gregorian: radeon driver takes 3.5 minutes to load GNOME. vesa does it in 5 secnds, but lower resolution. What do I do?
MostAwesomeDude: 21:10 < MostAwesomeDude> Are you actually testing from a cold boot? cire
MostAwesomeDude: 21:10 < MostAwesomeDude> Or did you just change xorg.conf and restart X? crib
MostAwesomeDude: 21:10 < MostAwesomeDude> GNOME takes forever to load, the first time after a cold boot. crimsonfl~
MostAwesomeDude: 21:10 < MostAwesomeDude> But it's all still in memory if you only restart X.
gregorian: xinit /usr/bin/xterm takes 1.5 minutes.
gregorian: I have pastebins
gregorian: for the xorg.conf, log from vesa and radeon
MostAwesomeDude: I know of no reason why radeon should take longer than vesa.
gregorian: And it's very slow
gregorian: Would you like to see the logs? I've put them on pastebin
MostAwesomeDude: Sure.
gregorian: one sec
gregorian: xorg.conf http://www.pastebin.org/68600
gregorian: radeon_log http://www.pastebin.org/68601
gregorian: vesa_log http://www.pastebin.org/68602
MostAwesomeDude: Try setting AccelMethod EXA in your Device section.
gregorian: I will try and return. Thank you.
gregorian: I set AccelDFS but it still didn't work.
MostAwesomeDude: Did you set AccelMethod to EXA? Does it say that it's using EXA in the xorg log?
gregorian: I didn't do that. I'll be back.
gregorian: I set it, it says that it's using it. Do these options have something to do with the performance once X has started or do they have something to do with increasing the speed at which X starts?
gregorian: It still takes a lot of time.
MostAwesomeDude: XAA sucks.
MostAwesomeDude: EXA should be a lot faster overall.
gregorian: But still even the worst can't take 3.5 minutes right?
MostAwesomeDude: No, it shouldn't.
gregorian: Was my radeon log file of any help?
MostAwesomeDude: Yeah. The only problem I saw was that you were using XAA before.
MostAwesomeDude: Well, also you're not using KMS, but I don't think that's related.
gregorian: Can you deduce anything by comparing it with the vesa log file?
MostAwesomeDude: Nope. Totally different code and HW paths. Nothing special.
gregorian: Any suggestions?
MostAwesomeDude: Not really. You could try waiting around for everybody else; I'm the only one that isn't really doing much for Christmas.
gregorian: When is this place usually busy?
MostAwesomeDude: Any time besides today.
gregorian: Alright, thank you very much for your time. I really appreciate it.
gregorian: Merry Christmas
lvecsey: Is there an example radeon xorg.conf I can see somewhere? Xorg is reporting lines like Blank CRTC 0 success, Disable CRTC 0 success. Blank CRTC 1 success, Disable CRTC 1 success. Output CRT1 disable success. Is there an xorg.conf option in the Device area to tell it to indeed use CRT1?
MostAwesomeDude: lvecsey: Those are for debugging; they have little real meaning.
MostAwesomeDude: Is your monitor not coming up properly?
lvecsey: Alright. I'm going to try and load an X application now. The monitor goes green for a while but I don't see a hash pattern in the background or anything. Maybe only xinit does that, and I was just running Xorg directly
MostAwesomeDude: The checkerboard is disabled by default; use -retro to enable it.
lvecsey: thanks will do
knoppers: Hello. How can I figure out wether 2D acceleration is really enabled? The windows, if big enough, sometimes need seconds to move along the line shown by the mouse pointer. So I think that 2D acceleration isnt enabled. Or is this normal? MV76
knoppers: On the wiki (/wiki/radeon) it says 2D acceleration (EXA) is stable for HD 2300 to HD 4890 and I have a HD 2600 Mobility
Ghworg: grep Accel /var/log/Xorg.0.log
knoppers: oh, cool, thank you
knoppers: (EE) RADEON(0): Acceleration initalization failed
knoppers: havent seen this line when I looked through the file
Ghworg: You probably need a newer kernel or libdrm to get it working
knoppers: hm
knoppers: I do not know enough about the interdepencies of the parts that are thrown together, e.g. xorg mesa radeon libdrm ...., to solve things by myself. Where can I find a description of this?
knoppers: I could install a new libdrm but I do not even know why this would solve my problem and what other parts I have to replace etc
soreau: See the link in the topic
soreau: If you can get kernel 2.6.32 it should be good enough, then libdrm, mesa and ddx should get you going
Ghworg: For 2D you need at least 2.6.30 kernel, 2.4.15 libdrm (I think) and 6.12.something xorg-ati
Ghworg: You should be able to get packaged versions of those from your distro if you don't feel adventurous enough to compile
knoppers: 2.6.31 I have
knoppers: hm
knoppers: How are all these elements layered? The kernel offers X, X is used by Y, Y by Z, ...
knoppers: dependency graph
soreau: Its not rocket science, just read the link in the topic
Ghworg: Kernel provides drm, libdrm interfaces with kernel to provide userspace api. radeon ddx (xorg driver) and mesa both use libdrm to get acceleration
knoppers: and where does dri get into
knoppers: soreau, It tells me how to build the driver but the second git command doesnt work. ;-)
soreau: Well we cant guess how it failed. Maybe it would help it you pastebinned the failed output
knoppers: http://pastebin.com/m10d55284
_KAMI_: HI
soreau: knoppers: Which distro is this and why are you running as root?
knoppers: This system is only to learn how I have to do everything. Then I create a new proper one where I dont usw root for this.
knoppers: I use ArchLinux
soreau: Well you didnt change into the directory you cloned
soreau: and your supposed to build libdrm first anyway
soreau: Im quite certain there are packages to help you with all of this
knoppers: but I want to learn it
knoppers: where to start
soreau: Build libdrm
knoppers: I build libdrm first because all other things I build rely on libdrm, right?
soreau: You could think of it that way
knoppers: is it true or not :D
knoppers: whatever I learn, I imagine I would have to teach it to others and they ask questions and I dont want to say "I dont know, it's just my esoteric explanation"
knoppers: hm
knoppers: should I build libdrm from git or download the source as an archive?
soreau: You should use git
soreau: Ghworg explained it pretty well when he said
soreau: ddx and mesa need to be built against libdrm compiled with kms support
soreau: You enable kms support by configuring libdrm with --enable-radeon-experimental-api
knoppers: ok
knoppers: thank you two
knoppers: but now I have to learn git first so I can do all this properly
knoppers: :D
soreau: git is not hard to use. typically, you dont need git init or anything, just git clone GIT_URL then optionally cd into the directory it created and git checkout BRANCH
knoppers: hm
soreau: You shouldnt have to checkout any other branch after cloning since it will automatically be on the master branch or the latest code
knoppers: git-clone sounds like it would copy all the files available (so also all branches)
soreau: You should think of it as checking out the latest code
knoppers: and what branch to take
knoppers: on the cgit website (http://cgit.freedesktop.org/mesa/drm/) I see now button to list all branches
soreau: Look at the bottom of the page where it says clone
soreau: You want the first link there and use it as your GIT_URL
knoppers: I already have
knoppers: cloned
soreau: ok, now cd drm
soreau: configure and build it
knoppers: make is done
mikkoc: http://pastebin.com/d2286f173
mikkoc: been getting it since 2.6.33-rc1 ^
mikkoc: seems random
taiu1:
knoppers: The programs are linking against the new libdrm and also the radeon driver is loaded but but this doesnt make 2D acceleration active: http://nopaste.info/dde3e314f6.html
knoppers: installed everything to /opt/xorg
knoppers: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed to open the DRM
knoppers: libdrm + xf86-video-ati should give me a 2D accelerated Desktop, right?
knoppers: Radeon HD Mobility 2600
knoppers: but I cannot figure out how to detect wether 2D accel is used
knoppers: I get error messages but I dont know if the errors could disable 2D
Guest84002: in mesa/drm/radeon/radeon_cs_gem.c on line 166 (in git), there is this line:
Guest84002: reloc->flags |= (flags & reloc->flags);
Guest84002: exactly what does that do? It clearly looks like a bug
Guest84002: (flags & reloc->flags) can never return bits that aren't already in reloc->flags. and these bits are then or'd back to reloc->flags? will not change nothing.
haeger: hello after update my kernel to 2.6.32.2 kms is broken... dmesg replies: [drm] radeon defaulting to userspace modesetting. Is it possible to force kernelspace modesetting?
knoppers: Should there be devices in /dev/dri?
dileX: hmm, mjamm knoppers
knoppers: dont eat me, I got the flu flu
soreau: Guest84002: My best guess is there used to be more to the line than that
soreau: haeger: Did you try booting with radeon.modeset=1 ?
soreau: knoppers: Yes, there should be card0 at the very least
haeger: Guest84002: no but i can try that... moment please will be back soon ^^
soreau: heh
knoppers: soreau, So is the kernel responsible if there is no card in dev/dri?
lordheavy: knoppers, i guess it's the radeon module that gives you dri entries, so the kernel
knoppers: or kernel module, hm
knoppers: ok
lordheavy: you can search in dmes for radeon module messages
lordheavy: *dmesg*
lordheavy: or /var/log/ entries
knoppers: shows an error: http://nopaste.info/7e0ca6a305.html
knoppers: but error -22 isnt so meaningful to me :D
lordheavy: cat /var/log/everything.log | grep drm ?
lordheavy: and cat /var/log/everything.log | grep radeon ?
haeger: Guest84002: hmm strange... after boot the console resolution setting is correct.... dmesg replies that kms ist enabled... but the boot screen isn't "modesetted"
Guest84002: haeger: are you talking to me?
haeger: furthermore with kernel 2.6.32.1 modesetting works without that modesetting kernel commandline option
haeger: Guest84002: yes you gave me the hint with the readeon.modeset=1 commandline option?
