cxo: agd5f, yeah airlied and i figured that out, cheers
cxo: so if dynpm is enabled, is there a governor written to switch modes when necessary?
agd5f: cxo: it's pretty simple right now. just checks number of pending fences
agd5f: shoudl probably implement some sort of variable timer so we don't constantly reclock when you are playing a game for example
agd5f: we'll also need some sort of bw calculator when we eventually enable mem reclocking to make sure we have sufficient bw available for all current clients
cxo: Considering unlike with a cpu, you have power mode "types" like "HD Video", "Low Battery" etc.. which (as far as I know) are not standardised, how will the driver know which clock profile to switch to when you want maximum performance? Is it just going to pick the highest clocks found in the powerplay table?
MostAwesomeDude: How is it unlike a CPU?
MostAwesomeDude: The CPU governors would probably work fine.
MostAwesomeDude: The biggest obstacle is that rapid switching is even more undesirable on a GPU than a CPU.
chithead: why? also conservative governor does not switch rapidly
cxo: MostAwesomeDude, because you have types associated with each profile
MostAwesomeDude: cxo: No, the types are inventions.
MostAwesomeDude: CPUs don't have Folding @ Home types.
cxo: yes i know that
cxo: thats why i asked
cxo: so will the linux driver just pick the fastest clock regardless of what the "type" is, when maximum performance is required?
MostAwesomeDude: The basic ideas laid out by the mainline governors are good guidelines. Always slow, always fast, as needed.
cxo: yes, but what i'm trying to clarrify is, what is deemed as "slow" and "fast" with the graphics driver
MostAwesomeDude: The slowest and fastest clocks.
cxo: ok, thats what i wanted to know :)
cxo: so even if the fastest mem clock and fastest core clock are not in the same clock profile, the driver would set the core and mem accordingly?
MostAwesomeDude: Are there such clock profiles?
cxo: i dont know, i was just wondering what the driver would do, it has to pick an index in the table somehow, what would the criteria be,
cxo: but i've seen some weird ones as well
MostAwesomeDude: Well, presumably you would clock the memory according to the actual demand on it.
cxo: oh you can do that?
MostAwesomeDude: As agd5f was saying.
cxo: that changes throughput, but that will affect latency wouldnt it? if you change the mem clock according to "demand"
MostAwesomeDude: Not sure. I haven't even looked at PM stuff.
cxo: ok, i was just curious thats all
Nightwulf|work: hi all
mcgregor: ahh linux released final 2.6.33 so probably in ~2weeks .34-rc1 will come out :)
mazzachre: Is .33 better for radeon cards? Or same same?
mcgregor: depends what drm you use before. i you used a standard .32 kernel, it brings you some improvement
mcgregor: *i=if
gschwind: hi
airlied: gschwind: I played around a bit more today but hit some roadblocks with the current code to get the ROM
gschwind: airlied, do you need some more things to unblock ?
gschwind: airlied, currently I'm block too, I could try to use an external bios and launch it manualy
okraits: hi guys
airlied: gschwind: no I'll probably rework the code tomorrow on a t500 and see if I can make it find the ROM
cxo: Looks like the first set of game studios have made dx10 a minimum requirement... you know what that means right? all the windows users are gonna be selling their R600 and lower cards cheap
MostAwesomeDude: oh noes
cxo: thats good thing man, for linux users
mcgregor: sorry .. *all* windows user? :) I suppose that would be perhaps 15% of them :)
okraits: short question: is it correct that the HD 5700 series (e.g. HD 5770) isn't yet supported by the radeon driver or does anyone already has a 5700 series card running with radeon?
cxo: now we'll be able to get better fps on glxgears for less money
airlied: okraits: from git it would at least modeset
okraits: airlied: what does that mean?
okraits: airlied: modeset
gschwind: airlied, Ok :) thanks, hold me in touch, I can try some patch in near direct live
airlied: okraits: the current development branch should be able to set modes
cxo: okraits, your screen, will produce colour
okraits: so 2d should work?
cxo: yeah, 2d would work with VESA as well
okraits: yeah
okraits: i'm using vesa atm
cxo: there is no 2d accel with r800
okraits: ah ok
okraits: thanks
okraits: i was surprised that i even got composite effects with vesa
cxo: http://www.x.org/wiki/RadeonFeature
okraits: the only problem i have with vesa (independent of composite on or off) is that my tty is broken after i get out of X
okraits: evergreen is rv810 to rv870, right?
cxo: that could be vesa framebuffer doing something perhaps
okraits: yeah
airlied: gschwind: btw can you make sure you have fbcon loaded
airlied: since tou mention you had no console once you loade di915
airlied: its wierd there is definitely somethin in your VRAM
airlied: gschwind: can you try not loading radeon, then cd /sys/bus/pci/devices/
okraits: anyway, keep up the good work, been using radeon for years on my 9600xt
airlied: grab http://people.freedesktop.org/~airlied/mmapr.c and build mmapr
airlied: then mmapr resource0 0 256 | xxd
airlied: and pastebin the dump
gschwind: airlied, I will try let me switch irc to another computer
gschwind: airlied, ok, I use a kernel 2.6.33-rc8 with switcheroo patch radeon and intel drm set as module
airlied: cool you can do the mmapr before or after loading i915 just don't load radeon
airlied: however make sure you load fbcon before either so you can see console
gschwind: I don't know mmapr :)
airlied: gschwind: just download http://people.freedesktop.org/~airlied/mmapr.c
gschwind: I use ssh, I don't need to see the console ^^
airlied: and gcc -o mmapr mmapr.c
gschwind: airlied, ok
gschwind: airlied, can I spam you with lsmod in private ?
airlied: gschwind: use pastebin or pm it
gschwind: airlied, ok
gschwind: airlied, I use mmapr on /dev/mem ?
airlied: gschwind: no /sys/bus/pci/devices/
airlied: get the pci id from lspci
gschwind: ok
airlied: mmapr resource0 0 1024 | xxd > dump
gschwind: airlied, I get full FFFF patern
gschwind: airlied, before load any module
airlied: oh echo 1 > enable first
airlied: in that directroy
gschwind: airlied, http://pastebin.com/cpngvcsP
airlied: gschwind: cool definitely no sign of a bios
airlied: looks like a GART table
airlied: or something like that
gschwind: airlied, What is that ?
airlied: ah well I'll try and write a patch to ask the ACPI tomorrow
airlied: gschwind: table the GPU uses to access main memory pages, it does translations
gschwind: ok
airlied: gschwind: can to try one more thing
airlied: mmapr /dev/mem 0xfc00 1024 | xxd
gschwind: ok
gschwind: only zero
airlied: actually try 0xfc000
airlied: I'm not very good at reading acpi
gschwind: I can give you a disasembled acpi ?
airlied: you have already
airlied: I'm reading it
gschwind: yes :) I downgraded the bios since I send you the last one
airlied: probably hasn't changed much
gschwind: airlied, the ATPX section havn't changed
gschwind: airlied, http://pastebin.com/YmTVid87
airlied: okay not a bios eiyther, ah well I'll write the patch tomorrow test it works on my bios
airlied: and then see if it works everywhere else
airlied: tv &
gschwind: airlied, ok thanks
gschwind: airlied, I give you data without loadin any module
gschwind: airlied, not acpi function was run :)
galtgendo: now, that 2.6.33 was released (where radeon modesetting was moved from staging to normal), will a release of xf86-video-ati follow shortly, or will we wait for xorg-server 1.8.0 ?
