, it looks like your 2nd attempt still detect monitor, not TV (105.947] (II) RADEON(0): Output DVI-1 using monitor section Dell 2408WFP
AndrewR: ). can you un-plug this Dell ?
rah: AndrewR: then there would be no monitors plugged in except the tv
AndrewR: rah, hm ..... but waht you have connected as DVI-0 ?
AndrewR: rah "105.885] (II) RADEON(0): Output DVI-0 using monitor section HP L1906
AndrewR: "
AndrewR: Rah, is this just one monitor connected via two cables?
rah: AndrewR: sorry, I'm a little confused; when you say "2nd attempt", what do you think I'm attempting?
rah: AndrewR: there are 2 X logs now; the first is without the VGA monitor connected, the second is setting the 'tv standard' property after the DIN output has been enabled (with the VGA monitor connected)
AndrewR: rah, first time you ran with just TV. Second time - with all two (or three) screens connected .... At least X thinks you have something on DVI-1
rah: AndrewR: right
AndrewR: Rah, can you try without any xorg.conf?
rah: no
rah: no, sorry I could
rah: but why would I?
rah: what difference would that make?
AndrewR: rah, not sure, sometimes user-defined things interfere wih autodetect in starnge ways
rah: I don't understand what you mean by "autodetect", or what that has to do with the 'tv standard' property not working
AndrewR: rah, driver can "sense" if there something connected to output. Sometimes it works ok. sometimes users disable this feature (due to bugs) and define things statically in xorg.conf. Mixing two ways of defining outputs may confuse driver
rah: *shrug*
AndrewR: hm .. so HP is monitor, and DELL is TV....
spstarr: glisse: if I can crash it i will give you my .config
glisse: spstarr: i hope it won't crash with X from rawhide
AndrewR: dell is monitor too .... http://www.tftcentral.co.uk/reviews/dell_2408wfp.htm
spstarr: it locked up
spstarr: glisse: same thing
spstarr: screen bleeded to white
spstarr: X version:
spstarr: X.Org X Server 1.7.99.901 (1.8.0 RC 1)
spstarr: Release Date: 2010-02-12
spstarr: Build Date: 16 February 2010 06:22:48AM
spstarr: Build ID: xorg-x11-server 1.7.99.901-2.20100215.fc13
spstarr: glisse: so I can reproduce the lockup on Fedora with stock mesa/libdrm and my kernel
spstarr: i will record video for you?
spstarr: stock ddx also
spstarr: glisse / airlied: with kdm, when logging in X dies and restarts once then logs in properly
spstarr: i noticed this in Debian also but it got fixed later on
airlied: spstarr: try with the stock rawhide kernel also
spstarr: airlied: ok one moment rebooting to latest .33 you have
spstarr: unless yours got built
spstarr: 2.6.33-0.44.rc8.git0.fc13.x86_64?
spstarr: or did build a newer one?
spstarr: yums u[
spstarr: up
spstarr: checks koji
airlied: not yet
spstarr: ok i will use my existing .33
spstarr: brb
rah: AndrewR: starting with an empty xorg.conf means I get no input devices
rah: AndrewR: but aside from that, it changes nothing
rah: AndrewR: the 'tv standard' property behaves exactly as before
spstarr: ok in
spstarr: nothing in syslog trying with fedora rawhide kernel...
spstarr: recording it also
AndrewR: rah, can you attach xorg.conf too?
airlied: wonders if pal is just broken
airlied: not sure how much r6xx tv out has been tested
rah: urg
rah: AndrewR: the 'tv standard' property behaves exactly as before
rah: AndrewR: what now?
glisse: daaammmm
AndrewR: rah can you look into /sys/class/drm ?
taiu: glisse: want to talk about it ;)
AndrewR: rah, false alarm, all files there just r-r-r (none writable)
rah: AndrewR: so what now?
spstarr: yes
spstarr: glisse / airlied: I can crash Fedora kernel, the only difference is instead of it fading to white, it goes black, I have it on video from two times I tried.
glisse: spstarr: does the kde konsole circling bug ?
spstarr: glisse: the trick is you *must* have composite OFF before you start X
spstarr: ie, make sure kwin has composite off, then start X, turn it on load 4 konsole windows and move mouse around over and over..
spstarr: it will eventually lock
spstarr: glisse: yes
thansen: DanaG: firepro M7740
spstarr: lemme put videos up.
thansen: and no, I'll be ordering in the next few days (supposedly an update is coming with usb 3.0)
DanaG: ah. HP's coming out with something with M5800 of some sort.
AndrewR: rah, not sure, i was hoping there was some writable files...sometimes even xrandr (small app) has bugs .... But, because all files there just r/o ... you can check some things ther, i think. But not change them :/
DanaG: Right now, the new EliteBooks are NV-only (bleh), but supposedly the ATI is coming soon.
DanaG: Eyefinity.... awesome.
spstarr: nothing in kernel log
glisse: obviously here it won't lock ..