Guest84002: haeger: nope soreau did
haeger: oh sry :D
idletask: Hello
Guest84002: np
haeger: ^^
haeger: modesetting without modesetting at boot time is a little bit wired for me :D
idletask: About that RLC firmware problem, what are the plans?
adamk_: idletask: What RLC firmware problem?
knoppers: grep -iE 'drm|radeon' /var/log/everything.log -> http://nopaste.info/6ff45ae5de.html
idletask: adamk_: from 2.6.33-rc1 on, it is needed to have KMS at least on r700 series, 2.6.32 didn't need that
idletask: adamk_: this very much sounds as a regression, because it requires that you either a) modify the kernel yourself or b) run an initrd
idletask: I don't run an initrd for performance reasons, hence I went the a) way
adamk: I've never considered that a problem. It's really up to your distribution to decide how they want to package the kernel and firmware, I guess.
adamk: And, if you're knowledgeable enough to compile your own kernel, you can decide for yourself.
idletask: Well, I do consider this a problem
adamk: I'm happy for you :-)
adamk: I don't know that the developers have anything planned. I also doubt there's much they can do as the firmware can't go into the official kernel tree any more.
idletask: It would be nice, at least, to have the kernel disable KMS if it cannot find the firmware instead of not booting as it does currently :p
uyf: is it possible to merge airlieds git tree with a stable kernel now?
idletask: uyf: define stable - 2.6.32.x or 2.6.32?
idletask: I have merged 2.6.32.2 with the drm-linus branch and some relatively trivial conflicts showed up
glisse: radeon microcode is painfull, sadly i don't think anyone is working on RE them so that one day we can as nouveau have our own open source microcode we can directly ship in the kernel
uyf: ok, i'm gonna try merging 2.6.32.2 then i can tell u how it went
idletask: glisse: there are quite a few "non binary" Radeon related firmwares in the kernel, though... Were they all reverse engineered?
idletask: Or did the specs given by AMD allow the creation of these?
idletask: glisse: oh, and BTW, I'm a happy rv790 owner, I wish to help trying out stuff, if you want me to try a given kernel or something, don't hesitate to ask
idletask: I will happily oblige
glisse: idletask: no, they went in before a policy change
glisse: kernel don't accept closed source firmware anymore
glisse: which is good i think
idletask: glisse: but which pretty much forces you to use an initrd if you want KMS on rv7xx series :(
soreau: glisse: How is nouveau's open and radeon's not?
glisse: you can build out of tree firmware
glisse: soreau: they are RE the microcode
soreau: RE?
adamk: reverse engineering.
glisse: reverse engineering
idletask: Reverse Engineering
soreau: glisse: So theirs is open and radeon never reverse engineered micro code?
idletask: Hmm, do they do this with a "clean room" design, like the b43 wireless driver did?
glisse: no
glisse: we don't need the clean room for gpu
soreau: glisse: If the radeon micro code is not reverse engineered, where does it come from?
glisse: because we don't dissasm the closed source driver
adamk: soreau: AMD :-)
soreau: Well that's kinda backward..
soreau: They release specs but not the firmware source?
glisse: we did like nouveau bacj in the day we read back the microcode after fglrx updated it but now we get it from amd directly
glisse: i think they don't own the IP of the various microcontrolleur behind each microcode
soreau: Ah
soreau: What is IP? :)
rehabdoll: ipv4 :)
glisse: intelectual property
idletask: Intellectual Property
knoppers: when there is no card0 in /dev/dri what can I make to make it appear? Do I have to recompile the kernel or only a module or what?
idletask: Or Inane Pettiness
glisse: patent
knoppers: Is DRI needed for EXA?
rehabdoll: atleast thats what it used to be, befor the lawyers took over :)
rehabdoll: before
soreau: glisse: That makes sense.. is this some of the same reason fglrx is also closed source?
idletask: soreau: they say so in the FAQ on their site
glisse: no, read the NVidia interview on phoronix
glisse: same reason apply to fglrx & AMD
soreau: knoppers: Pastebin your X log
adamk: knoppers: On most (all?) radeons, DRI is needed for accelerated EXA.
knoppers: I pasted something else and I think that this is showing the reason: grep -iE 'drm|radeon' /var/log/everything.log -> http://nopaste.info/6ff45ae5de.html
knoppers: xorglog: http://nopaste.info/64d747555c.html
soreau: knoppers: Did you ever finish building libdrm, mesa and ddx?
knoppers: yes, in the xorg log you can see it is used
knoppers: look for /opt/xorg
knoppers: (II) Loading /opt/xorg/lib/xorg/modules/drivers/radeon_drv.so
knoppers: and ldd on glxinfo for example shows that it binds to the libdrm from the /opt/xorg too
soreau: knoppers: Looks like you need to enable radeon drm in your kernel
adamk: According to the partial dmesg he posted, it's already enabled.
soreau: hmm
adamk: It doesn't really show why it's disabled, though.
knoppers: how to figure out
adamk: knoppers: Can you pastebin the full dmesg?
soreau: Well the tail end there doesn't look too promising
knoppers: ok
knoppers: full mesg: http://nopaste.info/d35551350b.html
adamk: Well I definitely don't know what's going on.
soreau: me neither :P
knoppers: :D
adamk: Perhaps a newer kernel would work? But that's just a guess.
soreau: knoppers: Can you pastebin the output of 'modinfo radeon' and the same for drm?
knoppers: yes
knoppers: http://nopaste.info/147bc3ba6f.html
soreau: This is kernel 2.6.31 from arch?
knoppers: yes, ArchLinux
soreau: Yea, I would say update to at least 2.6.32
lordheavy: knoppers, i'm on archlinux and currently i use kernel from git
knoppers: I have never created an own Kernel. Cannot I just dump the config used for 2.6.31 and use it to compile a 2.6.32 with it so that I get a compatible kernel?
knoppers: oh
soreau: is on gentoo and uses git kernel
knoppers: ok
lordheavy: and it's working well on an rs880 (hd 4200)
lordheavy: use the kernel26-git from AUR
knoppers: ok
soreau: knoppers: The config might be installed to the same location your kernel resides (/boot?) or you can try zcat /proc/config.gz
knoppers: ok
idle_task: Hmm, there's no clear tutorial on how to build an initramfs which can load firmware
knoppers: lordheavy, kernel26-firmware is empty, right?
lordheavy: if you got kernel-firmware installed, remove this dependency
lordheavy: it's the same files
knoppers: kernel26-firmware 2.6.31-1
lordheavy: yes
knoppers: but old version
lordheavy: but same files :-)
knoppers: :D ok
knoppers: cloning the git
lordheavy: knoppers: you you too add an entry in your /menu/grub/menu.lst
knoppers: it showas kernel configuration now
moobie: Hello folks
moobie: Does anyone besides me experience artifacts with fonts and some graphical things with latest oss stack + r700?
lmaowaffle: hi folks. I've got a question regarding 3d-enabled ati driver. I have a hd4850 atm on debian, and vo=xv is fine. If I were to compile and use git-sources, while 3d may be enabled, will 2d suffer in anyway?
