Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-12-05

Search This Log:


hnsr: weird, with drm-core-next (not sure if I should be using that branch) I keep getting a 'bus error' when I archive some data, then every programing refuses to run and dies with 'input/output error', doesnt seem to happen on a stock 2.6.31 kernel
hnsr: s/programing/program/
hnsr: though I guess I should try 2.6.32 without any radeon stuff and see if that does it too
luther32: I am here once again to seek anyone in here with a Lenovo Thinkpad W500 laptop in their hands ... running gentoo linux
luther32: Anyone in here know when/if full functionality is/will available for ATI Mobility Radeon 3650 , which I believe is r635 ?
luther32: and also, is glxgears a sufficient benchmark tool ?
luther32: I only get about 501 FPS on a good day. (and I am on a 3.0 GHZ core 2 duo machine with 6 gigs of ram and 512mb of video memory) what kind of graphics/glxgears performence should I expect ?
luther32: hears crickets
luther32: no one here ?
MostAwesomeDude: Well, I was doing other things.
roy_hobbs: no one here that knows the answer
luther32: heh
MostAwesomeDude: rv635 should work fine with new stuff.
MostAwesomeDude: rv635 should work fine with new stuff.
MostAwesomeDude: I recommend radeon + KMS, but whatever.
luther32: MostAwesomeDude, what do you think is the ETA for stable full funcionality
luther32: MostAwesomeDude, whats KMS ? is it an alternative for compiz
MostAwesomeDude: luther32: When It's Done. Alternatively, 2020.
MostAwesomeDude: (Alternatively, *now*, but with a few bugs.)
luther32: hmm maybe 6 months max. Ive never used compiz...so im not really missing much
soreau: luther32: KMS is Kernel Modesetting
soreau: Compiz is a window manager for the X windowing system
soreau: they are two entirely different things
luther32: oh I see...
luther32: soreau, man, I dunno if I'll ever wrap my head around all the different possibilites..KMS COMPIZ SSE KXH etc.
luther32: Ive come to terms that I know nothing about linux
soreau: luther32: If you want 3D acceleration for your card, you will want to install at least kernel 2.6.32, libdrm, mesa and xf86-video-ati all -9999 components
soreau: and compile xorg-server with VIDEO_CARDS=radeon
luther32: soreau, yea I know...but 3D 'sort of' works right? Its still in its experimental stages?
soreau: 3D should work fine
olaf: luther32, http://en.gentoo-wiki.com/wiki/Radeon
luther32: hmm
luther32: soreau, what kind of ranges should I expect/have with glxgears by the way with full functionality?
luther32: (what do people in general get.....including you)
soreau: luther32: glxgears is not a benchmark of any kind
soreau: You should not pay any attention to the FPS it reports
luther32: soreau, then what is it ? and what is a benchmark ? I mean how do I determine the FPS + performence
soreau: performance is measured by what the user experiences on their machine. That can be determined by a large number of factors
soreau: In other words, try it out and see for yourself
luther32: Oh I see.
Travis1: ok... so I'm back, and my video card still isn't working. Anyone have time to give it a crack? hehe
agd5f: Travis1: does this patch help? http://www.botchco.com/alex/xorg/limit_aper_256.diff
Travis1: uhh I can check. What do I need to do?
Travis1: sorry - total noob here
agd5f: Travis1: patch xf86-video-ati
agd5f: clone the git tree, apply the patch, built and install
Travis1: as in, that's what you want me to type into the terminal?
agd5f: Travis1: do you still have the git tree you checked out yesterday?
Travis1: I have no idea what a git tree is sorry man
Travis1: when I say 'total noob'... I really mean total noob
agd5f: Travis1: didn't airlied help you with this last night?
agd5f: Travis1: git clone git://anongit.freedesktop.org/xorg/driver/xf86-video-ati
Travis1: yeah but he was telling me commands to type into the terminal. I wasn't sure exactly what it was doing or remember what the commands were. Sorry mate.
Travis1: ok no worries I'll type that in
agd5f: cd xf86-video-ati
agd5f: wget http://www.botchco.com/alex/xorg/limit_aper_256.diff
agd5f: patch -p1 -i limit_aper_256.diff
agd5f: ./autogen.sh --prefix=/usr
agd5f: make
agd5f: sudo make install
Travis1: ok done
Travis1: next? Restart?
agd5f: Travis1: now restart X
Travis1: no worries. I'll be back with any log files if all goes badly.
Travis1: keep my Xorg.conf file in there?
Travis1: some people have been telling me to remove it, but it's been crashing when I do.
agd5f: Travis1: depends what's in it
Travis11: fail..
Travis11: want the log?
agd5f: sure
Travis11: http://pastebin.com/m495cf23f
agd5f: Travis11: that's from radeonhd
Travis11: that's from my Xorg.0.log file...
agd5f: Travis11: edit your xorg.conf and change the Driver line from radeonhd to radeon
Travis11: http://pastebin.com/m5ac5bb03 that's my xorg.conf file... is the only thing I need to do is change "radeonhd" to "radeon"?
agd5f: yes
Travis11: cool. restart again/
Travis11: back soon - fingers crossed!
Travis1: another fail... :(
Travis1: I got a black screen this time when loading X
Travis1: had to restart and load into recovery mode and restore my old xorg.conf file..
Travis1: want the Xorg.0.log contents? If I've tried to load twice, will that show you the info you need or do I need to post the contents of a different Xorg log?
Travis1: ok so this: http://pastebin.com/m88eeecb is from my Xorg.0.log.old file - which I'm pretty sure shows when I tried to load the "radeon" driver instead of the "radeonHD" driver
agd5f: Travis1: toshiba?
Travis1: yeah Satellite
airlied: agd5f: it seems to be one hell of a lying card
Travis1: haha
airlied: the whole R600 aper base/size regs just look like lies
Travis1: I can tell you exactly what the video card is...
airlied: the vram size may also be
airlied: though I do wonder if we are getting screwd by sign extension somewhere
agd5f: Travis1: I think you are hittign the broken mtrr issue
agd5f: lots of toshiba laptops don't set up the mtrrs right
agd5f: see this thread: http://lists.opensuse.org/radeonhd/2009-05/msg00250.html
Travis1: yeah sure
agd5f: http://lists.opensuse.org/radeonhd/2009-05/msg00040.html
agd5f: looks like the exact same problem
agd5f: you need to fix up your mtrrs which is a kernel issue
airlied: agd5f: why would the card report 4GB VRAM though/
airlied: I can't see mtrr affecting that
agd5f: airlied: I don't know. but running the kernel mtrr santizer seems to fix it
airlied: we should have that on already in fedora
airlied: Travis1: can you pastebin dmesg?
agd5f: airlied: it reports 4 gb of ram because mmio doesn't work
Travis1: yeah sure
agd5f: correctly
agd5f: which is also why the mc never idles and it eventually locks up
Travis1: http://pastebin.com/m88eeecb
Travis1: that's the dmesg report
airlied: Travis1: thats an xorg log ;-)
airlied: oh this is ubuntu so probably not mtrr sanities
Travis1: http://pastebin.com/m19dba11c
Travis1: sorry, that's the one
airlied: wonders how to add options to ubuntu kernel now
airlied: you should find an ubuntu support channel and ask how to add "enable_mtrr_cleanup mtrr_spare_reg_nr=0
airlied: to the kernel command line
agd5f: can't you add them in grub?
Travis1: I'm happy to try anything as long as I can come back from it... hehe
agd5f: Travis1: edit /boot/grub/menu.lst
airlied: agd5f: grub2
agd5f: ah
airlied: though not sure when they migrated to it
Travis1: so what's the command line?
airlied: enable_mtrr_cleanup mtrr_spare_reg_nr=0
airlied: at boot that needs to be added to the kernel command line in grub or grub2
Travis1: so I typed:
Travis1: sudo gedit /boot/grub/menu/lst
Travis1: it opened up a blank document.
agd5f: Travis1: that file no longer exists with grub2
agd5f: I've never used it so, I'm not familiar with how it works
taiu1: did some r600 fixes yesterday http://cgit.freedesktop.org/~andrem/mesa/log/
agd5f: Travis1: probably want to ask on an ubuntu channel
Travis1: yeah I'm asking in #ubuntu... not a lot of response but will keep trying and waiting.
Travis1: are we pretty confident that this change will fix it?
taiu1: dont' like them too much though yet, feel kinda incomplete...
airlied: I think /etc/default/grub add options to GRUB_CMDLINE_LINUX= variable
airlied: then run update-grub
airlied: Travis1: &
airlied: Travis1: ^^ even
airlied: https://wiki.ubuntu.com/Grub2
agd5f: taiu1: looks good. some of the smaller fixes should go in 7.6
Travis1: http://pastebin.com/m2f8275b5
Travis1: that's my current grub file
Travis1: can you guys just talk me through this. I want to be sure I don't put code where it's not meant to be and break something totally...
agd5f: Travis1: I can try, but I take no blame if I hose your system as I've never used grub2
agd5f: Travis1: I think changing this line: GRUB_CMDLINE_LINUX="" to GRUB_CMDLINE_LINUX="enable_mtrr_cleanup mtrr_spare_reg_nr=0" will do the trick
Travis1: haha sure. You've got more of a chance of doing the right thing than me. I can read the wiki on Grub2, but I honestly don't understand it as much as you guys might
agd5f: Travis1: if you can get to the grub menu during boot, You should be able to edit the command line there
soreau: Travis1: After you change that line agd5f said and save the file you would run update-grub as root
soreau: has never user used grub2 either oO
Travis1: hehe sure
Travis1: and I run update-grub within the /etc/default folder right?
agd5f: Travis1: shouldn't matter
Travis1: ok sure
Travis1: well.. here goes nothing. I saved a backup of the old grub file... but if I can't get into the recovery console I guess that won't matter...
agd5f: Travis1: you should still be able to edit the kernel command line from the grub menu at boot, so it's not the end of the world
Travis1: sure. Can I just check what this grub menu is by the way?
Travis1: is that where I choose to load normal or in recovery mode etc?
agd5f: yes
agd5f: see "Editing Menus During Boot" on https://help.ubuntu.com/community/Grub2
Travis1: ok the results of my editing and running update-grub: http://pastebin.com/m14849d1b
agd5f: Travis1: seems fine
Travis1: ok so are you guys saying I should reboot now and just let it boot as normal (unless there are any further problems)
agd5f: Travis1: sure
Travis1: haha love it
Travis1: back soonb
Travis1: I hope
Travis1: here's the latest Xorg.0.log.old file... http://pastebin.com/m58de1f1a
Travis1: it failed again..
Travis1: back to the drawing board?
airlied: Travis1: got the dmesg from that boot?
Travis1: http://pastebin.com/m7bf84d45
Travis1: that might be from the most recent boot tho?