DanaG: Keep an eye on hpfansite.com.
thansen: DanaG: HP's linux support has always been sub-par to me
thansen: probably would be fine though
DanaG: hmm, consumer and business are way different, though.
spstarr: glisse: putting up videos.
rah: oh well :(
thansen: DanaG: could be, I always hated the locked down bios for wifi cards etc
rah: no KMS for me :(
thansen: DanaG: I've been considering the Envy 15, but it doesn't have displayport on it for some reason
DanaG: My 8530w is locked down to Intel 5300... but that's not much of a restriction.
thansen: not really, but still silly
DanaG: yeah.
DanaG: Bigger issue would be wwan.... I can't "abuse" that slot for any other USB devices.
thansen: looks at 8530...
twnqx: DanaG: hotplug :P
DanaG: Oh, and the notebooks also have UEFI firmware, though the one I have has some UEFI-mode brokenness that's resolved by using BIOS-emulation mode.
spstarr: installs mencoder
twnqx: spstarr: while you're at it, could you answer my pm? :P
DanaG: And supposedly that's fixed in the new ones.
spstarr: im kinda busy at moment...twn
DanaG: Speaking of UEFI... how does radeon KMS interact with efifb?
thansen: DanaG: yeah, they're giving me a pretty good quote on the m6500 through my business account so I'll end up going with it
DanaG: eh. check the prices to compare, at least.
thansen: goes up to 16GB Ram and has dual hdds
DanaG: Dual HDD and CD too? spiffy.
thansen: yeah
thansen: goes to find the bed...
AndrewR: rah, you may try to hack driver itself (to see if PAL wors at all)
spstarr: glisse: videos ready
spstarr: glisse: http://www.sh0n.net/spstarr/radeon/rawhide-r600-crash1.avi
spstarr: glisse: http://www.sh0n.net/spstarr/radeon/rawhide-r600-crash2.avi
DanaG: eh, m6500 looks like a brick. =þ
spstarr: glisse: the crash #2 happened even easier, and i dont know why
spstarr: so we can confirm it affects stock rawhide
rah: is there a graphics card that has complete support with a free driver under Linux?
MostAwesomeDude: "Complete?"
mjr: complete is a fuzzy word
MostAwesomeDude: tdfx is pretty close.
rah: "pretty close" is not really "complete" is it?
rah: mjr: I don't see how
mjr: I'm pretty sure any basic VGA adapter is complete
MostAwesomeDude: Well, I have no 3dfx HW to test with.
MostAwesomeDude: But I think tdfx is complete.
MostAwesomeDude: cirrus is complete, if you're okay with classic FB.
rah: mjr: can you give me an example of a "basic VGA adapter"?
mjr: though of course, one might claim not since there's no support for funky things like Mode X in the X server
MostAwesomeDude: rah: cirrus
mjr: so, no, no complete support for even basic VGA adapters :I
AndrewR: rah, look into radeon_encoders.c in drivers/gpu/drm/radeon (kernel tree). May be you will find something or at least force driver into Always_pal mode .....
rah: MostAwesomeDude: I didn't know there were any cirrus adapters on the market
octoploid: spstarr: Maybe you should switch on some kernel debugging options like: PROVE_LOCKING, DETECT_SOFTLOCKUP, DETECT_HUNG_TASK, ...
MostAwesomeDude: rah: QEMU
rah: MostAwesomeDude: what adapter are you thinking of?
MostAwesomeDude: rah: QEMU's VGA chipset is a cirrus.
mjr: and by extension, since afaik modern cards carry vga compatibility still (corrections?) no card has "complete" support
rah: MostAwesomeDude: QEMU is not hardware
MostAwesomeDude: rah: Oh noes!
MostAwesomeDude: But, y'know, you asked the question, I answered.
spstarr: octoploid: GPU wedges dont usually record oops
rah: MostAwesomeDude: I don't think you did
MostAwesomeDude: rah: Oh, you're right. "Yes."
mjr: yeah but I did
rah: MostAwesomeDude: which card?
mjr: "no, because obscure old VGA features are not really used by anything in the Linux/X.org driver stack"
octoploid: spstarr: Just an idea
mjr: thus not complete
mjr: and see why I said complete is a fuzzy word now?
MostAwesomeDude: But that doesn't really count, since nobody uses those obscure old VGA features.
rah: mjr: "graphics" does not necessarily imply "VGA"
glisse: ok definitly can't reproduce it
glisse: i am done with lockup on r6xx/r7xx until i can actualy reproduce it
spstarr: gah
spstarr: airlied: you have that W500? :)
glisse: he doesn't but someone in the office has one
glisse: so he has to bribe people to install kde on it
spstarr: ugh
AndrewR: glisse, with atombios, everything low-level about tv-out setup handled by AtomBios itself? So, if rah just changed default mode - it have chances to work (display picture)? Or it was wrong idea ...?
glisse: AndrewR: he can force pal in the driver note that on some gpu you need to move a switch on the board to switch btw pal or ntsc
glisse: even thought i don't think any recent gpu has such switch anymore
mjr: rah, oh, you want to go older? I wouldn't know if Linux/X.org has full support for the original monochrome Hercules, that's true, but as said, vga is kinda implemented by everything modern still at least in the desktop space
mjr: MostAwesomeDude, ah, but htat's "as complete as anyone cares" which is very fuzzy :)
rah: I'm not interested in hacking the driver
mjr: yeah, I'll stop now
rah: scratch that
rah: I don't have time to hack the driver
AndrewR: glisse, i can't see any option like tv-norm in modinfo radeon output ... and sysfs has only r/o files ....