BioTube: lmaowaffle: shouldn't
chithead: moobie: there exist font corruption issues with r600 dri1. fixed in dri2
moobie: chithead, well I kinda got if with the char 't'
BioTube: lmaowaffle: just make sure to update the ddx or you'll lose 2d acceleration
lmaowaffle: BioTube: what's ddx?
chithead: xf86-video-ati
moobie: and sometimes the font is completely gone
lmaowaffle: chithead: oh thanks
knoppers: !expand ddx
lmaowaffle: chithead: you wouldn't happen to have this working on a debian system, do you?
lmaowaffle: where this=3d ati
moobie: agd5f, hello. Do you know af any fonts issues with kms + r700 + latest oss stack?
chithead: no, I do not run debian on desktops
BioTube: lmaowaffle: running squeeze here with my 3450
lmaowaffle: BioTube: hey there :) I've got a 2.6.32 kernel (zen-kernel.org) running just fine on my squeeze. I'm looking for instructions on installing the X packages too
lmaowaffle: it'd be nice if a amd64 package were available, but I'll look through the FAQ regardless
BioTube: lmaowaffle: the instructions in the topic are still good for the userland part
BioTube: but AIGLX won't load drivers except from /usr
lmaowaffle: Is that addressed under the "Configuring system to load mesa and libdrm from /opt/xorg" heading?
BioTube: lmaowaffle: that only works for direct rendering
BioTube: if you want AIGLX, you ahve to install to the hardcoded path
arch: Is there any news for support of the RS880 (Integrated HD4200) card?
chithead: arch: http://wiki.x.org/wiki/RadeonFeature look in r600 column
lmaowaffle: BioTube: so essentially --prefix=/usr for xf86-video-ati and Mesa?
chithead: arch: rs600 is not related to rs880. and do not pm before asking, this is considered impolite
BioTube: lmaowaffle: X can be configured to load the ddx from another locations
BioTube: lmaowaffle: but mesa does need to be installed to /usr for AIGLX
moobie: Anybody experience any of these problems with kms + r700?: Slow flash, video not really working and font issues
moobie: ?
moobie: How do I install irq support for r700 in ubuntu?
BioTube: you just need a drm-radeon-testing kernel and the firmware
moobie: BioTube, is this support in 2.6.33?
moobie: rc2
BioTube: probably
moobie: I don't want to compile a fresh kernel. So it might work with ubuntu's 2.6.33-rc2 kernel for lucid?
BioTube: i don't know, but I'm pretty sure linus has pulled since the irq support was added
lordheavy: moobie: irq is present in 2.6.33rc2
moobie: lordheavy, thx
lordheavy: just checked :-)
lmaowaffle: moobie: what's irq support for?
moobie: lmaowaffle, why I want irq support?
BioTube: irq allows vsync, among other things
moobie: Actually I have a lot of problems using the latest oss stack with r700
moobie: Where does I find the irq firmware?
BioTube: the mailing list for one
BioTube: agd5f(or was it airlied?) has it hosted somewhere
BioTube: just google for "R700_rlc.bin"
Ghworg: http://people.freedesktop.org/~agd5f/radeon_ucode/
chithead: maybe someone should put this link in the channel topic
moobie: thx
lordheavy: for archlinux, here http://aur.archlinux.org/packages.php?ID=33016
moobie: Darn. Openafs doesn't compile against kernel 2.6.33. :)
moobie: Hi. Just tried with kernel 2.6.33-rc2 on ubuntu
moobie: It starts gdm, but doesn show anything on the screen
moobie: doesn't*
sixtila: need help installing my grpahics driver
Wizzup: hmm
sixtila: cant install it
Wizzup: 2.6.33-rc2 is unable to load the firmware
BioTube: Wizzup: did you build the module into the kernel?
Wizzup: BioTube: No, I use modules
Wizzup: It worked fine on 2.6.32, and I used make oldconfig, so there shouldn't be a difference
BioTube: do you have the rlc firmware?
Wizzup: Hmm, is that new firmware? :)
BioTube: new since 2.6.32
Kurko: I have promblem: I update my radeon driver from xorg-edgers ppa and now X doesn't start anymore?
BioTube: Wizzup: people.freedesktop.org/~agd5f/radeon_ucode
Wizzup: BioTube: I may not have it. Where would I get this? I'm using gentoo with the x11 overlay
Wizzup: ah
BioTube: Kurko: any information in the log?
Kurko: I didn't found any errors in Xorg log
Wizzup: Have em now, let's reboot and see :)
Kurko: when I try start X then screen backlight blinking and then computer freezes
BioTube: Kurko: KMS?
Kurko: no
Wizzup: BioTube: Thank you! It works.
BioTube: Kurko: do you know when that driver was built?
lordheavy: 2.6.32.2 ?
Kurko: 2009-12-22
Wizzup: BioTube: Do you happen to know what the new firmware does? I'm kind off interested now
BioTube: Wizzup: it enables IRQs
Wizzup: Right
BioTube: Kurko: are Mesa and libdrm from after the twentieth?
Kurko: Mesa and libdrm are also built 2009-12-22
BioTube: does VESA work right?
Kurko: yes
BioTube: sounds like an error in the module itself
BioTube: ^kernel module
BioTube: try it with KMS
BioTube: modeprobe fbcon && modprobe radeon modeset=1
Kurko: okey
Kurko: it didn't work even KMS is enabled
BioTube: does it still freeze?
Kurko: yes
edgecase: anyone got firefox 3.5 to play html5 video using xv hardware overlay? mine seems to be software rendering in the browser
Wizzup: Somehow my wine thinks that direct rendering is not enabled
ferret_: you are probably missing the 32-bit mesa libraries and/or dependencies
ferret_: wine is a 32-bit program
Wizzup: I thought gentoo built both
ferret_: It would have installed precompiled binary libraries from the emul-linux-x86-* packages
ferret_: It's a very, very old version of mesa though
Wizzup: Any tips on getting the very latest ones for 32 bit? (Sorry, it's probably a noobish question)
BioTube: configure should have a --enable-32 option or somesuch
ferret_: Aye, mesa has a configure option for 32-bit
ferret_: It should work OK... I had wine running that way once
Wizzup: It doesn't seem to be in the use flags of mesa :-(
ferret_: no
ferret_: You can't do it in portage currently
ferret_: You would have to compile it outside of portage
ferret_: There is a project for compiling 32-bit and 64-bit packages properly in gentoo instead of the pre-compiled 32-bit binaries, but it's not finished yet
Wizzup: Alright then, I'll do a checkout on mesa I guess :)
BioTube: Seems adding a ENABLE_32_x flag wouldn't be that hard
Wizzup: will try
ferret_: BioTube: see "# the deps of the above" comment section in http://explodingferret.com/linux/package.use
ferret_: And that's just the list of packages to build that make a difference... there is a separate list of things that aren't 64-bit or 32-bit dependant
ferret_: portage doesn't currently have a mechanism to enable flags during the dependency tree calculation stage, so that's why I had to painstakingly list them in a file there :)
BioTube: ah
Kurko: BioTube: I start X over ssh and I get this error: /usr/bin/X11/X: symbol lookup error: /usr/lib/xorg/modules/drivers//radeon_drv.so: undefined symbol: radeon_cs_create
BioTube: strange
BioTube: sounds like somebody screwed up the package
Wizzup: ferret_: wouldn't it be possible to add "lib32" to global use flags?
Wizzup: hmm... that somehow doesn't seems smart
Hackus: Hey anyone here. ?
Hackus: :-)
Hackus: Going to attempt a rebuild of my X components, as I am on break, and I am dieing of bordom. :-)
Hackus: I have an 3870x2 and it fails because VGA abritration turns DRI off.
Hackus: So, hopefully, they have fixed the silly bug that turns off DRI just because you have two cards.
Hackus: :-)
Wizzup: if you have filed a bug, then they just might have
Hackus: I did file a bug, but at the time I thought I had a outdated version of libdrm.
Hackus: Also the error was different, with a version mismatch error.
Hackus: I have tried like three times to get a build to work, always end up wtih tons of merge problems.
Hackus: GIT repos always out of sync.
Hackus: Thats ODD. It would seem the repos for YUM are down at the moment.
Wizzup: ferret_: Are you using a specific overlay in conjunction with all those use flags?
evil_core: can I enable s3tc some way?
uyf: when i try to boot X with freshly compiled drm-radeon-testing X will not boot but hard locks my computer
twnqx: the radeon driver would be nice
twnqx: if it could avoid the occasional hangs >_>
spreeuw: uyf: are you sure it locks?
spreeuw: wait 30s
spreeuw: it may be the lack of r700 firmware
spreeuw: or a lacking fbconsole
twnqx: yeah, lack of firmware shouldn't freeze the kernel
twnqx: but does.
Hackus: Where did you put the firmware files for r700?
uyf: yes i have firmware, i'm starting X with startx from command line and it hangs, tried with official 2.6.32 kernel without modesetting, and it works but i don't get interrupts then
Hackus: On mine they go in: /lib/firmware/radeon (Fedora 12)
knoppers: What is the firmware good for? Is it loaded up to the graphics card? Why isnt it permanently stored on the graphcis card?
