Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2008-11-14

Search This Log:

spstarr: let's see if this works
spstarr: got it
spstarr: let's see if this video looks right and shows the corruption
spstarr: encoding
spstarr: airlied: i think i found something strange
spstarr: i will reencode you can see the corruption
spstarr: seems at 1400x1050 more vram is needed (given) at 800x600 its speedy fast
Pyr0Seraph: alright, I give, I've been at this for hours... please please please tell me someone can help me get the radeon drivers working..?
z3ro: Pyr0Seraph: well, you haven't told us the problem yet.
Pyr0Seraph: arg.. sorry about that. I truthfully don't know what the problem is really. I can't compile the x.org driver to even try it. and it's showing with direct -i and OpenGL -i that it's using mesa instead of ati.
z3ro: Pyr0Seraph: I guess you might be better off using a package from your distribution, instead of trying to compile it yourself.
Pyr0Seraph: I'm just trying to figure out something that'll work for it, preferably that'll do OpenGL. I've tried with how Fedora starts, I've tried the xorg driver, and the ati proprietary
Pyr0Seraph: when I try to do ./configure for the xorg driver it says that xorg server isn't here, but I just made sure that it's installed
z3ro: you might need to specify the path to the X server header files or something... I've never had that problem before.
Pyr0Seraph: alright, I'll keep working on it, thank you for trying
spstarr: airlied: http://www.sh0n.net/spstarr/corrupt-menus.avi
MrCooper: spstarr: is that using DRI2 direct rendering?
spstarr: MrCooper: DRI 1
spstarr: w/ direct rendering + composite is on
spstarr: MrCooper: this happens with xchat also when its offscreen and restored
MrCooper: well, it can't be direct rendering without DRI2
spstarr: garbled pixmap
spstarr: direct rendering: Yes
spstarr: or i that different for DRI 1 (glxinfo)
MrCooper: GLX_EXT_texture_from_pixmap only works with indirect rendering without DRI2
spstarr: ok so its indirect rendering
spstarr: what would be causing that?
MrCooper: no idea
spstarr: composite is holding stable so far with the newest drm bits in kernel
spstarr: i even switched resolutions with composite and it didn't break
spstarr: MrCooper: ive seen that corruption for 1 year +
spstarr: that sort of random corrupt pixmaps
spstarr: same for restore/minimizing apps
spstarr: corrupt initial display
MrCooper: 'initial' until what happens?
spstarr: got it.. another video coming
spstarr: you can see backbuffer corrupted onto the restored window
spstarr: ugh i need a lower resolution sec
Rabenklaue: hi, I just found on http://xorg.freedesktop.org/archive/X11R7.0/doc/html/radeon.4.html that R300 and R400 series cards aren't supported by the free radeon driver in xorg concerning the 3D acceleration
Rabenklaue: could anyone affirm that this means, that also the RS400 (my card is an 200m) series isn't fully supported?
MrCooper: Rabenklaue: all of those are supported
MostAwesomeDude: Rabenklaue: That's an old page, hence archival.
MostAwesomeDude: 3D acceleration is available for R300, R400, and R500 chipsets.
Rabenklaue: ah, that's great
Rabenklaue: thanx
Rabenklaue: than I'm able to check out compiz-fusion here on my system without fglrx.
Rabenklaue: cool
hifi: MostAwesomeDude: though you forgot to mention R500 acceleration is not released yet
MostAwesomeDude: hifi: Sure it is.
MostAwesomeDude: You mean R600 acceleration.
hifi: uh, 6.9.0?
MostAwesomeDude: Yep.
hifi: so whats with the unknown chip id?
hifi: mesa issue?
MostAwesomeDude: Probably, yes.
MostAwesomeDude: I didn't know you were having an unknown chip id issue. Which is your id?
hifi: 0x7146
MostAwesomeDude: Yep, it's present for both DRM and Mesa.
MostAwesomeDude: Looks like it was added before the 7.2 release, probably in the big batch.
MostAwesomeDude: Are you sure your Mesa is 7.1 final or newer?
hifi: 7.0.3-6, nope
MostAwesomeDude: Yeah, 7.0.x doesn't have any R500 stuff.
hifi: hmm, maybe I could get the 7.2 packages from experimental branch (debian)
MostAwesomeDude: That's what I did. Pinned experimental and then installed xorg-server from experimental. It'll pull in Mesa 7.2. Works great.
MostAwesomeDude: Just don't forget to do the pin with "aptitude -t experimental install xorg-server" rather than "aptitude install xorg-server/experimental".
hifi: could you pastebin your pin file?
