Ghworg: libdrm-radeon is going in to debian now anyway, they just closed the wishlist bug yesterday
amarsh04: excellent, Ghworg
hifi: now thats fast
amarsh04: has lots of open bugs in Debian... at least I find the bugtracker friendly
CartoonCat: http://pastebin.org/84789
CartoonCat: yippie
CartoonCat: anyone help me figure this out?
soreau: CartoonCat: Can you paste the entire log?
CartoonCat: yup
CartoonCat: http://pastebin.org/84790
soreau: It's selecting xaa by default which leads me to believe you have old ddx
soreau: I would build latest xf86-video-ati and mesa against libdrm-9999
CartoonCat: k
CartoonCat: ok so it seams to ahve worked. 3d is still slow and eselect opegl still only shows xorg. how to test if its working, glxgears is not installed as part of the package it seams?
Nille02: good morning everyone.
Nille02: i have an problem with kms. the console works now nice but if i want to switsch to the xserver the display gos in standby
Nille02: xorg.log http://noctarius.pastebin.com/m758a8ae1 dmesg http://noctarius.pastebin.com/d6166c45b
soreau: CartoonCat: What is the output of glxinfo|grep renderer
soreau: Also, eselect opengl list is only supposed to show xorg-x11
CartoonCat: bash: glxinfo: command not found
soreau: Install mesa-utils
CartoonCat: no match found, do you mean mesa-progs ?
soreau: Yes
CartoonCat: DRI yes, OpenGL renderer string: Software Rasterizer
CartoonCat: so something still isnt right
soreau: Right
soreau: Pastebin your X log
Nille02: has no an idea about my problem?
CartoonCat: http://pastebin.org/84807
soreau: CartoonCat: Did you configure your kernel with CONFIG_DRM_RADEON?
rothwell: 'lo. after a recent xorg upgrade, my 3D performance seems to have suffered... xorg seems to be using swrast_dri.so as opposed to the r300_dri.so i think it was using before
rothwell: xdriinfo reports "Screen "0" is not direct rendering capable."
soreau: rothwell: What is the output of glxinfo|grep renderer
rothwell: OpenGL renderer string: Software Rasterizer
soreau: Pastebin your X log
rothwell: http://waste.corolla.ath.cx/2010/02/01/Xorg.0.log
soreau: (II) RADEON(0): XAA Render acceleration unsupported on Radeon 9500/9700 and newer. Please use EXA instead.
rothwell: i'm afraid i don't know what that means
rothwell: i did have proper acceleration a few weeks back
soreau: You need to switch to EXA acceleration method by adding the following to the Device section of xorg.conf: Option "AccelMethod" "EXA"
rothwell: ok, thank you
rothwell: is this some recent development?
soreau: relatively, yes
soreau: freebsd is just a little behind..
rothwell: right
rothwell: i'll restart X and see what happens... one sec
rothwell: 'lo. well... ouch. my glxgears framerate has gone down to 14fps
soreau: What does glxinfo report as the renderer string now?
rothwell: OpenGL renderer string: Software Rasterizer
soreau: Pastebin your current X log
rothwell: http://waste.corolla.ath.cx/2010/02/01/Xorg.0.log.1
soreau: huh
rothwell: er, server's a little confused about the file suffix it seems... didn't serve it as text/plain
soreau: Can you pastebin it to pastebin.com please?
rothwell: yes, will pastebin
rothwell: http://pastebin.com/m5ea2f781
soreau: #
soreau: (EE) RADEON(0): [pci] Out of memory (-12)
soreau: (EE) RADEON(0): [pci] PCI failed to initialize. Disabling the DRI.
rothwell: erk
rothwell: that's extremely strange
rothwell: this machine has 12gb
rothwell: wonder if there's some sort of resource limit in effect
soreau: Is it a PCI card?
rothwell: yeah
rothwell: http://www.powercolor.com/global/products_features.asp?ProductID=7
rothwell: that's the exact model
soreau: Well this is a bit beyond my expertise
rothwell: ok
soreau: Maybe someone can help in a few more hours when folks start waking up
rothwell: ok, thanks
soreau: Since it has 256 VRAM, I really dont know what is going on
rothwell: i think rnoland_ will probably have a few ideas
soreau: Yea, wake him up ;)
rothwell: i'll google the error message for now
rothwell: got this sneaking suspicion i've seen it before... about 18 months ago
soreau: (II) RADEON(0): Will use 32 kb for PCI GART at offset 0x0fff8000
soreau: I think it is confused :)
rothwell: certainly looks it
rothwell: some hints from google as well that i may need to reboot. some people having issues with older cards that only work if X is started soon after boot
soreau: maybe, I never used a PCI card so I dont know
rothwell: thanks for the help anyway, i'll try a reboot now and see what happens...
rothwell: huzzah
rothwell: http://waste.corolla.ath.cx/2010/02/01/Xorg.0.log.2.txt
rothwell: certainly better than it was, no 'out of memory error' this time
rothwell: not sure if it's quite back to normal yet... trying to find out
soreau: What does glxinfo report as the renderer string now?
rothwell: as before
rothwell: "OpenGL renderer string: Software Rasterizer"
soreau: I dont see anything that would indicate failure there except that it is loading swrast
soreau: Can you pastebin the output of LIBGL_DEBUG=verbose glxinfo
rothwell: yeah, i initially wondered if it was just a configuration issue
rothwell: i do have r300_dri.so sitting there
rothwell: hm!
rothwell: heh, well that's very surprising output...
rothwell: http://waste.corolla.ath.cx/2010/02/01/glxinfo.txt
rothwell: "ibGL: OpenDriver: trying /usr/local/lib/dri/r300_dri.so
soreau: And, note the renderer string
rothwell: yep
soreau: This is a good sign..
rothwell: certainly not what i normally get out of glxinfo
soreau: I seem to recall someone saying you may need to tell it to enable dri via xorg.conf
soreau: Check the radeon man page and make sure to enable accel or dri or whatever the option is
rothwell: i've just noticed something else
rothwell: that glxinfo.txt was glxinfo output when executed with privileges
rothwell: http://waste.corolla.ath.cx/2010/02/01/glxinfo_nopriv.txt
rothwell: that's it without
Nille02: exsist an way to read the gpu informations? like temps?
soreau: rothwell: Is your user part of the video group?
rothwell: yep
rothwell: i don't see a "Mode 0666" line anywhere in xorg.conf... could this just be a permissions issue?
soreau: What owner/group/permissions for /dev/dri/card0 ?
soreau: Yes it could be
rothwell: 0660, root:wheel
rothwell: i'd better sort that out
soreau: huh
soreau: Is your user part of the wheel group? :P
rothwell: nope
rothwell: oh dear...
rothwell: this is the default configuration, too. i've not changed any /dev node modes/ownership
rothwell: no idea why it worked previously
rothwell: ok, restarting X again
rothwell: OpenGL renderer string: Mesa DRI R300 20060815 TCL
rothwell: success!
rothwell: thanks for patiently getting me to solve my own problem, hehe
rothwell: everything seems to be back to normal
soreau: Cool, glad I could help you help yourself ;)
amarsh04: has just installed libdrm-radeon1 2.4.17-1 in Debian unstable i386
loki_666: what's actually using gallium?
airlied: loki_666: noting major by default yet
airlied: vmware driver mostly
loki_666: so it's not possible to play openarean using gallium?
Wizzup: likes r300g
airlied: loki_666: gallium isn't a thing, r300g is the driver
airlied: and oa runs buts it slow
rnoland_: rothwell: the current code needs to find 32M of contiguous physical ram, which generally works after boot.
rnoland_: rothwell: but if the system has been up that can be hard due to memory fragmentation.
loki_666: airlied, ok
rnoland_: rothwell: i have reworked all of the scatter gather memory stuff yet again to remove that requirement....
airlied: loki_666: we know how to make it faster, just need to write the bits
rnoland_: rothwell: since we don't technically need contig ram.
rothwell: nice
rothwell: yeah, this machine had been on for weeks doing a lot of heavy work
rothwell: worked fine after boot
loki_666: airlied, cool, so the opengl st is ok, but not optimized yet
airlied: loki_666: its actually the r300 specific bit needs work
loki_666: what about the other st like xorg etc...
rnoland_: rothwell: i haven't committed that work yet... but since i broke nouveau yesterday already... i'll probably commit it soonish....
airlied: we need to do better on allocating vertex buffers and reusing them
airlied: loki_666: the Xorg st starts X, but I get glitches last time I ran it
rothwell: rnoland_: right!
loki_666: airlied, what need to be done to use the Xorg st? i guess, you need to change the xorg driver being used?
airlied: yeah it builds radeong_drv.so so you need to set xorg.conf to load that
loki_666: ok, so it build a xorg driver for each diffrent gallium driver? i though the goal was to have only one driver to rule them all :p
airlied: loki_666: no, never that
loki_666: what about the xorg-video-modesetting driver? it's just a basic xorg kms driver so?