Hackus: Because the firmware contains intellectual property bullshit.
Hackus: Pardon my french.
Hackus: :-)
BioTube: plus you can update it later
knoppers: but you can update a BIOS too :D
BioTube: RAM's cheaper than flash
knoppers: stored on an eeprom
knoppers: ok
Hackus: I mean, a video card is a REALLY simple device. I can see a router needing firmware, but not a grpahics card.
Hackus: Seems stupid.
mjg59: Hackus: ...#
amarsh04: how many transistors in a graphics card?
BioTube: Hackus: a GPU is simple. a graphics card is anything but
spreeuw: one billion
Hackus: Depends.
Hackus: I don't buy it.
Hackus: Implementation wise, firmware is stupid for a grpahics card, unless of course LAWYERS get involved.
Hackus: :-)
BioTube: Hackus: again, there's much more to a graphics card than the GPU
mjg59: Hackus: Not even slightly
Hackus: Same thing by the way for wireless devices.
BioTube: the crtcs, the interrupt controller, HDCP scrambler, etc
Hackus: Well, I guess we can debate the implementation. :-)
Hackus: MY implementation wouldn't use firmware in the same way as the Radeon cards. :-)
Hackus: If you go to the OPenGraphics card project and do some reading about firmware I think you will change your mind as they ponder full 3D functions.
BioTube: A cpu doesn't need firmware either, but those with it work better than those without
Hackus: Firmware most of the time, not always of course...there are exceptions due to real hard engineering implemention issues, but mostly=B.S.
Hackus: I don't mean a degree either. :-)
BioTube: besides, firmware=fixable hardware=broke for life
Hackus: Well, right. You hit the issue on the head Bio.
chithead: proprietary firmware is about as fixable as proprietary drivers
Hackus: Firmware should be used to compensate for hardware flaws, that a lgeitimate purpose.
BioTube: chithead: SOMEBODY can fix firmware
Hackus: But when you design firmware to make programming API's because of some intellectual property B.S., thats a crappy design.
Hackus: It actually makes the card slower.
Hackus: Anyway, if you want to read more about the debate, hit the open graphics card list.
Hackus: God, the entire fedoraproject.org domain seems to be either DOS'ed or servers are down.
idle_task: Question: why do most 3d-enabled programs, such as the venerable Quake 3, for example, disable compositing in the window manager?
idle_task: Are 3d and compositing incompatible?
BioTube: without KMS, yes
idle_task: Ah, without
idle_task: I am with
idle_task: And it happens nevertheless
idle_task: Does that mean that there's a bug/something not implemented somewhere?
BioTube: i believe it's just the entier architecture of UMS that prevents it
idle_task: UMS?
BioTube: userland modesetting
idle_task: hates knowing nothing about the subject at hand :(
idle_task: BioTube: and which component, among the throng that exist right now in the Linux graphics stack, is responsible for this userland modesetting?
idle_task: (well, when I say "most" above, in essence it's basically "all")
BioTube: the ddx does the actual modesetting
BioTube: I'm not sure how mesa talks to the card
idle_task: Mesa handles a lot of responsibilities, it seems
idle_task: Too many?
knoppers: Are ddx and mesa independent of each other?
BioTube: knoppers: to a degree
knoppers: but if I have a 3D desktop :D
idle_task: Side question, what does ddx stand for?
BioTube: device-dependant x
idle_task: x as in xorg?
BioTube: yep
idle_task: BioTube: OK, thanks
idle_task: A page explaining the graphics stack would be more than welcome :/ The dri-explanation.txt once seen is no more online
Hackus: cd drm
Hackus: woops...sorry wrong window. :-)
ppman: Hackus: is that another owner of a radeon card I see? :P
Hackus: Yes...I got my MESA, and 2.6.32 kernel with IRQ support built last night.
ppman: Hackus: I just did all that
Hackus: Now I am working on the drm libs and going to try and get DRI working again.
idle_task: is happy to report that compositing features of kwin are wonderfully smooth with his rv790, radeon and KMS
ppman: grabbed the drm-next kernel, mesa and xf86-video-ati -9999 versions
Hackus: You did?!
ppman: It's beautiful... 3d almost completely works
idle_task: ppman: gentoo? I'm a gentoo user too
Hackus: Do you have one or two graphics cards?
ppman: one
ppman: hd4830
ppman: idle_task: yup
idle_task: ppman: and define "almost completely"...
idle_task: Because here, there are several showstopper deficiencies
Hackus: Well, I have two Radeon 3870's (r600) and it fails miserably.
ppman: like, nexuiz will play great, unless you enable HDR, GLSL or Bloom
idle_task: Especially with wine
idle_task: ppman: one rv790 card here
ppman: foobillard works great as long as you don't do rendered ball reflections
ppman: little things like that
ppman: "works great as long as"
ppman: neverball/putt work great
Hackus: VGA arbitration is a bitch. :-)
idle_task: ppman: my guess is, the Mesa OpenGL implementation is not complete yet
Hackus: No really? :-)
Hackus: LOL
Hackus: Sorry.
ppman: idle_task: a few things still have to be done in software
ppman: I'd help out, but graphics is one of those things I'm still semi-clueless about
idle_task: Hackus: that's not really funny, especially since there's no status page on what is and is not implemented
idle_task: ppman: same here
ppman: I know everything about the original IBM vga card and earlier
ppman: but after that, my knowledge is lacking
idle_task: I cannot even claim that
idle_task: There's no program right now which clearly says, "that is hardware implemented, that is software implemented"
Hackus: Well, as I see things, desktop graphics for Linux is in its "embryonic" stages. I mean, we JUST got a graphics card manufacturer to come clean on the hardware. (ATI)
ppman: Hackus: intel's drivers are great
idle_task: Hackus: wrong, Intel has come clear long before that
ppman: it would be nice if their cards were good too
Hackus: Intel doesn't count guys and you know EXACTLY WHY.
ppman: not even cards, hardware
idle_task: Hackus: I don't, please explain
ppman: I'd buy an intel card if it existed and were pretty good.
Hackus: If you don't get what I mean, try playing games on a intel GMA graphics adapter.
Hackus: :-)
ppman: like the i740, but keep going
idle_task: Hackus: that's not really the question to my eyes
Hackus: Well, unfortunately it is EXACTLY the issue.
idle_task: Hackus: OK, say that it is, fine - but then, this still doesn't tell me where Mesa is about Intel chipsets
Hackus: If we want decent graphics on LINUX, we need GPU hardware to be opened up.
idle_task: Is the whole OpenGL API hardware accelerated on Intel?
Hackus: Intel stuff is nice for desktops, but otherwise it is crap.
idle_task: I have an Intel graphics adapter too, so if a program exists that tests for it all, I can report
Hackus: I would call "accelerated" on Intel GMA hardware a bit of a stretch.
Hackus: :-)
idle_task: It is especially frustrating not to have a single program that reports it
idle_task: In fact, glxinfo doesn't tell anything really useful
ppman: well, pretty soon, radeon will be a great driver
ppman: then nvidia gets to follow suit or die, pretty much
idle_task: Hackus: compositing desktops, even on medium Intel chips, run great, and the same cannot be told of other medium range chipsets from other manufacturers at the moment
idle_task: Compositing on my now ex X1350 was slow
Hackus: Thats my point. This idea of intellectual property is contrary to markets, how markets are formed and how they develop.
idle_task: Hmm, this turns into a more general debate, it seems
ppman: compositing for me works great on an nvidia 7025 and a radeon hd4830 :P
idle_task: ppman: the HD 4830 is not exactly low end though
ppman: ofc
Hackus: Yes, well general debate or not, it affects not just graphics but the entire idea of making LINUX mainstream.
Hackus: Anyone here know what the largest market for Windows operating system is?
idle_task: Hackus: yes, that too
Hackus: GUess.
idle_task: Thanks IP (Idiotic Pettiness)
Hackus: Give up?
Hackus: GAMES.
taiu1: yess all corporate workers playing solitaire!
idle_task: Hackus: wine is doing its best on this, and I'm afraid to report that I have much more success with fglrx than the radeon ddx for now :/
Hackus: Not exactly.
knoppers: lordheavy, I have created the new kernel and setupo grub to load it. I get this error on boot: http://nopaste.info/48999e402f.html
knoppers: line 698
knoppers: oh, 692
knoppers: sounds like what I have read here just an hour or so before
Hackus: If you use the radeon driver (proprietary) from ATI, its best to turn off updates for the kernel, mesa...etc.
Hackus: FYI, I am not sure how many here use the fglrx driver as well as the open source one.
knoppers: r600_cp: Failed to load firmware "radeon/RV630_pfp.bin"
Hackus: I put that in /lib/firmware/radeon on my Fedora installation.
knoppers: http://people.freedesktop.org/~agd5f/radeon_ucode/
knoppers: should rename the R600 to RV630 right
lordheavy: knoppers, did you have kernel26-firmware installed ?
