Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-6-30

Search This Log:

EruditeHermit: is it possible to just use the softpipe gallium driver at this stage?
airlied: is tempted to just write PreInitKMS and ScreenInitKMS completely separate to the current paths
nanonyme: airlied: Would there be something bad with that design decision? :)
airlied: nanonyme: well people would have to remember to fix both paths.
airlied: granted I'm not sure how much fixing in both paths would be necessary
airlied: the current scattering if (kms) and if (!kms) all over the place is fugly
nanonyme: Right.
nanonyme: has no opinion on the matter
nanonyme: Obviously.
adamk_: Anyone know if the ES1000 supports direct rendering?
nanonyme: adamk_: You know who probably would. :o
adamk_: You?
chithead: even the radeon manpage mentions the ES1000
nanonyme: Naw, I'm just mostly a generic end-user.
nanonyme: chithead: Wouldn't direct rendering be on Mesa side though?
chithead: mesa is involved too
nanonyme: adamk_: I was mostly thinking of bridgman since it's possible the card wasn't meant for 3D rendering in the first place and I've no idea about the hardware details. :)
chithead: es1000 is an r100 card
nanonyme: Oh, right.
nanonyme: So it's not even technically on the business chip branch?
nanonyme: *shrug* Carry on. :)
nanonyme: goes back to having fever
adamk_: It's designed for servers, apparently.
adamk_: It's definitely supported in the 2D driver, because I'm looking at a log file with it detected and supported.
adamk_: DRI is not enabled, but I'm assuming that is normal.
chithead: I think dri should work with that card
chithead: iirc it was disabled in older xf86-video-ati due to crashing bugs, but that was resolved in recent releases
wiwar: Hello, how to switch off "Mirror Screens" option in xorg.conf ?
wiwar: I found it in "System"->"Preferences"->"Screen Resolution"
jan__: ping airlied or agd5f
jan__: i have an modesetproblem with my laptop, it has an x700 mobility and an 1280x800 panel, running modeset produces an whitescreen on startup
jan__: i put logs with drm.debug=1 at http://pastebin.com/m4783b50f
jan__: running fedoras kernel, however running linus lattest git still has the problem
jan__: if you or any other has time it would be nice to get this fixed. thank you
glisse: jan__: open a bug please
glisse: attach log for lastest linux kernel
jan__: glisse: ok will do so, will take some time to compile. thank you
rnoland_: agd5f: where is the latest r600 drm code?
agd5f: rnoland_: http://cgit.freedesktop.org/~agd5f/drm/?h=r6xx-r7xx-3d
agd5f: latest stuff for 3d
rnoland_: agd5f: ok, that is what i was afraid of....
agd5f: although the official tree is linus
rnoland_: it is still based on drm git... and my code is based on drm-next
rnoland_: so merging is painful....
agd5f: rnoland_: don't worry about that. the only difference is one ioctl
agd5f: once 3d is working better, I'll merge the new ioctls into drm-next
agd5f: the only file you have to worry about is radeon_cs.c
rnoland_: agd5f: so just import radeon_cs.c and merge the ioctl?
agd5f: yeah
rnoland_: ok, cool
Zajec2: agd5f: "once 3d is working better" is there anything working already?
Zajec2: anything like... glxgears or sth?
Zajec2: did I miss sth? :)
agd5f: Zajec2: Richard and Cooper sorted out the major issues with r6xx-rewrite last week
jan__: glisse: bugreport is at http://bugzilla.kernel.org/show_bug.cgi?id=13683
glisse: jan__: thanks adding it to my list
spstarr: agd5f: nice
jan__: glisse: thanks
nike: How can I tell if I am using KMS?
nike: Is there something in the glxinfo output or some other command to absolutely verify?
Weiss: what do you get from "cat /proc/fb"?
nike: Weiss: Nothing.
Weiss: hmm.. if you have KMS, it should say "0 radeondrmfb". pretty sure that's an absolute, but not 100% so
nike: Oh, nuts.
nike: Sorry I just got a new laptop.
nike: My old one had an r500
nike: This one has an intel 4500 MHD I guess.