AndrewR: glisse, nouveau has it tv_norm
glisse: it's somewhere
glisse: can't remember what the name is
glisse: never used tv-out
AndrewR: glisse, may be it was dropped from drm-radeon-testing? I can see only tv (enable/disable)
AndrewR: glisse, code seems to create drm_property for tv_norm, but ..... how to access it?
obscurant1st: howhave the radeon kernel module load with modeset=1 in ubuntu 9.10 fresh installation?
obscurant1st: somebody?
soreau: I sent obscurant1st here because I can't remember what file you have to edit in /etc on ubuntu to have radeon load with modest=1
glisse: /boot/grub/menu.lst
obscurant1st: but does grub2 use menu.lst?
soreau: No, it does not
soreau: obscurant1st: AFAICT, you create /etc/modprobe.d/radeon.conf and add the contents "radeon modeset=1"
soreau: but I could be wrong
obscurant1st: ok np, as far as i can boot into ubuntu.. :D
obscurant1st: i can right?
soreau: yes
obscurant1st: :D
soreau: obscurant1st: If you do try it, of course don't add the kernel parameter and let me know if that works
obscurant1st: ok, ll let u know
soreau: You will know because 'glxinfo|grep renderer' will say DRI2
soreau: and cairo-dock will be working without the black and flickering of course
cire: hmm, checking dmesg, I get "You have old & broken userspace please consider updating mesa"
cire: I have drm from radeon-testing, ddx and mesa from master
DanaG: soreau: actually, there are two ways:
DanaG: as a kernel parameter, it's radeon.modeset=1
DanaG: as a modprobe.d/whatever.conf file, it's "options radeon modeset=1"
spstarr: hmm
spstarr: glisse:
spstarr: uh oh
spstarr: [ 3594.329412] radeon 0000:01:00.0: z/stencil with no depth buffer
spstarr: [ 3594.329416] radeon 0000:01:00.0: r600_packet3_check:1055 invalid cmd stream 142
spstarr: [ 3594.329418] [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !
spstarr: [ 3594.336703] Querying an unemited fence : ffff88001373ca40 !
spstarr: fsck
spstarr: I started firefox.. noticed some text glyph corruption, the kerneloops applet popped up then
spstarr: this is with my kernel with your fixes too :(
spstarr: i note, this is with default fedora ddx/libdrm/mesa
spstarr: so unless trunk has some newer stuff, i did not see this before when in Debian
spstarr: the fence issue... i want to try today to reproduce once i finish rebuilding my $HOME dir and compile second life.
spstarr: ok i see
spstarr: the glyph corruption is triggering the error
spstarr: [ 3989.334668] radeon 0000:01:00.0: z/stencil with no depth buffer
spstarr: [ 3989.334671] radeon 0000:01:00.0: r600_packet3_check:1055 invalid cmd stream 142
spstarr: [ 3989.334674] [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !
MrCooper: more likely the other way around, as a rejected command stream means the rendering encoded in it is dropped on the floor
spstarr: ah
spstarr: but i never encountered that with trunk libdrm/mesa/ddx and X server 1.7.4
spstarr: Fedora is using X 1.7.99.x
spstarr: so it might be that I just never saw it because of newer X server, I suppose.
spstarr: glisse: you are running rawhide right?
spstarr: glisse: I am using rawhide X server, mesa/libdrm/ddx
obscurant1st: soreau, u there?
obscurant1st: after restarting wht should i check?
glisse: spstarr: yeah using rawhide
spstarr: ok
spstarr: glisse: hmmm
spstarr: glisse: so now im seeing new stuff I didn't see before
glisse: i can't reproduce the z/stencil error here neither
glisse: was it with composite on ?
spstarr: no
spstarr: plain kwin w/o composite
spstarr: i had firefox opened
spstarr: went to about:plugins to look at what plugins were there (after I installed flash DSO)
glisse: oh so it's likely flash related
glisse: flash use gl
spstarr: then i saw some glyphs were corrupted (showing chinese or japanese unicode)
spstarr: well i didnt load a flash website
glisse: weird
spstarr: just checked if the plugin was found by firefox
spstarr: now the kerneloops thing is flooding me with this over more
spstarr: glisse: this is using the same kernel code i used in Debian before
spstarr: I just recompiled it on fedora but didn't git pull
soreau: obscurant1st: cairo-dock?
obscurant1st: yep,
obscurant1st: i mean its working
obscurant1st: fine
soreau: obscurant1st: If 'glxinfo|grep -i dri2' gives any output, it worked
obscurant1st: but u asked me to add something to radeon.conf right
obscurant1st: i hv added it
obscurant1st: now wt to ceck after restarting?