Hackus: yeah, or if you build a kernel, it should be in your build directoy.
knoppers: lordheavy, yes, I told you
agd5f: knoppers: no. don't rename it
knoppers: pacman -Qs kernel26-firmware -> local/kernel26-firmware 2.6.31-1 (base)
lordheavy: ls /lib/firmware/radeon ?
lordheavy: got this file here
idle_task: Hackus: games, heh
Hackus: Yes, I normally copy the build firmware to that directory.
knoppers: there is no radeon folder or file there
idle_task: I'd love wine to catch up, mind you
Hackus: Well, depends on the distro. :-)
Hackus: I am using Fedora 12.
lordheavy: local/kernel26-firmware 2.6.32.2-2 (base) from testing
Hackus: all firmware for stuff goes in that directory.
idle_task: In fact, I wonder whether it is not time to scratch X and start over
idle_task: With 3D in mind from the get to
Hackus: You can find the latest firmware in the kernel build root: /firmware/radeon/RV635_me.bin
idle_task: get go, sorry
knoppers: lordheavy, but I have 2.6.33
Hackus: try and use 2.6.33-rc2 if you want to try irq support.
lordheavy: knoppers, try to install kernel firmawre fro maur
lordheavy: from*
lordheavy: aur
idle_task: confirms that IRQ support does work with this kernel, but after a fashion
knoppers: fatal: Unable to look up git.kernel.org (port 9418) (Temporary failure in nam
knoppers: resolution
lordheavy: knoppers, have you the /lib/firmware dir ?
knoppers: yes
Kurko: can somebody help me to build mesa?
idle_task: I hate it that you basically need an initrd for that, unless you modify the kernel :(
Hackus: damn another bad build.
Hackus: :-(
honk: idle_task: you do?
idle_task: honk: hate it? Yes. Modify the kernel? Yes
idle_task: I hate initrds
idle_task: They are a hindrance for boot time performance
honk: just make it a module and load it a little later o.O
Hackus: Ok, going to try again, fixed my libdrm build issue. Wish me luck. :-)
honk: I just copied the firmware to /lib/firmware and it worked w/o any changes to either the kernel or my initrd (which I need for my raid anyway ;P)
idle_task: honk: I try to avoid that as well - I use RAID too and all RAID is compiled in
idle_task: That makes a 3 seconds difference on start to system booted
idle_task: Not negligible
honk: yeah well.. wont work with my setup ;P
honk: 3 seconds? wtf
honk: ydiw ;P
idle_task: wtf I get it, but ydiw?
honk: anyway, that I'm using initrd is irrelevant
honk: you're doing it wrong =)
honk:
idle_task: honk: I don't agree
honk:
idle_task: I don't WANT to use an initrd AT ALL
honk:
honk: then dont
idle_task: And I want a fast boot
honk: I didnt say: "you need initrd" I said: "you dont need an initrd"
idle_task: honk: OK, but I cannot compile radeon in the kernel (which I currently do) if I don't modify the kernel to include this firmware
honk: well.. then dont
honk:
er0x: hey, who wants to post link where to get r600_rlc.bin
idle_task: honk: I know I can, I just don't want to... And up to 2.6.32.x, this worked perfectly
idle_task: Now it doesn't
honk: and you're complaining about.. what?
idle_task: honk: a regression in functionality
honk: it's not a regression if you just dont get new features, is it?=
idle_task: honk: it is, in the sense that what worked before doesn't work anymore
idle_task: make allyesconfig won't provide a bootable kernel
honk: errh.. are you sure your kernel is up to date?
idle_task: 2.6.33-rc2
honk: I'm pretty sure that was fixed at some point
honk: yeah well, try radeom-drm-testing or something
idle_task: honk: tried it, same problem
idle_task: At the _very_ least, in this case it should disable KMS
idle_task: But it doesn't
idle_task: It just hangs for some seconds and then panic
honk: ehh.. :]
honk: I'm pretty sure that that shouldnt happen =)
idle_task: I git fetch'ed a few hours ago and there were no changes, so I didn't bother to recompile
idle_task: honk: rv790 fwiw
honk: well, I may be wrong ;)
honk: though I dont think I am =)
idle_task: Well, my experience tells me that you are :p
honk: "doesnt work on my system" != "doesnt work"
idle_task: Well, yes
honk: anyway, bring it up when someone with a clue is on =)
BioTube: er0x: http://people.freedesktop.org/~agd5f/radeon_ucode/
honk: and till then.. module =)
idle_task: The opposite of "for all x, p(x)" is "there exists x such that p(x)"
idle_task: I am that existing x
idle_task: Such that !p(x) sorry
idle_task: x being "radeon chipset" and p(x) being "kernel boots with chipset x, with or without modeset"
honk: idle_task: no it's not
idle_task: Yes it is
idle_task: By pure logic it is
idle_task: This is basic mathematics
honk: yeah well, this is real life though, not maths ;P
bridgman: here's the commit that handles missing rlc ucode - it's in drm-radeon-testing, just looking now to see if it's gone any further :
bridgman: http://git.kernel.org/?p=linux/kernel/git/airlied/drm-2.6.git;a=commit;h=779720a3209849be202ac36a811e934865c50971
honk: or if you wanna stick to it, all x is not "all people in all situations at all times" ;P
idle_task: honk: computing is an exact science which incorporates a lot of maths though :p So I stand to the point that I am right :p
honk: bridgman: cheers
honk: idle_task: are you sure you were on the testing branch? ;P
idle_task: honk: testing branch of what? The drm tree?
honk:
idle_task: honk: as I said above, _yes_ I have tested this kernel
honk:
honk: ^^
idle_task: Grrr
honk: well, like bridgman just said.. it's supposed to work
honk: (or rather: fallback gracefully)
honk: so either you were not running code with that commit or it really is broken :]
idle_task: http://pastebin.com/m364ef25c <-- there
honk: if in doubt, blame the user
idle_task: As you can see, I know what I'm doing
idle_task: I AM on the drm-radeon-testing branch, my tree KNOWS of this commit
idle_task: And the resulting kernel FAILS to boot
idle_task: I'm not inventing anything
honk: I'm not saying you're inventing it ;P an incompetent user is more likely than a bug though ^^
idle_task: Well, as you can see, I am NOT incompetent
honk: anyway, try poking the one with a clue now that he revealed himself ;)
honk: idle_task: indeed
Hackus: Wow. EPIC fail. :-)
idle_task: I'll happily run a netconsole and report what I see if this can help
idle_task: But it does happen, unfortunately :(
Hackus: I also noticed that acceleration is turned off even for none DRI stuff. (Seems like EXA ceases to function.)
idle_task: With the firmware in kernel though, there's no problem at all
bridgman: the change is also in the kernel mainline; maybe that will work for you ?
bridgman: EXA needs drm on 6xx and higher
honk: bridgman: he tried -rc2 :]
bridgman: it's all implemented on the 3d engine, the 3d engine needs a ring buffer, and drm manages the ring buffer
Hackus: Yeah well, the VGA arb code kills way too many things with two cards.
bridgman: looks like there were no more changes in the kernel tree since tagging rc2, so that should have all the changes
idle_task: bridgman: if this is to me you were making that comment - I haven't tried -rc2 at all with a non loadable firmware yet. I have applied the necessary modifications to the kernel given to me by penguin42 on this channel and this is the kernel I am currently running
Hackus: I wonder if this affects just r600 people or if all persons with r600/r700 with two cards do not currecntly get DRI?
honk:
honk:
honk: grml ;P
honk: anyway, just listen to whatever bridgman says =)
bridgman: idletask, which kernel fails to boot then ?
Hackus: Anyone here have two r600 or r700 cards in thier box?
idle_task: Namely, these changes --> http://dpaste.com/138005/
Hackus: Need to perform an EXPERIMENT. :-)
idletask: bridgman: the drm-radeon-testing kernel without the above changes applied
idletask: bridgman: I have not captured the failure but I can on request
Hackus: All of this kernel building has made me ungry for pizza. :-)
Hackus: *hungry.
idletask: Hackus: this is the end of year, pickup something better
idletask: Like a tournedos Rossini
Hackus: Last year I gave up my car guys to get more excercise. I must walk like 50 miles a week around campus.
Hackus: I AM NOT GIVING UP MY GLASS NIKEL PIZZA.
Hackus: Mine as well install Windows too if that happens. :-)
wsnelr: I finally got the radeon driver working with an HD 2400 Pro card. The last thing I had to do was fix the monitor definition for my sony crt display, and Xorg -retro worked ok and I got xeyes to run too. The last thing I'm going to try is to switch things over to the TV Out with maybe the ATOMTvOut option.