MrCooper: you don't really need one
MrCooper: just something like one of the aptitude lines above
hifi: but I'd still need to source the experimental branch, what happens to other packages when I upgrade?
hifi: or rather dist-upgrade
MostAwesomeDude: hifi: Just don't source experimental except when you need packages from it.
hifi: hehe, I'll try
MostAwesomeDude: And when experimental packages migrate down, aptitude will sort them by version and do the Right Thing.
hifi: Couldn't find any package whose name or description matched "xorg-server"
MostAwesomeDude: xorg-xserver, maybe?
hifi: nope :)
hifi: theres a xserver-common package
hifi: http://rafb.net/p/uU9wLF13.html does that look ok?
spstarr: MrCooper: the same for the menus, when they are initially displayed you get garbled pixmap or some old buffer memory, then it repaints it properly.
spstarr: MrCooper: same for xchat window content it looks like that, but then is repainted a few 1-2 seconds after
MrCooper: MostAwesomeDude, hifi: the default APT pin priority of experimental is so low that packages from it are never automatically selected
hifi: hmm, rebootin X
MostAwesomeDude: MrCooper: Yeah, I know. I have to pin *up* for experimental and *down* for unstable to get the stuff I need.
MrCooper: not really, just select it explicitly in aptitude
MrCooper: it'll still get upgraded from experimental thereafter
hifi: hmm, Xorg booted with minor xorg.conf tweaks
hifi: haz accelerated gears \o/
hifi: MostAwesomeDude: thanks
MostAwesomeDude: hifi: NP.
hifi: cool, open source accelerated openarena
hifi: btw. how mature the R500 acceleration is today? should it be considered stable?
MostAwesomeDude: It's stable although there's a lot of deficiencies.
MostAwesomeDude: Notably only OGL 1.3 support.
hifi: feature matrix anyone? it'd be interesting to follow :)
hifi: well, more complete one
hifi: ok, it had more stuff in it than I remembered
MostAwesomeDude: Well, there's not much interesting, really.
MostAwesomeDude: OGL 2.0 support is dependent on some very big things.
MostAwesomeDude: And OGL 1.4 and 1.5 have some very specific dependencies which are pretty focused.
hifi: damn it, I need a AGP R500 to play with
spstarr_home: MrCooper: i got the video
spstarr_home: MrCooper: http://www.sh0n.net/spstarr/corrupt-restore.avi
spstarr_home: the corruption in this video is when i view firefox after its not visible for a while
spstarr_home: but its the same thing as the other video basically.. the pixmap buffer is bleeding from the active pixmaps in use
spstarr_home: you see that black garble which we know is coming from the gnome-terminal
spstarr_home: sleep.. will see any responses with spstarr_work
Kano: hi agd5f , airlied
Kano: how to patch xrandr extension that it does not flicker every time?
Kano: i got a new tft (currently connected via vga port due to kvm) and it still flickers, just like a crt
Kano: whenever xrandr is used explicitely or implicitely by kde4
Kano: every 10s
jcristau: fix kde then. it shouldn't do that
Kano: well i want a working driver, the old one from xorg 7.1 does not have that issue, i would only need a dcc patch for this...
Kano: and my rv410, is it possible to backport it?
marcheu: yeah fix kde... if they keep probing xrandr that's bound to happen
marcheu: it blinks because you can't probe and output stuff at the same time
marcheu: so fix kde, since that's going to be a problem on other cards also
Kano: marcheu: i am no kde dev
Kano: btw. fglrx nor nvidia has that issue
marcheu: well complain to kde people instead of radeon people
Kano: it is specific to radeon driver
marcheu: neither does xrandr
marcheu: no it's not, it'll happen on intel too
Kano: well when i remove my nv card i could test intel too if needed
Kano: but how about giving me a patch to disable what i dont want?
marcheu: ask kde
Kano: i want a radeon driver patch, and here is the channel for it...
marcheu: the problem is in kde, and you want a radeon patch ? sounds weird...
Kano: changeing a driver is simpler
Kano: why not adding a detect option to xrandr?
Kano: btw. did you notice some bad lines when you use xrandr -o 2
Kano: 1/3 seem to be completely broken
Commie: hi
Commie: can i get some help with ati drivers?
adamk: Commie, If it has to do with the open source drivers, just ask your question.
Commie: i dont think it is about open source drivers
Commie: i get error while installing drivers for ATI
kristian_: since yesterday my ubuntu box freezes when entering X, I'm using a git version of the driver compiled myself that hasn't been touched since in a long time. anyone that has some good suggestions how to find out what's causing the lock ups?
spstarr_work: airlied: did you get a chance to see those videos?
airlied: spstarr_work: not yet... its the weekend.
spstarr_work: ok
oga: do dri clients actually need the mapping of the mmio registers?
spstarr: hmmm
spstarr: where's the glow changes?