airlied: loki_666: thats gone
loki_666: ah
airlied: plan is to use libkms as a fallback sw driver if thehw gallium driver can't start
loki_666: ok
loki_666: another quick question: does r600 kms driver supports hdmi sound?
airlied: loki_666: 2.6.33 one does
airlied: forgot to push patch to fix audio after s/r
loki_666: ok great
glisse: airlied: did you try to reproduce the kwin garbled window bug ?
glisse: i haven't been able to reproduce it here
airlied: glisse: which one is that?
glisse: http://bugzilla.kernel.org/show_bug.cgi?id=15186
glisse: been trying with desktop effect on or off in kde on rv710 or rv770
glisse: haven't a rv730
mokoloko: It happens quite randomly
MrCooper: glisse: apparently the CPU isn't reading the correct contents from the 1x1 pixmaps traditionally used for solid RENDER pictures, or something like that
glisse: isn't the white area somethings transparent in kde
glisse: like somekind of window shadow
Nille02: if i try to start the xserver with kms the monitor shutdown. dmesg http://noctarius.pastebin.com/d1f652bfc xorg.log http://noctarius.pastebin.com/d637c125a
mokoloko: open couple of apps and switch between them and you should see it at one point
airlied: glisse: wierd
Nille02: on the console kms working fine
airlied: glisse: sounds like we were relying on buggy behaviour somewhere
glisse: yeah
MrCooper: airlied: what do you mean by that?
glisse: well the flush with my patch should happen right after ib
glisse: so it seems that before we were working because the flush happened randomly after the ib
MrCooper: are we positive that writing to the register from the CP is (always) effective?
glisse: shouldn't change a thing, but we don't know the cp microcode
airlied: MrCooper: I'd expect we'd see more problem than just the 1x1 pixmap though
airlied: though then again probably not
MrCooper: and that the GPU write caches are properly flushed etc.
glisse: mokoloko: you are the bug reporter ?
glisse: how much app are you talking about ?
glisse: if i open 10apps and do nothings with them beside switching windows does it happen ?
taiu: probably this was the reason i had all piglit tests working with some kernel versions
taiu: now some of them fail reading pixels, havent bisected but suspect the same commiyt
glisse: well according to doc it seems that if we want to flush gpu write to pci/pcie/agp space we need to make the gpu read back some value in the pci/pcie/agp space
glisse: wonder if wait_reg_mem would be enough
airlied: glisse: probably
MrCooper: glisse: but the commit in question should only make a difference for CPU reads from VRAM?
glisse: yeah
taiu: glisse: maybe piglit fp-abs-01 can be used - it fails under kms reliably for me
Nille02: if i try to start the xserver with kms the monitor shutdown. dmesg http://noctarius.pastebin.com/d1f652bfc xorg.log http://noctarius.pastebin.com/d637c125a
taiu: and I'v tried adding http://pastebin.ca/1774195 to every SyncSurf - didn't seem to help but i'm also known to produce failed tests in the past ;)
glisse: Nille02: why do you have vga16 ?
Nille02: i dont know this are the pakages from my distribution
Nille02: its save to blacklist it?
glisse: do you have any weird kernel parameter ?
Nille02: no
Nille02: only radeon.modeset=
Nille02: 1
glisse: nothing else at all ?
Nille02: no
glisse: to which output your monitor is connected ?
glisse: well for the white pixmap i will try to add a gpu flush before each ib too
Nille02: DVI-
glisse: see if that helps
Nille02: 1
glisse: Nille02: well this kernel is weird
glisse: the vga framebuffer device might mess in kms back
glisse: and it maybe outdated
Nille02: and if i modprobe -r the vga module i have the samne
glisse: we have a lot of fixes since 2.6.32
Nille02: latest ubuntu 10.04
glisse: if you modprobe -r vga you need to rmmod radeon and reload radeon again
Nille02: with .33-rc6 i have no output in the dmesg
Nille02: ok
taiu: hmm, R600_OUT_BATCH(CP_PACKET0(0x5480,0)); R600_OUT_BATCH(0x1); from SyncSurf makes it run
glisse: well if 2.6.33-rc6 doesn't work then open a bug against 2.6.33-rc6 with kernel log
glisse: taiu: i think such packet will be rejected by kms
taiu: glisse: no, passes, but I don't have you lastest checker bits
glisse: even without my checker
glisse: what is 5480 ?
Tommeh: glisse: I've just recompiled 2.6.32.7 for EXTRA_FIRMWARE etc. but would it be better to use 2.6.33-rc6?
Tommeh: (Just wondering. I might test both.)
taiu: glisse: R_005480_HDP_MEM_COHERENCY_FLUSH_CNTL, i just hardcoded it
glisse: taiu: where did you insert those packet ?
glisse: at begining of ib ?
taiu: mesa, in r700SyncSurf, well that's the wrong place, it gets called too often for each ib
glisse: so it gets inserted before draw cmd
glisse: will try to do that in the kernel and ask the bug reporter to test
taiu: and after it (it syncs CB and DB)
glisse: ddx cmd might be: create 1x1 pixmap write color to it, schedule ib refing this pixmap
glisse: and hdp isn't flushed soon enough and gpu endup using random data from vram, though i would expect that bug reporter would not see constant color
taiu: it's just dummy test - the register seems to work, but we should emit or wait for it some other place ....
glisse: well not only white, yellow or blue
mokoloko: glisse: No Im not but I see that too. Though it's practically vanished with latest mesa, but sometimes when I open or switch to different app the blue window outline thing turns yellow, red, white or something and is like inch thick
mokoloko: it was much more common like two weeks ago
mokoloko: oh btw would it matter that im on qt4.6 and kde 4.4rc2
Nille02: with .33-rc6 i get this dmesg http://noctarius.pastebin.com/d5302dc83 xorg http://noctarius.pastebin.com/d78254c85
Nille02: CONFIG_FB_RADEON=m CONFIG_DRM_RADEON=m
Nille02: lsmod http://noctarius.pastebin.com/d570e1b5b
Nille02: and the screen is garbaged
fudje: Nille02: You don't want CONFIG_FB_RADEON
Nille02: this is the option from the distibution
fudje: Are you using KMS?
Nille02: this is my goal
fudje: then config_fb_radeon=m is useless to you ... but might serve to confuse things.
Nille02: with .32 kms is "working" but if i start the xserver the monitor shut down
fudje: The good news is you don't seem to be loading radeonfb
mokoloko: glisse: Curious it happens more often if you switch between kde and gtk app. With kde app and wine app it happens even more often.
fudje: I don't see any drm messages in your dmesg either though :/
Nille02: i try rc5 myabe its only a bug
fudje: from the x.org messages, I'd suspect either you didn't build libdrm with kms, or you didn't build xf86-video-ati afterwards.
schmankerl: could somebody explain me the differences between dynamicpm and clockgating?
Nille02: same userspace only with .32 ;) http://noctarius.pastebin.com/d637c125a
Nille02: but then the monitors shut down
fudje: That seems to be a completely successful X.org startup :/
Nille02: i hear the startup sound but the display is off
Nille02: and i can witsch to tty etc
taiu: glisse: you were right, i went to sw rasterizer due to typo and the register is rejected,
fudje: Do you get a high resolution console?
Nille02: yes
fudje: Okay.
fudje: What bits did you build yourself?
taiu: glisse: it's about 3'rd time i fail like this and produce invalid results, sorry
Nille02: only the ddx but with the ddx from the dirto i has the same
Nille02: distro*
fudje: So mesa, libdrm, and the kernel are all from your distro?
glisse: Nille02: remove CONFIG_FB_RADEON
Nille02: yes and all with kms if i trust the messages from ddx
glisse: really test with lastest 2.6.33
Nille02: glisse: for this i have to recompile the kernel or not?
fudje: What distro is this, btw?
Nille02: ubuntu 10.04
fudje: wonders why anyone would ship a prebuilt kernel since 2.6.32 with CONFIG_FB_RADEON....
Nille02: so what is better for test linux-next or rc6 from mainline?
fudje: Nille02: what's the contents of /etc/initramfs-tools/modules?
fudje: also, do you get the high resolution console with 2.6.33-rc6?
Nille02: i dont know if i try to
Nille02: switch to the console the system freze
Nille02: the fire is more or less empty
Nille02: file*
fudje: you should be able to tell during bootup...