Hackus: Mine as well install Windows too if that happens. :-)ot a updated firmware.
Hackus: Speaking of xorg.conf definitions.
Hackus: Anyone got a nice list of options for the r600 and r700 with an example config listing everything they do?
idletask: Hackus: I run X without an xorg.conf and it works perfectly
bridgman: Hackus; something more than
honk: Hackus: what kind of experiment? I might plug in my second gpu again ;p
Hackus: Oh, I didn't see the man page. :-)
Hackus: <-Such a silly man.
idletask: Hmm, the manpage doesn't report rv790 as supported, but it is
Hackus: Yeah Honk.
Hackus: Try throwing in a second GPU, then perusing your Xorg.0.log and see if your DRI still works.
Hackus: Honk: what sort of GPU do you have?
MrSmith: I updated on the latest version of the ati-driver; Compositing doesn't work (It worked before) I have an ati mobility radeon x1400
Hackus: I guess I got two RV670's.
honk: 4850
honk: and mwll is coming out today, so I need the extra power for the first time in years anyway ;)
MrSmith: I updated on the latest version of the ati-driver; Compositing doesn't work (It worked before) I have an ati mobility radeon x1400
Hackus: VGA abrbitration is a bitch. :-(
Hackus: Oh well, rebuilt all of the latest drm, still doesn't work.
honk: Hackus: gimmeh a few hours while I try to fit it into the case *lol*
Hackus: :-)
Hackus: Are the 4850's big form factor cards?
Hackus: Might need a small 800 watt power supply. :-)
honk: errh, the case is kinda full
honk: and yes, the 4850s are big, too
honk: they pretty much hit my harddrives
Hackus: Sorta like Wisconsin women. :-)
Hackus: LOL
Hackus: Oh well, don't make a mess of your box for this experiment Honk.
Hackus: I think I will file a more detailed bug report.
honk: I was planned to plug it in anyway
honk: I'm just pulling it ahead a few hours
Hackus: Alright. Well, I think I am going to play some SupCom for a while and get something to eat. I will be back later.
Hackus: www.freedesktop.org
honk: ^^
Hackus: Yeah I do that alot.
Hackus: I got too many windows open and I forgot to focus my cursor.
Hackus: :-)
honk: heh
Hackus: Wow...graphics cards are going to get a bit smaller guys. :-)
Hackus: "http://science.slashdot.org/story/09/12/26/1256217/Scientists-Create-First-Functional-Molecular-Transistor"
Hackus: CUL8r.
honk: o.O
honk: that was actually a _lot_ easier than expected#
honk: 3d accel is working just fine with 2 cards (only one connected to a monitor though)
edgecase: ok with 2.6.33-rc2, my s-video only has mode 800x600, it used to also include 640x480...
mjt: hm. XAA, EXA, now UXA.. what next?
mjt: (and before that there was something too)
idletask: mjt: "stop using acronyms and just make the whole damn thing work"?
mjt: yeah, exactly ;)
idletask: Honestlt
idletask: Honestly, sorry
idletask: Just kill X
idletask: Now, the infrastructure is in place to do just that
mjt: and use windows, i see ;)
jcristau: mjt: "now"? intel forked uxa from exa in 2008.
mjt: jcristau: i never used intel cards so far, but come arcoss this term just now
yangman: if we had a good patch for everytime someone mentions "just kill X", we'd have a truely awesome implementation by now >.>
mjt: in http://wiki.debian.org/XStrikeForce/CurrentProblemsInUnstable
idletask: yangman: probably, but truthfully I see this as a worthy goal
mjt: i'm not complaining really (however idletask expressed some common sense i think, in his first statement), just.. oh well.
idletask: How much of the X API is used nowadays?
yangman: X isn't an API
mjt: btw, i use remote X everywhere
mjt: diskless workstations running a WM on server
idletask: yangman: but it defines one
mjt: and i also remember a project to implement "Y"
jcristau: idletask: not much
idletask: yangman: 3d in X is an afterthought - and that's a problem
idletask: A HUGE problem
yangman: go ahead and solve that problem by time travelling into the past when X was being designed, and 3D was still years away
knoppers: so is there already a replacement for X?
jcristau: no
idletask: knoppers: there have been tentative replacements and still are
idletask: Berlin, Y, and more recently, Wayland
mjt: yeah i remember berlin too
idletask: In fact, the main problem as far as I'm concerned is why are Mesa and X separated at all
mjt: wayland: "a tiny something" :)
idletask: I know that the goals of Mesa extend beyong X
idletask: beyond
idletask: But they shouldn't
idletask: 3D should be in the display server, and that's the end of it
yangman: sighs
idletask: EXA is in X, OpenGL in Mesa
knoppers: cannot there be other places other than display server where "3D" might be useful?
idletask: Where's the sense in all that?
idletask: yangman: sigh all you want, this is a fact
yangman: do you tell NASA how they should be designing rockets, too?
yangman: "why have a separate fuel tank? dont they have enough space in the boosters?"
idletask: yangman: this has nothing to do
yangman: it's a parallel that too many people don't seem to ge
yangman: *get
idletask: When designing a new rocket, the NASA doesn't have the burden of legacy
yangman: "patches welcome"
mjt: . o O { client-side rendering }
mjt: Now I know why firefox and thunderbird as so f**g slow when run remotely...
knoppers: idletask, What are the disadvantages of today's design and how would your design rougly look like to imporve the current design?
idletask: Right now, it just seems that the Linux graphics stack does
idletask: knoppers: 2D hardware rendering and 3D hardware rendering are separating as you go up the top layer (the screen), this should be united
knoppers: maybe the separation does have advantages, too :D
idletask: knoppers: I fail to see any
idletask: knoppers: the radeon ddx has to cope with the kernel, Mesa and X
idletask: What if the Mesa and X were one and not two?
stikonas: in the future exa will be implemented in Mesa
stikonas: so Mesa will be an acceleration library
idletask: stikonas: OK, let's go further, then - Mesa doesn't provide OpenGL but 3D accelerating primitives
idletask: This way, we can have a native port of (yes, I know) DirectX/Direct3D too
idletask: Instead of having to emulate Direct3D via OpenGL
_Groo_: hi/2 all
_Groo_: any dev alive?
idletask: After all, Windows supports both natively, why not Linux?
stikonas: idletask, mesa can do that
stikonas: there is a project called gallium3d
idletask: stikonas: hmm, I thought gallium3d only had to do with software rendering?
honk: idletask: yeah go ahead and write a d3d frontend for gallium :)
idletask: honk: I don't have the skills to do that - wine, however...
idletask: They can do that
stikonas: wine refused to do that
stikonas: they only care about proprietary nvidia drivers
honk: they've been implementing d3d on top of opengl for several years already
airlied: idletask: when you know how to design sw stacks, get back to us
honk: and they're nowhere near complete
honk: gives you some kind of idea what you're asking for ;P)
airlied: you sound like you have a Slashdot comments 101 view of the X.org stack
honk: /. doesnt really post stories about X ;P
honk: or at least they didnt recently on the front page ^^
idletask: airlied: well, the xorg stack just seems to be the problem, to my eyes
FxChiP: I'm coming in late here, but problem with what?
mjt: what's a "compositor" or "composite wm" ?
airlied: idletask: thats because you don't seem to know what the problem is
idletask: airlied: to my eyes? Legacy
AStorm: honk: on top of gallium might actually be easier
AStorm: opengl is quirky ;)
FxChiP: mjt, a window manager (or alternately composite manager in the case of xcompmgr, which I don't think replaces the WM) that uses the X.org Composite extension
honk: AStorm: not that quirky =)
AStorm: different enough from directx
AStorm: while on gallium, you could work more directly
airlied: idletask: i.e. you don't know what the problem is
idletask: So, what is gallium exactly?
airlied: you are just parroting blame at the same thing other ppl blame
mjt: found http://www.freedesktop.org/wiki/Software/CompositeExt
AStorm: idletask: your google skills need work
airlied: idletask: its an architecture for writing acceleration drivers
AStorm: and more
AStorm: it should be able to expose direct specialized LLVM code to apps
AStorm: so you could program cards using LLVM
airlied: it doesn't really change the stack, it more makes designing parts of it easier
airlied: but in sw terms, its pretty much just another abstraction layer
AStorm: OpenCL on top of that? almost trivial
idletask: So, does that mean that any 3D API can be programmed on top of it? And that for a card to support gallium, you just have to implement the gallium API?
idletask: (and also 2D)
airlied: idletask: mostly
idletask: Well, that looks very interesting
idletask: From what I've read so far, nouveau rendering is entirely gallium based, correct?