Weiss: surely you know for sure what GPU you have? :)
_Groo_: hi/2 all
_Groo_: airlied: ping
_Groo_: agd5f: ping
agd5f: _Groo_: pong
_Groo_: agd5f: hi alex :)
_Groo_: agd5f: are you very busy? my poor rs485 could use some dev love
_Groo_: agd5f: mesa master breaks kwin 3d and ddx still is a mess with rs485
agd5f: _Groo_: sorry, I haven't had time to look at that yet
_Groo_: agd5f: the last know good commit to mesa is a18e209edb5348eb167e9d7184597031bbbbe622
_Groo_: agd5f: after that is breaks kwin3d
_Groo_: agd5f: so im basically stuck with old code till you guys have the time to fix rs4xx...
_Groo_: agd5f: do you guys have some eta so i dont have a heart attack waiting? :) just out of curiosity
agd5f: _Groo_: did you file a bug yet?
_Groo_: agd5f: do i have to? both you and dave already know rs485 is broken... seems kinda pointless..
MostAwesomeDude: _Groo_: File a bug please.
agd5f: _Groo_: yeah, but I can't remember all the issues and it would be nice to consolidate the known facts and what versions of things you are using
agd5f: etc
_Groo_: agd5f: ok, gonna open in half an hour.. should i include all in one bug report (since its broken in ddx, mesa and kms?
_Groo_: ddx = doesnt work, mesa = works but kwin is broken and no fbo, kms = no modesets
agd5f: _Groo_: if the ddx doesn't work, then how are you testing mesa?
_Groo_: agd5f: using an older ddx..
MostAwesomeDude: _Groo_: Are you completely certain that the commit you mentioned is last good? The commits around that area don't affect r300 at all.
agd5f: _Groo_: what's the last working ddx?
_Groo_: MostAwesomeDude: for mesa? i didnt pinpoint the EXACT commit, but after that one its broken, i just reverted till it worked again.. i dont know how to use git bissect yet
_Groo_: agd5f: e1c50d8fe4c620933ce25f730eb996ddcb767717 last ddx that works for me
_Groo_: agd5f: around 16 jun
agd5f: _Groo_: from which tree?
MostAwesomeDude: _Groo_: Bisect, please. There have been several *hundred* commits, and I can't track them down for you.
_Groo_: agd5f: kms-support
_Groo_: MostAwesomeDude: i dont know how! could you point me to a wiki or something?
suokko: _Groo_: git bisect --help
_Groo_: suokko: you must be kidding me ¬¬
MostAwesomeDude: Here, let me Google that for you. :T
MostAwesomeDude: http://www.kernel.org/pub/software/scm/git/docs/git-bisect.html
MostAwesomeDude: http://blogs.atlassian.com/developer/2009/01/git_bisect.html
agd5f: _Groo_: the commits after e1c50d8fe4c620933ce25f730eb996ddcb767717 should be irrelevant for you
suokko: _Groo_: git has very good helps
adamk_: Heck, even the man page for git-bisect is very good.
agd5f: _Groo_: there are only two and one is BE only, and the other only affects Xv
_Groo_: agd5f: well i assure you that kms-support isnt working anymore
agd5f: _Groo_: did you change your drm?
_Groo_: agd5f: i get colored stripes when X starts and everything hangs
_Groo_: agd5f: drm from master, mesa from master
agd5f: drm from master of linus tree?
agd5f: _Groo_: I suspect you aren't building the right trees
_Groo_: agd5f: yep, still crashes, but after my tests this morning i went back to your 2.6.29 tree
_Groo_: agd5f: ok, lets start with that...
agd5f: _Groo_: http://airlied.livejournal.com/66958.html
_Groo_: agd5f: drm = http://cgit.freedesktop.org/mesa/drm
_Groo_: agd5f: master
agd5f: _Groo_: yes, but did you configure with --enable-radeon-experimental-api
agd5f: also, what drm kernel modules are you using?
agd5f: you should be using linus tree now
_Groo_: agd5f: from config.log $ ./configure --enable-udev --prefix=/usr --enable-radeon-experimental-api
nanonyme: didn't know --enable-udev existed
_Groo_: agd5f: linus tree http://www.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
agd5f: ok
_Groo_: agd5f: mesa from master, ddx from kms-radeon branch
agd5f: _Groo_: it's probably a drm kernel module issue. can you pastebin your dmesg?