Nille02: and if i stop the gdm from shh the screen goes white
Nille02: and how?
fudje: well it'll be a highres console ;)
fudje: The other option is to run `cat /sys/class/vtconsole/vtcon1/name' and see if you get "(M) frame buffer device" back
fudje: or a file not found error
Nille02: mh ok but first i compile my own kernel mybe its was an fault in the kernel from ubuntu
fudje: although, your lsmod shows 71 users of the fbcon module which suggests that yes, it's loaded
fudje: and in use.
fudje: Hmm. Well this is embarrassing.... I log on to ask about a display corruption issue I'm having and it vanishes....
fudje: oh no, there it is....
fudje: What I'm getting looks like incorrect texture geometries...
fudje: Or sometimes the X11 colour shows through and the rendered surface doesn't show at all.
fudje: Of course then there's always the option of freezing the entire system...
glisse: for those having issue with r6xx/r7xx and strange corruption please test patch attached too http://bugzilla.kernel.org/show_bug.cgi?id=15186
Nille02: so what are the important kernel options for kms ? i i understand it now right config_fb_radeon off
Nille02: need i only config_drm_radeon and config_drm_raeon_kms?
soreau: Nille02: You only need config_drm_radeon enabled with a kernel that has radeon kms. Alternatively, you can set config_drm_raeon_kms to have kms on by default
Nille02: thx you
amarsh04: ok, I have kernel modesetting working - is there any guide to what is good about it?
chithead: fast vtswitch, clean suspend/resume, kernel memory management
Nille02: i hope this can help
Nille02: http://www.phoronix.com/scan.php?page=article&item=kernel_modesetting&num=1
Nille02: o know its old ^^
amarsh04: thanks, I've gone back to software rendering though /-:
Nille02: mh post your xorg log maybe someone can say why
adamk: Most likely, you your libdrm or ddx isn't compiled with KMS support.
amarsh04: ah, probably
amarsh04: libdrm/libdrm2/libdrm-radeon1 is 2.4.17 from Debian unstable
amarsh04: and ddx is: xserver-xorg-video-radeon 1:6.12.4-3
chithead: 6.12.4 is too old for kms
amarsh04: ouch, power went out for a half a minute
suokko_: amarsh04: you need a newer ddx
amarsh04: yes, I'm just looking at the URL in the /topic to find out how to download and build
Nille02: amarsh04: http://www.x.org/wiki/radeonBuildHowTo
Nille02: oh nvm
fredrikh: glisse: that patch doesn't fix the problem for me
fredrikh: glisse: but it looks to me like it's a duplicate of http://bugs.freedesktop.org/show_bug.cgi?id=25370 which was filed back in november
amarsh04: yes, building now
glisse: fredrikh: their maybe several interwined bugs here
fredrikh: yeah
fredrikh: 25370 seems to be a problem with solid fills
Curan: Just out of curiousity: will the R300 branch (even though it is in maintainance mode) receive OpenGL 2.0 support (of which the hardware is capable AFAIK)?
adamk: Actually, the hardware isn't really capable of it. There was a discussion on dri-devel about that.
Tommeh: Is there any way to tell if I've loaded the extra R600 firmware correctly?
adamk: They can, however, still advertise opengl 2.0 and they do advertise opengl 2.0 (or maybe 2.1, actually, if you use the gallium3d driver.)
adamk: Tommeh: The new IRC one? dmesg should say, iirc.
Tommeh: k :)
Tommeh: teh@teh-desktop:~$ grep -i firmware /var/log/dmesg
Tommeh: [ 9.164184] intel_rng: Firmware space is locked read-only. If you can't or
Tommeh: [ 9.164186] intel_rng: don't want to disable this in firmware setup, and if
Tommeh: erk :(
adamk: Check for the name of the firmware.
kdekorte: Tommeh, "dmesg | grep radeon" or "dmesg | grep drm"
Curan: adamk: thanks, though I'd say the wiki page should be fixed then (there the OpenGL line reads 1.5/2.0 with a footnote, that one extension is not supported)
adamk: Curan: Not sure I understand that needs to be fixed?
adamk: In any case, it's the nature of wiki's to be out of date :-)
Curan: adamk: the row „OpenGL Compliance (Driver/Hardware)“ reads for R300 „1.5/2.0¹“ on http://wiki.x.org/wiki/RadeonFeature
Curan: adamk: and you just said, that the hardware isn't capable of OpenGL 2.0, so the „/2.0“ is wrong, isn't it?
adamk: That's what the footnote is for.
Tommeh: adamk: http://pastebin.com/d32661d0
Tommeh: From the look of that, I don't *need* the rlc firmwares
adamk: Tommeh: I don't believe it's necessary, unless you want support for IRQs.
Tommeh: Ah, erm, OK :)
adamk: Curan: Does it make more sense now? The hardware supports opengl 2.0 except complete support of ARP NPOT textures, which can be implemented in the driver anyway, somehow.
adamk: That's presumably how the proprietary driver handles it.
fudje: well, handled it.
Tommeh: So, is there a way for me to debug the enabling/disabling of compiz under Ubuntu?
Tommeh: I'm using the xorge-edgers PPA to get all the latest X/Mesa/radeon/drm drivers
amarsh04: after building and intalling the 6.12.99 radeon ddx, I get the error: (EE) AIGLX error: Calling driver entry point failed
Tommeh: s/drivers/package.
Tommeh: I'm wondering if it's a problem with mesa - glxgears just churns over my CPU time (although it's about 600FPS when it does it.)
fudje: Tommeh: what exactly about it do you wish to debug?
Tommeh: Enabling it :)
fudje: glxgears will always do that.
Curan: adamk: yeah, I understood that, and my initial question was more: is this kind of support (as much as possible) going to come or not (e.g. because the R300 branch is in maint. mode)
adamk: Curan: As I said, the gallium3d driver supports it now.
Tommeh: In Ubuntu terms, I try to enable 'desktop effects' but it sits there searching for available drivers and then simply tells me that desktop effects could not be enabled - quite annoying.
adamk: Curan: It's not perfect, obviously, but opengl 2 is advertised in that driver.
adamk: Tommeh: Does 'compiz --replace &' in a terminal start compiz?
Curan: adamk: is Gallium stable enough to use it now, or still very experimental?
fudje: Tommeh: You could try, at the terminal, running `LIBGL_DEBUG=verbose compiz --replace'
adamk: amarsh04: Pastebin the full output of 'LIBGL_DEBUG=verbose glxinfo'
adamk: Curan: I've not used it recently. When I tried, it wasn't really that usable. I do know that there are some folks that use it regularly, but I would still expect any number of things to notg work.
adamk: I believe soreau uses it all the time these days, so he might be able to shed some light on what works and what doesn't.
Curan: adamk: thx, then I'm waiting patiently ;)
soreau: Curan: r300g lies about what it gl extensions supports. But it is at least good enough to run compiz and some games
amarsh04: ok adamk http://paste.debian.net/58338/
Curan: soreau: does KDE's compositing run too?
fudje: soreau: but does it fallback to softpipe for the extensions it doesn't support?
soreau: Curan: I dont really see why it wouldnt but then again I dont see any reason to run kwin, either ;)
soreau: fudje: yes
adamk: amarsh04: IF I had to guess, I'd say tht either Mesa and libdrm are not in sync on your machine, or one or more of your components is not compiled with KMS support.
amarsh04: ok, thanks adamk
Curan: soreau: thx for the info; so Gallium is atm a mix of hw acceleration and software rasterization?
amarsh04: probably mesa
Curan: Then I think I'll stay with the standard driver for some while
soreau: Curan: fudje: r300g reports it supports gl2.1 and most all extensions even though the implementation may not be fully complete. So some things run better than others visually and performance wise. It depends on what extensions the app is using
Tommeh: adamk/fudje:
Tommeh: $ LIBGL_DEBUG=verbose compiz --replace
Tommeh: Checking for Xgl: not present.
Tommeh: xset q doesn't reveal the location of the log file. Using fallback /var/log/Xorg.0.log
Tommeh: Detected PCI ID for VGA:
Tommeh: Checking for texture_from_pixmap: not present.
Tommeh: Trying again with indirect rendering:
Tommeh: Checking for texture_from_pixmap: present.
Tommeh: Checking for non power of two support: present.
Tommeh: Checking for Composite extension: present.
Tommeh: Checking screen 1Comparing resolution (2560x1024) to maximum 3D texture size (4096): Passed.
Tommeh: Checking for Software Rasterizer: present.
Tommeh: Software rasterizer detected, abortingaborting and using fallback: /usr/bin/metacity
BioTube: Tommeh: PASTEBIN
Tommeh: (Sorry for spam - killing that there and then has totally screwed up Gnome, so I'm going to have to restart :)
adamk: Tommeh: Ack... *Pastebin*
Tommeh: ^
soreau: Tommeh: Use a pastebin service like pastebin.com !
Tommeh: I KNOW
Tommeh: One-off people, with an apology too
Tommeh: Doesn't need THREE of you to tell me, either.
soreau: It doesnt make it ok if you know better and still flood
Tommeh: I have little choice at the minute.
Tommeh: kills x
adamk: Anyway, your drivers are clearly screwed up.