MrCooper: galtgendo: xserver shouldn't matter
Mazzachre: Does others have problems with applications that freeze X? Symptoms are: No input works, except for mouse arrow still moves... Ctrl-alt-backspace does nothing, ctrl-alt-f1 does not change to vt1. Causes: firefox change tab, xchat change tab, change windows in window manager... All inconistent (not all the time), firefox+flash consisten (all the time), though Chromium + flash works
mazzachre: Is this a driver problem? And in that case, how to fix it?
kjeldahl: Mazzachre: Yes, I'm seeing it as well. Moved from FF to Chrome because of it. It seems to lessen the problem, but it still happens.
gschwind: mazzachre, does your Xorg driver are build against the new kernel ?
kjeldahl: I'm on an updated Ubuntu Lucid (Kubuntu flavour) x64 distro.
mazzachre: It is build again .32 kernel, I was considering upgrading to new .33 kernel, would that help?
gschwind: mazzachre, I'm not sure but you should rebuild xorg driver and hald at first place
mazzachre: I am not running HAL
mazzachre: So, rebuild drivers?
gschwind: mazzachre, yes, if you xorg.conf was right before update
gschwind: mazzachre, build mesa can help too
mazzachre: my xorg conf is possibly wrong... I will paste it.. 2 sec
gschwind: mazzachre, if you don't have keyboard it could
mazzachre: http://dpaste.com/164605/
mazzachre: I do have keyboard... in normal operations...
gschwind: mazzachre, ok try rebuild xorg drivers :)
gschwind: mazzachre, and check if /usr/src/linux is linked to your new kernel before
galtgendo: actually, my question was more along the lines of: "do you consider 2.6.33 release as a good enough reason to do an official release of KMS driver ?"
glisse: yeah we likely should do that
galtgendo: and a one more note on 2.6.33: while driver was moved out of staging, you didn't update help text in menuconfig (at least it looks like it needs to be updated for that reason)
mazzachre: gschwind, I am compiling and installing new kernel now... After install should I compile drivers before reboot? Or no matter?
gschwind: mazzachre, you should I think
mazzachre: gschwind, ok
mazzachre: gschwind, Did you look at my xorg-conf file? Is it all bad?
gschwind: mazzachre, I made a kick check but if it worked, it should still work
mazzachre: I works... even if I have no idea how the input section works by now... Last time I had to do something with xorg.conf on a machine it was back when all modelines and input devices had to be configured manually... And USB support was sketchy and event device driver was not invented if I remember correctly
mazzachre: I have manually (by-path) setup the keyboards... and the synaptics input... And have generic /dev/input/mice support that seems to work with the rest of things I have added... (I have 3 mice now :D)
gschwind: mazzachre, rebuild udev too
gschwind: mazzachre, check if event dev name still valid :)
naquad: hi
naquad: what's the difference between propietary driver and radeon-hd?
glisse: well opensource or not
naquad: is radeon-hd able to provide direct rendering or big desktop?
naquad: ?
[Enrico]: naquad: radeon is able to
naquad: ok
naquad: tried to configure fglrx for about an hour, but that sh*t doesn't work as it supposed and doesn't set up 2 monitors properly
glisse: radeon & radeonhd have the same featureset from user pov
naquad: glisse: i'm hope radeon hd is better supported and has easier configuration
naquad: i already want to throw away this stupid video card :(
jcristau: radeon should mostly "just work"..
naquad: it would be perfect
Nightwulf|work: can second that
naquad: any ideas how to delete fglrx?
naquad: does it have some unistall
naquad: ?
naquad: found it
mazzachre: What is the difference between radeon and radeonhd drivers now? I get conflicting info from google...
glisse: naquad: there is 2 different driver radeon & radeonhd
glisse: naquad: note that from feature point of few opensource is several years late against fglrx
naquad: years?!
dileX_: if you have an old radeon card (r100-r500) fglrx closed source driver is *not* an option
dileX_: for r600+ radeon OSS driver development is still work-in-progress
naquad: i've got radeon hd 4870
dileX_: naquad:
naquad: dileX_: it seems to support my card
dileX_: if you build from GIT yes
dileX_: and have fitting kernel and mesa
dileX_: see topic wiki/howto and read
glisse: naquad: feature point of view, iirc we only support gl 1.5 so far
glisse: gl 2.0 is wip
dileX_: is for renaming "radeonBuildHowTo" -> "radeonWIPdevelopment"
kjeldahl: naquad: FWIW, I'm running a Radeon 4650 successfully under Ubuntu Lucid alpha, with GL compositing (read eye-candy) successfully. Tried getting early driver up on <32 kernel 3-4 months ago usuccessfully, but now it works with Lucid alpha.
naquad: a stupid question: how do i disable xrandr in xorg.conf?
dileX_: dunno, but why?
twnqx: did you try Section "Extensions"\n Option "xrandr" "disable"\nEndsection ?
twnqx: (just a guess)
naquad: twnqx: didn't work :(
dileX_: seems to be: Option "RandR" "false" (Section "ServerFlags")
mazzachre: How to set my screens via RandR in xorg.conf? Currently the first thing I do when I log in is to start a terminal and type...
naquad: dileX_: didn't help either :(
dileX_: naquad: did you look into Xorg.log or where do you know?
naquad: dileX_: i'm running aticonfig --list-pair and it says randr is enabled
dileX_: restarted X?
naquad: dileX_: many times
dileX_: normally you use xrandr command
naquad: dileX_: geez... fglrx says there's randr enabled, xrandr says there's no randr enabled....
dileX_: fglrx is closed source and not supported here, try #ati
naquad: i remember, just thought it's easy to make
dileX_: naquad: as I said try #ati for fglrx (and read topic of channels you join)
adamk: mazzachre: http://www.thinkwiki.org/wiki/Xorg_RandR_1.2 and http://wiki.debian.org/XStrikeForce/HowToRandR12
bragon: Hi
bragon: I'll test the 2.6.33 with r600 with KMS.
bragon: (it compiling)
bragon: but it's ok with the OpenGL now ?
adamk: bragon: If your libdrm, mesa, and ddx are new enough, yes.
xming: it's ok-ish
adamk: 2.6.33, alone, doesn't provide opengl acceleration.
bragon: of course
bragon: i need to recompile libdrm/ddx and mesa
Terman: chcat: detecting retransmissions in a mostly correct fashion is neither hard nor easy. Using the wireshark infrastucture would probably be easier
mazzachre: Hmm... still seem to have stability problems
mazzachre: Upgraded drivers, X and mesa from git
mazzachre: and kernel to .33
adamk: mazzachre: You'll probably have to describe the problems better than that :-)
edwin: I just realised why wine was showing all corrupted graphics. It needs the 32-bit dri!
adamk: edwin: Yeah, it's a 32-bit application :-)
edwin: yeah but I didn't know the DRI is loaded by each app, I thought it is loaded by the X server
edwin: which is 64-bit
adamk: The client applications link to opengl libraries and drivers.
adamk: The X server does load the DRI driver, too, but only for AIGLX>
edwin: perhaps the radeonBuildHowto wiki should be updated to mention that you have to build a 32-bit libdrm and mesa, and put into /opt/xorg/lib32 (and add that to LIBGL_DRIVERS_PATH as well)
adamk: If you build 32-bit mesa properly, you don't need to set LIBGL_DRIVERS_PATH
edwin: I don't install to /usr, since the next time I upgrade the Debian package it'll overwrite it
adamk: My point still stands.
adamk: It doesn't matter where you install Mesa. If you build it with the correct options, you don't need to specify LIBGL_DRIVERS_PATH
mazzachre: adamk, Ya, I know... The problem is, apparently when an application tries to draw itself, X dies... sometimes... It is hard to diagnose...