Travis1: I just typed dmesg then.
airlied: hmm looked like it didn't help
Travis1: I'm really sorry this is proving to be so difficult.
airlied: might need to look at it some more, any bios updates available?
airlied: tv &
Travis1: I can only see windows bios updates... on the toshiba website.
Travis1: are we running out of options?
Travis1: airlied or agd5f do you guys have any further thoughts?
Nightwulf: hi all
Travis1: ok so I'm at the point where I'm willing to do a clean install of Ubuntu if anyone thinks that will help at least with seeing the logs and going through the process from scratch
TBBle: Did you get any where last night after I left?
Travis1: airlied and agd5f have both been helping today as well as last night. We installed a patch. We installed an update to my grub file today
Travis1: the thinking was that I need to fix up my mtrrs which is a kernel issue. So we tried something, but it didn't work
chithead: Travis1: I think both are asleep now
Travis1: ah right...
Travis1: is there anyone else with some know-how and patience that might be keen to go through this process with me?
TBBle: Did they by any chance leave any clues as to what might be wong with the MTRRs? Not that I know anything about it. I was frankly here to try and learn something by osmosis.
Travis1: agd5f: Travis1: I think you are hittign the broken mtrr issue
Travis1: (19:38:28) agd5f: lots of toshiba laptops don't set up the mtrrs right
Travis1: (19:38:48) agd5f: see this thread: http://lists.opensuse.org/radeonhd/2009-05/msg00250.html
TBBle: Oh, there's a thread on it? Neat
TBBle: And did you try that?
Travis1: I wasn't quite sure what the thread was saying. So those guys told me to add a line of code to my grub file and try again
Travis1: I tried. It didn't work :(
Travis1: agd5f: http://lists.opensuse.org/radeonhd/2009-05/msg00040.html
Travis1: (19:40:04) agd5f: looks like the exact same problem
Travis1: (19:40:23) agd5f: you need to fix up your mtrrs which is a kernel issue
TBBle: Did you try taking memory out of your laptop? Down to 2gB?
Travis1: uhhh nope.
Travis1: I wouldn't know how.
Travis1: besides, don't I want to be able to use my full 4g?
TBBle: Well, it'd confirm the diagnosis.
Travis1: what's the diagnosis?
Travis1: I still don't understand... hehe
TBBle: The aforementioned MTRR problem. According to the link in the first link you posted, it only affects machines with more than 2gB of RAM.
TBBle: What were the grub changes you needed to make?
Travis1: agd5f: Travis1: I think changing this line: GRUB_CMDLINE_LINUX="" to GRUB_CMDLINE_LINUX="enable_mtrr_cleanup mtrr_spare_reg_nr=0" will do the trick
TBBle: And then ran update-grub, right?
TBBle: Did you check for the CONFIG_MTRR lines in your /boot/config-`uname -r`?
Travis1: yep ran update-grub
Travis1: nooo I don't think I checked those lines...
Travis1: shall I check those lines now?
TBBle: Yeah
Travis1: so what's the exact code I type into my terminal?
TBBle: grep CONFIG_MTRR /boot/config-`uname -r`
TBBle: Also, cat /proc/cmdline
Travis1: CONFIG_MTRR=y
Travis1: CONFIG_MTRR_SANITIZER=y
Travis1: CONFIG_MTRR_SANITIZER_ENABLE_DEFAULT=0
Travis1: CONFIG_MTRR_SANITIZER_SPARE_REG_NR_DEFAULT=1
Travis1: does that look right?
Travis1: BOOT_IMAGE=/boot/vmlinuz-2.6.32-020632-generic root=UUID=75f7f92b-a7c8-4c1d-a49e-4bd1ff32f9ec ro enable_mtrr_cleanup mtrr_spare_reg_nr=0 quiet splash
Travis1: and that
TBBle: Yup, that all looks right.
Travis1: ok so now I should restore the xorg.conf file that keeps breaking and restart?
TBBle: Did you get something in /var/log/kern.log that looks like the dmesg excerpt in http://lists.opensuse.org/radeonhd/2009-05/msg00070.html?
TBBle: The ? isn't part of the URL...
Travis1: haha it comes up normally anyway
Travis1: I'm not sure if I did or not. I could check the kern.log now? Would it come up?
TBBle: Yeah, should come up every boot, with those options set.
TBBle: I think. I kinda skimmed the details a little.
Travis1: haha no worries. I'll check the log now
Travis1: [ 0.000000] original variable MTRRs
Travis1: [ 0.000000] reg 0, base: 0GB, range: 2GB, type WB
Travis1: [ 0.000000] reg 1, base: 2GB, range: 1GB, type WB
Travis1: [ 0.000000] reg 2, base: 4GB, range: 1GB, type WB
Travis1: [ 0.000000] reg 3, base: 4094MB, range: 512KB, type WP
Travis1: [ 0.000000] total RAM covered: 4096M
Travis1: [ 0.000000] Found optimal setting for mtrr clean up
Travis1: [ 0.000000] gran_size: 64K chunk_size: 64K num_reg: 3 lose cover RAM: 0G
Travis1: that's what mine says
Travis1: doesn't look exactly the same.
TBBle: It doesn't continue with "new variable MTRRs" like the post did?
Travis1: [ 0.000000] New variable MTRRs
Travis1: [ 0.000000] reg 0, base: 0GB, range: 2GB, type WB
Travis1: [ 0.000000] reg 1, base: 2GB, range: 1GB, type WB
Travis1: [ 0.000000] reg 2, base: 4GB, range: 1GB, type WB
Travis1: [ 0.000000] e820 update range: 00000000c0000000 - 0000000100000000 (usable) ==> (reserved)
Travis1: [ 0.000000] e820 update range: 0000000000002000 - 0000000000006000 (usable) ==> (reserved)
Travis1: yeah sorry that's after the first part
Travis1: some small differences with the reg 2 and not having the other reg's. Does that matter?
Travis1: do you really think that my laptop having 4GB of RAM is affecting my video card not displaying?
TBBle: The people who wrote the relevant RadeonHD troubleshooting wiki entry, and those mailing list posts think so, yeah.
Travis1: haha sure.
bridgman: Travis1; how about using the mem= boot parameter to simulate having 2GB and see if anything changes ?
Travis1: so the next part is finding a way to fix it :D
Travis1: sure I'll try anything. Can you be a bit more specific about how to do that?
TBBle: I suspect that won't help, it's the BIOS which is mis-building the MTRRs, if I'm reading this correctly.
bridgman: hold on... I think it's just adding something like "mem=2048M" to your boot string, but let me check
TBBle: But certainly, your MTRR maps look sensible before and after the sanitiser, so it might _not_ be this bug, or it might be some other instance of it. But having over 2GB might make a difference, since it removes the 3gB hole from the picture.
TBBle: Are you on a 64-bit OS, BTW? And/or have a 64-bit CPU?
Travis1: I have a 64 bit CPU but I'm running a 32bit OS. I didn't think it mattered, and wasn't sure where or how to get the 32bit Ubuntu OS
Travis1: if you think that will fix the issue, I'm happy to re-install a 64bit OS instead of this one
bridgman: yeah, mem=2048M should do it; let me say up front that this will only help if the problem *is* mucked up mtrrs above 2gb
Travis1: so if I change that, will it do something horrendous to my comp if it fails? Or will it just look like normal?
bridgman: the worst it should do is "not be any better"
Travis1: haha sure. And if it does work, will it mean I won't be using my full 4Gb of RAM?
bridgman: the boot strings only affect the way the os runs this time, when you reboot the settings are gone
Travis1: or will it just be tricking the system?
bridgman: while troubleshooting, yes
bridgman: you are telling the kernel not to use all the available ram
Travis1: sure. OK so where do I enter mem=2048M ?
chithead: kernel command line in grub
TBBle: You _could_ try a 64-bit live-CD, and see if that has the same problem. I had some idea that the AMD-64 setup didn't use MTRRs, and doesn't necessarily have the 3gB hole.
chithead: you could also try memmap=1024M$0x80000000 so only the area between 2048 and 3072 mb is unused
TBBle: Same place you added the enable_mtrr_cleanup stuff
Travis1: ok I'm downloading the 64 bit version. I'll try this mtrr thing, but if we're not getting anywhere, I'll try installing the newer version
Travis1: I know with Windows the 64 bit version is better at recognising higher volumes of RAM
chithead: "better at recognising"? it is architectural limit of 32 bit x86 kernels that only about 3 gb ram can be used. disregarding ugly things like pae
Travis1: riiight...
Travis1: so I'm being penalised for having a better computer? hehe
TBBle: You're not being able to take full advantage of your 64-bit capable system under a 32-bit operating system, yes.
chithead: if you have more than 2gb ram and don't depend on 32 bit proprietary software, then 64 bit is the way to go
bridgman: more like you're not getting full advantage while the world isn't ready to supply 64-bit versions of everything
Travis1: yeah I understand that. but even if I run a 64bit OS, I can still run 32bit apps on it can't I?
bridgman: as long as you have 32-bit libs on the 64-bit os as well; some distros don't like to do that
chithead: in principle yes, but it can be a bit more work to install certain 32 bit software
bridgman: you sort of need a hybrid 32/640 OS
bridgman: and the hybridization is never perfect
Travis1: so regardless of this fix that you're suggesting, I should really just install the 64bit OS anyway since I have 4GB of RAM..
Travis1: and there could potentially be less hardware and driver issues doing that...
bridgman: as long as you're going to be running 64-bit apps or 32-bit apps which work ok on a 64-bit system, yes
Travis1: yeah sure. Well at least Adobe are supporting 64bit linux updates before windows and mac with their 64bit flash
Travis1: suck on that microsoft.
TBBle: If Microsoft cared about that, they'd probably have bothered with a 64-bit Silverlight.
chithead: apart from flash, adobe are not so into 64 bit. eg. adobe reader + browser plugin only exist in 32 bit
Travis1: hmmm
chithead: and it took years of bugging by users to get browser plugins for 64 bit
Travis1: so the ubuntu software center... will that then suggest only 64 bit software if I'm running the 64bit OS?
TBBle: Hmm. I'm surprised there's no open-source PDF reader browser plugin, poppler-based or something.
chithead: TBBle: mozplugger
TBBle: Ah, that makes sense.
Travis1: yeah I think I was running 64bit vista prior to windows 7 rc and I didn't have too many problems with that. Considering it could run 32bit apps just fine. But there may be a difference with Ubuntu and the way it handles that. Not sure
chithead: 64 bit vista is somewhat like 32 bit windows 95, where a large part of the userland was still 16 bit
Travis1: I'm not sure what that means...
Travis1: I'm trying my hardest to learn comp & os things, but I'm quite rare still... hehe
Travis1: I really hope that one day I'll be able to jump on these chats to help people out like you guys have done for me.