Tommeh: And back
Tommeh: For what it's worth (for those whom haven't ignored me), it's now pastebin'd: http://pastebin.com/d56448442
adamk: As I said, your drivers are clearly screwed up. You're using the software rasterizer.
Tommeh: I'm still none the wiser as to what the hell is wrong given that I have recent DRM, firmwares, Mesa, X, all from the same package repo.
Tommeh: adamk: that's what I'd wondered, but I've no idea how to ensure it doesn't use the software rasterizer
adamk: Pastebin your /var/log/Xorg.0.log file.
[Enrico]: Tommeh: paste your /var/log/Xorg.0.log there can be some clue on why you are using software rasterizer
soreau: No matter how bad your computer is screwed up is never a reason to not use a pastebin service
Tommeh: soreau: Whatever - nothing's good enough, so I can't win ;)
soreau: just make sure to use pastebin this time
Tommeh: I do know how to use it!
soreau: Coulda fooled me!
Tommeh: And I did just now. Blind more than fool ;)
Tommeh: adamk: could be some clue: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch.
soreau: It doesnt help to pastebin after flooding
adamk: Can we just drop the discussion about pastebin'ing please.
[Enrico]: Tommeh: pastebin the whole Xorg.0.log pls
Tommeh: kk
soreau: Pastebin the log. We dont care to see the bits you might think are interesting
adamk: Tommeh: Sounds like your libdrm or DDX isn't compiled with KMS support.
[Enrico]: seems to be
[Enrico]: i mean i agree with adamk :D
[Enrico]: Tommeh: well you can try disabling kms at boot
Tommeh: http://pastebin.com/d274d50fb
Tommeh: Somewhere in there, it does say KMS was enabled..
Tommeh: KMS is enabled in the config file, rather than via grub.
[Enrico]: Tommeh: mhm seems you are using an old kernel
Tommeh: teh@teh-desktop:~$ uname -a
Tommeh: Linux teh-desktop 2.6.32.7 #1 SMP Mon Feb 1 11:10:09 GMT 2010 x86_64 GNU/Linux
[Enrico]: or well xorg thinks it
adamk: Right, the issue is what I said :-)
adamk: Most likley, anyway.
Tommeh: Ah
[Enrico]: Tommeh: yeah try to disable kms with radeon.modeset=0 in grub
[Enrico]: nomodeset should work too
BioTube: (II) [KMS] drm report modesetting isn't supported.
adamk: BioTube: Good catch.
Tommeh: Ok, added that to grub
Tommeh: I'll try another reboot in a bit. Need to do some work :)
Tommeh: Thank you!
dante: so after turning kms off, I am still getting a problem with X using up all CPU every day or so.
dante: sometimes it says the dreaded MQ overflowing in the log file.
soreau: is glxinfo still reporting software rasterizer?
CartoonCat: hellos
CartoonCat: ok booting into the right kernel helps a bit =)
CartoonCat: but kde/x still is using software raster
dante: soreau: was that to me?
dante: oh, nm
CartoonCat: latest xorg log http://pastebin.org/84941
suokko_: CartoonCat: Your ddx is not compiled with KMS support
CartoonCat: uhg, ok what did i miss? i followed the instructions in the link in the topic
suokko_: You need to check that configure of xf86-video-ati reports KMS enabled
suokko_: It might be that configure is finding wrong headers for libdrm which don't include KMS support
CartoonCat: suokko: i dont see any mentions of KMS in the output of ./configure
suokko: For me it reports in end "Kernel modesetting: yes"
CartoonCat: does the fact that im amd64 matter?
suokko: no
CartoonCat: http://pastebin.org/84944
CartoonCat: no mention of kms / modesetting / etc
suokko: What version of ddx is that?
CartoonCat: ums =)
suokko: Looks like very old configure script at least.
CartoonCat: thats odd, that is in the git repo from the link
looonger: hi
looonger: is it a known issue that 3d apps may flicker with xf86-video-ati driver and kwin effects enabled?
looonger: or maybe i didn't setup up everything right?
looonger: btw, I have radon hd 3850 card
fudje: See, mine has checking for LIBDRM_RADEON... yes
CartoonCat: mmm
fudje: doesn't say anything about KMS
fudje: and git says that's the most recent mainline....
CartoonCat: are you having issues too, or does your work?
fudje: I'm having some odd corruption issues, but otherwise mine works well
fudje: CartoonCat, where'd you get your source tree from?
CartoonCat: im used to corrution issues with ati lol
looonger: with kms enabled I have strange artifacts in kde and lower fps, with kms disabled i have no artifacts but I have flickering in 3d apps with desktop effects enabled
CartoonCat: fudje: the link in the topic
CartoonCat: I followed those steps to the letter =\
fudje: CartoonCat: can you try the configure again with --enable-kms ?
CartoonCat: fudje: will do
CartoonCat: configure: WARNING: unrecognized options: --enable-kms
suokko: CartoonCat: What does "git branch" report for you?
CartoonCat: 6.12 branch master
suokko: Which one is selected?
CartoonCat: 6.12
suokko: ok
suokko: do this:
fudje: gti checkout master
fudje: git*
suokko: git checkout master && ./autogen.sh && ./configure ....
suokko: 6.12 is the stable driver without KMS support
CartoonCat: doh
CartoonCat: yea just saw the LIBDRM float by
suokko: looonger: flickering is known problem witohut KMS
CartoonCat: ok so after that, make and make install ?
suokko: yes
looonger: suokko: are artifacts with kms a known problem too? :)
suokko: AGP card?
looonger: suokko: actually i prefer flickering cause artifacts appear with desktop effetcts enabled, without them the driver is perfect :)
suokko: If yes then you could try Option "EXANoDownloadFromScreen" "no" in xorg.conf
suokko: oops that should be "yes" :)
suokko: I hate options that have "No" in the name
looonger: suokko: actually i don't know what this agp is :)
fudje: looonger: obsolete technology
looonger: suokko: I'm quite a newbie to radeon
fudje: Accelerated Graphics Port
fudje: From the pre-PCI Express days....
looonger: it's radeon hd 3850, this much I know
suokko: yes. AGP is old connection interface between graphics card and motherboard
pepee: hi
fudje: Why, I remember when it was new (and already considered dodgy).... Back in my day, you had an ISA connection to the graphics card....
suokko: looonger: you can check xorg.log and see if it says AGP there
fudje: Does anyone make an AGP 3850? Sounds unlikely
suokko: There is even 4xxx AGP cards :)
fudje: I was lucky (in a way) to get an AGP 2600...
pepee: is the radeon driver compatible with the ATI driver? I mean, can I install both at same time?
fudje: pepee: You mean Catalyst?
pepee: yep
fudje: The most recent versions of catalyst will load happily over the top of radeon providing KMS is not enabled.
fudje: However the libGL's are incompatible
suokko: pepee: In theory you could install both but only one can be used and switching them in running system is not supported
pepee: hmm ok
pepee: I wanna install ati stream sdk
looonger: suokko: there's something like MC_AGP_LOCATION in my xorg.log
pepee: well, I'll install the ati driver
pepee: thanks people
suokko: "(==) RADEON(0): Using AGP 8x" from my log
suokko: I think also that MC_AGP means AGP card
pepee: which is the correct way to uninstall the radeon driver in ubuntu/debian?
looonger: suokko: is it good or bad? ;)
looonger: suokko: if it's agp?
suokko: So you should try to disable the down load from screen
suokko: looonger: I would say bad :)
maligor: pepee, you don't need to do that, assuming you're using the standard packages
suokko: agp is nice idea but it tries t ogain too much performance in expense of easy of implementation and stability
maligor: the catalyst packages will divert the libGL when you install it
looonger: suokko: how to make sure it's agp card?
suokko: looonger: Can you find "Using AGP Nx" from your log?
looonger: suokko: it used to flikcer too with older versions of catalyst too, but it didn't with the newer
Droste: if it's using a agp mode it surely is an agp card
pepee: ahh ok
suokko: looonger: Or "AGP card detected"
looonger: suokko: There's nothinkg like Using AGP Nx or AGP card detected in my log
fudje: looonger: if you do an `lspci | grep AGP' you should get a line reading something like "01:00.0 VGA compatible controller: ATI Technologies Inc Rsomenumbers maybePRO AGP [Radeon HD 3850 maybePRO AGP]
fudje: if you don't, it's not AGP
Sarvatt: anyone having problems with floods of errors like this? CS section size missmatch start at (../../src/radeon_textured_videofuncs.c,R500DisplayTexturedVideoCP,2700) 12 vs 14
maligor: fudje, are you sure it'd say that? the cards use a AGP to PCIe bridge
maligor: they don't make native agp chips anymore
fudje: maligor: I'm aware of that
pepee: thanks maligor fudje suokko
fudje: maligor: however, what you get is what the PCI bus reports, which is an AGP card.