edwin: adamk: without LIBGL_DRIVERS_PATH it tells me it can't find r600_dri.so
edwin: ./configure --prefix=/opt/xorg --with-dri-drivers=radeon,r600,swrast --disable-gallium --enable-32-bit --libdir=${exec_prefix}/lib32
edwin: thats what I did
edwin: and of course the application looks for libGL in /usr/lib
edwin: so it won't find the new libgl in /opt/xorg
suokko: edwin: Unless you add /opt/xorg/lib to ld.so.conf
adamk: Use --with-dri-driverdir
adamk: Then the libGL you build will look in the correct location.
mazzachre: Firefox seems quite consistent, when it starts a page with flash content on it, as soon as the flash content starts to load, X freezes. That is, output is usually shown (but sometimes goes to black) mouse can be moved around, but clicking does not do anything, keyboard (either of them) does nothing... Ctrl-alt-backspace or ctrl-alt-f1 does nothing
suokko: adamk: oesn't help if ld doesn't load corrent libgl.so
edwin: well I need to set LD_LIBRARY_PATH then
adamk: suokko: Right, but LIBGL_DRIVERS_PATH doesn't have anything to do with the location of libGL :-)
adamk: edwin: As suokko, add the path to the new libGL to /etc/ld.so.conf
edwin: I did
edwin: ldd still shows /usr
adamk: edwin: If you listed it first, it would link against the newer libGL
suokko: edwin: You need to restart to make ld reload configs
adamk: (After running 'ldconfig' of course)
mazzachre: DynamicClocks is not working for radeon r600?
adamk: Errr... You do not need to restart.
suokko: or ldconfig
mazzachre: (just tried to enable it now)
edwin: yeah I rebooted already
edwin: and ldconfig is looking in /opt/xorg/lib and /opt/xorg32/lib
edwin: but that still doesn't override /usr/lib
agd5f: mazzachre: doesn't do anything for r6xx+
edwin: since the application I am running is in /usr/bin
adamk: edwin: It doesn't matter where the application is that you are running. If ld.so.conf is setup properly, it will always use the newer libGL.so file.
mazzachre: agd5f, ok. thx
adamk: edwin: If it's not, then /etc/ld.so.conf is not right.
edwin: well /usr/lib doesn't appear in ld.so.conf at all
edwin: it only has /usr/local/lib, /lib/x86_64-linux-gnu, /usr/lib/x86_64-linux-gnu, /opt/xorg/lib, and /opt/xorg/lib32
mazzachre: But more annoying than firefox+flash that I know is unstable is XChat, which have AFAIK always been rock stable and never misbehaves. Sometimes when I change "tab" (channel) the same behaviour... It have also happened a few times for my windowsmanager (fluxbox) when switching between large windows...
edwin: adamk: it could be due to prelink
edwin: that it loads from /usr/lib
edwin: checking
edwin: libGL.so.1 (libc6,x86-64, OS ABI: Linux 2.4.20) => /usr/lib/libGL.so.1
edwin: libGL.so.1 (libc6,x86-64) => /opt/xorg/lib/libGL.so.1
edwin: libGL.so (libc6,x86-64) => /opt/xorg/lib/libGL.so
edwin: glxinfo links to libGL.so.1
edwin: hence it gets /usr/lib
suokko: edwin: Looks like OS ABI tagwas removed from the latest mesa which is causing the wrong load order
edwin: looks like Debian builds with --enable-glx-tls
edwin: trying with that now
edwin: libGL.so.1 (libc6, OS ABI: Linux 2.4.20) => /opt/xorg/lib32/libGL.so.1
edwin: libGL.so.1 (libc6, OS ABI: Linux 2.4.20) => /usr/lib32/libGL.so.1
edwin: yep looking good now
edwin: ./configure --prefix=/opt/xorg --with-dri-drivers=r600,swrast --disable-gallium --enable-32-bit --libdir='${exec_prefix}/lib32' --with-dri-driverdir=/opt/xorg/lib32 --enable-glx-tls
adamk: Woohoo.
adamk: I've not had to use --enable-glx-tls myself, but I was sure it could work if it was built properly.
edwin: hmm dri-driverdir should have been /opt/xorg/lib32/dri
edwin: it links to correct libGL now though :)
adamk: Ahhh, yeah, you need the full path to the driver directory.
edwin: btws progs/glsl doesn't obey -m32
edwin: I had to manually add it to LDFLAGS there
suokko: edwin: You could make patch and send it to mesa3d-devel
edwin: suokko: yeah will do that later
edwin: its still not picking up the dri though
edwin: libGL error: unable to load driver: r600_dri.so
suokko: git send-email is nice tool for patches
adamk: Check the output of 'LIBGL_DEBUG=verbose glxinfo' to see if it's checking the correct path or throwing another error.
Ronis_BR: is anyone using kernel 2.6.33 + drm-radeon-testing?
mazzachre: Ronis_BR, Probably not on same distribution, but I am using 2.6.33 and radeon driver from git
Ronis_BR: mazzachre: wondering if there are many conflicts between kernel 2.6.33 and drm-radeon-testing
mazzachre: Ronis_BR, It seems to run about as stable as it can be currently... I had more problems with yesterdays build of driver and kernel .32
Ronis_BR: i understand
mazzachre: Ronis_BR, Have only had 1 crash so far with this build...
Ronis_BR: good
Ronis_BR: artifacts?
dileX_: Ronis_BR: there are several merge conflicts - maybe its good to wait till there is a drm-next GIT branch on top of .33
Ronis_BR: dileX_: that is my fear
Ronis_BR: but ok, I'll wait
edwin: adamk: this is first error from LIBGL_DEBUG=verbose glxinfo libGL error: unable to load driver: r600_dri.so
edwin: well maybe I should restart my X
edwin: yep that was it, I had to restart X to make it pick the new dri up
edwin: adamk: all good now without sett LIBGL_DRIVERS_PATH
edwin: suokko: do I need to be subscribe to send patch to mesa3d-devel?
suokko: edwin: no. It jsut takes longer to arrive to ML because of moderation
suokko: edwin: sorry, it is mesa3d-dev
edwin: suokko: k subscribed
edwin: suokko: sent
fufler: hi all. is there any way to downclock my ati hd 3870? i use xf86-video-ati driver on gentoo x64
suokko: fufler: If yo uhae 6.12.99 then yes. ForceLowPowerMode (see man radeon)
fufler: suokko, i have 6.12.4-r1, it's from portage. should i upgrade it ?
xming: will drm-radeon-testing be rebased on 2.6.33? Or can I pull drm-radeon-testing on top of 2.6.33?
edwin: adamk: could you (or someone else) update the wiki with the information from here: http://paste.debian.net/61468/
edwin: I described how I built mesa on my Debian, as 64-bit and 32-bit without LIBGL_DRIVERS_PATH
edwin: how to check that its the correct one thats picked up, etc.
Neo_The_User: I have a freedesktop wiki account
Neo_The_User: what wiki are you talking about?
Neo_The_User: also when you cross compile (./configure --prefix=/usr --libdir=/usr/lib32 CFLAGS="-m32 -O2 -g") two things. its usually just /usr/lib and second, its best to specify an actual target
suokko: edwin: It is wiki so everyone can edit it ;)
suokko: You just need to register
edwin: this wiki http://wiki.x.org/wiki/radeonBuildHowTo
edwin: ah ok will register
Neo_The_User: -m32 never works for me though. might want to change that. ;)
edwin: Neo_The_User: what error do you get?
suokko: Neo_The_User: -32 requires multilib (same for -m64 in 32bit system)
suokko: -m32 that is
Neo_The_User: what about the CHOST flags?