TBBle: I was running XP x64, was relatively happy with it, but Windows 7 x64 was better. During that whole period, got more mileage and success from my Debian/unstable AMD64 though. Even with the ia32-lib-tools "incident" earlier this year.
Travis1: so to ask my question again, will the Ubuntu software Center supply me with 64bit apps?
sxy: Hi guys!
sxy: Can someone please help me with a problem?
[Enrico]: sxy: ask :D
sxy: I use latest radeon drivers with my x1950xt board
sxy: lol [Enrico] :)
[Enrico]: yeah i'm here too :D
sxy: and when I try to run openGl game I get this error: GL_ARB_occlusion_query not found
sxy: glxgears work, DRI works, hardware 3d acceleration works
TBBle: Travis1: Dunno. Maybe ask on #Ubuntu? Or heck, their website might tell you. I _expect_ that on 64-bit it'll either supply 64-bit apps, or 32-bit apps with appropriate dependencies to pull in 32-bit librarie.s
sxy: glxinfo | grep GL_ARB_occlusion_query
sxy: outputs nothing
TBBle: Does glxinfo claim that ARB_OCCLUSION_QUERY is supported?
TBBle: Ah, right.
sxy: other GL_ stuff is okay
TBBle: Then it's not supported by that driver, I guess.
sxy: as I thought :S
sxy: now
TBBle: Are you using latest mesa?
sxy: Yes I do
TBBle: What does xdriinfo say you're running?
sxy: mesa-7-5-1
sxy: Screen 0: r300
sxy: is this okay?
sxy: after all my card is r580
sxy: (not sure if r300 refers to the name of the chip)
TBBle: I'm not sure if 7.5.1 is recent enough.
TBBle: r300 is the DRI driver name, that's the mesa driver you're using. It's correct.
adamk: sxy, A newer Mesa should provide GL_ARB_occlusion_query
sxy: I'm on gentoo, kernel is 2-6-30-r5
sxy: x86_64 system
hnsr: where should I get the "radeon/R700_rlc.bin" firmware? I don't run a free distro (gentoo), but I just switched to airlied's drm-radeon-branch
TBBle: You need mesa 7.6 or later for GL_ARB_occlusion_query in r300
TBBle: hnsr: On the dri-devel mailing list, I got them.
hnsr: TBBle, ah, thanks
sxy: I'm upgrading to mesa-7-5-2
sxy: (it is still masked in the portage)
amarsh04: here is the link hnsr: http://marc.info/?l=dri-devel&m=125960757404659&w=2
TBBle: sxy: For reference, ARB_occlusion_query is commit 60587182d4ade36df75ee13edf8df6b529fbb0f1 in mesa git.
TBBle: 7.5.2 is _not_ 7.6 or later.
adamk: Heh :-)
sxy: TBBle: It is not in the portage :(
sxy: (I'm on gentoo)
hnsr: thanks amarks_
hnsr: oops, amarsh04
adamk: sxy, Whether it's in portage or not does not change the fact that you need it for OQ :-)
TBBle: sxy: Well, maybe google will turn up a guide on getting mesa 7.6, 7.7-prerelease or master into Gentoo? I don't use it, so dunno off hand. I thought sometimes recipes for newer package versions showed up in Gentoo's bugzilla?
hnsr: sxy, the gentoo X11 portage overlay has mesa 7.6.1/7.7_rc1
sxy: thank you guys!!! :)
hnsr: and also a live ebuild (-9999) which you can use to get stuff from git
sxy: I have to learn how to add portage overlays
hnsr: sxy, you should, they're pretty handy if you like to run cutting-edge stuff
hnsr: use layman by the way :p
sxy: layman -a x11 && emerge mesa
Wizzup: sxy: I'm pretty sure you'll have to unmask mesa too
Wizzup: mesa-9999 that is, if you want the latest
sxy: okay
sxy: I have to emerge git first :S
Travis1: ok guys, gonna go install 64bit Ubuntu to see if it will solve all my problems... wish me luck!! :)
wirry: good luck :P
TBBle: There goes another brave warrior, off to the front lines in the 64-bit revolution. *tear*
Travis1: ok so I go to install the Ubuntu 64bit OS... and the screen goes blank!
Travis1: I select to install the OS... it starts loading... blank.
TBBle: Well, I guess consistency is nice...
Travis1: haha depends on the consistency
Travis1: It was the same when I was installing this OS. I had to choose 'safe graphics mode' at the ubuntu install menu and I figured I would sort out the graphics later...
Travis1: but I guess now after all this, I figure if I did that again I'd prob end up in the same boat as now. Or do you think there might be something different to it next time?
TBBle: Have you tried an older version of Ubuntu?
TBBle: Nah, if it's failing on boot, it'll fail in-system as well.
Travis1: like I'll need to go into the OS and install the proprietary drivers before it actually starts working?
TBBle: The proprietary drivers didn't work for you either, did they?
Travis1: not in here they didn't. But is that because of the 32bit issue?
TBBle: I don't suppose you had the Xorg.0.log from that attempt, by any chance?
Travis1: no it just failed on me...
TBBle: I doubt it. If going to 64-bit would fix the fglrx drivers, it probabby would have fixed the radeon driver too.
Travis1: and it would've done that during bootup?
TBBle: Yeah, I think so.
Travis1: I haven't tried any older versions of Ubuntu... happy to give that a go as well tho. Would you recommend one that you think would work? And also link me to where I could download it from... hehe
sxy: okay.. where to put mesa-9999 ?
sxy: in the /usr/portage/profiles/package.unmask ?
sxy: or somewhere else?
MNZ: has agd5f's drm branch been merged into upstream yet?
chithead: sxy: please read gentoo documentation, http://www.gentoo.org/doc/en/handbook/handbook-x86.xml?part=3&chap=3 how to unmask
TBBle: MNZ: Which branch?
Travis1: could someone recommend a version of Ubuntu that should work with my radeon 3650 and where I could download it from?
MNZ: TBBle, is there more than one? well I meant the r600 development repo he has
uyf: hi everyone, is s3tc compression supposed to work with kms mode. i can't find any info that it shouldnt work
TBBle: Travis1: I'd start with whatever's before the current one (jaunty, I guess) and see if that works.
TBBle: uyf: Isn't s3tc a mesa concern, not a driver-specific concern?
TBBle: An r600 drm development tree I wasn't aware of? I'm off to find that. ^_^
MNZ: TBBle, git://anongit.freedesktop.org/~agd5f/drm
uyf: TBBIe: ok just thought it might be the driver since it works for me with fglrx but not with oss with yesterdays gits.
TBBle: Oh, libdrm. I'm thinking of kernel drm.
Travis1: ok getting Jaunty now
TBBle: uyf: fglrx doesn't use mesa, they provide their own GL stack, which I presume includes a licensed s3tc implementation, or whatever it is that causes it to not be in mesa in the first place.
MNZ: TBBle, it's the kernel drm code.
BioTube: MNZ: patent issues are the same
TBBle: Oh, pre-rearrangement.
uyf: TBBIe: yes of course, now i understand. so noone has working s3tc with r600 drivers then. i was suspicous that there was something wrong with my card because i have had strange problems with fglrx drivers too.
MNZ: BioTube, what are we talking about exactly?
TBBle: uyf: Can't you do whatever the normal procedure is to enable s3tc? (I think there's a seperate file you compile as a mesa plugin?)
BioTube: MNZ: got confused
uyf: TBBIe: ok i will try that
BioTube: i remember somebody saying it doesn't work with r600 or multitexture
MNZ: TBBle, rearrangement ?? I did see a huuuuge commit on the drm repo a while ago but never investigated
TBBle: MNZ: The kernel drm code no longer lives in the drm repositories on fd.o, it's now living in the various linux kernel trees (and appropriate other operating system kernel trees) with the drm tree containing only libdrm.
MNZ: TBBle, :/ so how do I get the latest drm code without downloading the entire kernel? or is that not possible any more?
TBBle: But I seem to have dri running on my rv770 card, so something equivalent must have been merged, since it looks like the branch you're talking about is the actual implementation, not just fixes.
TBBle: Yup, entire kernel. You want to grab stuff from airlied's drm-2.6 repository on git.kernel.org. There's a mail on dri-devel describe the branches there, you probably want drm-radeon-testing for the bleedingest edge radeon stuff.
MNZ: It's been fun using packaged kernels for a few months. Oh well. Is KMS working for r600/700 now?
BioTube: MNZ: is for me
TBBle: And me.
TBBle: Although someone was here last night saying it wasn't for them.
BioTube: latest radeon code does need the interrupt firmware to work, IIRC
TBBle: Oh yeah. ^_^
MNZ: um where do I get this firmware from?
TBBle: But with latest drm-2.6 kernel, mesa, libdrm, and a couple of patches, you get R6xx/r7xx + KMS + DRI + HDMI audio
TBBle: dri-devel list. (23:54:58) amarsh04: here is the link hnsr: http://marc.info/?l=dri-devel&m=125960757404659&w=2
Travis1: so is there a reason that an old OS install would work better than Karmic?
Travis1: shouldn't Karmic have all the same drivers as the old Ubuntu's? And work the same as them?
Travis1: or better...
TBBle: 'Should', yes.
Travis1: hmmm. That would make sense to me
TBBle: But if we're bouncing off something that the older code handled differently, more poorly of course, but that happens to work with your card, then you'll have a working system and we'll have a datapoint.
Travis1: hmmm
hnsr: 2D seems to be prety fast now with KMS
hnsr: pretty*
uyf: i think its blazing fast
uyf: compared to fglrx
BioTube: everything is
TBBle: Oh, libdrm_radeon's API is stable?
MNZ: airlied/drm-2.6.git <-- this is the one I need right?
MNZ: TBBle, the flag for it is still "--enable-radeon-experimental-api" so I guess not haha
BioTube: MNZ: yes
TBBle: Oh, and now libdrm 2.4.16's tagged, the xorg-radeon configure script needs to depend on it for DRM_MODE, in that mode it uses one of the new APIs.
TBBle: Well, the Debian 2.4.16 libdrm package is likely to have libdrm_radeon packaged, so if the API's going to change such that a soversion bump is needed, I hope upstream will mirror the bump...
Ivanovic: hiho
Ivanovic: using kernel 2.6.32 and latest mesa/libdrm (git from master) i got some annoying flickering when starting "sacred gold"
Ivanovic: i got a radeonhd 3850 (rv670) card
Ivanovic: the flickering is somehow as if there was a black screen drawn
Ivanovic: that is: game screen, black screen, game screen, black screen, ... where you can see each redraw
Ivanovic: in general it seems to be fast enough for playing, but the flickering makes the eyes hurt real bad
TBBle: Is this vanilla 2.6.32 or 2.6.32 + one of drm-2.6's drm-radeon-* branches?