soreau: lol
spstarr: sleep time
obscurant1st: it didnt work actually
obscurant1st: soreau, i hd to do the -c thing
obscurant1st: :(
soreau: obscurant1st: Oh well, I was hoping someone here knew how to do specify modeset=1 for the radeon module through the file in /etc.. but the other way to make it permanent is to add it to the boot loader. See /msg ubottu grub2
soreau: or ask in #ubuntu
soreau:
soreau:
soreau:
soreau: obscurant1st: I missed DanaG's comment from earlier^^
soreau: So I missed the 'options' part at the beginning it seems
obscurant1st: so i hv to add options radeon modest=1
obscurant1st: instead wht u gv me?
soreau: yea
obscurant1st: oh k
soreau: obscurant1st: Make sure it's modeset=1 and not modest
obscurant1st: options radeon modeset=1
soreau: correct
obscurant1st: this is wt i gv in radeon.conf
dileX: soreau: http://bgoglin.livejournal.com/19346.html
soreau: dileX: That's not very descriptive WRT a file name, but thanks
soreau: I guess it would have to be understood that you put it in a .conf file in that dir
obscurant1st: soreau, its not working.. :(
soreau: huh
soreau: obscurant1st: Well i guess you will have to ask in your distributions support channel
obscurant1st: yeah, am doin it.. :D
obscurant1st: thx for helping though,
obscurant1st: u made it working.. :D
obscurant1st: soreau, cn u pls tell me wht was that parametr again if you dont mind?
obscurant1st: boot parameter?
dileX: soreau: add new line "radeon" to /etc/modules and (create new file) /etc/modprobe.d/radeon.conf have to contain "options radeon modeset=1"
MrCooper: glisse, spstarr: as a shot in the dark, do you guys have sub-pixel text anti-aliasing enabled? spstarr, does Option "EXANoComposite" or "EXANoDownloadFromScreen" or "EXANoUploadToScreen" work around the problem?
dileX: soreau: I agree - description in that blog could be a bit clearer
dileX: soreau: fee free to comment :-)
dileX: feel*
AndrewR: feels very stupid about modern TVs ....
soreau: obscurant1st: The kernel param to tell radeon to load with modesetting is radeon.modeset=1
soreau: but please see what dileX said above^^
obscurant1st: i tried it with no luck
obscurant1st: should i try the same with option radeon.modeset=1
obscurant1st: ?
soreau: dileX> soreau: add new line "radeon" to /etc/modules <-- Did you do this obscurant1st ?
obscurant1st: /etc/modprobe.d/radeon.conf this line is there, i added it
obscurant1st: :(
soreau: obscurant1st: There's two different things you need to do apparently
soreau:
obscurant1st: yeah i hv done both, soreau
soreau: Well I don't know what else to tell you other than to ask in your distros support channel
obscurant1st: i hv asked there, no help till now.. :(
soreau: I doubt removing vowels from your words is helping :P
obscurant1st: :|
dileX: obscurant1st: could depend on distro and its kernel
obscurant1st: its ubuntu 9.10 fresh install dileX
dileX: for example you could test in grub-line which param you need
obscurant1st: how to test it?
dileX: reboot and test in grub(2) bootmenue
obscurant1st: dileX, there is a file which says we can add custom entries to it, so how can i get the full entry for booting my ubuntu, i am tired of restarting, :(
dileX: unloading radeon kernel-module after loaded with kms-support is tricky (see topic radeon-buildhowto), e.g. unbind vt-console
AndrewR: found some info about connectors .... http://en.wikipedia.org/wiki/Mini-DIN_connector
rah: AndrewR: what next?
AndrewR: scratching head .... and looking at bug again
glisse: MrCooper: well i have been through ddx code and i am pretty sure we always emit the 3d state
glisse: this is why this message puzzle me
glisse: i didn't notice a code path where we were lacking this initial setup
glisse: need to recheck the blit/video path thought as i have overlooked them
rah: AndrewR: why do you think this function returns prematurely?
rah: (ie, radeon_connector_set_property() )
AndrewR: rah, i don't know. If they run normal - something wrong up there .. (X, libdrm .... ?)
rah: AndrewR: what next?
AndrewR: rah, not sure, i can't even find files where driver send xommands into drm .... need some sleep/
AndrewR: rah, look into drmmode_display.c in xf86-video-ati, but this is only beginning ... well mid-way down to kernel.
AndrewR: rah drmmode_output_set_property(), drmModeConnectorSetProperty() ....
rah: saying function names isn't of much help
rah: what am I supposed to do with these functions?
rah: AndrewR?
AndrewR: rah, try to ptint out values ...i'm not sure where this drmModeConnectorSetProperty located ....
rah: AndrewR: what values; for what purpose?