Neo_The_User: in my experience... _just_ using -m* doesn't do the job. I have to tell it literally everything
edwin: in Debian it does do everything if you have gcc-multilib installed (and probably binutils-multiarch too)
edwin: depends how you built your gcc
Neo_The_User: ah so in debian they change that.
Neo_The_User: i build my own gcc
edwin: did you build --enable-multiarch?
Neo_The_User: yes
Neo_The_User: and --enable-multilib is on by default
edwin: well thats all that should be needed
edwin: although Debian tends to use multilib different than everyone one
Neo_The_User: ohhh.
edwin: Debian has 64-bit /lib and /lib64
edwin: and 32-bit /lib32
suokko: Neo_The_User: You will also need all 32 bit depedencies
edwin: Fedora has 32-bit /lib and /lib32, and 64-bit /lib64
Neo_The_User: suokko: how do you get the 32 bit depenencies on a custom distro built from scratch then?
suokko: Neo_The_User: Compile everything twice. once with -m64 and once with -m32
Neo_The_User: because i dont have that. which explains it now
Neo_The_User: oh ok
dileX_: edwin: one request for wiki-updates: please put a comment. thx.
edwin: dileX_: what comment?
Neo_The_User: the edit summary
edwin: ah ok
dileX_: edwin: you will see if you edit
Neo_The_User: edwin: your KMS dependancies are off
Neo_The_User: you want to turn off stuff too. radeonfb etc.
edwin: it was never on :)
dileX_: edwin: I would add a separate paragraph for "CONFIG_DRM_RADEON=y, CONFIG_DRM_RADEON_KMS=y +++ #CONFIG_FB_RADEON is not set. maybe below "Prereq KMS" section if you like
edwin: I'll make the edit incrementally: first the ld.so.conf stuff
edwin: then the kernel stuff if you want to
edwin: and then the Debian specific notes
edwin: ok?
Neo_The_User: also its /lib/firmware in most distros
Neo_The_User: and you want r660_rlc.bin too
edwin: yeah but I built radeon into the kernel, not as module
Neo_The_User: same here
Neo_The_User: but r600_rlc.bin is still required
Neo_The_User: http://pastebin.com/Trb5Ji7P thats _my_ list
Neo_The_User: some optional things may be checked but thats what i use
edwin: depends on your card, mine is RV730
edwin: anyway I'd rather not update the kernel stuff
edwin: it depends very much on my card and the way I build my kernel
Neo_The_User: i have a r700 too
Neo_The_User: r600_rlc.bin is yet again... still required
edwin: I don't have it
fufler: suokko, can you help me? what is engine clock? i want to decrease memory and core clocks. how can i do this?
dileX_: edwin: dunno, if 64/32-bit building *and* distro-specific is so interesting for others. IIRC the howto should held as much possible not distro-specific.
dileX_: edwin: what about a troubleshooting section at the end about 64/32-bit thingie?
suokko: fufler: engine is core. Memory can be done only if you build special development kernel
adamk: It's probably not a bad idea to mention that 32-bit applications (such as doom3 and wine) require 32-bit libraries.
dileX_: edwin: --with-dri-driverdir=/opt/xorg/lib32 missing in your paste?
adamk: It's not unusual for people to pop in hear saying that they have 3D acceleration working, but wine says they don't.
dileX_: (forget it - its included)
fufler: suokko, when i play games (for example, warsow), after some time my system hangs: static image and cycled sound. i think this caused by vpu recovery feature of ati cards. under windows i use rivatuner to reduce core/memory clocks. i want to do the same under linux. should i set LowPowerModeEngineClock to desired core frequency?
edwin: dileX_: added the ld.so.conf stuff
edwin: dileX_: will add the 32-bit thingy at troubleshooting
Neo_The_User: fufler: i think VPU recovery code only exists within FGLR... oh no my eyes.
fufler: Neo_The_User, i don't exactly know. but i hope that clock reducing should help
Neo_The_User: AMD Overdrive doesn't support open source DDX. :(
Neo_The_User: i wonder if ATIPower does
suokko: fufler: Or better chasis fans ;) But I think you could set engine clock to what every you wich if you force the low power mode and set the engine clock in xor.conf
Neo_The_User: doesn't look like it
Neo_The_User: suokko: do you know the syntax for overclocking in xorg.conf?
Neo_The_User: or can you only underclock?
fufler: suokko, i think, that my card has "bug". there is no thermal sensor, but when i touch it with my finger, i doesn't seem so hot. i've changed settings, now rebooting
suokko: Neo_The_User: None has added any OC features. And unless there is coming some OC fan to contribute it is very low priority.
edwin: dileX_: http://wiki.x.org/wiki/radeonBuildHowTo#Configuringsystemtoloadmesaandlibdrmfrom.2BAC8-opt.2BAC8-xorg http://wiki.x.org/wiki/radeonBuildHowTo#Troubleshooting32-bitapplicationson64-bitsystems
edwin: so far I haven't added my Debian specific stuff
Ghworg: Can dynpm be switched on after the module is already loaded or does it need to be a kernel/module param?
suokko: Ghworg: It should be possible to change from sysfs if there was interface
Ghworg: suokko: But there isn't an interfae currently?
suokko: points at the kernel source
Ghworg: was trying to avoid that ;-)
suokko: lovely assertion in libc: test: malloc.c:3074: sYSMALLOc: Assertion `(old_top == (((mbinptr) (((char *) &((av)->bins[((1) - 1) * 2])) - __builtin_offsetof (struct malloc_chunk, fd)))) && old_size == 0) || ((unsigned long) (old_size) >= (unsigned long)((((__builtin_offsetof (struct malloc_chunk, fd_nextsize))+((2 * (sizeof(size_t))) - 1)) & ~((2 * (sizeof(size_t))) - 1))) && ((old_top)->size & 0x1) && ((unsigned long)old_end & pagemask) == 0)' failed.
edwin: suokko: did you corrupt the heap?
suokko: edwin: yep
fufler: suokko, thnx, warsow not hangs anymore
Neo_The_User: warsow?
Neo_The_User: nevermind
Neo_The_User: its exactly like openarena
fufler: Neo_The_User, i've never played openarena. is it pretty cool? )
Neo_The_User: its OK i guess. more recently I've been dual booting with windows 7 x64 playing DiRT 2, and Operation Flashpoint Dragon Rising
suokko: openrena=quake 3 clone
fufler: i've played just first flashpoint. that was cool )
Neo_The_User: yeah codemasters does a pretty good job
Neo_The_User: does anybody see white artifacts on 1080p flash videos on youtube at the top?
Neo_The_User: im using stock fedora 12 (not modified)
kdekorte: Neo_The_User: Are they white rectangles that are about 1/4" long and appear in a pattern?
kdekorte: I see that sometimes in glxgears
Neo_The_User: more like thin lines
kdekorte: I believe that problem has been fixed in the drm-radeon-testing kernel
kdekorte: yeah, sounds similar
kdekorte: which fedora kernel are you using?