Ivanovic: that is: there is no problem with playing world of goo windowed on my system, everything perfectly fine, sacred gold seems a little more problematic
Ivanovic: vanilla 2.6.32
Ivanovic: already had this flickering some time ago when testing rc7 and the likes
Ivanovic: no graphics related message in any logs
TBBle: Well, it's possible you're missing some page-flipping changes in the kernel, but I don't see anything startlingly relevant in the git logs.
TBBle: Oh, there's the page-flipping ioctl support in drm-2.6/drm-core-next. But I don't know how its absence is handled by libdrm master.
Ivanovic: okay, when starting in full screen there is no flickering
Ivanovic: next prob when starting the game itself is that the interface is shown, but the "game view" itself is not
Ivanovic: (as in i load a game, see the HUD, but not the map with the char running around, that area is just plain black)
Ivanovic: but it is really great to see the progress of the open drivers, lovely to be able to start the game already
TBBle: Huh. The person who's enabling libdrm_radeon in Debian is here, so I'll assume it's all sorted.
edt: with an rs780 based r600 with the firmware posted in: http://marc.info/?l=dri-devel&m=125960757404659&w=2 built into the a 2.6.32 kernel with airle's drm-next from dec4 10pm and latest libdrm, xf86-video-ati and mesa iterrupts are working here and, so far, no corruptions or problems noticed. THANKS.
edt: with KMS/DRI2 enabled
edt: airlied of course right after posting all was good I managed to get the kernel to stall. Want the console?
edt: see: http://pastebin.ca/1703191
edt: agd5f you may also be interested in the above stall
edt: see: http://pastebin.ca/1703194 & with an additional comment
edt: pm me if you want a bug opened for this.
edt: TYPO see: http://pastebin.ca/1703193 & with an additional comment
ossman: osiris_, ping
osiris_: ossman: pong
ossman: osiris_, did you ever test progs/perf/copytex?
ossman: it provokes some nasty bug in the radeon code
osiris_: ossman: yes, I did
ossman: and it didn't crash and burn for you?
osiris_: ossman: nope
ossman: odd... from what I can tell, the problems are in the common code
ossman: the basic issue is that the radeon code tries to fiddle with the texture that's the target for glCopyTexImage
ossman: and since Data is NULL at that point, things crash
ossman: I commented out the texture mapping in radeonSpanRenderStart(), which avoids the crash
ossman: osiris_, do you have any insight into how these pieces should fit?
osiris_: ossman: yes
osiris_: ossman: please report a bug, I'll look into it tomorrow
ossman: osiris_, k
ossman: osiris_, bug 25463
osiris_: ossman: thanks
eosie: interesting, if I set a scissor rectangle of min=(0,100),max=(0,99) (all pixels culled), it dumps a CS... maybe we shouldn't invoke a draw call at all in this case
flyback: http://www.youtube.com/watch?v=JGjZLqR2Jf8&feature=related <--- this is the main reason that wind and solar power never took off till recentely. It was not "evil oil" it was a pain in the ass to achieve this before modern cheap electronics http://www.youtube.com/watch?v=JGjZLqR2Jf8&feature=related
flyback: every time the wind speed changed you could lose sync and crash the grid
felipec: anyone else has problems with Xv in R500?
flyback: needs to start working on his xmas tree electronics hack today
flyback: bbl
flyback: doesn't have to go food shopping, so going to work on cleaning his room, won't be around much for helping people, so leave message and I will get back to you when I can
ossman: osiris_, I'm thinking the problem is that the texture is marked complete, when it really isn't. The problem is that we can't just check Image->Data as the driver might have copied it to something internal. Is there a callback or something to see if this texture has actual data?
osiris_: ossman: yes, that's what I think too. I don't think there's a callback, but for radeon you just check if all radeonimage->bo, radeonimage->mt, radeonimage->base.Data are null then there's no data
ossman: osiris_, yeah, but the check is in mesa core so it has no knowledge of radeon specific things
osiris_: ossman: for this specific bug, I think it should be enough just to s/else/ else if (image->base.Data)/ in migrate_image_to_miptree
osiris_: ossman: such situation will probably take place only when we're falling back to software for something (i.e. radeonSpanRenderStart is called)
ossman: indeed
ossman: but it feels like that would be a workaround
ossman: perhaps a new callback is needed to test completeness?
osiris_: ossman: I don't think so. try analyzing why we need to map textures in radeonSpanRenderStart at all
osiris_: ossman: probably it isn't needed at all
ossman: osiris_, I'm not sure I'll be able to do that. From the looks of the code, it's there to guarantee an assumption by some other part of the code. So short of reviewing everything, I don't know how I'd be able to check that
ossman: osiris_, the code was touched not that long ago by airlied
ossman: osiris_, the code was added a year ago by glisse. commit e5d5dab8c03f72097ec3e5b465fe93b6e369bb2d
spreeuw: r600_texstate.c:920: error: too few arguments to function ‘radeon_update_renderbuffers’
spreeuw: make[6]: *** [r600_texstate.o] Error 1
spreeuw: make[6]: Leaving directory `/home/pheldens/src/build/mesa/src/mesa/drivers/dri/r600'
spreeuw: present git
airlied: edt: yup file that please doesn't look good
hagabaka: wow
hagabaka: mplayer in framebuffer looks just like in X
spreeuw: hmm tried -vo fbdev and it sort of worked
spreeuw: it wasnt scaled fs
spreeuw: and it was on all terminals
spreeuw: overwriting xorg
spreeuw: ah fbdev2 works better
MNZ: I'm getting spreeuw's error too
airlied: bridgman: OpenCL we should talk :-)
turmlos: airlied: Regarding the sideport memory detection issue you mentioned the other day, does it only affect KMS?
airlied: turmlos: at least for suspend/resume yes
turmlos: airlied: Ah, haven't gotten that far. Maybe my problems are unrelated.
turmlos: airlied: Any ideas? http://sturmartillerie.org/linux/radeon-kms.png Xorg log and dmesg are in parent dir.
adamk: What is the package name on F12 that provides the r600 driver?
adamk: Never mind... Don't know why I didn't just google it first.
MostAwesomeDude: mesa-dri-drivers-experimental, from $ yum whatprovides /usr/lib/dri/r600_dri.so
adamk: Yeah, I just plugged in 'fedora 12 r600 driver' in google and got the name. I always forget how helpful google can be sometimes.
airlied: turmlos: wierd, maybe sideport does bad things to us normally
turmlos: airlied: Should I open a bug?
airlied: turmlos: yeah, any bios options to disable SP?
turmlos: airlied: Negative.
ossman: MostAwesomeDude, what would be the best way to construct r600 fragment and vertex programs from inside the r600 driver?
ossman: Using the assembler? Or the shader interface directly?
MostAwesomeDude: ossman: For what purpose?
ossman: MostAwesomeDude, implementing an accelerated blit
ossman: so it's fairly simple programs
MostAwesomeDude: ossman: Write it by hand.
MostAwesomeDude: I'd only use the assembler if you think that it'll give a speedup, or if you think that the hand-written prog would be too tough to maintain.
ossman: MostAwesomeDude, what routines do the actual loading of the progs?
MostAwesomeDude: ossman: Couldn't tell you; I haven't dug deeply into that code yet.
ossman: k...
ossman: seems like this would be the first user of progs outside the direct GL interface
ossman: might not be any helper routines yet...
Zajec: i am looking for cheap notebook with ATI card
Zajec: I see many with HD3200 with shared memory
Zajec: should I expect any problems becase of that?
Zajec: it seems they only have shared memory...
ossman: agd5f, ping
MNZ: No Xorg with KMS on :(
MNZ: http://pastebin.com/m4763e282 <-- xorg.log
airlied: MNZ: make sure you modprobe radeon before X starts
airlied: also make sure -ati is built with kms support it doesn't loook like it
MNZ: airlied, isn't radeon modprobed automatically?
Lebenskuenstler: hi agd5f. i tried your patch to fix flickering in high resolutions over DVI. I am sorry report, that it did not work for me.
agd5f: ossman: just allocate buffers for vp/fp and copy hand written shaders to them
agd5f: Lebenskuenstler: did you try both patches?
ossman: agd5f, there doesn't seem to be any decent defines though, so readability will be horrible
ossman: I'm looking at using the structures for R700_Shader
ossman: bad idea?
agd5f: ossman: they are only about 6 commands, hex code is fine
ossman: k
agd5f: ossman: you can even use the ones from the kernel
ossman: ?
Lebenskuenstler: agd5f: you gave me this one: http://www.botchco.com/alex/xorg/0001-drm-radeon-kms-pll-fixes.patch BUT I think i did something wrong. I will try once again.
agd5f: Lebenskuenstler: http://www.botchco.com/alex/xorg/0001-drm-radeon-kms-tighten-the-pll-range-when-selecting.patch
agd5f: ossman: we accelerate bo moves in kms using a 3d blit as well
ossman: agd5f, got a filename? :)
agd5f: r600_blit* in drm tree
Lebenskuenstler: ag5df: which one should i try first?
ossman: agd5f, I'll give it a go
agd5f: Lebenskuenstler: the first one
Lebenskuenstler: adg5f: ok, I will give you feedback. thanks.
ossman: agd5f, that huge "default state" thing, is that something I need to pay attention to here as well?
agd5f: ossman: that's the stae, you need, but it's easier to just refer to set_default_state() in r6xx_accel.c and R600DoPrepareCopy() in r600_exa.c
DanaG: hmm, lower-end notebooks with ATI.... does this count as low-end? http://jkontherun.com/2009/11/19/lenovo-thinkpad-x100e-netbook-info/#comments
agd5f: and then use the state tracking in mesa to set the right bits
DanaG: doesn't yet say what gpu.
ossman: agd5f, I'm afraid my knowledge of mesa state tracking is non-existent :/
agd5f: ossman: in the r700 hw struct we track the state of all the 3d regs
agd5f: so in the blit function just set them up like you need them for the blit and set all the state as dirty and emit it
ossman: agd5f, radeon->hw.is_dirty ?
agd5f: ossman: yeah or all_dirty
ossman: the difference being?
agd5f: don't recall the difference off hand
ossman: k
DanaG: Actually, I'm curious... could one theoretically use coreboot on AMD M880G?
DanaG: http://h10010.www1.hp.com/wwpc/us/en/sm/WF06a/321957-321957-64295-3929941-3955552-3935050.html
DanaG: "ATI Radeon HD 4200 with up to 256 MB dedicated video memory -- how can it be "up to"?
agd5f: DanaG: probably marketing speak
agd5f: referring to how much system ram is reserved for fb
DanaG: =P
gimzo: I jave hd4200 with 128MB dedicated and can take from system ram to add more
gimzo: *have
gimzo: but when the gpu is disabled those 128MB can't be used anywhere else
ossman: agd5f, r700InitState should do what I want?
agd5f: ossman: no. you'll have to manually set all the state like the ddx does
agd5f: in set_default_state()
ossman: it looks like they should share things at least. like the sq setup
agd5f: yeah some stuff is common
Zajec: i am looking for cheap notebook with ATI card
Zajec: I see many with HD3200 with shared memory
Zajec: should I expect any problems because of that?