_Groo_: agd5f: when i activate the broken ddx?
_Groo_: agd5f: or in "normal" operacion?
agd5f: _Groo_: when you load radeon with modeset=1
_Groo_: agd5f: ok, but im using your tree today... let me look for the linus one
_Groo_: agd5f: what should be the correct drm output date? (so i can grep it faster)
agd5f: _Groo_: I don't know off hand
_Groo_: agd5f: arent you using it?
agd5f: _Groo_: not at the moment. working on r6xx/r7xx stuff
_Groo_: agd5f: ah ok..
_Groo_: agd5f: gonna compile the latest kernel from linus tree from today, and update the broken ddx and mesa (i have separate branches)... when im done im gonna reboot and see if it changes anthing...
agd5f: _Groo_: it's probably drm related. so file a bug and include your dmesg from linus tree
_Groo_: agd5f: ok, gonna do the tests and ill open up the bug subsequently
agd5f: _Groo_: thanks
_Groo_: agd5f: btw ist normal not having modesets by default?
_Groo_: agd5f: im the one who should be thankful, you need the patience to put up with me
agd5f: I think that's the default for now
_Groo_: agd5f: can i force it?
agd5f: _Groo_: yeah, radeon.modeset=1
suokko: btw, Is there any common tree where you all push changes before they enter Linus' tree?
_Groo_: agd5f: its a pain because xorg isnt respecting xorg.conf modesets.. only xrandr can change thme on the fly
agd5f: on your kernel command line
suokko: (Just so I don't have apply all the patches from mailing list) :)
_Groo_: agd5f: lol, no sorry, i knopw that and i can turn it on by default in linus kernel
_Groo_: agd5f: i mean mode sets.. you know 1024 x768, etc etc
agd5f: suokko: airlied usually aggregates them through his drm-next branch
_Groo_: agd5f: xrandr only shows me one.. the max for my lvdi
agd5f: _Groo_: the kms code doesn't currently add non-native modes for lvds like the ddx does
agd5f: just hasn't been done yet
suokko: So that means I have to apply the patches manually because no updates in the tree
_Groo_: agd5f: ah ok.
_Groo_: agd5f: but shouldnt at least have the native modes? 1024x748, 800x600? or are those non native also?
_Groo_: 768
agd5f: _Groo_: lvds panels are fixed size. they only have one mode
agd5f: one native mode
_Groo_: agd5f: ahh ok :)
agd5f: everything else uses a scaler
_Groo_: agd5f: now it makes sense :)
_Groo_: agd5f: but shouldnt xorg respect my monitor settings? or is it diferente when using xrandr --add-mode
agd5f: _Groo_: it does. but the mode has to exist before you can use it
_Groo_: agd5f: ahh and since it doesnt, only xrandr can create and use them, since its on the fly
agd5f: yeah. you can also specify modelines in your config. http://wiki.debian.org/XStrikeForce/HowToRandR12
_Groo_: agd5f: thats the problem i told you.. i specify them in xorg.conf in the monitor section, but xrandr/xorg doesnt show them, i can only use them if i call xrandr after X is up
_Groo_: agd5f: by use i mean, xrandr -add-mode and xrandr --use-mode
agd5f: _Groo_: are you linking the monitor section to the output properly and such?