pepee: bye
maligor: shouldn't boards with AGP list a 'Chipset AGP Bridge'
maligor: and those would be very unlikely to have pcie
looonger: isn't lspci part of pciutils? I have no such command
looonger: oh, I have it
fudje: actually I've seen pcie boards with "legacy" AGP slots
looonger: fudje: lspci | grep AGP doesn't give any results
looonger: lspci shows this "02:00.0 VGA compatible controller: ATI Technologies Inc RV670PRO [Radeon HD 3850]"
fudje: sounds like PCIe to me :)
maligor: you can also check if you have devices there with something like 'PCI Express Port' if you feel like it :P
looonger: ok, now how does agp or its lack refer to my problem? :>
suokko: looonger: DFS is often broken in AGP systems but you could try to disable DFS still to see if it helps you
suokko: Also it would help if you showed screenshot which shows the rendering problems
adamk: looonger: 3D flickering? Are you using KMS?
looonger: adamk: with kms it didn't flicker
looonger: adamk: but there were artifacts appearing on kde4 even with kwin desktop effects disabled
adamk: Gotcha.
suokko: Sarvatt: Looks like change to R500 video code and not adding extra bytes to reserved space
looonger: adamk: did you have any problems with 3d flickering?
suokko: Without looking the code safe fix looks like adding 2 to reservation count in the line which error message complains about
looonger: actually it's not a huge problem, open source driver is much better than catalyst anyway but it would be nice if everything worked fine ;)
adamk: looonger: I do not.
looonger: adamk: did you enable kms?
adamk: I do have it enabled, yes.
looonger: adamk: do you use kde?
adamk: Yes, I do.
suokko: Sarvatt: hmm. Not exactly that. But close
looonger: adamk: no problems so far?
suokko: looonger: It would help if you could make a screenshot of artefacts. There is some hw quircks that need to be fixed in driver which open drivers don't know about yet.
adamk: looonger: I've been using that combination for a while now without any issues.
suokko: Sarvatt: http://sprunge.us/fBPB
Rabenklaue: hi, I'm getting (EE) DoSwapInterval: cx = ... errors when trying to run any application using clutter (opengl library) or any other opengl app in general, which displays text.
Rabenklaue: Instead of characters there is only garbage displayed (black boxes). This appears only, if I use compiz.
Sarvatt: suokko: thanks for the pointers, rebuilding a package with that patch for the person having the problem to see if they have any luck with it. here's the log they have getting flooded with multiple gigs of errors https://bugs.edge.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/515548/comments/2
Rabenklaue: My graphics card is a ATI Radeon 200M
stikonas: Rabenklaue: this card is evil, had it before.
Rabenklaue: And I'm using the latest mesa/libdrm and radeon drivers from git
stikonas: though you can check if some compiz extension causes black boxes
stikonas: maybe blur
stikonas: turn of compiz plugins, and check if it works
Rabenklaue: stikonas: I know, ... but atm I do not have a chance switching it
Rabenklaue: I mean my laptop regarding a better graphics card
stikonas: I understand, that's why I gave some tips regarding compiz plugins
Rabenklaue: stikonas: Nope, just deactivated all plugins instead of window decoration and window moving and this doesn't do it. But thanks for the advice.
suokko: Sarvatt: Looks like you may want to update the ddx from git ;)
zhick: w00t
zhick: HoN started working with r300g :D
zhick: well, kinda :>
MostAwesomeDude: Really? :3
amarsh04: is recompiling mesa - nearly 3 million lines of code
Dr_Jakob: HoN?
zhick: well, havent tried ingame yet :>
zhick: but the menu shows up :)
zhick: http://img109.imageshack.us/i/snapshot1n.png/
zhick: will try ingame now... probably brb after horrible X crash :>
mcgregor: erm, wait, before you crash, what is hon? ;)
zhick: Heroes of Newerth
zhick: see heroesofnewerth.com for more details
chithead: hon reportedly does not work properly
zhick: yep, just noticed :)
zhick: well it didnt take X with it
chithead: glsl support needs to mature more
mcgregor: ehh should I know the game? ;) never heard of it before :)
zhick: well, it's just the best game available for linux :>
zhick: admittedly also the one with the most shitty community
Dr_Jakob: zhick: thats a very objective statement
zhick: it's currently in closed-beta, so that might be the reason youve not heard of it yet
Dr_Jakob: I would say that Spring RTS with BA is the best game :)
zhick: Yeah I used to play Balanced Annihilation as well
zhick: youre right that's also a very nice game
chithead: spring rts mostly works but still looks different from fglrx or intel
zhick: it's pretty hard to convince your windows-buddys to play it with you though, HoN has the advantage there.
Dr_Jakob: zhick: huh why? Sprint RTS works on windows as well.
zhick: ya, but it looks very much last-gen. thats a big - for windows-gamers.
mcgregor: oh the game has nice gfx , well, looks a litle like warcraft3+++
zhick: well, the environment yes
zhick: but the models are butt-ugly :>
zhick: whoops?
zhick: hope he didnt take that as an offense or sth...
MostAwesomeDude: Windows gamers are not exactly a bright crowd. :3
MostAwesomeDude: Dr_Jakob: The HoN guys sponsored a LAN I was at a week ago. Decent game.
zhick: well, two of my buddys are rts-nerds and they played a few games of spring and liked it, but they lost interest quickly.
zhick: MostAwesomeDude: So you've got an incentive to make HoN work. Good to hear. ;)
WhiteRabbit56: I have to admit HoN on the radeon driver would be awesome!
MostAwesomeDude: zhick: Actually, my current targets are TF2 and Civ4, games I enjoy much more.
Dr_Jakob: MostAwesomeDude: class?
MostAwesomeDude: Dr_Jakob: Pyro or Medic, usually.
zhick: MostAwesomeDude: Oh well, I guess by the time those two games work (via wine I take it) HoN will be working fine as well. :)
Dr_Jakob: MostAwesomeDude: Ah Medic or Medic myself (yes not a typo).
MostAwesomeDude: Dr_Jakob: Well, it fits your nick. :3
Dr_Jakob: With the occasional sprinkle of Heavy, Eng or Spy.
Dr_Jakob: Hehe :)
MostAwesomeDude: Never could get the hang of Spy. It's a tough class.
Dr_Jakob: Yes, its very hard.
Dr_Jakob: But if you get lucky its hard not to giggle like a maniac playing it :)
MostAwesomeDude: Heh, yeah.
WhiteRabbit56: Real quick let me get this straight before I go and try to build this again:
WhiteRabbit56: I have to build libdrm
WhiteRabbit56: then mesa
WhiteRabbit56: the ddx
WhiteRabbit56: right?
MostAwesomeDude: libdrm, then mesa & ddx
MostAwesomeDude: The important thing is that libdrm is built first.
WhiteRabbit56: alright... i'm gonna try this again as soon as im back at my linux box... the university changed my ip so i can't remote in at the moment...
WhiteRabbit56: mad: do you have any helpful "hints" on fedora 12 to make sure I get this to build right
airlied: WhiteRabbit56: if you have F12 just use updates-testing
airlied: or updates
airlied: and you should only nee dto build mesa if you want newer bits
WhiteRabbit56: updates-testing?
WhiteRabbit56: and then just build mesa?
AstralStorm: which branch is the latest devel drm for radeon and 2.6.33-rc?
amarsh04: does mesa need to be built against libdrm 2.4.17 or later?
WhiteRabbit56: alright i can try that
airlied: WhiteRabbit56: actually F12 updates should have latest libdrm etc
AstralStorm: does drm-radeon-testing go clean into 2.6.33-rc?
WhiteRabbit56: airlied aren't you the maintainer for the mesa-dri-drivers-experimental package?
airlied: WhiteRabbit56: partly
WhiteRabbit56: airlied: i thought i saw your name last time i looked up who maintained that...
WhiteRabbit56: airlied:i just ran git clone git://anongit.freedesktop.org/git/mesa/mesa
WhiteRabbit56: and then ./autogen.sh
WhiteRabbit56: then ran ./configure --prefix=/opt/xorg --with-dri-drivers=radeon,r200,r300,r600 --disable-galliun &&make && sudo make install
WhiteRabbit56: added Section "Files" ModulePath "/opt/xorg/lib/xorg/modules,/usr/lib/xorg/modules" EndSection
WhiteRabbit56: to my xorg.conf
WhiteRabbit56: and rebooted... and it won't boot into X
WhiteRabbit56: what am i doing wrong on fedora 12
Dr_Jakob: WhiteRabbit56: Dunno about the ',' in the module path.