Neo_The_User: 2.6.31.12-174.2.22.fc12.x86_64
Neo_The_User: its ancient
Neo_The_User: maybe i should go rawhide
kdekorte: Yeah that is old... might try this one: http://koji.fedoraproject.org/koji/buildinfo?buildID=158328
kdekorte: but you need to ensure you have libdrm, plymouth and mesa all updated
melkor: 2.6.31 is ancient?
kdekorte: and probably the xorg-x11-drv-ati
kdekorte: for F12 it is
kdekorte: one of the first kernels for it
kdekorte: all 4 months ago... :)
Neo_The_User: melkor: everything is ancient when you compile everything from git, cvs, and svn for several months on a hand rolled distro ;)
Neo_The_User: i switched to fedora so i could just use it, rather than just build it in infinite loops
kdekorte: Neo_The_User: a nice way to get koji builds: yum install koji
kdekorte: then "koji download-build --arch=[x86_64|i686] --arch=noarch [build#]
kdekorte: so you will need to select your arch either x86_64 or i686 and then the build # which is 158328
Neo_The_User: ah ok thanks!
kdekorte: and then when the files are downloaded use "yum install *"
kdekorte: that will update / install the kernels as needed
kdekorte: I usually do koji downloads in a clean folder
dileX_: edwin: nice. any changes to ddx to mention?
Neo_The_User: kdekorte, will this cause selinux to flip out?
Neo_The_User: like without creating a policy
Neo_The_User: also the kernels arent in menu.lst
melkor: So does radeon drivers work with compize at all?
suokko: melkor: of course it work
Neo_The_User: yeah its great
Neo_The_User: big buck bunny 1080p trailer still chokes for me though
suokko: Neo_The_User: 64bit system?
melkor: how about xorg-server 1.7 ?
Neo_The_User: yes
suokko: :/
Neo_The_User: xorg servr 1.7 works fine for me
Neo_The_User: with krh's invalidate tree
Neo_The_User: *flushinvalidate
Neo_The_User: i think...
suokko: Neo_The_User: Can you give me the link to the video? I will later chekc what is slowing down the playback
Neo_The_User: http://mirror.cessen.com/blender.org/peach/trailer/trailer_1080p.mov
edwin: dileX_: I just installed xserver-xorg-video-radeon 6.12.99+git20100201.a887818f
Neo_The_User: i have yadif x2 deinterlacing though
edwin: dileX_: didn't build any DDX/EXA/whatever drivers myself
Neo_The_User: to really do a stress test
Neo_The_User: suokko: you can also download it: http://www.bigbuckbunny.org/index.php/download/
Neo_The_User: i did the 1080p mp4 one
kdekorte: Neo_The_User: Unknown about selinux... I don't use it
Neo_The_User: yet you use fedora?
kdekorte: Yup, I almost always disable it... doesn't give me anything
Neo_The_User: /etc/selinux/config?
kdekorte: disabled
Neo_The_User: ok
Neo_The_User: thanks :D
kdekorte: That Buck Bunny trailer really beats up the machine
Neo_The_User: haha yeah
suokko: Also it beats up my bandwidth :/
suokko: take for ever to download
kdekorte: I'll have to try it on my intel chipset
kdekorte: my intel g45 machine plays it in gnome-mplayer using xv. The cpu in that machine is a hair faster, but both 64bit fedora12
Neo_The_User: try using yadif x2 deinterlacing with vlc
Neo_The_User: or smplayer i forget
Neo_The_User: one of them has it
kdekorte: Neo_The_User: gnome-mplayer had yadif enabled on the ATI machine, the intel one did not. I enabled on the intel and got a little stuttering as the tree was zoomed in on.. CPU was pretty much 100% on both
Neo_The_User: what cpu you have in there?
Neo_The_User: my 2.2 Ghz core 2 studders like mad
kdekorte: Neo_The_User: Q6600 on the ATI , T9400 on the Intel
kdekorte: 2.4Ghz and 2.53Ghz respectively
Neo_The_User: oh ok
suokko: Do you use multithreaded decoder?
kdekorte: suokko: nope just mplayer from rpmfusion on the laptop and mplayer from svn on the Q6600
melkor: so if I want to enable the radeon drivers in the simplest way that gets 3d hardware support is the wiki in the title going to get me there?
melkor: (I'm using ubuntu
melkor: I'd sure like to avoid git if possible.
Neo_The_User: use the x edgers ppa then
Neo_The_User: it might be ancient though
Neo_The_User: git is the best way to go
BioTube: mesa >= 7.6.1 supports 3d
BioTube: for r600 cards
suokko: melkor: xorg-edgers ppa should point to lauchpad if you google for it
suokko: +mainline kernel is good idea
melkor: BioTube: how about for hd3650
BioTube: melkor: that's r600
melkor: It looks like xorg edgers is at 7.8
Neo_The_User: ill be back in a few hours :)
BioTube: melkor: that includes OpenGL 2.0 supports
melkor: I'll give this git a try once, last time I put xorg edges into my repos I had to start from scratch.
melkor: whats the difference between radeon and radeonhd ?
suokko: melkor: There is purge-ppa script
BioTube: melkor: radeonhd is deprecated
melkor: great, that makes the choice simple
melkor: okay so I built/installed xf86-video-ati, now I need to modify my xorg.conf, I'm not sure what I should put in my 'device' section
kdekorte: melkor: try removing xorg.conf, you most likely will not need it
kdekorte: if you need a driver name use 'ati' or 'radeon'
melkor: okay I'm working on the mesa stuff now, is there a way I can check to see if the xf86-video-ati installed correctly?
melkor: I'm also a bit confused by which branch of the mesa drivers I should build. I want 3d support so should I git checkout -b mesa_7_8...
melkor: or should I use the glxinfo and git checkout that (version 7.4).
suokko: melkor: Just use the master
suokko: But if you hve mesa 7.4 now you will need new libdrm first
suokko: assuming libdrm is that old too
melkor: suokko: There is a libdrm that is in the std. ubuntu repos.
melkor: Where should I get the newer version?
suokko: melkor: either from git or xorg-edgers
suokko: there is link in build how to
melkor: k I think I found it.
ghepeu: hello
ghepeu: soreau, I upgraded today to 2.6.33. I tried KMS (on a r300 card) and apparently there's something wrong with vsync?
ghepeu: (not soreau, **** autocompletion)
ghepeu: with 2.6.32 and 2.6.33 rc2 or rc3 there was no tearing in compiz, for example when moving a console window horizontally, now there's tearing again
ghepeu: somebody mentioned http://bugs.freedesktop.org/show_bug.cgi?id=26240, but I also found that a xf86-video-ati commit should've fixed the problem...
suokko: glisse: I'm creating radeon_bo_manager and then calling radeon_bo_open with CPU domain set. That fails without any dmesg entry. Is there any other drm calls that I should make before bo open?
suokko: That one returns EACCESS :/
glisse: suokko: you are likley missing #define _FILE_OFFSET_BITS 64
glisse: before including anythings
suokko: glisse: yes
glisse: oh this should only affect mmap iirc
suokko: but still EACCES from kernel
glisse: suokko: you are root ?
suokko: no. But I'm part of video group
glisse: i think bo create shouldn't need root
suokko: And open drm device works
suokko: I check the return values
glisse: suokko: my r600_winsys as whole code for standalone init and it's working
suokko: I will look from there what I'm missing
glisse: well if you remove the modesetting stuff it should work as non root
glisse: it's all in radeon.c
suokko: I jsut want GTT and VRAM buffers for memory speed testing ;)
glisse: i did have such test prog
glisse: i think it got overwritten by r600winsys
suokko: glisse: I have write a assembler memcpy that at least can outperform memcpy a lot when using cached source and destination
suokko: 320M/s for memcpy while my assembler version cn push cloe to 500M/s
glisse: libc should already provide such
suokko: glisse: It doesn't use block optimzation that gains huge speed up for large copies
suokko: at least not one in my system
suokko: My libc doesn't even use mmx or sse optimizations that would provide small performance boost too
cxo: if there is ever a scenario where mmx or sse is needed and wasn't available, i think the performance boost from having mmx or sse would give you a VERY LARGE performance boost
suokko: cxo: memcpy is not going to benefite much from mmx or sse because old fashioned movsl instruction is well optimized in cpu
suokko: It is more about trying to improve the cpu prefetecher to gain performance boost
edwin: suokko: I think you might get a performance gain from movntq, since the nontemporal moves don't keep the data in the cache
edwin: which you don't want for video stuff anyway
edwin: (keeping in cache stuff you wrote to video mem)
suokko: edwin: yes
suokko: Too bad there was null pointer dereference in kernel :/
suokko: glisse: runnig as root allowed me to run the test app
suokko: But the test application found null pointer deference in pool allocator
suokko: (Yes. I had writen that naively to allocte and deallocate buffers all the time)
suokko: How does mesa gain right to call radeon_bo_open? :/
melkor: If I want to use kms I need to build libdrm with kms support?