airlied: it doesn't change the X stack thuogh
airlied: idletask: just their 3D drvier
idletask: airlied: but it allows it to be changed
airlied: so far only the VMware svga 2D driver is gallium based
idletask: It allows !X
_Groo_: airlied: hi airlied, googleearth is still bleending all over the place with latest dri2/ddx/drm/mesa and rs485
airlied: idletask: I think KMS matters more for that
idletask: airlied: I have watched a video of the 2009 LKS where Keith Packard told just that
airlied: idletask: gallium just lets you in theory design the core of the 3D driver once and plug other accel features into it
idletask: Though I fail to see why, since I don't have the technical knowledge to understand that
idletask: airlied: "just" is an understatement, I'd say!
idletask: Well, thanks for the enlightenments
knoppers: KMS,UMS,Gallium,Mesa,Xorg,DDX,EXA,...1001 terms, all related somehow. Need document with a bit history describing how it worked in the past with the problerms of the past, then describing today's state with problems and solved problems, and planned future developement
knoppers: idletask, you write the document
idletask: And, I don't see the road to the end of X that hard, tbh
idletask: When you look at the two main toolkits nowadays, they have clearly separated the drawing engine from the toolkit engine
idletask: knoppers: I'd like to, but expect it to be online in x months with x > 1 :p
AStorm: knoppers: want a chart? ;p
knoppers: hehe :D
airlied: just google for "state of linux graphics"
airlied: it describes in a slightly biased way a few years ago state
idletask: airlied: been there, done that, explanations are obsolete (as you said, it dates back a few years ago)
idletask: http://www.inf.ufpr.br/prz05/links.txt <-- this one is interesting
idletask: But you need to collect, gather, summarize
spreeuw: knoppers: wikipedia should help
spreeuw: and for the stack theres already some descriptions flying around
idletask: airlied: more technical question now...
idletask: airlied: I have an rv790, and tried to boot the drm-radeon-testing kernel with AGP and DRI builtin, and modeset on by default
idletask: airlied: but it failed to boot with a crash, which I haven't gathered (it didn't try to revert to non KMS)
idletask: airlied: commit 779720a3209849be202ac36a811e934865c50971 supposedly fixed it, but it didn't in my case
AStorm: idletask: checked /topic yet?
AStorm: you have the firmware right?
idletask: AStorm: that is the thing
AStorm: someone should say R600 and R700 in the topic
idletask: AStorm: the firmware wasn't in the kernel, and according to the commit above, KMS should have failed and dri fall back to non KMS
idletask: But it didn't
AStorm: yes, it doesn't
AStorm: unfortunately.
AStorm: so, grab the firmware, build it into the kernel
idletask: AStorm: is that a known bug, then?
AStorm: (note: the firmware names should be radeon/foo.bin - so path /lib/firmware is the usual)
idletask: I know how to build it into the kernel, I have done it with 2.6.33-rc2
idletask: (and I have not tested 2.6.33-rc2 without the firmware added, btw)
AStorm: hehe
AStorm: good.
AStorm: I mean irq firmware
AStorm: that's the new one
AStorm: right?
AStorm: there should be a way or option to push KMS init after initramfs
AStorm: right now it's impossible to run it built-in w/o firmware built-in as well
idletask: Well, I grabbed http://people.freedesktop.org/~agd5f/radeon_ucode/R700_rlc.bin and commited it in
AStorm: and did what?
AStorm: /proc/config.gz:CONFIG_EXTRA_FIRMWARE="radeon/R600_rlc.bin radeon/R700_rlc.bin"
AStorm: /proc/config.gz:CONFIG_EXTRA_FIRMWARE_DIR="/lib/firmware"
AStorm: something like this?
AStorm: :)
idletask: AStorm: http://dpaste.com/138038
idletask: These are my two commits on top of 2.6.33-rc2
idletask: I have not tried yet to apply them on top of drm-radeon-testing
AStorm: funny, but pointless
AStorm: there's an option for extra firmware for a reason ;p
idletask: Uh
idletask: Even if you have dri built in?
idletask: Ah, CONFIG_EXTRA_FIRMWARE_DIR
idletask: This option doesn't show up in menuconfig
AStorm: it does.
AStorm: /proc/config.gz:CONFIG_FIRMWARE_IN_KERNEL=y - need that as well
AStorm: use /
idletask: I have that set to y
AStorm: to find the options
idletask: And I have used /
idletask: Oh
idletask: PEBKAC
idletask: I didn't look in the correct submenu
idletask: Well, that gives me another thing to try, then
idletask: drm-radeon-testing with that option set
AStorm: I'd like an option like radeon.delayload=1
AStorm: that delays it till / is mounted
AStorm: or maybe radeon.waitroot=1
mjt: what's wrong with modular radeon?
mjt: oh i think i know...
idletask: mjt: say, you don't want to use an initrd (or initramfs as it should be called now, I guess)
airlied: you can just let udev load it
mjt: airlied: with tons of screen mode switching...
idletask: airlied: udev won't load it if you have dri built in, unless I'm mistaken
mjt: as i have here
airlied: mjt: should just switch once
airlied: idletask: if its built-in no need to load it then
mjt: it starts with 80x24 text, at 1/3 of the boot process switches to manyXmany and tiny font (when /etc/modules are loaded), next at 2/3 boot process a correct font gets loaded...
idletask: airlied: except that if you don't have the firmware in the kernel and want KMS, it crashes at boot - at least with the drm-radeon-testing branch
edgecase: just updated, now openarena hangs with black window
airlied: idletask: the plan is to not crash but turn off accel
airlied: but we haven't fixed it al lyet
idletask: airlied: accel altogether? Not just KMS?
idletask: I was talking about the r700 rlc firmware here
airlied: idletask: kms will work, no accel
mjt: no firmware - no accell
idletask: Aieee
mjt: -l
idletask: Oh well, at least there is a way to build a kernel with dri built in and the firmware built in without having to modify the kernel itself
edgecase: does resizing glxgears influence the benchmark results? ie when people quote fps, is that at the size of window it starts with?
idletask: That's something already
mjt: (is there a way to load font _before_ loading radeon?)
knoppers: I am now using Mesa 7.8 and kernel 2.6.33rc2 etc with ATI Mobility Radeon 2600 HD and 2D accel and Xv is working. in glxgears I get 500 fps, Sauerbraten 5 fps and graphics errors
idletask: I meant, the kernel _source tree_
airlied: mjt: don't think so
mjt: that suxxx ;)
idletask: airlied: OK, so from now on, I'll test the drm-radeon-testing tree of yours (fwiw, I have an rv790, so if I can do anything to help, let me know)
mjt: screen flickering and losing 1/3 of boot process output twice is not good
mjt: idletask: is it built-in 790g or something else?
idletask: mjt: I can't make out your question
idletask: Og
idletask: No
airlied: mjt: what makes you lose it the second time?
idletask: This is not a builtin chipset
mjt: airlied: loading the right font
mjt: airlied: (i understand the two can be done together)
idletask: I have replaced my rv370 with that, three days ago
airlied: you can probably built font in or something
edgecase: airlied, just tried 2.6.33-rc2, "scaling mode" is *much* better on this rv200, although not perfect
mjt: airlied: i need cyrillic which is not good for generic kernel :)
mjt: airlied: but the problem really is with font _size_
edgecase: scaling mode "none" on LVDS has top-left gravity, right overscan is black, but bottom and bottom right overscan has some junk
mjt: airlied: when loading radeon.ko it switches to default 16point font and large resolution, later on larger font gets loaded...
edgecase: has anyone tried openarena since 21st build?
idletask: I don't have openarena but I have plain Quake 3
idletask: I can test
edgecase: mine seems to be using 20% cpu polling something + timer delays
edgecase: freedoom crashes when you pick start game
edgecase: what other linux games are native 3d?
edgecase: nm the web is telling me
idletask: edgecase: nexuiz
edgecase: yeah it's d/l now :)
knoppers: sauerbraten
edgecase: i wonder if any of the games have a screencast mod
edgecase: let people view from browser, a game in session
edgecase: heh there's a use for a GPU in a server!
nfrs: I know this is offtopic, but someone's gotta know: does amdcccle touch any files under user's home? which?
edgecase: run it with $ strace -eopen amdcccle
edgecase: that would output every file it opens
edgecase: you might need to add -f
edgecase: if it's a multi-threaded app
idletask: Uh
idletask: airlied: anyway, in spite of our disagreements - I wanted to thank you for the job you did
idletask: airlied: two years ago, I'd have set back on an old card... Today, I have bought one of the latest and greatest without a thought, and _you_ are partly "responsible" for that
idletask: This is no mean feat
nfrs: edgecase: it shows a lot of files, but all unrelated
cxo: [ 54.070215] [drm:r600_irq_process] *ERROR* Unhandled interrupt: 21 5
cxo: known or unknown?