WhiteRabbit56: it's whats on the http://wiki.x.org/wiki/radeonBuildHowTo page
Dr_Jakob: Ok nm then..
airlied: WhiteRabbit56: why touxh xorg.conf?
airlied: it won't pick up the mesa driver from anywhere else
airlied: its hardcoded
WhiteRabbit56: just tried what the wiki page said... so i should remove that line then?
airlied: doesn't use xorg.conf all generally
glisse: damm those rv730 is nasty
WhiteRabbit56: so what do you suggest i do then?
airlied: WhiteRabbit56: what are you trying to do?
WhiteRabbit56: well i guess removing "Section" in xorg.conf lets me boot into X again
airlied: remove xorg.conf
WhiteRabbit56: I want the extension ARB_texture_non_power_of_two
airlied: we don't need one in Fedora
WhiteRabbit56: and i asked in here awhile back and someone told me it had been implemented
airlied: on what hw? r600? then just build mesa and copy the r600_dri.so into /usr/lib/dri/
WhiteRabbit56: yah its an r600
WhiteRabbit56: alright ill try that then
WhiteRabbit56: and ill delete my xorg.conf...
WhiteRabbit56: ah... i had to uninstall mesa-dri-driver-experimental...
WhiteRabbit56: thanks airlied!!!
WhiteRabbit56: it's much appreciated!
arnej: glisse: is the rv730 the radeon hd 4570?
arnej: i have the latest kernel from linus' tree and the latest xorg-stack from git master. when i start my pc with kms enabled everything works fine except of a deadlock when i close my laptop lid. (i have a radeon hd 4570). does anyone know what might be the problem?
WhiteRabbit56: nope never mind... that's using the software rasterizer now...
agd5f: arnej: probably some funky acpi interaction
MostAwesomeDude: airlied: Finally fixed imm inlining, so at this rate, I should have the buffer handling fixed up this time next year. :3
arnej: agd5f: yes it looks like but i can't figure it out. i have had that for quite a long time now. it's difficult to get logs when the machine had a deadlock
WhiteRabbit56: wow i'm stupid
WhiteRabbit56: x64 distro means i need to put it in lib64/dri
WhiteRabbit56: and it works great now
WhiteRabbit56: thanks for all your hard work everyone!
WhiteRabbit56: and thanks for helping me to get it working!
Vash63: I'm having firmware issues with R700_rlc.bin, I have it in both /lib/firmware/radeon and in /usr/src/linux/firmware/radeon but it's still not loading.
Vash63: Is there something else I have to do now in 2.6.33 other than copying the firmware files?
adamk: If you compile radeon into the kernel, or add it to your initrd, you need to compile the firmware into the kernel or add that to your initrd.
FIReun: I wonder if thats why mine is segfaulting
Vash63: Probably my issue...
microm_: I have corrupt glyphs with my AGP card, will using bleeding edge help?
adamk: It was my impression that if you didn't do that, the driver would still work, but it just wouldn't use the new firmware and, therefore, you wouldn't have IRQ support.
adamk: But I can't be sure of that.
FIReun: hmm
BioTube: adamk: it drops all accelleration if it can't find rlc firmware
BioTube: (if it doesn't just crash)
Vash63: Heh, after like a minute yeah.
Vash63: It should really timeout sooner.
Vash63: I thought it had crashed, was going to copy the error.
Vash63: And it continued.
BioTube: I'm pretty sure the timeout's in another part of the code
BioTube: s/code/kernel/
Vash63: Yeah, that's probably not in the radeon code, just generic firmware loading code or something.
Vash63: Just seems that 60s is excessive.
Vash63: I don't think any systems, even slow embedded ones would need anywhere near that long.
Vash63: For like.. any firmware.
eichi: damn ;) without a xorg, only using hal my video driver works great, but with xorg.conf it freezes my system
eichi: without xorg.conf i mean
adamk: Modesetting on r8xx?
airlied: adamk: coming soon
airlied: or the basics are in UMS now
airlied: KMS coming soon
MostAwesomeDude: New firmware needed?
dmb_: airlied, how about 3d :P
dmb_: i heard it wasn't very different then r6xx
airlied: dmb_: no idea yet
FIReun: the future is 2d!
FIReun: 3d was just a fad
FIReun: vrml!
bridgman: I'm pretty sure the Evergreen chips will need new firmware for each of the blocks
Jonimus: airlied: any Idea why my mouse can go off the screen with my dual mon setup with different sized monitors
agd5f: Jonimus: xrandr limitation
MostAwesomeDude: bridgman: Haha, I have you now. Know anything about the MSAA on r300-r500?
agd5f: you actually have one big rectangular desktop
MostAwesomeDude: Or should I just keep experimenting until things start to look smoother?
Jonimus: agd5f: really, because before my mouse was at least stopped at the top of the screen on the smaller mon, now it just keeps going
bridgman: let us know where you are and we can ask our driver devs questions
MostAwesomeDude: bridgman: Well, I figured out how to use AA_CONFIG and MSPOS.
bridgman: we tried to get all the registers for msaa covered but we didn't do a writeup so there are probably holes
MostAwesomeDude: bridgman: I'm investigating AARESOLVE, which I think is the key.
MostAwesomeDude: My current hypothesis, which I'm gonna test tonight, is:
bridgman: yeah, that's the part I haven't looked at yet; I kinda know how the rendering works but not how all the samples get blended back together
MostAwesomeDude: Set the number of samples with AA_CONFIG, set the actual positions with MSPOS.
MostAwesomeDude: Turn on AA resolve, and put an extra-large buffer there.
bridgman: then a miracle occurs ?
agd5f: MostAwesomeDude: there aren't any missing regs as far as I know, but I'm just not sure how they are used
agd5f: Jonimus: depends where the
bridgman: if he was sure you would have seen a Phoronix article with a picture of an antialiased triangle ;)
agd5f: dead area is
MostAwesomeDude: The card copies the super-sampled fragments to AA resolve, and then the blender blends the fragments from AA resolve to the actual colorbuffer.
bridgman: the copy seems redundant, doesn't it ?
MostAwesomeDude: Er, maybe not copy. Maybe the frag shader's outputs are written to AA resolve.
Jonimus: agd5f: the dead area is above the smaller of the two monitors
MostAwesomeDude: I know for sure that if AA resolve isn't set up, then the frag shader writes too many pixels, and ends up clobbering past the colorbuffer bounds.
bridgman: I guess that's a nice simple question we can ask tomorrow - "on 3xx-5xx, how do the MSAA buffers get combined ?" ;)
MostAwesomeDude: Yeah, there's also AA compression on r500, but I don't know how to use that or if we want to use that.
bridgman: I thought there were N colour/depth buffers when running MSAA, and the shader put potentially different values out to each
MostAwesomeDude: The other thing that's interesting to me is whether there are any undocumented vertex attribute formats, but I'm betting that there aren't.
bridgman: our devs recommended staying away from it - colour compression, right ?
bridgman: sorry, colour buffer compression
MostAwesomeDude: Dunno if it was actually color compression or if it only affects AA rendering.
MostAwesomeDude: But either way it's not something we've ever touched.
MostAwesomeDude: Right now I'm in fairly uncharted territory: we have *never* done MRTs before.
bridgman: I think it helped in cases where the same colour was going to be written into >1 samples
MostAwesomeDude: And a lot of the blending regs are untested.
bridgman: just checking something here - AFAIK MRTs and multisample targets are different things, right ?
bridgman: everything made more sense once I learned that ;)
MostAwesomeDude: The other thing that would be appreciated, since I can't find the revenge dumps, is whether there are any recommended positions for MSPOS.
MostAwesomeDude: bridgman: Yeah, MRTs and MSAA are unrelated in theory.
bridgman: ok, we should be able to get that
MostAwesomeDude: But they're both set up using regs that we've never really used.
bridgman: you're not very good at staying on the main roads, are you ?
MostAwesomeDude: Heh.
bridgman: this should be a lot easier to find out than the power management stuff
MostAwesomeDude: Lemme see; there was also something about D3D rasterization centers, but I think I saw that already.
airlied: MostAwesomeDude: just plug in fglrx + revenge ;-)
MostAwesomeDude: airlied: How do I get a revenge? I don't have that tree copied and it was in z3r0's home.
airlied: I have it ;-)
airlied: I even have some dumps
MostAwesomeDude: airlied: If you're feeling generous, you could push a revenge tree. Extra-generous would be a dump with the various smoothing/MSAA stuff enabled.
airlied: I've no idea if had that test
bridgman: well then, the race is on
MostAwesomeDude: I'm nearly certain there was a GL_POLYGON_SMOOTH test.
airlied: MostAwesomeDude: http://people.freedesktop.org/~airlied/scratch/revenge_all.tar.bz2
airlied: git tree + some dumps I had locally
MostAwesomeDude: airlied: As usual, you rock.
MostAwesomeDude: bridgman: Okay, 10.4.6 in the newest r500 docs answers the rasterization center question.
MostAwesomeDude: As well as that hilarious "The hardware will convert 7 into 8 internally."
shadowmaster: will UMS get deprecated at some point soon or can I still continue pulling and building the driver without KMS without worries ?