AndrewR: rah, if you are shure hardware driver (-ati) receives info from X server and send it down to kernel correctly ..... not sure, try your own ideas .....
rah: AndrewR: I am not sure of anything; I have no ideas
kdekorte: glisse: are you here?
kdekorte: glisse: ping when you get in, I found how to create the z/stencil crash... use gtkperf -c 500 and then choose the GtkDrawingArea - Circles test
AndrewR: rah, with your modified kernel - does it print newly added by you lines DRM_DEBUG at xrandr prorty change? If yes - ddx is working ...and you need to track not jut event but actual values in-kernel ... i have no sane idea how to do this.
glisse: kdekorte: will try that
AndrewR: rah, if everything work down to AtomBios tables and scripts .... try to find Luc, he know more about this area .... (radeonhd author)
kdekorte: glisse: I have a feeling the problem might be in the DDX... BTW I'm running gnome with compiz on Fedora 12..
rah: AndrewR: I didn't try it "at xrandr prorty change"
kdekorte: glisse: kernel, libdrm, mesa and ddx are all from git master, except kernel it is drm-radeon-testing
rah: ffs
kdekorte: glisse: let me know if you can duplicate the crash that way
agd5f: AndrewR: tv standard is an xrandr property
mentor: Compiling shaders gives me 'glsl-fs-fragcoord: shader/slang/slang_compile.c:665: parse_layout_qualifiers: Assertion `0 && "Bad layout qualifier"' failed.' with master from yesterday. I can't see anything wrong with parsing layouts, so I think the previous token(?) may be overflowing.
MostAwesomeDude: mentor: You should send that to the ML, as Vinson (the guy doing the new shader compiler stuff) isn't on IRC.
mentor: OK, thanks.
glisse: agd5f: btw your patch for fixing bpe warning still apply ?
glisse: i don't seem to get warning here
glisse: with the cs checker in drm-radeon-testing
agd5f: glisse: yeah
glisse: what is your gcc version ? :)
kdekorte: glisse: were you able to duplicate the z/stencil issue with the gtkperf test?
glisse: kdekorte: haven't yet tested
glisse: i am working on other stuff
glisse: will test later
kdekorte: ok
kdekorte: I saw a couple of bugs about slow flash playback in firefox, but this one seems really slow to, and it is not flash based. It is an ogv video in firefox 3.5.6 using the native firefox decoder. Maxes out the cpu just loading the video in the browser: http://www.0xdeadbeef.com/~blizzard/weblog-videos/2010-02-14-difference-engine/VID00212.ogv
Shuren: kdekorte, same here, but i think it's not related to radeon driver... theora videos are slow played through html5
kdekorte: Shuren: I played that video on my laptop with an intel card and it plays fine with only a 30% cpu load. So I think the video driver is involved
Shuren: :/
Shuren: maybe you're right
svenstaro: File radeon_dma.c function radeonReleaseDmaRegions line 344
svenstaro: Leaking dma buffer object!
svenstaro: ah!
Obscene_CNN: places the bit bucket under the leak
svenstaro: Well that fixed it
xming: gives Obscene_CNN some free bits from /dev/random
Obscene_CNN: :)
xming: Obscene_CNN: any news?
Obscene_CNN: I haven't had time to port my patches to rc8 yet. I did look at it and I don't think that very much of my patches needs to be rewritten.
xming: so it's good news :D
Obscene_CNN: yes
xming: and bad news :p
Obscene_CNN: the bad news is I'm quite busy at work :(
cxo: drm-radeon-testing doesnt merge with rc8 nicely anymore
Obscene_CNN: a darn customer keeps changing things on my circuit board designs
xming: I agree, customers should be eliminated
cxo: especially ones that require fglrx
xming: cxo: go mess with your kvm :p
Obscene_CNN: It wouldn't be so bad if it was a completely different change. But they are undoing changes they had me make and I have to redesign the board each time.
xming: customers are always like that
xming: I usually get so fustrated in the meeting that I just leave and refuse any further meeting requests :p
xming: better let project manager/sale manager to solve it
cxo: seriously, PMP is such an overexaggerated profession, might as well make them work for it
xming: cxo: idd
spstarr: hmm
codin: I would like to try this patch: http://lists.x.org/archives/xorg-driver-ati/2010-February/013775.html
codin: where can I get a copy of the code ?
airlied: codin: its in the ati driver git master
codin: airlied, can you give me a link ? I am new to this
codin: thank you
airlied: codin: in the topic
twnqx: airlied: can you add a "how to get drm-radeon-testing kernel" section to the wiki?
twnqx: :(
spreeuw: nice, I never noticed the slow Xv
agd5f: codin: code is already in git
agd5f: spreeuw: you might not have. it depended where the buffer ended up
codin: airlied, thank you. I'll give it a try
spreeuw: how's the 3d memory > 256 comming along?
spreeuw: and *cough cough* s3tc?
xming: is switching back and forth between fglrx and radeon os stack
soreau: Is pbuffer supported on r3xx?
glisse: soreau: with dri2 ie kms yes
otaylor: soreau: not sure about pbuffers. fbo's are, which is the more modern way of doing things
soreau: Ok, thanks glisse, otaylor
soreau: Ok, I'm trying to get cairo-dock to run with gallium but I think I might be having some libGL problem.. it still runs with classic mesa
soreau: LIBGL_DRIVERS_DIR=/opt/xorg/lib/dri/ glxinfo|grep nGL shows OpenGL renderer string: Gallium 0.4 on RV350 \n OpenGL version string: 2.1 Mesa 7.8-devel
soreau: but when I try LIBGL_DRIVERS_DIR=/opt/xorg/lib/dri/ cairo-dock -o, it says that it's using - OpenGL renderer: Mesa DRI R300 (RV350 4150) 20090101 x86/MMX/SSE2 TCL DRI2
soreau: So I tried cp'ing /opt/xorg/lib/libGL* to /usr/lib/ but this did not help. It still runs without gallium
soreau: tries cp'ing r300_dri.so over..