Zajec: it seems to be only shared memory in these... ?
DanaG: hmm, define "cheap", also.
agd5f: Zajec: no
agd5f: they work fine
Zajec: DanaG: the cheapest with ATI and not being Acer or Aristo :P
Zajec: DanaG: its 1500-1600zl in Poland, our unit :)
gimzo: what's wrong with acer ?
DanaG: http://h71016.www7.hp.com/html/hpremarketing/daily.asp#SMB10 -- refurbs. Link way above was about Lenovo.
DanaG: Are you specifically looking for R600 or newer?
Zajec: gimzo: i got one and saw enougt of them in life :)
Zajec: gimzo: mostly broken :P
Zajec: DanaG: no, just some cheap card capable of compiz
Zajec: compiz + maybe
Zajec: maybe some simple 3D games, but nothing special
Zajec: it's weird by the way that I was not able to find a one notebook with Intel X3100/X3500/X4500 with HDMI...
Zajec: don't they have HDMI outputs by design?
gimzo: Zajec: I have one here, that I got because the battery died, this one survived some heavy usage, it has a few broken keys, but works ok (except the battery)
Zajec: agd5f: great, thanks :)
DanaG: http://www.notebookreview.com/default.asp?newsID=5015&review=hp+probook+4510s
DanaG: hdmi.
DanaG: Intel.
DanaG: But plastic, not metal like the higher-end ones.
Zajec: DanaG: oh, so they do :)
DanaG: And I think that one is Broadcom. :(
Zajec: i actually need Broadcom ;)
MNZ: I'm trying to enable KMS on r700. So far enabling it causes xorg to fail and I get something about an undefined symbol (radeon_vbo_init_list). Disabling KMS, xorg starts up, black screen, and I can't switch to consoles.
agd5f: MNZ: re-run autogen.sh
Zajec: gimzo: my got many white pixels, but no guaranty because... not enought in central square of panel when deviding it into 9 pieces... ! damn, can you believe that?
agd5f: and rebuilt your ddx
gimzo: Zajec: Interesting, haven't heard of that before
Zajec: gimzo: then hinges on my panel broke (common in Acers)
MNZ: agd5f, ok will try that.
Zajec: gimzo: google for acer accidents, you can see many amazing stories ;)
gimzo: ok
Zajec: HPs generally look fine for me
Zajec: but there is one horrible thing about them...
DanaG: Whitelist?
Zajec: BIOS with wirelss whitelist
Zajec: DanaG: bingo
DanaG: At least non-Elitebook is non-UEFI, so it's more hackable, I've heard.
Zajec: DanaG: i don't like compfortable reHEXing bios or hacking wireless card for PCI ID and then driver for same...
DanaG: And EliteBook UEFI turns out to be just messed up.
Zajec: *like -> feel
DanaG: Oh, and notebookreview.com -- handy.
elbeardmorez: hi. i'm trying to force dual-screen with xrandr on an old 9200 pci but both monitors always show the same section of the screen, inconsistent with what verbose output tells me. http://pastebin.ca/1703528 ..any ideas please?
Zajec: elbeardmorez: --left-of --right-of etc./
DanaG: Biggest thing I'm waiting for: KMS + PM going into some git branch, so I can keep up to date without manual patches.
Zajec: DanaG: i've few days of hard working on school projects
Zajec: DanaG: will back to that soon
Zajec: DanaG: will see if I can apply glisse's suggestions
DanaG: I'd even give you a cookie, or a cake, or something. =þ
elbeardmorez: Zajec: thanks, but yes i've tried that. the pastebin shows that xrandr thinks it has correctly setup the displays ..but it hasn't!
Zajec: sorry, opening pastebin now :)
DanaG: Zajec: oh, and check PM, unless you have those blocked.
DanaG: by "UEFI messed up", I mean, the pointer to the framebuffer in GraphicsOutputProtocol seems to give the address as zero. And things tend to get trampled on and corrupted, and all that.
elbeardmorez: Zajec: so A = B clones, A right-of-B shows RHS of dual-screen, A left-of-B shows LHS of dual-screen ..despite what the 'pos XxY' output states for the connected displays.
ossman: agd5f, I might be in a bit over my head here, so don't consider this issue handled if you feel like looking at it yourself
Zajec: DanaG: yup, got message :) but that HP... eh, not sure
Zajec: DanaG: maybe I'll get some Asus
Zajec: DanaG: i saw some with AMD Turions and HD3200
Zajec: DanaG: what do you mean by UEFI?
Zajec: http://en.wikipedia.org/wiki/Extensible_Firmware_Interface ?
DanaG: yup.
Zajec: elbeardmorez: yeah, xrandr looks fine... even with CRTCs set correctly
agd5f: ossman: no worries :)
Zajec: elbeardmorez: don't know, maybe driver sets something incorrectly actually
MNZ: agd5f, that worked. Though now with KMS on compiz dies on me
Zajec: elbeardmorez: do you use KMS?
elbeardmorez: yes
Zajec: elbeardmorez: did you try without?
soreau: MNZ: How does it die?
MNZ: Solefald, white screen. then a bunch of flickering (apps show, though just white rectangles) then more white screen. Then I just killed it from the console
MNZ: woops
MNZ: I meant soreau
soreau: Sounds like swrast
soreau: What is the output of 'glxinfo|grep renderer'?
elbeardmorez: Zajec: hmm. nope. does VGA arbiter work without KMS?
MNZ: swrast :/
soreau: Can you pastebin X log?
Zajec: elbeardmorez: err, do you need VGA arbiter ?
Zajec: elbeardmorez: dual head works fine, does not need VGA arbiter
MNZ: soreau, http://pastebin.com/m3ffe1bb4
MostAwesomeDude: vgaarb kicks in on all cards, but is turned off by KMS.
elbeardmorez: Zajec: yes ..two identical PCI cards ..part of a 'been waiting for years to get it working' 4-screen setup ..so that could also be an issue. i'll try without KMS though.
soreau: MNZ: Have you built latest libdrm and compiled mesa and ddx against it?
MostAwesomeDude: elbeardmorez: Works for me with KMS.
MostAwesomeDude: elbeardmorez: Could you perhaps pastebin your dmesg?
MNZ: yeah I have. though not entirely sure about mesa.... also mesa stopped compiling a few hours ago after a git pull.
soreau: MNZ: Can you pastebin the output of the failed mesa build?
MNZ: soreau, r600_texstate.c:920: error: too few arguments to function radeon_update_renderbuffers
MNZ: another guy was getting the same thing on the channel, and airlied asked to report it
soreau: MNZ: Well I guess you could try git checkout HEAD~1 to revert one commit at a time and see if you can find a good one
soreau: Or git log and checkout a commit that looks suspicious
soreau: one before it even
DanaG: I wonder why compiz can't work under swrast.... I'd think it should work, though perhaps 1 FPS.... not just be white.
elbeardmorez: shocking, there is an error message there sorry.. http://pastebin.ca/1703586
evocallaghan: rnoland: Are you around?
MostAwesomeDude: elbeardmorez: Do you have two fbs? /dev/fb0 and /dev/fb1?
MNZ: soreau, I checked out 1 commit before and it compiled alright.
elbeardmorez: MostAwesomeDude: yes i do.
soreau: MNZ: Cool. Install it and restart X then see who made that commit and complain to them ;)
MNZ: it works!
soreau: MNZ: Also, you can disable AIGLX to get direct rendering
MostAwesomeDude: elbeardmorez: Sweet. Both of them should be radeondrmfb DRI-capable, and you should be able to multicard/multiseat on those.
elbeardmorez: MostAwesomeDude: there's a revelation ..DISPLAY=:0.1 xrandr --output DVI-1 --above VGA-1 ..works as it should do. so it's only the first card that has the problem.
MostAwesomeDude: elbeardmorez: To get a multicard xorg.conf, use Xorg -configure.
MNZ: what is AIGLX for actually?
MostAwesomeDude: MNZ: Acceleration of indirect GLX contexts.
soreau: MNZ: Accelerated Indirect GL X
MostAwesomeDude: elbeardmorez: Oh, you're already in X?
soreau: MNZ: So right now, you are running compiz with --indirect-rendering
evocallaghan: Accelerated Indirect GLX
elbeardmorez: MostAwesomeDude: yes.
soreau: MNZ: Alternatively if you want to disable KMS and use DRI1, you will need to have AIGLX enabled for 3D to work
MostAwesomeDude: elbeardmorez: Alright, so you're just trying to get the screens in the right positions.
MostAwesomeDude: elbeardmorez: There's some non-obvious fun here; you have to have your xrandr terminal on the card you're trying to configure.
MNZ: where does the option for AIGLX go? under Device?
MostAwesomeDude: So you'll need *two* terminals in order to get both cards configured. After you've got it, I suggest you copy it to xorg.conf.
soreau: MNZ: It would be Option "AIGLX" "off" in the ServerFlags section
DanaG: hmm, how does kms deal with multiple cards?
DanaG: Different console on each card?
DanaG: Where does keyboard go?
elbeardmorez: MostAwesomeDude: yes. i can happily send xrandr commands to the correct screen ..i can set clone mode ..turn off crts ..i am affecting the correct screen/monitors, it's just that it's not setting the screen the way it should.
MostAwesomeDude: elbeardmorez: Hm, not sure.
rnoland: evocallaghan: whats up?
evocallaghan: rnoland: Hi, hows it going? ; I just wanted to know if you started working on GEM and how far you may of got with that?
MNZ: reverted to swrast with AIGLX off. Also when I turned it back on I noticed the renderer is now Mesa DRI R600, but direct rendering is off, because it looks for the drivers in @DRI_DRIVER_SEARCH_DIR@/r600_dri.so
soreau: MNZ: Can you pastebin the X log with AIGLX off?
rnoland: evocallaghan: i have a branch with it started.... but it is in now way usable yet.
MNZ: isn't @DRI_DRIVER_SEARCH_DIR@ Makefile syntax or something?
soreau: somethin like that
evocallaghan: rnoland: sure sure, possibly interested to help. Where may I ask?
evocallaghan: rnoland: looks like GEM has made it into opensolaris at last. Possibly use some of that?
rnoland: um, well i don't have my repo public... but....
rnoland: i use a git version of the freebsd repo....
elbeardmorez: MostAwesomeDude: thanks anyway.
MostAwesomeDude: elbeardmorez: Sure. Xrandr's not that tough to get; there's plenty of help if you Google for it.
evocallaghan: rnoland: sure, maybe quicker if you just generated a patch of your current work and emailed me it?
rnoland: um, yeah... i don't mind doing that...