BioTube: 2.4.18 builds with radeon KMS support by default
BioTube: for 2.4.17, you need to pass --enable-radeon-experimental-api
melkor: BioTube: so when it says libdrm will be compiled with... with the options the fact that it says libdkms no, is not refering to radeon kms support?
BioTube: no
BioTube: libkms is an abstraction meant for simple users and doesn't support radeon yet anyway
melkor: how about dri2proto? I have version 1.99x what is a good place to update that. I tried to d/load it and build it from source and I need xorg macros which elude me.
cxo: git://anongit.freedesktop.org/xorg/util/macros
Curan: Is there some schedule on when a DDX with KMS support will be released? (AFAIK it's currently only in Git, or am I mistaken?)
kdekorte: Curan: depends on your distro, it is in Fedora 12
Curan: kdekorte: hm, it is in Debian Experimental too, I meant more: a stable DDX release like the current 6.12 branch
cxo: stable is just in name, in reality no one does the type of testing to ensure a particular version is "stable", its more of a cornerstone for a particular feature of sorts
Wizzup: http://paste.pocoo.org/show/182700/ Issue remains... :-(
okraits: does radeon do some kind of fan management?
edwin: is there a temperature sensor on Radeon HD4650, and can the driver see it?
mjg59: kdekorte: No
edwin: sensors-detect doesn't find it
mjg59: kdekorte: Oops, sorry, mistab
mjg59: okraits: No
okraits: mjg59: but catalyst does?
mjg59: I have no idea
mjg59: I've never used it
Wizzup: soreau: didn't you tell me that issue would probably be resolved in 33? I was wondering if you perhaps knew the bug id
okraits: mjg59: do you know if there's any possibility to control the fan?
mjg59: okraits: If fglrx doesn't, then I don't think so at present
okraits: mjg59: because i have a hd 5770 (which isn't supported by radeon now anyway) and the fan drives me crazy - it's no problem with catalyst though
suokko: okraits: There should hw control of fan if you reduce gpu clocks
okraits: i think that catalyst controls the fan dependant of gpu usage
okraits: even under linux
okraits: at least it's quiet with catalyst
Obscene_CNN: it should be temperature driven
okraits: yeah
Obscene_CNN: just because the clock is slowed doesn't mean the chip isn't hot
okraits: but for catalyst i have to downgrade xorg-server
okraits: Obscene_CNN: of course
kdekorte: okraits: that is the radeon dynpm option, it is still work in progress on r6xx and r7xx and not started on r8xx from what I understand... so if the fan is loud it might be a few months until it is fixed in radeon.
okraits: kdekorte: ok, thanks
spstarr: glisse: still waiting to test your mega patches
spstarr: against 2.6.33
melkor: is there any chance that the radeon drivers are going to work with 2.6.33?
agd5f: melkor: they work fine
melkor: great.
melkor: mesa is finally building for me I should have it up and running shortly.
odinsbane: Okay I've followed the directions built mesa, and installed it make/make install and my glxinfo hasn't changed.
odinsbane: I'm using ubuntu, and they mention the 'distribution packages' often have patches.
BioTube: if you installed it to a different prefix than your distro does, it might be linking against the wrong libGL
odinsbane: well I let it install in the default prefix, so I assumed it would be in the system one. I'll look for 'libGL's
Neo_The_User: LibGL is either in /usr/$libdir/dri or /usr/$libdir/xorg/modules/dri usually
odinsbane: yeah it looks like the system version is in /usr/ and it looks like the default version is in /usr
odinsbane: er /usr/local
BioTube: Neo_The_User: no it's not
BioTube: that's where the dri drivers live
Neo_The_User: ah right/ usr/$libdir
Neo_The_User: */usr/$libdir
Neo_The_User: got mixed up. happens a lot to me. sorry
odinsbane: well I changed the prefix and now it has replaced my glxinfo.
odinsbane: er my glxinfo shows the updated driver.
Neo_The_User: if you run "LIBGL_DEBUG=verbose glxinfo | grep libGL" you should find it all the way at the top too
Neo_The_User: but im not on a unix based system right now so i cant verify that
melkor: okay everything appears ok, but I still cannot enable desktop effects (compiz). where should I look next?
BioTube: melkor: compiz and the like require AIGLX; check Xorg.0.log to see what's going on there
Wizzup: Does 2.6.33 need a specific libdrm version?
BioTube: shouldn't
Wizzup: ok, then I'll downgrade
Neo_The_User: why would you want to downgrade drm?
BCMM: is it possible to check the temperature of my r600?
BioTube: BCMM: not yet
Wizzup: Neo_The_User: Because the current libdrm+mesa+ddx simply doesn't work with kms
Wizzup: For me at least
BCMM: oh dear
Neo_The_User: BCMN: unless you have bad fans / cooling, you should be fine
Wizzup: Neo_The_User: http://paste.pocoo.org/show/182700/ Can't get rid of this
Neo_The_User: looks like you need newer xserver
Neo_The_User: http://bugs.archlinux.org/task/17022
soreau: BioTube: compiz doesn't need aiglx with dri2
Neo_The_User: but compiz does need AIGLX
soreau: Neo_The_User: No, it doesn't
Neo_The_User: well howcome it doesnt work for me when AIGLX is hosed without xrender?
Neo_The_User: or xrandr..
soreau: no idea, but compiz can run with direct rendering just fine
melkor: maybe I have the same issue my xorg is 1.6
Neo_The_User: i use krh's flushinvalidate tree and i have no issues
soreau: melkor: Does 'glxinfo|grep renderer' not report software rasterizer?
melkor: soreau: it does
Wizzup: Neo_The_User: newer than 1.7.5?
soreau: melkor: Well, compiz wont be able to start then. (or if it does, it will have major issues)
soreau: melkor: Can you pastebin your X log?
Neo_The_User: http://cgit.freedesktop.org/~krh/xserver/log/?h=dri2-invalidate
Neo_The_User: its not exactly stable
Neo_The_User: 1.8 RC1
BCMM: Neo_The_User: bad cooling is a possibility
Neo_The_User: Wizzup: "AC_INIT([xorg-server], 1.7.99.901"
BCMM: Neo_The_User: got it from an ebay user who unscrewed the heatsink, presumably while trying to get the card out of the machine
melkor: http://www.pastie.org/843244 <--Xorg.0.log
BCMM: where should i ask about how to check the temperature under windows?