spstarr_home: airlied: the high CPU is coming from EXA use
spstarr_home: im back to XAA + composite and X is using about 10-20% cpu instead of 50
spstarr_home: though XAA w/o offscreen pixmaps will eat CPU
spstarr: wedge
spstarr: ugh
spstarr: XAA + no composite, this should hold stable
spstarr: interesting
spstarr: airlied: other people triggering race
spstarr: Bug #18465 etc
spstarr: looking at the fixes coming and seeing how
spstarr: KDE 4 has a workaround now
spstarr: Make sure the GLX context is destroyed before the window it's been made current with. Even though it should be allowed to do it afterwards it seems this can crash AIGLX
spstarr: disables AGP and tries composite
ssieb: I've been noticing an issue with fast user switching and gl performance
ssieb: the initial logged in user has full gl performance, but if you switch users, the second user seems to be at software rendering performance, even though glxinfo still says direct rendering is on
oga: ssieb: swrast reports it's direct rendering
ssieb: ah, so how can you tell then?
ssieb: is this a known bug?
ssieb: didn't think to check the X log and does that now
jcristau: ssieb: the renderer string in glxinfo
ssieb: got it!
ssieb: (EE) [drm] DRM was busy with another master.
ssieb: (EE) RADEON(0): [dri] DRIScreenInit failed. Disabling DRI.
ssieb: known bug? if not, where should I file it?
adamk_: ssieb: Not a bug, just a limitation.
adamk_: ssieb: Only the first user gets a lock on the DRI device.
ssieb: is that going to be changed in future?
adamk: Only time will tell :-)
adamk: I imagine it's on someone's to-do list, but who knows.
ssieb: :-/
ssieb: it's kind of painful... does the fglrx driver have the same limitation?
adamk_: ssieb: I believe it does.
ssieb: so it's a kernel limitation?
adamk_: I believe it's a limitation of the direct rendering infrastructure.
adamk_: *believe* being the keyword. I'm certainly not an expert on this.
ssieb: ok, thank you
ssieb: it's unfortunate, because my wife likes to stay logged in so it's easier to access, but then the kids can't play the games that use gl
airlied: its a drm problem
MostAwesomeDude: I can't remember whether or not that's addressed by DRI2...
airlied: F10 should or soon after F10 will do it...
airlied: its not DRI2 related.
spstarr: interesting
spstarr: kwin workaround seems to stop that crash now
airlied: I had it working on Intel last year fine, I just haven't had time to play with it.
airlied: you just need a memory manager really. modesetting helps also.
spstarr: im in exa , w/ exa composite mode off and DFS accel off
lkro: agd5f: ping.
ssieb: airlied: excellent! that's good to hear :-)
spstarr: airlied: im a little confused with EXA.. when we're in GL composite, why do we still need EXA to draw pixmaps ?
airlied: spstarr: GL doesn't draw pixmaps
spstarr: but it can render them onto textures ?
airlied: spstarr: nope.
airlied: spstarr: compositing just takes pixmaps and puts them onto the screen via the compositor
airlied: how the compositor does that is up to it. GL or Xrender or whatever.
spstarr: ok so then why is EXA still needed? or is it only now used for download/uploading from screen for offscreen stuff?
spstarr: (but wouldn't any offscreen pixmaps be handled by the kernel memory manager now?)
airlied: spstarr: http://people.freedesktop.org/~ajax/fennec-fox.jpg
spstarr: ha
airlied: spstarr: to quote ajax, you are so missing the point I can't figure out where to start, here's a picture of a cute baby fox.
spstarr: well, it just seems very odd when moving windows around etc is accel, yet displaying pixmaps is slow
airlied: you need to understand what compositing actually does, you are missing first principle knowledge
airlied: its like you read the words but never figure out what the meanings behind them are.
spstarr: I am, yes
airlied: compositing just takes windows and puts them on the screen, it has nothing to do with making the contents of the windosw.
spstarr: ok, but 'takes windows and puts on the screen' being the coordinates only not the actual contents of the window
spstarr: ?
spstarr: if so,why isn't the two combined together?
spstarr: i would think, we want to blit the whole window, contents and all store those coordinates in offscreen area, no separate the content and window itself
spstarr: EXA gets the contents with DFS and returns it with UTS
spstarr: if a unification of 2D and 3D together happens,wouldn't the representation of what 2D is just be represented with 3D textures ?
spstarr: airlied: what im getting at is, I know X' is unique in its display system, but extra interfaces to render pixmaps, textures should be come one method ?
spstarr: brb
airlied: spstarr: you are missing how window system works.
airlied: how you put contents into windows is independent of the how you composite windows.
airlied: however you can use similiar methods to do both.
airlied: i.e. you can use Xrender to draw windows and Xrender to composite
airlied: or you can use GL to draw to windows and GL to composite.
spstarr: so then EXA and Xrender orthogonal?
airlied: nope.
airlied: EXA accelerates Xrender
spstarr: I see
spstarr: brb EXA w/o exa composite is so damin slow i cant use this :/
damentz: brb
damentz: brb
spstarr: sighs well, at least with XAA im back to being able to use eyecandy with stability as from OLS 2007