MNZ: soreau, or is it automake/autoconf/etc syntax... in any case, I should actually have a path in place of that variable right?
evocallaghan: rnoland: cheers,
soreau: MNZ: Obviously
rnoland: it may actually still need to have two branches merged right now...
evocallaghan: @ auroraux.org
rnoland: evocallaghan: drop me a mail and I'll make a patch...
soreau: MNZ: You could probably set LIBGL_DRIVERS_DIR to tell it where to look, but that's just a workaround
rnoland: it is really pretty raw right now though....
evocallaghan: sorry, I forget your mail.
rnoland: rnoland@freebsd.org
evocallaghan: rnoland: sure, just wanted to know where your at.
MNZ: soreau, bunch of grepping later, looks like I needed to run autogen.sh again for mesa :/ it's now set properly (in the conf files at least, haven't compiled yet)
soreau: MNZ: Yea, I always make a habit of make clean, configure/autogen then build
evocallaghan: rnoland: p.s. do you have any currently known issues with X core dumping with intel cards?
soreau: Typically avoids potential issues
rnoland: evocallaghan: not that i know of...
evocallaghan: mine does it a *lot*
MNZ: I always make clean, but never run autogen. Second time this has happened to me today haha.
elbeardmorez: MostAwesomeDude: but it has to be a quirk with dual-card / KMS / VGA arb. right, i can't be screwing up simple xrandr commands - my right hand two monitors 'screen 1' 4 above 3 works fine! hence i imagine Google won't help. I've built the .32 kernel (as opposed) to rc6 (running) ..see if that helps. cheers.
evocallaghan: rnoland: it does so on solaris also
evocallaghan: I got a trace from there
evocallaghan: digs that out.
rnoland: then i would guess it might be a userland issue...
rnoland: but my g45 seems to be ok....
rnoland: evocallaghan: so are you familiar with the freebsd vm system at all?
rnoland: because that is kinda a prerequisite for doing gem or ttm....
evocallaghan: rnoland: only the basics, but happy to go off and learn anything in my way. don't worry about me. your a.) get no patches or b.) get patches :)
rnoland: evocallaghan: hehe, ok...
evocallaghan: rnoland: here's that bug I was talking about, not sure of the code path, seen anything like it before? http://defect.opensolaris.org/bz/show_bug.cgi?id=12801
rnoland: patches are better than not patches, but....
evocallaghan: lol
agd5f: mesa issues should be fixed now
rnoland: evocallaghan: this trace appears to point to libdrm as the issue...
rnoland: or libdrm_intel specifically.
MNZ: http://pastebin.com/m6a14606b <-- after disabling AIGLX. Compiz white screens me. Renderer is Mesa DRI, direct rendering is enabled.
rnoland: evocallaghan: perhaps a refcounting bug in intels code
evocallaghan: rnoland: sorry, can you explain a bit more, just want to make sure I know what your thinking.
rnoland: evocallaghan: so, libdrm provides the userland api to drm
evocallaghan: Your ref to 4: /usr/X11/lib/X11/xserver/amd64/libdrm_intel.so.1'drm_intel_bo_unreference+0x10 [0xfffffd7ffebe3c60] right?
soreau: MNZ: Does /usr/lib/dri/r600_dri.so exist?
rnoland: intel has their own bits... libdrm_intel.so
agd5f: MNZ: rebuild mesa
rnoland: evocallaghan: yes
agd5f: I jsut fixed the issue
MNZ: soreau, yes it exists.
MNZ: agd5f, confirm, it builds now.
evocallaghan: rnoland: ok.
MNZ: so, everything seems to be fine now. except I'm still getting the compiz white screen with AIGLX off :/
chithead: compiz white screen usually points to software rendering
MNZ: chithead, nope, direct rendering is on, and the renderer is Mesa DRI R600
evocallaghan: rnoland: yea ok, that's what I was thinking.
evocallaghan: rnoland: tried to explain that to a Sun eng. :p
evocallaghan: wonder what is happening on freebsd then?
MNZ: 0_0 something weird just happened. I started compiz from the command line to see if I get some sort of error. it worked. it doesn't work from the fusion-icon though. ok so I guess this is not a #radeon issue anymore
evocallaghan: the core file, I can't get a trace from it as it says it can't load/access an address
evocallaghan: I assume its the 'stuff' on the gpu side that is non addreeable in the core its self.
soreau: MNZ: I think it's looking in strange paths.. rebuild latest mesa and ddx then reboot
p_bp: hmm, in freebsd the newest driver is: Port: xf86-video-ati-6.12.4
DanaG: oh, handy resource for xrandr: google for "debian xstrikeforce".
DanaG: how DOES kms deal with multi-card-ness and consoles and keyboard and all that?
MostAwesomeDude: DanaG: It just kind of works.
DanaG: I mean, if you have only one keyboard, does it give each screen a different console, or make all screens have the same console?
MNZ: found the problem. it's not running compiz (a shell script), it's running compiz.real and passing --indirect-rendering to it
soreau: MNZ: You're supposed to check the output of glxinfo for swrast..
MNZ: soreau, ? what?
sixtila: i get a black background with my cairo dock - please help
MNZ: Everything works now. Thank you everyone for all the great help and great job on the drivers!!
sixtila: 01:00.0 VGA compatible controller: ATI Technologies Inc Mobility Radeon HD 3650
airlied: MNZ: dependds on whether kms is enabled by default
airlied: DanaG: kms has nothing to do with keyboards
airlied: it just sets up graphics cards
DanaG: I mean, what does the console do?
DanaG: KMS gives a framebuffer, of course... but how does that deal with multiple cards?
hatseflats: evening everyone
hatseflats: I'm running into a rather weird problem with my hd3850, I think it's related to X or xf86-video-ati, perhaps someone can help me
hatseflats: the problem is that every now and then, for no appearant reason, the output on X will become permanently very 'laggy'
hatseflats: 'low fps', gamers would say, the mouse skips tens of pixels per stroke, that sorta stuff
hatseflats: the only way to fix it is to restart X alltogether, internally restarting the WM isn't helping either
hatseflats: there's no 100% looping process anywhere, there is no real reason the system should start lagging
airlied: DanaG: first card gets the console
airlied: same as current multi-fb set ups
DanaG: ah, so the rest would be blank? Interesting.
airlied: DanaG: pretty much, the kernel console layer isn't that complex
DanaG: http://anandtech.com/weblog/showpost.aspx?i=659
DanaG: no Larrabee retail..
MostAwesomeDude: And nothing of value was lost.
hatseflats: we don't really know
hatseflats: do we?
zhasha: MostAwesomeDude: This conversation intrigues me. Carry on
felipec: anyone else has problems with Xv in R500?
MostAwesomeDude: zhasha: No Larrabees for consumer markets.
zhasha: felipec: what kind of problems?
hatseflats: felipec: I suppose many people may at some point have had problems with Xvideo on the r500, unless you tell us more, we can't give a meaningful answer
zhasha: MostAwesomeDude: you mean we don't get to play with the awesome OpenCL WPA cracking?
felipec: zhasha: the output is very strange when rescaling to certain framesizes... small framesizes are ok, not big
zhasha: strange how? do give a picture
felipec: zhasha: mostly it's like a mesh of squares with weird colors, it depends on the framesize, if I increase it or decrease the number and sizes of squares increases or decreases
zhasha: MostAwesomeDude: on an unrelated note, you now want one of these: www.openpandora.org
MostAwesomeDude: zhasha: I have one of these: http://beaversource.oregonstate.edu/projects/cspfl
MostAwesomeDude: And one of these: http://phones.verizonwireless.com/motorola/droid/
zhasha: nice, almost the same hardware
MostAwesomeDude: Yeah.
zhasha: what're the dimensions on the oswald?
MostAwesomeDude: It's like a first-gen DS.
MostAwesomeDude: We call it the "Gameboy on steroids."
zhasha: so also almost the same size
MostAwesomeDude: Or the Ozzie.
MostAwesomeDude: Yeah, they're all the same size, roughly, *except* the Droid.
MostAwesomeDude: The Droid's *small*.
zhasha: how long does the battery last and what does it cost?
felipec: I have an N900 and a beagleboard :)
zhasha: is droid also Cortex A8 based?
DanaG: Do those things have open-source video drivers?
zhasha: Basically I like the Pandora over the competitors for it's gaming controls and ridiculously hi-res screen
DanaG: wonders if anyone makes an ARM board with a PCI slot.
zhasha: DanaG: SGX has open kernel modules, closed 3D drivers
MostAwesomeDude: Droid's Cortex A9.
felipec: zhasha: yeah, all OMAP3
MostAwesomeDude: Stupid SGX>
zhasha: but the SGX 3D driver is pretty damn good
zhasha: doesn't even need X to run
felipec: MostAwesomeDude: A9 isn't out yet
DanaG: Display size: 3.69" × 2.21" (9.37cm × 5.62cm) = 216.97 PPI, 0.1171mm dot pitch
DanaG: hmm, what DOES it run, then?
DanaG: how do you run X apps on such a beast?
zhasha: DanaG: start X?
MostAwesomeDude: felipec: Hm, sure? I must remember wrong.
zhasha: all this OMAP3 hype is pretty damn awesome
felipec: Android doesn't even run X does it?
zhasha: devices coming out right and left with hardware superior to my laptop
MostAwesomeDude: Yeah, Android's not X-based.
Travis1: so I've installed Karmic 64bit now
Travis1: should I install the updates and the proprietary drivers now?
Travis1: or does someone want to take a look at my Xorg.0.log first?
felipec: the N900 is also OMAP3430, and uses X, but I think the user-space driver is closed (SGX)
hatseflats: ~
MostAwesomeDude: felipec: You're correct.
MostAwesomeDude: Goddamn SGX. :C
zhasha: MostAwesomeDude: but the driver is pretty damn good, so why are you complaining?
MostAwesomeDude: zhasha: Because SGX is the same core that's in Poulsbo.
zhasha: didn't intel recently release a proprietary poulsbo driver with gallium at its core?
Travis1: so should I install the proprietary graphics driver, or should I go through the online community guide for my radeon hd video card
Travis1: at the moment it's running in safe graphics mode
zhasha: Travis1: just install the proprietary driver through the restricted drivers thingy
zhasha: this is not a help channel for the prop. driver
chithead: iirc, n900 has sgx 530, poulsbo has sgx 535
MostAwesomeDude: zhasha: Yeah.
zhasha: MostAwesomeDude: and nobody has made a GPL claim?
MostAwesomeDude: I'm assured that the SGX 530 and 535 are only marginally different.
MostAwesomeDude: zhasha: Gallium's BSD/MIT, and SGX are bastards.