Wizzup: Neo_The_User: Where on the page did you find that? I didn't see it on that page
Neo_The_User: http://cgit.freedesktop.org/~krh/xserver/tree/configure.ac?h=dri2-invalidate
Neo_The_User: right there ;)
Neo_The_User: line 29
Neo_The_User: BCMM: within the ati driver
Neo_The_User: go to ati overdrive
Neo_The_User: or whatever its called
soreau: melkor: I don't see anything that indicates why accel init is failing. Have you installed firmware for your card? Do you see anything interesting in dmesg (or can you pastebin it?)
BCMM: Neo_The_User: is that in control panel? sorry, hardly use the windows partition
Neo_The_User: you have to go to amd.com and download the drivers. if you want further instructions, please PM me
Wizzup: Neo_The_User: ok. Sorry, I'm a bit tired but I don't see what that has to do with my bug
Neo_The_User: well the arch guys say to bump your xserver
Neo_The_User: so it _might_ help
soreau: melkor: Now I see it -> (EE) RADEON(0): [dri] RADEONDRIGetVersion failed (libdri.a too old)
melkor: so I should probably reinstall libdri ?
melkor: I've never done anything with firmware to my knowledge
soreau: melkor: It's possible that you may need to upgrade Xorg server or rebuild it
soreau: melkor: Or it might be a stray file.. try mv/rm'ing it
Wizzup: Neo_The_User: yeah, but they said it was fixed like way back no?
Wizzup: [2009-11-07 14:45] upgraded xorg-server (1.7.1-1 -> 1.7.1.901-1)
melkor: soreau: I was looking at cgit, and I d/loaded xorg-server, but the dependencies are killing me, is there a way to get all of them?
soreau: melkor: Which distro are you using?
melkor: ubuntu
melkor: The build-dep won't get me new enough versions
Neo_The_User: ah
Neo_The_User: melkor: http://neo-technical.wikispaces.com/xorg-latest
Neo_The_User: that should help
soreau: If libdri.a doesn't exist on the system I would make sure all libdrm, mesa and xf86-video-ati components are lastest
BioTube: isn't .a a static library?
soreau: Actually, I'd make sure those components are up-to-date regardless
melkor: soreau: I think the xf86-video-ati is out out date, it is probably the default one.
soreau: melkor: I believe the xorg-edgers repo provides latest for these components and different versions of ubuntu
melkor: soreau: thats what I'm working on now, I figure Ill get the xorg build deps from there.
melkor: and if its the same xorg, Ill just use that.
soreau: You wont need very latest X server but a more recent version wont hurt either
Neo_The_User: can anybody here confirm that a 5830 will have a caseless and small variant similar to http://xfxforce.com/ecms.ashx/e56cf024-9184-475f-8056-be9f57aab468/RelatedGraphicCardSeries_RelatedGraphicCardModels1/HD5830354x312.jpg like the 5750?
kdekorte: Kristian, just killed a bunch from mesa
Neo_The_User: woohoo!
Neo_The_User: less legacy junk
airlied: misses miniglx already, spent a lot of time on that, but I suppose I got paid well
Neo_The_User: does krh have direct git access?
Neo_The_User: or did someone agree to this whole thing?
BioTube: it was mentioned on the mailing list
Neo_The_User: be back tomorrow. gotta finish my LFS system.
Neo_The_User: have a great night all!
JyZyXEL: is there brigthness/contrast control support?
agd5f: JyZyXEL: knobs on your monitor, or adjust your gamma
JyZyXEL: :(
JyZyXEL: wouldn't controls like that be easy to reverse engineer from the official driver?
spstarr: attempts to build kernel with drm-radeon-testing/drm-core-next
shadowmaster: ohh, Linux 2.6.33 released!
spstarr: nope
spstarr: drm-radeon-testing still not merging into .33
spstarr: resets -1
spstarr: hmm
tstellar: Is kms supported with the xpress 200M cards? When I have kms enabled, running glxgears causes X to freeze.
BioTube: tstellar: it's been support on r300-r500 cards since 2.6.31
BioTube: s/support/supported/
BioTube: and is tis an x-only freeze or a total lockup?
BioTube: s/tis/this/
tstellar: Just X freezes, I can still access the system via ssh.
BioTube: then the error's likely in mesa
BioTube: (or ddx)
spstarr: builds new mesa/ddx
cxo: let it snow, let it snow, let it snow
Hackus: Hey, thats off topic.
Hackus: Moderator!!!
Neo_The_User: hi. back earlier than i expected. is winsys required fo gallium?
Neo_The_User: *for
airlied: its part of gallium
naquad: hi
Neo_The_User: hello!
Neo_The_User: how are you? how can we help?
naquad: i'm trying to set up radeon driver, i've took version from git and tried to start x, but i've got error message: (EE) module ABI major version (6) doesn't match the server's version (2)
naquad: what is that and how do i fix it?
Neo_The_User: xsever might be too old?
Neo_The_User: *xserver
naquad: vcard is radeon hd 4870, xorg is 1:7.3+20
Neo_The_User: 7.3 yuck
Neo_The_User: thats why
Neo_The_User: you need to bump Xorg to 7.3 to 7.4 basically
airlied: sounds like you built against and old or newer version also
naquad: i've got debian testing, but there's no newer xorg :(
Neo_The_User: i rememeber i built dbus from git and xserver gave me the same error
mancha: xorg 7.3?
Neo_The_User: way back when..
naquad: thats the repos version, i don't have any other :(
Neo_The_User: try sid ;)
mancha: what versioning system are you folks talking about?
naquad: squeeze/sid
Neo_The_User: squeeze is not sid
naquad: thats my debian version
naquad: ok
naquad: doing dist-upgrade
Neo_The_User: sid = unstable. squeeze = testing. ;)
mancha: oh i see, you all mean 1.7.3. and 1.7.4
naquad: it'll be very offending if after all that messing around i won't have normal dual monitor setup
Neo_The_User: we can help you if it doesnt work
cxo: is using 1.6.4 with a HD4870 and the lastest git just fine
cxo: naquad, did you build drm,mesa and x86-video-ati?
airlied: naquad: you might have an old headers installed or something
mancha: yes, i seriously doubt if you have 1.7.3 you have too _old_ a version
mancha: the problem lies elsewhere
Neo_The_User: he doesnt have 1.7.3
Neo_The_User: thats Xorg. not xserver. look at it closer 1:7.3+20 not 1.7.3-20
cxo: whats the difference?
Neo_The_User: xorg = bunch of X libs, headers, and stuff. xserver = 1 part
Neo_The_User: thats difference
mancha: hah
mancha: i'm checking out of this clever discussion :)
Neo_The_User: haha
cxo: btw did any of you guys see the microsoft spy guide?
Neo_The_User: no
Neo_The_User: does it teach you about how the NSA watches you on windows 7 or whatever?
cxo: they've been sending "Cease and desist" orders all day
cxo: but its already replicated too much on the net for them to stop it now
cxo: http://www.cryptome.org/
cxo: if you scroll down you'll see a directory listing
Neo_The_User: i fail to see what that has to do with microsoft.
cxo: dont worry about it then :) this is way off topic anyways :)
Neo_The_User: haha
Neo_The_User: cxo: were you the one that wrote the radeon reverse engineering BIOS utility?