MostAwesomeDude: shadowmaster: It's already deprecated, but it won't go away for a while.
bridgman: goes to read the latest doc again
Jonimus: yeah like on my SIS chipset Laptop, the dev who worked on the graphics driver gave up long ago :P
airlied: shadowmaster: not really soon
MostAwesomeDude: airlied, glisse : I will write up new patches for the kernel. We need to bump minor version for MSPOS: It's required for correct D3D rendering too, so it'd be nice to have it enabled as soon as possible.
shadowmaster: airlied: well, but UMS is still preferable for 3D I guess?
Jonimus: shadowmaster: the difference is very small if nonexistent at this point
bridgman: 11% of R5xx_Acceleration_v1.4.pdf has been downloaded...
airlied: shadowmaster: not really, I'd expect KMS to surpass UMS in a few places
airlied: and be slower in others
shadowmaster: well, I'd better wait for the 2.6.33 kernel before messing with KMS anyway I guess :P
BioTube: speaking of that, what's the plan concerning libdrm and xf86-video-ati?
BioTube: just curious when libdrm_radeon's going to be enabled by default and when 6.13's going to get released
airlied: now that KMS is out of stating
airlied: we can remove the experimental flags
airlied: does so now
agd5f: MostAwesomeDude: there are no undocumented vertex formats
MostAwesomeDude: agd5f: Thank you very much. :T
MostAwesomeDude: 's life gets tougher
agd5f: MostAwesomeDude: the point/line/poly smooth stuff should be in the latest version of the r5xx guide. you can use coord stuffing and add extra fragment instructions to handle it
bridgman: 52% of R5xx_Acceleration_v1.4.pdf has been downloaded...
bridgman: this is pitiful
Jonimus: bridgman: just wait till you have to open a file of that size :P
bridgman: OK, I've read that "hardware will convert 7 into 8" a half dozen times and still don't get it
bridgman: if you want 7 what do you write ?
bridgman: maybe you just can't have a distance of 7 subpixels
bridgman: and yes, I did proof-read the changes before they went out ;(
FireBurn: Is anyone working on HDMI audio on the newer cards?
Jonimus: FireBurn: there is already some basic support for it
FireBurn: Jonumus: Not for my card
soreau: Hmm. With latest r300g, neverball just segfaults
FireBurn: In fact I have to pass audio=0 just to stop getting a green screen with the latest kernels
Jonimus: FireBurn: hmm well it has shown up as an alsa interface for me for a long time so IDK
FireBurn: Jonimus: Yes but no audio is pumped out until the DRM kernel driver supports it
FireBurn: Or not using KMS and using RadeonHD instead
MostAwesomeDude: bridgman: Yeah, that was what I was thinking. You can't have 7. :3
MostAwesomeDude: soreau: I'll take a look when I get home.
Jonimus: FireBurn: hmm well I never actually tried outputting sound from it.
Jonimus: goes to try
soreau: MostAwesomeDude: I'm updating latest drm-radeon-testing to see if it helps anything
MostAwesomeDude: soreau: make clean, then gdb and get a backtrace, I might be able to spot it from here.
microm_: there is heavy corruption with EXA for my mouse pointer, and most icons on my desktop.
Telek: Radeon X1250 still doesn't support 3d acceleration?
Jonimus: Telek: it should
Jonimus: http://www.x.org/wiki/RadeonFeature check there
Telek: Ok
Jonimus: I have a spare Radeon x300 SE, would using that maybe give me better performance in some cases as it is a better supported card, or will my HD4850 kicks its but no matter what?
BioTube: you need an r600 or r700 for OGL 2.0
chithead: Jonimus: it will work better for s3tc because that is not yet supported for r600
BioTube: unless you want to play with gallium
chithead: but for all other purposes the hd4850 is better
Jonimus: Ok I'll just stick to the HD4850 for not then, unless I get another monitor.
peteweez: I just installed the radeon driver and it works great—much better than radeonhd and much less of a headache than fglrx, but there's one thing that doesn't work right. I can suspend my computer and revive it once, but if I try again without rebooting, it will freeze before it can suspend itself again. What could cause this?
Jonimus: peteweez: thats better than my laptop that doesn't restart the fans upon resume, my laptop was up to 105C before I noticed :/
peteweez: Ouch!
peteweez: I had a problem like that once
Jonimus: yeah, I assume its a dell bios issue, but I'm really not sure
peteweez: The fans would spin up, but the temp sensor would give ridiculous readings
peteweez: I sent it into ASus (it was under warranty), and it's fixed now. :-)
peteweez: Your /etc/default/acpi-support file might not be correctly tailored to your system.
Jonimus: I have another laptop which the ACPI Table just flat out lies about the battery level
peteweez: nice
Jonimus: peteweez: I don't think that is the issue but I'll take a look
peteweez: Well with mine it would give abnormally low readings
peteweez: like 2 or 3°C
peteweez: then it would spike to 127° or 128°
peteweez: obviously a sensor fault
peteweez: try installing a temp monitor on your panel
Jonimus: but yeah I looked through the ACPI table for the battery formula and it just had some random adjustment to make the out put seem sane but it actually had no clue and would jump all over the place
Jonimus: peteweez: I did
peteweez: That's strange
peteweez: Probably a software fault, but no way to tell without trying it out on Winblows.
Jonimus: peteweez: windows did the same
chithead: possibly it reads only the lower byte from the sensor
Jonimus: chithead: I'm really not sure, I only learned enough acpi to try and fix it, but now the battery is totally dead so I matters not
peteweez: Hardware fault probably
peteweez: send it in for repairr
peteweez: hopefully it's under warranty
peteweez: So anyone know what's up with my box>
peteweez: *?
peteweez: I can suspend and revive the box once, but if I try twice, it hangs
peteweez: Suspend worked under fglrx
peteweez: didn't work at all with radeonhd
peteweez: and works once with radeon
microm_: is there a manual for the radeon GPUs to help driver development?
BioTube: http://www.x.org/docs/AMD/
microm_: ty
CartoonCat: Hellos
CartoonCat: so suddenly my system will not boot a kernel with KMS enabled
Jonimus: woot resume is still working :)
soreau: CartoonCat: I thought you had it working yesterday
CartoonCat: any ideas on what might be the issue, and on what to do to get 3d now?
soreau: What happened?
CartoonCat: soreau: me too, afaik
CartoonCat: same kernel i was using yesterday
soreau: Which distro are you using?
CartoonCat: today, it hits the "enabling radeon modesetting"
CartoonCat: another 2 lines then poof, black screen and it never comes back from it
CartoonCat: this would be spot where, i think, it moves to a higher res mode and goes on booting
soreau: Maybe fbcon is not being loaded
MostAwesomeDude: soreau: At home. Which prog wasn't working for you? Neverball?
soreau: MostAwesomeDude: Yes
CartoonCat: i have no idea, and i dont know how to get a dmesg from a failed boot to have someoen look at it
soreau: MostAwesomeDude: I am about to boot into latest drm-radeon-testing here in a sec
soreau: MostAwesomeDude: Yea, it's still the same. running neverball with classic mesa works. With 300g, it segfaults
MostAwesomeDude: soreau: Works here.
soreau: CartoonCat: It shouldn't stop working all of the sudden. You must have done something.. perhaps inadvertently
MostAwesomeDude: soreau: Did you have to trigger it?
soreau: MostAwesomeDude: No, it segfaults immediately
CartoonCat: soreau: your right, i shut down and booted win7
soreau: MostAwesomeDude: I can get some gdb going though, probably
soreau: CartoonCat: I certainly would hope that win7 didn't do anything to alter your linux system, but I've been wrong before ;)
MostAwesomeDude: soreau: Yes, please.
CartoonCat: soreau: well, that is exactly ALL i have done, once kms was working and part of the driver was loading, i went back to working, I needed win7 today, and booted it up, jsut got home and came back to gentoo, and a black screen
soreau: MostAwesomeDude: I suck at this. How should I invoke gdb? :P
MostAwesomeDude: soreau: $ gdb neverball
MostAwesomeDude: (gdb) r
MostAwesomeDude: ...
MostAwesomeDude: (gdb) bt full
MostAwesomeDude: And pastebin.
soreau: MostAwesomeDude: I need to pass LIBGL_DRIVERS_DIR=/blah/blah
MostAwesomeDude: soreau: Put that before the gdb.
soreau: ok
soreau: MostAwesomeDude: Now it has gone to black screen and doesn't do anything beyond that
soreau: Everything's still running though
MostAwesomeDude: soreau: You got the bt?
soreau: MostAwesomeDude: Maybe I should run gdb from ssh
MostAwesomeDude: soreau: Wait, did it lockup?
soreau: nope
MostAwesomeDude: Are you running fullscreen?
soreau: It just took a sec to start, went to black screen with cursor
soreau: Yes, fullscreen
MostAwesomeDude: Oh, you haven't gone back to the gdb.