soreau: It still runs with classic mesa somehow..
soreau: wtf *scratches head*
codin: who is reading the EDID data ?
dileX: whats RPB?
dileX:
airlied: glisse: btw what you planning on doing if there isn't enough space for VRAM between PCI BAR and end of meomry
airlied: hasn't read patches totallt to see if they handle that
glisse: airlied: the patch limit to pci bar size
airlied: so it'll need another rework when we want to use all VRAM?
glisse: airlied: my point is that no gpu below r6xx should hit that
glisse: well i am wondering about r5xx
airlied: glisse: I tihnk you can get 512MB cards
airlied: they can hit easily
airlied: there are X800 512MB
glisse: hhhmm
airlied: we also have 128MB bar limits with 256MB cards
glisse: that should work well
glisse: solution is then to move vram lot lower than aperture
airlied: well I'd rather we tried to get it all right, everytime we touch the MC setup I get scared
glisse: airlied: then if you want i can move everyone to vram at 0 execpt igp
airlied: isn't that what we have now?
glisse: anyway right now most of the gpu have vram at 0 and work fine
airlied: though not sure why rs480/rv620 don't resume
glisse: they doesn't with rc8 neither
airlied: but I suppose the patch makes less lines of code
airlied: which can't be a bad thing
glisse: airlied: well let me redo a version where i put vram at 0 for r4xx,r5xx and newer
glisse: this should work well as this is what we do now
glisse: that way we shouldn't have anyissue with 512M or more VRAM
airlied: sounds good
glisse: also i might a special case in r6xx to keep puting igp on top of aperture
glisse: agd5f: do we really need igp vram to be on top of aperture ?
kdekorte: glisse: I have an rv635 with 512MB would that be a problem?
glisse: i guess i need to ask for 1g gpu ;)
glisse: i doubt my mutt will ever fill it
agd5f: glisse: IGPs need to point wherever the stolen mem is
glisse: even for rs780
glisse: or newer ?
glisse: because right now we don't do that on rs780 or newer
glisse: we put it at the aperture
suokko: glisse: Do you have any idea why we are realocating gtt pages all the time whe flash video is playin? I tough that there was some pool scheme notto free pages immediately
agd5f: glisse: I thought we treated it the sames as the older IGPs
agd5f: glisse: right now we just read back the value from MC_VM_FB_LOCATION
suokko: Pool sizes is only 1M probably that kills the performance for larger allocations
glisse: agd5f: oh yeah
glisse: well i can do that for rs780 and newer igp
suokko: Maybe pool sizes should be dynamic and some timestamp freeing policy. (Free pages that has been unused for more than ~5 seconds)
glisse: suokko: maybe we need some statistics
suokko: glisse: yes. It would help if we saw how much there was bo allocation and freeing from gtt. But implementing dynamic pool sizes and time policy isn't very hard
glisse: yeah freeing after couple of second sounds good
suokko: glisse: I can add that to my todo list. If it improves allocation performance I will send patch the patch for review
glisse: suokko: cool :)
glisse: airlied: send updated patch obviously not well tested yet
glisse: this rs780 is weird
glisse: sometimes it resumes
glisse: sometimes it doesn't
glisse: i am wondering if it's related with what the gpu was doing
codin: how can I update a git tree to the current version ?
twnqx: "git pull"
suokko: git pull --rebase (safer if you have patches in tree)
kdekorte: codin: might need git clean -df if you get merge errors
cxo: agd5f, thanks for zaphod :)
soreau: In the output of glxinfo for gallium it says: Mesa warning: couldn't open libtxc_dxtn.so, software DXTn compression/decompression unavailable
soreau: What package provides libtxc_dxtn.so or why is it looking for this file?
suokko: soreau: google
suokko: There is patented technology
adamk: It provides s3tc support, when it works.
soreau: suokko: Ok thanks
soreau: adamk: What do you mean, when it works?
stikonas: soreau: when driver supports s3tc
stikonas: and it will still be this way for some 10 years until patent expires...
adamk: soreau: Last time I tried it, it had issues with applications that use multitexturing (ie. doom3, ut2004) on r300.
adamk: soreau: It worked fine on Neverwinter Nights, though. It's never worked on r600.
chithead: soreau: if you have a license for s3tc, you can find packages for libtxc_dxtn for various distros if you look around a little
Neo_The_User: does xserver have to be installed to solve the ./configure: line 11526: syntax error near unexpected token `XINERAMA,' ./configure: line 11526: `XORG_DRIVER_CHECK_EXT(XINERAMA, xineramaproto)' problem?
Neo_The_User: like to compile -ati DDX?
adamk: Neo_The_User: You need the proto package.