_Groo_: agd5f: yep, i can paste.bin it if you want
agd5f: pastebin your log and config when using kms
_Groo_: agd5f: http://pastebin.ca/1479766
_Groo_: agd5f: and i only use kms/dri2
_Groo_: agd5f: for some time now :)
_Groo_: agd5f: since your original post anyway
rnoland: agd5f: is the _dri.so module from r600_rewrite enough, or do i need to install libGL as well?
agd5f: _Groo_: you haven't associated your monitor section with an output
agd5f: rnoland: _dri.so should be enough
rnoland: agd5f: ok, cool
_Groo_: agd5f: http://pastebin.ca/1479767
_Groo_: agd5f: compiling linus tree (my poor laptop gonna explode today), brb
nanonyme: _Groo_: You probably knew you don't have to actually compile the kernel on the same computer you want to use it on? ;)
nanonyme: (Especially useful for laptops with slow CPU's to compile what you need on another computer)
agd5f: _Groo_: change the Identifier in your monitor section to "LVDS" rather than "Configured Monitor"
suokko: nanonyme: Do you know that people like me don't have alternative computers to the poor laptop? :)
nanonyme: Awww. :)
nanonyme: would personally never compile anything on his laptop if cross-compiling to PPC wouldn't be so troublesome
ajax: ppc? there's your problem, fishbulb.
nanonyme: Well, I don't want to throw away a perfectly working laptop. :)
MostAwesomeDude: ajax: You have revealed yourself!
mjg59: Like many predators, ajax is drawn to snark
MostAwesomeDude: ajax: Any comments on that email I sent?
ajax: you probably sent it to the mailbox i don't check as much...
ajax: (because, well, the machine hosting is a sparc with not a lot of io)
MostAwesomeDude: Hm. I used the one you use on the mailing lists, but maybe I should have used the RH one.
MostAwesomeDude: Resending.
mcgreg: hi
mcgreg: just did a "git pull" for radeon. I see there are a lot of changes , some deletations and a lot of insertions. are there any new features/improvements?
yangman: mcgreg: read the logs ;)
mcgreg: radeon: initial preparation for kms patch. then . ok thx
rnoland: agd5f: hrm... it is still trying to load r300_dri.so....
_moep_: hey i've some trouble with my xorg.conf: if I use vesa as driver all works but if I use radeon(hd) as driver i get an error like no device
ajax: what device do you have?
_moep_: Radeon HD 4350
_moep_: should be RV720 i think
ajax: lspci -n calls it 1002:954f i assume?
_moep_: 01:00.0 0300: 1002:954f
ajax: then your build of radeon is just old
ajax: try 6.12.2
_moep_: i used the squezze package, to old, too?
ajax: i don't know what that is?
_moep_: debian squeeze
ajax: oh is that their new name
ajax: if it's 6.12 or newer then there's probably something else going wrong, and you should pastebin your X log from failing to start with radeon
_moep_: hm nope its 6.9 ok lets get the git :)
_moep_: git clone git://anongit.freedesktop.org/git/xorg/driver/xf86-video-ati <- thats the right isnt it?
ajax: yep
agd5f: rnoland: you need an updated ddx. either 6.12-branch or master
rnoland: agd5f: ah, ok... will do.
agd5f: rnoland: 2888dd9ae9689b1cd72115dc0ceea1f5957299b0 for refernece
_moep_: http://mypaste.ja-s.de/1882 <- whats missing
rnoland: i'll just build master
agd5f: _moep_: needs a quick fix or a newer x server for that. hold on fix on the way
_moep_: ok
agd5f: _moep_: should be fixed now
_moep_: agd5f: seems good compiling works
_moep_: but then only black screen :D
agd5f: _moep_: pastebin your xorg log
_moep_: wait...
_moep_: agd5f: http://moep.cmkr.nl/pc/Xorg.0.log
agd5f: _moep_: still using vesa
_moep_: oh wait then its the wrong log
_moep_: i switched too vesa again for chating but let me eat brb in few min :)
_moep_: agd5f: try again
_moep_: updatet logfile
agd5f: _moep_: log looks fine. screen is blank?
_moep_: yes only here the "beep"
_moep_: (EE) AIGLX error: dlopen of /usr/lib/dri/r600_dri.so failed (/usr/lib/dri/r600_dri.so: cannot open shared object file: No such file or directory) but i think thats not rly good
agd5f: doesn't matter
_moep_: ah ok
agd5f: does your monitor have a VGA cable as well? if so does connecting that work?
_moep_: i have VGA but i've no connector for VGA
agd5f: _moep_: what connectors does your have? log indicates DVI-I and VGA is that correct?