airlied: hdmi audio related, can be ignored
Ghworg: HDMI audio causes unhandled interrupts, don't know if that is one of those or not though
icewaterman: is there going to be UXA support on radeon? i really like compiz to work together with 3d apps like googleearth
airlied: nothing to do with UXA
airlied: and that already works fine
icewaterman: airlied: hmm, gonna try
airlied: at least on my machine
edgecase: heh lots of developers omit "at least on my machine", kudos for honesty/accuracy :P
idletask: edgecase: s,omit,admit,g? ;)
icewaterman: airlied: seems to work indeed
cxo: so how come nvidia doesnt give out shader and instruction set documentation like ati does?
icewaterman: cxo: they dont have to. neither did amd, but their problem was, that intel does and amd didnt have anything competitive to offer
icewaterman: ati drivers sucked from the beginning (the closed ones), while intel drivers were open and pretty good at least for business-desktops
twnqx: and there's the fact that nvidia's driver actually works, something you can't say about fglrx
airlied: twnqx: not in the markets thay pay the money
airlied: i.e. fglrx for CAD/compute is pretty good, just not wine/games
twnqx: a drivers that can't even *enable* my laptop's display is as good as no driver at all.
twnqx: and i would never even consider an amd card for the desktop...
icewaterman: twnqx: well, i'd definately go for intel
icewaterman: because they have the best open driver - just not for playing games
twnqx: sure, i'd turn off the amd in this laptop
icewaterman: twnqx: :-)
twnqx: if the intel was connected to the display port
twnqx: but no, apparently only the ati is. or intel'd driver just doesn't support it.
icewaterman: anyway, older cards work fine with radeon driver
twnqx: this is a HD3650
icewaterman: newer cards work partially or not at all.
twnqx: i'd call that junk age
icewaterman: my guess is that your box doesnt have an integrated intel graphics at all
twnqx: it has
twnqx: if linux drivers would support it i could switcj
icewaterman: anyways, christmas is over here, gtg to bed
icewaterman: by
icewaterman: e
twnqx: bye
twnqx: btw, is it normal that 2D crawls to a halt if i run a 1680x1050 internal display with a 1920x1200 external?
airlied: icewaterman: we support displayport now
airlied: just not sure how well tested the chip in lenovo is
FxChiP: fglrx worked pretty well for me
FxChiP: 'til I had to upgrade
FxChiP: now I'm on the experimental radeon stuff... and am surprised to find that visualizations (Goom, specifically) are actually *faster*
FxChiP: Darwinia has an odd glitch-type thing that happens on startup "where it says Program: Darwinia" but I don't know if that's part of the game or not; it then goes to the 'stage select' screen on pressing Esc, but crashes after a choice is made. C'est la vie, but the select screen has some beautiful performance
FxChiP: Uplink runs a little slower, but I wasn't thorough by any means
FxChiP: Videos play a *lot* better
FxChiP: I might admit to a little bit of placebo effect, there, but
FxChiP: Maryo Chronicles is missing libGLU, that's probably my fault :)
AStorm: FxChiP: experimental?
AStorm: which version of mesa?
FxChiP: git
AStorm: FxChiP: Arch?
AStorm: since missing libGLU is AUR mesa-git problem
AStorm: :P
FxChiP: x86_64
FxChiP: and yeah, Arch
AStorm: http://omploader.org/vMzM0Yg - try this pkgbuild
AStorm: it worked for me - not in AUR due to problems with split pkgbuilds.
FxChiP: seriously? fuck, it already had problems copying GL header files
evil_core: why qemu is using kms and not X11?
evil_core: it opens framebuffer
AStorm: hmmh?
AStorm: might be SDL
AStorm: or that console support
AStorm: fine either way
evil_core: AStorm: its not know, it was very stressfull situation
AStorm: heh.
evil_core: it was rendering to FB, but didnt grabbed keyboard. I was checking grub modules to remapping bios memory, and default OS started because I wasnt unable to change
evil_core: not not worked, but anything works like enter
evil_core: hopefully 5 secs was enough to make hard shutdown
evil_core: I was scared about file system. Some years ago I destroyed them in identical way
evil_core: From git://anongit.freedesktop.org/git/mesa/mesa
evil_core: e81fe08..331e910 mesa_7_6_branch -> origin/mesa_7_6_branch
evil_core: Already up-to-date.
evil_core: So I got old mesa_7_6?
evil_core: remotes/origin/gallium-s3tc - do I need that to get s3tc with multitexturing?
bridgman: evil_core; I think the changes from that s3tc branch have already been merged to master
evil_core: bridgman: so should s3tc works fine on mine r500?
bridgman: not sure; there's still discussion of a bug with s3tc and multitexturing but I think that's just for compressing textures with the external library
evil_core: I was reading that libdxtn has broken multitexturing, but I dont understand ow all it works - shouldnt ut be used only if GPU doesnt support it?
evil_core: hmmm...any known patches or replacemnts for that lib?
bridgman: I think the lib was for compressing textures, and was made optional so it could be omitted for countries where patents were an issue
bridgman: not sure what the current status is... hold on, let me check bugzilla
dmz001: soreau: I've built a driver (following the instructions) It created a driver 6.12.4. That didn't help.
soreau: airlied: With latest drm-radeon-next, I suspend with pm-suspend and everything goes well on resume except there's no signal sent to S-video even though xrandr behaves as if it is. I turn it --off and back oon to no avail, there is no signal output whatsoever
dmz001: soreau: The version reported by the fedora install is 6.12.99...
soreau: dmz001: What are you trying to do today?
dmz001: soreau: Trying to get 1920x1200 to work.
dmz001: We left off with me having to build a driver to see if that would fix the problem.
dmz001: So now I'm looking for more suggestions or something that would narrow down the problem.
soreau: dmz001: I remember someone said upgrade your driver. Looks like what you've done in effect is downgrade it
dmz001: Well, I followed the instructions as per the HowTo. Well can you tell me the correct version and what I need to tell git?
soreau: dmz001: When you first clone ddx, it will be master be default
soreau: You don't need to checkout another branch, just build what it already checked out
soreau: For your current clone, you can say 'git checkout master' to get the latest code
bridgman: looks like s3tc/multitexturing patches are still "up in the air" :
bridgman: https://bugs.freedesktop.org/show_bug.cgi?id=8056
bridgman: scroll to end
dmz001: soreau: Off to the mines.
dmz001: soreau: git checkout master -- reports Already on 'master'
soreau: airlied: The only way I have found so far to get tv-out working again after resume is a reboot
soreau: dmz001: Are you sure you installed to /usr? I believe the default is /usr/local
evil_core: bridgman: omg, wqtd is that? UT2K4 works perfectly for me
dmz001: soreau: I installed to /opt/xorg as per HowTo. Modified "Files" and Xorg.0.log reported that it picked up the rebuilt module
soreau: dmz001: And the same problem persists i take it?
evil_core: problem is quake3, and biugger tc-elite which segfault with enabvled, and makes ~2-5fps while s3tc is disabled
dmz001: Yes.
soreau: dmz001: Well I guess the next logical thing to do is file a bug report
dmz001: I checked up on the manuals for the various parts. Monitor handles 1920x1200x60Hz KVM handles 2560x1600x60Hz (DVI-D)
evil_core: bridgman: what means "in the air?
dmz001: And this worked under my windows box.
dmz001: Okay. Where and How?
soreau: evil_core: That means nothing definite or undecisive
bridgman: evil_core - uncertain status
bridgman: eg "throw it up in the air and wait to see where it lands"
dmz001: Hmm, Should I try the ati catalyst driver? If that works definitely bug.
bridgman: bug report at bugs.freedesktop.org
soreau: dmz001: If a mode your monitor can handle causes a black screen, it is already a bug
evil_core: bridgman: so it means: not applied upstream?
soreau: You don't need another driver or os to tell you that
bridgman: we don't know - the last post in the bug report was someone asking if it had been applied upstream
bridgman: and that was yesterday ;)
soreau: evil_core: http://www.google.com/search?hl=en&client=firefox-a&rls=com.ubuntu:en-US:unofficial&hs=Mrj&defl=en&q=define:up+in+the+air&ei=cM02S5ncCIbZngfa8r2wDg&sa=X&oi=glossary_definition&ct=title&ved=0CAcQkAE
evil_core: GL_EXT_texture, GL_EXT_texture3D, GL_EXT_texture_compression_s3tc,
evil_core: GL_S3_s3tc
evil_core: soreau: give me link to rfc ;)
dmz001: Why is Option "Monitor-