Neo_The_User: for xinerama? got it
Neo_The_User: xproto, x11proto and every other proto is installed too
adamk: Perhaps you need a newer version (just a guess).
spreeuw: Neo_The_User: try using git xinermaproto
Neo_The_User: have that too
soreau: x11-macros?
DanaG: hmm, is radeon UMS supposed to support the BACKLIGHT property in xrandr?
Neo_The_User: got it from git
unimatrix: does blur work yet with radeon?
soreau: It did until someone(TM) broke it again
DanaG: hmm, is radeon UMS supposed to support the BACKLIGHT property in xrandr?
soreau: It only works (or has worked) with r3-5xx cards though
unimatrix: funny, it worked in fglrx to, until they broke it...
soreau: DanaG: We heard you the first time
Neo_The_User: haha!!
unimatrix: the new 10.2 Catalyst is a complete disaster
Neo_The_User: i like it
unimatrix: soon everyone will be forced to use radeon
soreau: Neo_The_User: I am fairly certain that error is caused by not having x11-macros installed, installed correctly or the latest version
DanaG: hmm, for some reason, my lvds is NOT showing backlight property.
Neo_The_User: unimatrix: why is that bad?
Neo_The_User: loves AMD and is proud of it
unimatrix: Neo_The_User: it's not bad, everyone should ditch that piece of shit proprietary driver of theirs
DanaG: And ATI HQ in Canada looks pretty damn awesome.
DanaG: http://www.hardwareheaven.com/insideati/index.htm
Neo_The_User: unimatrix: im only using catalyst 10.2 right now so i can play dirt 2 on windows using my new force feedback wheel. for linux, open source all the way and FTW!
unimatrix: Neo_The_User: that's great, but i'm not using windows, 10.2 breaks compiz on linux and causes kernel hardlocks, while radeon is still in experimental stage for 3D
unimatrix: basically at this point i'm considering selling the card and going nvidia
stikonas: why radeon is in experimental stage for 3D?
Neo_The_User: i see no issues with -ati DDX for 3d
Neo_The_User: its fine
stikonas: yes, it works fine for me also
unimatrix: maybe i should try it again
Neo_The_User: unimatrix: i beg of you to not become an nvidiot
DanaG: Radeon KMS on R600 works damn well for me; it just uses too much power without powermanagement.
DanaG: I have to go UMS for now, with forcelowpowermode.
unimatrix: last time i tried KMS on my R700 it caused a white screen
DanaG: I'll try those PM things again once they end up in drm-next. or wait, are they already there?
soreau: Neo_The_User: lol
Ghworg: Radeon is still experimental for >r500 in the sense that it isn't packaged in most distros
soreau: unimatrix: That's because you did it wrong
Neo_The_User: unimatrix: doll
Neo_The_User: wait
chithead: white screen often means you are trying to use opengl compositing with the software rasterizer
Neo_The_User: wrong button
unimatrix: chithead: it happened at GDM login screen
soreau: unimatrix: White screen is usually caused by compiz running on incorrectly installed graphics drivers
Neo_The_User: you need to follow the right guide.
DanaG: ah, remnants of fglrx?
unimatrix: DanaG: like what? i removed everything i could
DanaG: Did you reinstall libgl1-mesa-glx and libgl1-mesa-dri and such, too?
Neo_The_User: fglrx never leaves until you reformat i think
DanaG: Speaking of which... the fglrx installer really should check for legacy hardware... and not allow it to even try to install on that hardware.
unimatrix: DanaG: i may not have done that
unimatrix: speaking of guides, is there a "right one" for Ubuntu 9.10 ?
Neo_The_User: yes
Neo_The_User: mine
Neo_The_User: http://neo-technical.wikispaces.com/radeonhd-3d
unimatrix: lovely, let's see
kdekorte: unimatrix: yeah compiz runs pretty good on my rv635 chip
soreau: sudo git clone/pull... ?
unimatrix: Neo_The_User: i don't wanna compile anything, it will just mess up the system and i won't be able to revert
kdekorte: unimatrix: there are ppas available... xorg-edgers or something
Neo_The_User: yeah
soreau: unimatrix: If you are on ubuntu, you can use xorg-edgers repo and a kernel from ppa
unimatrix: yeah i tried those last time... well i guess i can give it a shot again
Neo_The_User: sudo dpkg -i whatever.deb
Neo_The_User: except it might be a bit out of date
unimatrix: which kernel should i use?
unimatrix: 32 or 33 ?