_moep_: yes
agd5f: _moep_: file a bug and attach your xorg log and config
agd5f: https://bugs.freedesktop.org
_moep_: but i've no VGA->DVI adaptor for my tft ;)
_moep_: ok
agd5f: ah, your monitor is DVI only
_moep_: yes :)
nike: Where can I ask general X.org questions... like how to use xrandr to have two monitors not cloned, but one left of another.
MostAwesomeDude: nike: http://wiki.debian.org/XStrikeForce/HowToRandR12
_moep_: agd5f: is my problem a xorg bug?
agd5f: _moep_: probably a driver bug
_moep_: ok
nike: MostAwesomeDude: Thanks, but... "xrandr: screen cannot be larger than 1920x1920"
nike: Is that a hardware limitation?
agd5f: nike: you need to add a virtual line to your config
MostAwesomeDude: nike: Virtual line
agd5f: nike: see section II.5.
nike: Section IV.1. says: "If you enable a larger virtual screen, DRI will be disabled and some problems may appear. "
nike: VI.1 excuse me
agd5f: nike: depends on the hw
nike: agd5f: Can I not specify the virtual screen on the command line? my version of X.org does not even ship with a xorg.conf
agd5f: nike: has to be in the xorg.conf. IIRC the gui tool for screen config will create one if you try to enable dualhead
suokko1: nike: sudo dpkg-reconfigure -phigh xserver-xorg
suokko1: that one works also in deb based distros
nike: So adding a virtual screen size of 3360 1080 should do it? My internal LVDS is 1440 x 900 and my external LCD is 1920 x 1080
nike: By setting a virtual screen that adds the horizontal together, the hardware limitations can be bypassed via this virtual screen setting?
nike: (and I'm guessing I have to restart X11)
agd5f: nike: that should do it. the virtual line is bypassing software limitations
agd5f: if it were hardware limitations, it wouldn't work :)
_moep_: agd5f: wow i disabled the DRI settings in the config (cause if xv device) and now i see twinkle
_moep_: and now black again :D
_moep_: agd5f: bugreport is open
nike: agd5f: Does AMD still employ you?
agd5f: nike: yes
nike: rms is curious is AMD is helping the radeon driver
nike: Who else are they employing?
agd5f: we have two other folks, Richard and Cooper
agd5f: working on open source stuff
stuckey: Hello
stuckey: I use a x1300 with the radeon driver; my mesa version is 7.4. I opened a game and the stuff that was supposed to be coloured was just grey and it didn't have any texture to it.
nanonyme: Whoa, steady there. ;) Which game was this?
stuckey: Springta
stuckey: springrts.com/
stuckey: nanonyme: springrts.com/ --- TA-SPring
DanaG: http://mjg59.livejournal.com/111853.html
DanaG: interesting.
_Groo_: hi/2 all
_Groo_: agd5f: ping
_Groo_: airlied: ping
_Groo_: any devs alive
_Groo_: just did a batch of tests with latest linus tree, drm, mesa and ddx...
_Groo_: with an evil rs485
_Groo_: like before, mesa is broken for kwin3d and ddx is broken for rs485 (stripped colors and X crash)
redondos: hi. is there any free driver that'll give me 3d on an r600?
redondos: am I better off with radeon or radeonhd?
sgcb: redondos: you can try this guide: http://www.phoronix.com/forums/showthread.php?t=17603
redondos: thanks, sgcb
debio264: is it a known problem that on an ATI XPress 200M, OpenGL canvases are always drawn above all other windows and are not subject to compositing?
debio264: this is using the Radeon driver, of course, not the ATI proprietary driver
airlied: debio264: yes,
ssieb: debio264: that means you're not using KMS, right?
ssieb: ok, I tried my other card now. it's an RV515 / X1300 and it seems to perform reasonably well as long as I don't use KMS. other than the GL windows being on top of everything :-)
debio264: ssieb: KMS?
debio264: ssieb: Looked it up, no, I'm not using it
debio264: does it help anything?
debio264: gah, I gtg sleep
debio264: I'll be back tomorrow
dawid: hi there, im googling around but i cant find the reason why my radeon is so slow at 2d and 3d