MostAwesomeDude: It segfaulted, so gdb stopped the program.
MostAwesomeDude: This is your chance to get the backtrace.
soreau: and since I didn't run gdb from ssh, I'll have to do it again
MostAwesomeDude: Use "q" to quit gdb when you're finished.
soreau: ps ax shows neverball is still running
MostAwesomeDude: That gdb session is still going. You'll have to kill it.
soreau: killing gdb brought the screen back
MostAwesomeDude: When gdb pauses the program, X doesn't deallocate the window.
MostAwesomeDude: So the window stays there until you're finished.
MostAwesomeDude: You really should switch to windowed mode. :3
soreau: MostAwesomeDude: http://pastebin.com/m6039db3d
MostAwesomeDude: soreau: Haha, I was afraid of this. Just to confirm, could you run with RADEON_DEBUG=fp,vp and pastebin the entire console log? You don't have to use gdb.
soreau: Of course
soreau: MostAwesomeDude: http://pastebin.com/m4f407c28
soreau: MostAwesomeDude: FWIW, before this mesa update (from maybe last week), I have getting the hex dumps in console when running neverball
soreau: It still works, but it drags ass
soreau: but now of course, with the segfaulting and all
MostAwesomeDude: soreau: Yeah, I fixed the hex dumps in another commit. Could you locally revert 9a1bf52c184b6c0393543fe4bb03c790630b9e21 and try again?
soreau: CartoonCat: So you don't have a backup kernel installed you can boot? Or can you boot with nomodeset, does it work?
soreau: MostAwesomeDude: Yes, give me a few
CartoonCat: soreau: I am in my backup kern right now
CartoonCat: soreau: well, my backup backup kern lol
CartoonCat: i booted my backup, went and changed the modeset line, recomiled, and booted that kern
soreau: MostAwesomeDude: So to clarify, try git revert
MostAwesomeDude: soreau: Yeah, and then make. You don't have to make clean.
soreau: CartoonCat: It should work as long as you have fbcon loading (probably best to have it compiled in) and firmware installed (if you have >=r6xx)
soreau: MostAwesomeDude: ok
CartoonCat: Ii have a rs690
soreau: CartoonCat: And which kernel version again?
CartoonCat: rs690m rather
CartoonCat: gentoo 2 6 31 r6
soreau: hm.
soreau: CartoonCat: Can you try latest vanilla kernel or at least 2.6.32? I suspect it's just a kconfiguration issue though
CartoonCat: well i can up my config.gz if you would like, or look for something specific
soreau: MostAwesomeDude: Yup, it works
MostAwesomeDude: soreau: Okay, yeah, that's what I thought. Dare I hope you have valgrind? I think it's an r300-specific path, but I might be able to nail it down from here.
CartoonCat: soreau: gentoo says 2 6 32 is masked for amd64, but yea i can give it a go
MostAwesomeDude: soreau: Hey, nevermind, valgrind hits it here, too.
soreau: CartoonCat: All's you really need is CONFIG_DRM_RADEON and CONFIG_FRAMEBUFFER_CONSOLE (and possibly CONFIG_FRAMEBUFFER_CONSOLE_DETECT_PRIMARY)
soreau: MostAwesomeDude: Yes I have valgrind here if need be
CartoonCat: soreau: ill check
soreau: of course the agp stuff if it's agp..
CartoonCat: ok config framebufferconsole is set, but detect pri is not (and of course in this one drm_radeon is not)
soreau: CartoonCat: Where you the one with the PCI card?
CartoonCat: its pcie
CartoonCat: nope
soreau: ok
CartoonCat: well, i have it, but im not the one who kept saying AGP when it was pci
soreau: CartoonCat: Do you have more than one output plugged into the card?
CartoonCat: i do not
soreau: remebers *oh yea, the guy with the pci card was using freebsd*
soreau: CartoonCat: Why it would mysteriously stop working is confusing, but try updating to at least 2.6.32 or get a later one from vanilla-sources package and try it
soreau: The only thing I can suspect is it's a kernel related issue
CartoonCat: or hardware ;-)
CartoonCat: my faith in mobile ati is lacking
soreau: Well I don't think rs690 is one of the least problematic cards, but let's hope it's still in working order ;)
CartoonCat: it works in windows, so, the hardare does work
soreau: CartoonCat: and fwiw, did you try cold booting the machine? to power down the card completely after running any other driver on it (like another os driver)
CartoonCat: but that makes my google earthng with out 3d is
CartoonCat: soreau: yea, i had to do that just to turn it off (after the first black screen)
CartoonCat: I even pulled the battery
soreau: That should do it, I expect
soreau: CartoonCat: So MagicSysRq didn't work? it was a lock?
CartoonCat: soreau: correct
CartoonCat: after 3 of them, i booted my working kern, changed the drm setting, recompiled and booted that, works fine otherwise
MostAwesomeDude: soreau: Probable fix pushed.
soreau: MostAwesomeDude: Funky, eh? :)
soreau: MostAwesomeDude: Yup, that did it
MostAwesomeDude: soreau: Awesome.
soreau: MostAwesomeDude: And fwiw, I tested OA today, seems to me that g is just as fast as c without any cons
soreau: I would like to see enemy territory visual errors fixed however
MostAwesomeDude: I don't think I have that game. It's not free, right?
soreau: Seems the game is up to speed but there's several problems with it. In classic, works fine except the sky not being 'updated'. With gallium, it looks like a dark rainbow fantasy, like the game is tripping acid
soreau: MostAwesomeDude: It is totally free
soreau: One of the main reasons I play it
MostAwesomeDude: soreau: Alright, I'll look into it.
MostAwesomeDude: I also need to investigate Spring.
soreau: MostAwesomeDude: This installer file can be slippery if you're trying to get your hands on it et-linux-2.60.x86.run so I will provide a link here http://etfiles.spacedump.net/
soreau: MostAwesomeDude: eosie reported the game working fine on rv530 IIRC, but here on rv350 it looks like this http://omploader.org/vM2Y4dQ
soreau: With classic mesa, the only minor issue is the sky updating, leaving a smearing effect
Jonimus: lol wow, nice soreau
Jonimus: its almost like abstract art :P
soreau: otherwise, both drivers work at a fine speed
soreau: Jonimus: I call it 'a dark rainbow fantasy' ;)
MostAwesomeDude: soreau: Oh wow. That's awesome.
soreau: I certainly thought so :)
MostAwesomeDude: soreau: Anything in dmesg when that happens?
soreau: MostAwesomeDude: That's a good question
soreau: Let me test it
Jonimus: I still can't play torcs on either of my cards at a decent framerate, and changing the quality settings changes nothing, :(
Jonimus: but UT2004 is playable at leas
Jonimus: t
Jonimus: as for bugs and corruption I haven't been able to find any so far :)
soreau: Oh great, fail fail fail
soreau: MostAwesomeDude: I hadn't copied radeong_dri.so to r300_dri.so, so I was testing neverball with classic. Fortunately, it works with gallium as well now :P
soreau: So far as enemy-territory goes, it's still looking not-so-good
soreau: MostAwesomeDude: Absolutely nothing when running the game
soreau: MostAwesomeDude: It is noteworthy that the menu works fine. It's only the loading screen and in-game graphics that are a problem
soreau: I don't recall the box textures loading ok before, but they are now so looks like a bit of progress
soreau: s/box/crates
MostAwesomeDude: soreau: Huh.
soreau: MostAwesomeDude: Absolutely no output from dmesg, that is
MostAwesomeDude: soreau: Alright.
soreau: oh wait
soreau: I might have lied
soreau: MostAwesomeDude: Is it different doing dmesg and tail -f /var/log/messages
soreau: ?
MostAwesomeDude: soreau: Yeah.
soreau: ok let me try again then
dileX_: "r300g: Fix funky segfault." fixes glxgears segfaulting
MostAwesomeDude: dileX_: Ah, good.
soreau: yay
soreau: MostAwesomeDude: All's I see are a few [drm] not detecting due to 00000004
MostAwesomeDude: soreau: Huh, then that's not it. Interesting.
soreau: MostAwesomeDude: Maybe if you have any r3xx you can see what I mean. Like I said, I believe eosie reported this particular game working fine with rv530
soreau: MostAwesomeDude: The good thing about this game is it gives somewhat verbose info in the console output about what it's detecting for graphics
Jonimus: ok this is weird, if I just run glxgears I get no window decoration corruption but if I run glxgears -info the window decorations are corrupted
soreau: Which WM? KMS? gallium?
Jonimus: XFCE
Jonimus: KMS
Jonimus: no Gallium
Jonimus: xfwm4*
soreau: screenshot?
soreau: no corruption happening here