Neo_The_User: i use airlied's tree
Neo_The_User: you can always change kernels using menu.lst or grub.cfg so you can sort of "revert"
Neo_The_User: so its safe compiling his awesome kernel code. drm-radeon-testing is great
soreau: unimatrix: Since you don't have to worry about the open driver not being compatible with some latest component (as is the case with fglrx) and since there are constant bug fixes and new features going into latest components, you probably should use the latest you can
Neo_The_User: i gotta go install windows 7 on my dad's netbook. bye all
soreau: uses drm-radeon-testing kernel as well
unimatrix: alright let's try 33-rc8
philipp64: agd5f: dumb DPM question for you. what's the proper behavior of an Energy Star monitor when the signal source goes away? it should eventually go into "off" mode, right?
unimatrix: lovely now virtualbox modules are failing
chithead: virtualbox (and vmware, xen or any other out of tree virtualization solution) is always behind kernel development
xming: vbox modules are fine here with .33rc8
mrmcq2u: how do I check if my gfx card is under r300?
eviljames: lspci
eviljames: eg: 01:00.1 Audio device: ATI Technologies Inc RV620 Audio device [Radeon HD 34xx Series]
mrmcq2u: 01:00.0 VGA compatible controller: ATI Technologies Inc Radeon Mobility X1400
chithead: that is r500
mrmcq2u: nothing about really
mrmcq2u: whoops
mrmcq2u: thought it was r300
unimatrix: (WW) RADEON(0): Color tiling is not yet supported on R600/R700
unimatrix: (WW) RADEON(0): Direct rendering disabled
chithead: these two messages are unrelated to each other
suokko: unimatrix: Do you have firmware
mrmcq2u: right this is slightly confusing
unimatrix: what firmware?
mrmcq2u: OpenGL renderer string: Mesa DRI R300 (RV515 7145) 20090101 x86/MMX/SSE2 TCL DRI2
mrmcq2u: so it uses r300 but its rv515?
suokko: unimatrix: see the topic
unimatrix: so xorg-edgers didn't put this in?
BioTube: mrmcq2u: r300-r500 are closely related enough to share a lot of code
mrmcq2u: cool
suokko: unimatrix: You need also new kernel from mainline ppa
unimatrix: suokko: like 2.6.33-020633rc8-generic for example?
suokko: unimatrix: And if you have problems you should always remember to check dmesg too. Large part of driver is in kernel
unimatrix: ah there's also this: (II) RADEON(0): GPU accel disabled or not working, using shadowfb for KMS
cjb: could someone tell me whether an ATI ES1000 is pre- or post-avivo, please?
suokko: unimatrix: Did you check the dmesg?
unimatrix: suokko: yeah, it looks like all is ok there
suokko: unimatrix: Can you paste both logs?
chithead: cjb: es1000 is r100 based
unimatrix: suokko: sure thing
unimatrix: suokko: http://pastebin.com/m32c6079a and http://pastebin.com/m402b7003
cjb: chithead: cool, thanks
suokko: unimatrix: #
suokko: [ 2.241382] r600_cp: Failed to load firmware "radeon/R700_rlc.bin"
unimatrix: why did i not see this
suokko: yes. mainline ppa doesn't include that yet :/
chithead: you need to install the irq microcode, see the wiki link in the channel topic
unimatrix: suokko: ok i've put it in /lib/firmware/radeon, but something tells me that's not going to be good enough right
chithead: the wiki says what else to do
unimatrix: yes, recompile the kernel
suokko: unimatrix: recreating initrd should be enough
chithead: if you build radeon drm as module, then it is enough. if you have it built-in, you need to set kernel firmware options in menuconfig
unimatrix: ah yes, xorg-edgers build it as a module
suokko: update-initramfs -u -k `uname -r` and reboot should work
unimatrix: ah i've put it in the wrong folder
unimatrix: it should be /lib/firmware/`uname -r`/radeon
unimatrix: w00t, it works :)
unimatrix: compiz animations are lagging a bit though
unimatrix: hmm, but mostly when firefox is maximized in the back
mrmcq2u: Whats the best way to enable the gallium3d changes in mesa?
mrmcq2u: using mesa 7.8 from xorg-edgers
BioTube: mrmcq2u: I'm pretty sure gallium needs to be decided at build time
mrmcq2u: crud
mrmcq2u: would be nice if there was a build on xorg-edgers with it enabled, like mesa-gallium or something
spstarr: airlied: built second life (i really do hate multilib madness)
spstarr: airlied: should I build mesa/ddx/libdrm from git master or just follow rawhide ?
airlied: mesa it probably all thats changed much
spstarr: ok
spstarr: now i can build less from git since changes are going into koji rapidly
spstarr: kernel + mesa and thats pretty much it unless the ddx is changing more
spstarr: for now im gonna stick to koji kernels and UMS
spstarr: airlied: UMS is broken on Fedora :/
spstarr: airlied: [ 58.197] (EE) RADEON(0): Acceleration initialization failed
spstarr: it turned off all acceleration
spstarr: i dont know why
spstarr: oh
spstarr: platform r600_cp.0: firmware: requesting radeon/RV635_pfp.bin
spstarr: r600_cp: Failed to load firmware "radeon/RV635_pfp.bin"
spstarr: [drm:r600_do_init_cp] *ERROR* Failed to load firmware!
spstarr: odd...
spstarr: modeset turns off this?
spstarr: oh
spstarr: kernel-firmware
spstarr: doh
spstarr: n/m
spstarr: my custom kernel ate the firmware dir ;)
shadowmaster: did it find it delicious?
spstarr: all good
spstarr: in UMS now
spstarr: i'll remain in a holding pattern for now
Telek: Has anyone heard word of a PCI based HD5450 card?