Travis1: what you say the restricted drivers thingy... is that just through the System - Administration - Hardware Drivers? Or do you mean something else zhasha?
zhasha: what? I distinctly remember someone talking about how it was GPL'd and all components were forced to go GPL as well (old stuff)
zhasha: Travis1: that might well be it
chithead: gpl mostly affects kernel code
Travis1: ok so when I did that on my 32bit system, I got a blank screen when X was loading..
Travis1: I'll let you know if this works any better...
zhasha: I think it was zackr in that talk...
Lebenskuenstler: agd5f: I tried both patches. If I did not something wrong during building the custom kernel, both have no effect. I can use only resolution up to 1680x1050
felipec: zhasha: ah, I found that the glitch happens when I resize to 2x or more
felipec: it was working fine before, so I guess I'll have to bisect *sigh*
Travis1: is it best to install the open source radeon hd drivers?
Travis1: like, is it best for me to just go through the guide here: https://help.ubuntu.com/community/RadeonHD
felipec: Travis1: open ati > radeonhd
DanaG: wishes radeonhd would go away.
DanaG: And leave radeon alone.
chithead: sometimes, radeonhd works where radeon fails and vice versa. but in general you should stay with what your distro provides you unless you have reason to switch
DanaG: s/alone/the winner/
felipec: do people still use radeonhd? why?
chithead: I have one 4350 whose dvi output does not work properly with radeon, only with radeonhd
Travis1: so Ubuntu is recomending that I activate my ATI/AMD proprietary FGLRX graphics driver. Are you guys saying I should just do that?
Travis1: or should I use EnvyNG?
MostAwesomeDude: Travis1: Don't use Envy, also go to #ati if you're using fglrx.
spstarr: interesting
spstarr: w KMS off DRI1 secondlife works faster, I can even use ultimate mode, but I also did tweak with driconf a little
spstarr: I just cannot use shaders yet
spstarr: Use hardware TCL as first TCL pipeline stage (the r6xx has no TCL? anyway?)
spstarr: method to limit rendering latency -> "Bust wait for the graphics hardware"
spstarr: oh it does have TCL
spstarr: but we should never use software pipe if we can use hw
Travis1: so I just installed the proprietary drivers. I got this result: http://pastebin.com/m67526c7c
soreau: Read the /topic
Travis1: you talking to me soreau?
MostAwesomeDude: Travis1: #ati
zhasha: MostAwesomeDude: how goes the r300g project?
MostAwesomeDude: zhasha: Haven't been working on it as much as I like.
MostAwesomeDude: I need to figure out the draw_buffers stuff and push it.
zhasha: what's wrong with draw_buffers
MostAwesomeDude: I had a scissors bug, but MrCooper discovered and pushed it first.
MostAwesomeDude: ARB_draw_buffers doesn't work right.
zhasha: is that multiple render targets?
Travis1: is there a guide for me to install the open-source ati driver?
zhasha: Travis1: ubuntu? it comes with it stock
Travis1: yeah running ubunty.
Travis1: *ubuntu
zhasha: disable fglrx and you're golden
Travis1: nothing is working stock. I'm running in safe graphics mode at the moment because nothing has installed as it should
zhasha: MostAwesomeDude: that's multiple render targets as in render the same thing to 4 different buffers right?
zhasha: Travis1: disable fglrx, show Xorg.log
zhasha: we can't help you with fglrx
Travis1: yeah sure. Do I disable it through the Hardware Drivers?
MostAwesomeDude: zhasha: Yeah.
Travis1: ok so I removed the driver. Do I need to restart to show you the Xorg.log?
soreau: yes
zhasha: MostAwesomeDude: what's the limit on renderbuffers on r500?
eosie: I've put together hopefully complete to-do list for r300g: http://paste.pocoo.org/show/155154/
zhasha: eosie: have I ever expressed how much I love you?
eosie: zhasha: do you have big tits? ;)
MostAwesomeDude: zhasha: Same as r300: 4 MRTs.
zhasha: sadly, I don't have tits at all
MostAwesomeDude: XD
eosie: better grow some ;)
eosie: anyway, I forgot to add NPOTs, which is rather st-related than driver-related
zhasha: are you sure it's st related? it seems more like something we should compensate for
zhasha: also, don't we already have accelerated clear?
MostAwesomeDude: eosie: I'm scribbling on that list right now.
eosie: concerning NPOTs, we can accelerate these cases: repeat and mirroted repeat, generate mipmaps, clamped lod to only one mipmap (with any mip filter); we cannot accelerate mip filters
eosie: zhasha: no, I think util_clear is mapping the renderbuffer and uses memset or a for cycle
eosie: *maps the renderbuffer
MostAwesomeDude: http://paste.pocoo.org/show/155157/
MostAwesomeDude: Oops, http://paste.pocoo.org/show/155158/
MostAwesomeDude: util_clear calls util_surface_fill, which is unaccelerated.
MostAwesomeDude: That's why OA sucks balls./
MostAwesomeDude: I'm glad that only one or two of these are "fix hardlock". :3
MostAwesomeDude: At any rate, should I switch video cards? We might be able to nail one or two of these to the wall.
eosie: zhasha: I working on accelerated clear (MAD: yes I am gonna use mostly things from utils except the draw call, which will be one maximized point), I also have a patch for point sprites but they doesn't work and I don't know why
MostAwesomeDude: eosie: Doing a quad instead of a point isn't a big deal.
MostAwesomeDude: A point saves a couple dwords of verts, but costs a couple dwords of ga state.
MostAwesomeDude: Points are internally converted to quads.
MostAwesomeDude: So there's no real speed advantage; we did it that way because that's what fglrx did.
Dr_Jakob: MostAwesomeDude: about edge flags, bug keithw I have no idea about them.
MostAwesomeDude: Dr_Jakob: Will do.
airlied: eosie: I can't read your TODO
zhasha: MostAwesomeDude: points are converted to quads in fglrx?
airlied: in hw
MostAwesomeDude: zhasha: In HW.
MostAwesomeDude: airlied: Lemme copy it to our wiki.
eosie: airlied: do you mean the server is not responding?
airlied: eosie: yup
eosie: and I thought I picked the right one
MostAwesomeDude: Pocoo's usually a good server.
MostAwesomeDude: Copying...
[Enrico]: ehi guys phoronix Linux Graphics Survey 2009 reports that xf86-video-ati has more share than fglrx this here. Congrats to all :D
[Enrico]: radeon 3117 count vs 2770 for fglrx!
[Enrico]: this year*
chithead: it is questionable wheter the survey results are meaningful, but still
MostAwesomeDude: airlied: http://dri.freedesktop.org/wiki/R300ToDo
MostAwesomeDude: What troubles me about that survey is the large chunk of people still stuck on rhd.
zhasha: MostAwesomeDude: michael larabel will be over this like a buzzard on a dead body
chithead: on the computers I maintain, I also have one 4350 with radeonhd as radeon does not work with it properly
airlied: chithead: bug number/
airlied: ?
airlied: last time you tested?
[Enrico]: chithead: well ok, but it is still an hint at least :D
MostAwesomeDude: zhasha: Yeah, that's life.
[Enrico]: MostAwesomeDude: that's right
chithead: airlied: it uses latest archlinux. I will upgrade to 2.6.32 with kms and report a bug if the issue persists
[Enrico]: MostAwesomeDude: i think a lot of people thinks that only radeonhd supports hd cards
[Enrico]: damned novell :P
zhasha: radeonhd needs to give all the neat features to the kernel, then quickly die
[Enrico]: chithead: well i can say that with archlinux zen-kernel works a lot better then the arch one
[Enrico]: chithead: with radeon i mean
[Enrico]: zhasha: well it should be merged at least
[Enrico]: i don't see the point of radeonhd since AMD officialy support radeon
[Enrico]: supports*
eosie: MostAwesomeDude: now it's MAD's todo?
MostAwesomeDude: eosie: I'm cleaning it up right now.
zhasha: [????] Workaround rendering limit of hardware (r300 on ATI 9600 => maximum width 2650): https://bugs.freedesktop.org/show_bug.cgi?id=8357
zhasha: is that even remotely possible?
airlied: zhasha: in theory but messy
airlied: kinda needs to be done at the GL level
airlied: in theory two passes with all the geometry and some crazy viewport hacks
MostAwesomeDude: I still advocate just buying a second card.
MostAwesomeDude: Especially since multicard's rock-solid these days.
eosie: MostAwesomeDude: weird colors - not sure what's causes this; "Bad CS" - swrast passes these, so not S3TC-related; clipFlat - again, it's not clipping; centroid and flat GLSL modifiers - it's rs-block-related
eosie: MostAwesomeDude: back-face output colors in VS - they are standard output colors in VS, rs-block-related, there are already XXXs in all places where the code should added; depth writes in FS - yes, r300c passes this
eosie: that's all from me
MostAwesomeDude: Whoo, finished.
MostAwesomeDude: eosie: Five bucks says swrast passes S3TC stuff too.
DanaG: accidentally marked both systems on one instance of the survey.
MostAwesomeDude: S3TC textures don't pass the kernel checker. It's a kernel bug I need to take a closer look at.
DanaG: both the nouveau one and the sometimes-radeon-sometimes-fglrx one.
Travis1: ok I removed my fglrx divers... does someone want to take a look at the Xorg.0.log file now?
eosie: MostAwesomeDude: softpipe passes texCombine, so it's not S3TC related
eosie: there isn't support for S3TC in softpipe
MostAwesomeDude: eosie: Oh?
Travis1: http://pastebin.com/m506aec35
MostAwesomeDude: Alright.
zhasha: MostAwesomeDude: what's the max CS length?
zhasha: Travis1: that Xorg.0.log is fairly useless. There's no info about falling back, it's just a clean start of VESA
eosie: MostAwesomeDude: BTW, gl_FragCoord=fragment.position=WPOS
eosie: MostAwesomeDude: Fog already works
MostAwesomeDude: eosie: Oh, I was thinking of FogCoord.
MostAwesomeDude: zhasha: 64 kibis.
Travis1: well I got told to remove my fglrx drivers... so I did that.
zhasha: which in dwords would be 16384?
Travis1: happy to follow whatever steps you guys recommend will produce the appropriate Xorg.0.log to give you more info
MostAwesomeDude: zhasha: IIRC yes.
MostAwesomeDude: 64 * 1024 / 4.
chainsawbike: Travis1, do you have the radeon driver installed?
[Enrico]: Travis1: Xorg.0.log is produced automatically when Xorg is started
zhasha: MostAwesomeDude: and every packet can have a max of 16384 dwords as well, right?
soreau: Travis1: You just have to pastebin /var/log/Xorg.0.log file
Travis1: yeah I did that
Travis1: that was the: http://pastebin.com/m506aec35 file
chainsawbike: Travis1, it says its using the vesa driver - do you have the radeon driver installed?
MostAwesomeDude: zhasha: I think so, yes.