Neo_The_User: or am i thinking of somebody else?
nike: I just added a Radeon HD 4670 to my desktop and there is definitely no acceleration whatsoever (dragging windows is very laggy)
nike: I am on 2.6.32 with Debian unstable
cxo: firmware
Neo_The_User: sudo apt-get install linux-firmware
Neo_The_User: :D
Neo_The_User: wait.. linux-firmware-non-free for unstable
Neo_The_User: let me check
nike: Neo_The_User: the package firmware-linux-nonfree is isntalled
Neo_The_User: post /var/log/Xorg.0.log to pastebin then
Neo_The_User: so we can be of further assistance
mancha: 2.6.32 is quite old now
cxo: dmesg even
mancha: not necessarily talking about radeon stuff but there have been lots of security related patches btween 2.6.32 and 2.6.32.8
airlied: if not using kms it should be same as 2.6.33 mostly
cxo: i've actually found linux get faster over the last 6 months, 2.6.32 included
cxo: s/get/got/
naquad: upgrade fails =\
cxo: i havent done any tweaks, but this ubuntu-9.10 userspace with 2.6.33 boots up in under 30s, with mysql, postgres, samba, jackd and vsftpd all loading at boot
naquad: loves unstable...
nike: Neo_The_User: http://sorry.google.com/sorry/?continue=http://groups.google.com/group/alt.drugs.ketamine/browse_thread/thread/1993ef2792148f1f
nike: whoa
nike: ignore that please
nike: Neo_The_User: http://pastebin.ca/1811321
Neo_The_User: hahah!!!
cxo: s/\(.*\)\(es\)/\1e i\s/g
spstarr: cgit.freedesktop.org is down?
cxo: does obscure regres on purpose to see who gets it
spstarr: airlied: do you know when drm-radeon-testing is going to be merged into drm-linus for .34?
Neo_The_User: i dont understand 1337 speak that well
cxo: <> echo "loves unstable" | sed -e "s/\(.*\)\(es\)/\1e i\s/g"
cxo: love is unstable
Neo_The_User: nike: DRM isnt working for you
Neo_The_User: neither is Atombios (i think..)
Neo_The_User: you dont have swap configured?
airlied: spstarr: yes
spstarr: airlied: will I able to test this next week?
Neo_The_User: (II) RADEON(0): [drm] Closed DRM master.
airlied: well I'm just pushing drm-next now
spstarr: oh great!
airlied: so it should mirror out in a few mins
airlied: I haven't exactly booted it yet
spstarr: i'll start kernel build once you've done so
spstarr: is this with glisse's patches or just from -testing at the moment?
spstarr: i dont know if the thomas has given green light yet
airlied: just -testing at the moment
spstarr: ok
nike: Neo_The_User: What do you mean about swap?
mancha: does oldconfig work w/o issues from 2.6.32.8 to 2.6.33 ?
Neo_The_User: nike: nevermind about swap
airlied: am trying to fix gpu switcher for merge now
Neo_The_User: post LIBGL_DEBUG=verbose glxinfo too
airlied: since I kinda forgot to finish it
spstarr: airlied: I's also like to test this tonight too
spstarr: airlied: curious to see how you use it, I assume have to kill X to do so, but better than going to bios each time.
airlied: yeah you write a file in debugfs then kill X
spstarr: thats not bad
spstarr: how much work would be required to have X figure out which GPU is in use? or it is too big of a change?
cxo: what does gpu switching do?
spstarr: even with KMS now controlling video modes
airlied: spstarr: you mean dynamic or xorg.conf less?
airlied: spstarr: at the moment it works fine without an xorg.conf
spstarr: airlied: dynamic
spstarr: yeah but you have to kill X still right now
airlied: dynamic is about 3-6 months work in the X server
airlied: maybe more
spstarr: thats not 'too' bad
Neo_The_User: cxo: my guess is that it lets you change GPUs
airlied: on crazy laptops
Neo_The_User: so you can choose between 945gm or 5890, etc.
spstarr: airlied: i will test vgaswicheroo once you push it in to drm-next
airlied: I doubt I'll do ynamic too much X hacking
airlied: also the next gen of switchers don't actually switch
spstarr: maybe just document what might need to be done on wiki and someone might scratch the itch for you?
airlied: they can just runs apps on each gpu but only one gpu does outputs
airlied: spstarr: I think only ajax can do it, and he knows
spstarr: ok
airlied: you pretty much need to rewrite X
airlied: along with DRI
spstarr: is it because how pScreen is used?
airlied: no its because X is 20 years old
airlied: how to migrate a bunch of data from one driver to another is an open question
nike: http://pastebin.ca/1811337
nike: Neo_The_User:
spstarr: ugh, that does not sound pretty at all
nike: That is my verbose glxinfo output
spstarr: assuming, ajax wants to do it
spstarr: otherwise, we just have gdm/kdm do it or some other component in GNOME/KDE to do it
airlied: well considering these laptops have been around and there is about 100 ppl on a ubuntu bug , and it still took me to bother writing the 400 lines of code
spstarr: [Switching GPU] [ Which GPU do you which to use.. [X] [Y] select and click restart X button
spstarr: or something
airlied: I'm not waiting for a community member to finish it
Neo_The_User: nike: hate to say it but you might need to do some recompiling
spstarr: airlied: I was not expecting it to be something with ACPI, i thought something more proprietary
spstarr: and considering Lenovo was not wanting to release info on this (on their web forums they did not want to mention this IP)
spstarr: guess they have egg now
airlied: the DSDT contains nearly all the info
nike: Which part of the stack is not working?
nike: DRM, kernel, mesa?
spstarr: so just a matter of dumping the DSDT info, neat
Neo_The_User: drm
Neo_The_User: do you have KMS in your kernel?
airlied: spstarr: dumping + understanding
spstarr: yes
airlied: lots of ppl dumped
nike: libdrm-radeon1 is version 2.4.18-2
spstarr: you'd have to know how to read DSDT table info
airlied: the lol bug is https://bugs.launchpad.net/ubuntu/+source/xorg-server/+bug/312756
nike: libdrm2 is the same version
cxo: isnt wayland or whatever meant to be the solution to world peace and multi-gpu crazyness?
nike: Is there a newer one available?
airlied: cxo: no
Neo_The_User: no
Neo_The_User: @nike
spstarr: wayland is more of a concept thing, like a concept car ;)
airlied: cxo: unless you believe phoronix
spstarr: ouch that's a lol bug
Neo_The_User: nike: it might be the kernel side if KMS is disableed
Neo_The_User: *disabled
Neo_The_User: looks like it is since KMS isnt even trying to load in libdrm
Neo_The_User: *libdrm_radeon isnt loading
Neo_The_User: nike: can you post /boot/config-2.6.32.8-dsa-amd64 or whatever its called?
cxo: well like Luc said, the only thing holding linux back from desktop prime time is the graphics. so lets just hope wayland is the solution
spstarr: i think if wayland can run X apps within it, it will become a solution to use
spstarr: will->could
airlied: cxo: you ever tried to play a sound? ;-)
nike: Neo_The_User: http://pastebin.ca/1811352
cxo: doesnt get the joke
airlied: if there was money in desktop prime time I suspect Linux would work fine
airlied: cxo: if graphics was all that was wrong with the Linux desktop it would be grate
airlied: great
cxo: one step at a time airliedy
cxo: btw, do any of you AMD people know if the X6 cpus coming out next month will work on current AM2+/AM3 boards?
spstarr: drm-next pushed
spstarr: gets
Neo_The_User: its off
Neo_The_User: thats nots good
spstarr: merge completed
dileX: airlied: I updated the radeon build-howto the last days. once upon... beyond your daily work you might have a look into it.
Neo_The_User: nike: can i /msg you really quick?
nike: sure
spstarr: airlied: you merged switcharoo into this or not yet?
airlied: spstarr: no its only getting finished now
spstarr: oh, ok
airlied: I think I've found the bug that breaks it once it goes back to Intel the 3rd time
spstarr: odd
cxo: has like a 3-way merge going on in radeon_ib_pool_fini
airlied: woot it works