[Enrico]: Travis1: since you are using xorg.conf check that the radeon driver is selectect (or no driver is selected then radeon will be used by default)
Travis1: " Driver "vesa"
soreau: Travis1: Delete or comment out that line
[Enrico]: Travis1: well changing it to radeon or comment the line is worth :D
soreau: Then restart X
MostAwesomeDude: "Give the name of a [popular] GPU manufacturer?"
MostAwesomeDude: puts S3
flyback: trident cyberFAIL
flyback: *CANUCKED*
flyback: *CANUCKED*
flyback: *CANUCKED*
flyback: *CANUCKED*
spstarr: agd5f: ping
spstarr: agd5f: testing your changes to DRI right now (building)
spstarr: agd5f: you are aware of the stalling with irqs used?
flyback: although
flyback: suprisingly
flyback: it was very smooth as a thin client for rdp session
flyback: I was impressed
[Enrico]: MostAwesomeDude: rofl
DanaG: spiffy: http://www.amazon.com/Compaq-Presario-2701US-Laptop-Pentium/dp/B00005QERW
DanaG: Mobility Radeon M6 LY
DanaG: And two CD/DVD drive bays.
MostAwesomeDude: switches GPUs
spstarr: MostAwesomeDude
spstarr: agd5f: (--) RADEON(0): Chipset: "ATI Mobility Radeon HD 3650" (ChipID = 0x9591) is not right
spstarr: agd5f: The FireGL is a 3650 but its a more powerful GPU in terms of features for OpenGL?
spstarr: thats what lenovo says
spstarr: http://forum.lenovo.com/t5/W-Series-ThinkPad-Laptops/Help-W500-graphics-card-question/m-p/68973#M1356
spstarr: so im confused
spstarr: "Although based on the same GPU, the Mobility FireGL V5700 can use extended workstation features and logic present in the M86 GPU that are not available to the consumer level ATI Mobility Radeon HD3650."
spstarr: is this correct? If so, shouldn't the DDX be reporting ATI Mobility FireGL V5700 ?
spstarr: bridgman ^^^^^^^^^^^
eosie: I've heard that FireGL is fake, the GPUs are identical, but drivers (read fglrx) might be more optimized for CAD/CAM applications
DanaG: I'd agree with that.
DanaG: As it is, FireGL is about the same as Radeon when in Win7.
DanaG: ... and using apps like DWG TrueView (read-only AutoCAD 2008).
spstarr: eosie: not what Lenovo says
spstarr: I'd hate to have 3D without the full extentions if they exist, but for now, we just need to get things working
DanaG: On my laptop, all it takes is a name change: "HP EliteBook 8530w" <=> "HP EliteBook 8530p".
spstarr: I'll see what bridgman says
spstarr: in the meantime I am checking the bios for any updates to report different PCI ID
DanaG: Former has FireGL (1002:9591, subsys 3604103c), latter changes subys to 30e7103c.
spstarr: 01:00.0 0300: 1002:9591
spstarr: subsys is
eosie: spstarr: Lenovo says whatever it wants you to believe
spstarr: Subsystem: Lenovo Device 2127
DanaG: 01:00.0 VGA compatible controller [0300]: ATI Technologies Inc Mobility Radeon HD 3650 [1002:9591]
DanaG: Subsystem: Hewlett-Packard Company Device [103c:3604]
MostAwesomeDude: spstarr: FireGL chipsets have more pixel pipes. They go faster.
MostAwesomeDude: That's all.
spstarr: MostAwesomeDude: we turn those on?
spstarr: or not yet
MostAwesomeDude: spstarr: Yeah.
spstarr: :)
DanaG: Hmm, since I can change mine back and forth, I'll check it out some time.
DanaG: With fglrx, firegl == radeon.
spstarr: i have the newest bios
spstarr: System BIOS version 1.18 and later change the subsystem ID section of the Hardware ID’s for the ATI Mobility FireGL V5700 to 2126 for switchable graphics mode and 2127 for discrete graphics mode.
spstarr: since i have switchable mode off right now only 2127
DanaG: should try diff'ing xorg log from radeon to fglrx.
spstarr: MostAwesomeDude: how are we checking if its a FireGL? by subsys?
MostAwesomeDude: spstarr: We aren't.
spstarr: so then how do you know how many pipes are available?
MostAwesomeDude: We don't change those parts of the engine.
spstarr: is that drm side?
MostAwesomeDude: Nope, that's just not programmed.
spstarr: isn't that contradicting? :)
MostAwesomeDude: It's like asking why a P4 3.2GHz is faster than a P4 2.0GHz.
MostAwesomeDude: It simply goes faster.
spstarr: ok, so you're telling me, the we just dont enable the extra pipes right now
MostAwesomeDude: All pipes are enabled, regardless of how many there are.
DanaG: No, wouldn't it be more like celeron versus non-celeron?
MostAwesomeDude: DanaG: Technically, it's like P4 vs. Prescott.
spstarr: oh there IS a bios update
spstarr: brb
spstarr: i have 3.09
spstarr: there is 3.10
spstarr: boots to windows
DanaG: I should do "gpu caps viewer" with one ID, then change it to the other ID.
DanaG: hmm, but there's no officially certified recent driver for the thing, anyway.
DanaG: Recent as in 9.11.
flyback: goes out for a redbull and a pack of twizzlers
turmlos: airlied: Submitted fdo bug #25469.
spstarr: bios updated
spstarr: bridgman: hello
spstarr: bridgman: is it just pipes that make the FireGL better than the regular mobile?
bridgman: hi spstarr
bridgman: mostly different driver paths, different testing, different clocks etc...
spstarr: I see
bridgman: normally same # pipes but the optimization is for a different class of apps
spstarr: do we take that into account in the DRI drives?
spstarr: drivers?
bridgman: see carmack's quote about stupid driver tricks for quake vs stupid driver tricks for cdrs
bridgman: not really; current mesa is somewhere in between AFAIK
spstarr: oh, ok
DanaG: bridgman: how are you talking, yet not visible on the right side?
spstarr: hes on my irc list
bridgman: I don't show up in photographs either
spstarr: heh
spstarr: come next week, i want to help debug why I cannot use VBOs on the r6xx
spstarr: installs debug builds of mesa trunk
spstarr: bridgman: I've had 3 interviews last week, waiting for this week to find out if im finally out of the unemployment lines
spstarr: had another person call me friday, so monday I will be busy talking to them
bridgman: cool, will cross fingers again
bridgman: it's hard to type so they aren't crossed normally ;)
spstarr: uh oh
spstarr: that DRI change broke textures...
spstarr: reboots
spstarr: yeah
spstarr: mesa regressions
DanaG: Once Zajec gets that power-savings into shape, and into radeon officially, I'll jump to it, and even build my own kernel for it.
spstarr: hmm
spstarr: lots of stalling
spstarr: bridgman: how to debug radeon to determine what the stalls are coming from?
spstarr: i wonder
spstarr: it seems like the radeon is doing an IRQ broadcast storm
spstarr: its causing network to stall
bridgman: only happens when IRQs are enabled ?
spstarr: yes
bridgman: my debug technique, carefully refined over the last couple of years,
bridgman: is to ask agd5f
spstarr: RES: 76247 75498 Rescheduling interrupts
spstarr: if im not mistaken thats bad
spstarr: it keeps rising
bridgman: rising is probably OK, guessing that's offset into the ring buffer
spstarr: even if i throw out secondlife, glxgears shows stallouts also
spstarr: heh
spstarr: it locked up good there
spstarr: now in DRI1
spstarr: catches VBO crash
bridgman: thinks about crashing
dmb: hi bridgman !
bridgman: hi dmb
spstarr: i can crash secondlife with VBOs on but gdb shows no mesa calls
spstarr: only SL code crashing with vertex
spstarr: 0x0000000001eeb492 in LLVertexBuffer::mapBuffer(int) ()
bridgman: I did a quick Google search on llvertex buffer, nearly all the hits were secondlife
bridgman: sorry, secondlife + crash
bridgman: seems to be a problem on windows as well
spstarr: but it only happens if i enable VBOs
hnsr: anyone here getting a very dark picture with ioquake3? as if the gamma or brightness is lowered
spstarr: which means two things: SL is not properly figuring out the VRAM buffer size from xorg log file (investigating)
spstarr: or VBOs is broken in r6xx in reporting buffer info
dmb: bridgman, can I annoy you in #ati about something?
bridgman: sure
bridgman: the fix seemed to be pretty recent, not sure if it's in a sl release yet
bridgman: what version of sl are you running ?
spstarr: Emerald build 1.23.4.904
spstarr: i can get the latest code from snowglobe or their main viewer code
spstarr: bridgman: there is a fix?
bridgman: there seem to be a bunch of different crashes in llvertexbuffer, the early fixes (april) seem to be in 1.2.3
spstarr: 1.21 is old :)
spstarr: hmm
spstarr: trying snowlgobe 1.3.0
spstarr: (trunk from 2 days ago)
spstarr: crashed in snowglobe
spstarr: ERROR: mapBuffer: glMapBuffer returned NULL (no vertex data)
spstarr: hmmmm
spstarr: http://www.spec.org/gwpg/gpc.static/vbo_whitepaper.html
spstarr: orly
bridgman: ok, that's better ;)
spstarr: so is that the DRI being bad?
bridgman: no idea
bridgman: is thinking more about sleeping than about debugging
spstarr: heh
MostAwesomeDude: bridgman: Not missing the good ole days of having to look up dozens of little r500 tidbits for us? :3
bridgman: I am actually, but back then I was *really* missing sleep as well
bridgman: and these days I don't seem to be able to function without it ;(
bridgman: *those* good old days were a *long* time ago ;(
dmb: bridgman, try my 10.5 hours of sleep last night :)
dmb: and I still feel tired
bridgman: you're probably doing it now - work on something for 3 days straight and then wake up in the middle of the living room floor
MostAwesomeDude: Not really.
bridgman: oh good
dmb: its a really bad feeling waking up at a computer desk
MostAwesomeDude: I have occasional insomnia.
bridgman: that can be really rough
dmb: MostAwesomeDude, its seems like I have insomnia until I actually get to sleep, and once I get to sleep I never wake up
MostAwesomeDude: Yeah, I'm trying to pace myself.
bridgman: my problem is that I consistently get sleepy around 6 AM
bridgman: but sooner is hit and miss
dmb: is that when you are just waking up, or just going to sleep?
bridgman: that's when I can reliably go to sleep
bridgman: also roughly the time I have to get up
bridgman: either one on its own is fine, but together not so fine ;)
bridgman: anyways, feeling vaguely sleepy now so going to run off an try to catch it before it goes away
bridgman: ttyl
spstarr: *********************************WARN_ONCE*********************************
spstarr: File radeon_dma.c function radeonReleaseDmaRegions line 343
spstarr: Leaking dma buffer object!
spstarr: hmm
spstarr: ew