AstralStorm: that makes it fairly useless ;p
taiu: well ... :)
AstralStorm: goes back to downgrading the shaders
taiu: for glsl 1.2 also all seemed core mesa changes, and as it can claim 1_2 for sw and intel? i thought we should be also ok
taiu: but I'm not at all at home in this part of the code
taiu: oh, one thing if someone wants to start hacking with the driver is to add gl_FragDatax handling
taiu: should be about +20 lines ;)
suokko: AstralStorm: Hacking driver is not that hard ;) Specilly when you know gl already from outside
AstralStorm: suokko: unfortunately I'm pressed for time on those 50 shaders
AstralStorm: (and 50 is a rough estimate)
agd5f: AstralStorm: you could always run the proprietary driver in the meantime
Nightwulf|work: hi all
AstralStorm: hmm, does the feedback buffer even work here?
AstralStorm: I mean on RV670
AstralStorm: or is some extension missing?
AstralStorm: that's technically OpenGL 1.5
AstralStorm: I get no errors, but do get -1 from glRenderMode(GL_RENDER) after it
AstralStorm: oh wait, I did GL_3D with lights. foolish of me
AstralStorm: so, is feedback buffer supported?
agd5f: AstralStorm: EXT_transform_feedback?
AstralStorm: none here, I'm screwed
agd5f: if so, not likely as brian just added support for that today
AstralStorm: so I should pull this? great
agd5f: AstralStorm: added support to mesa, not the drivers
AstralStorm: even funnier
AstralStorm: so I won't have any
AstralStorm: well, it's not EXT_transform_feedback apparently
AstralStorm: it's mentioned in OpenGL 2.1
AstralStorm: (the function glFeedbackBuffer that is)
agd5f: AstralStorm: looks like a GL3 feature
agd5f: and it requires streamout buffers which aren't implemented on r6xx yet
AstralStorm: // Selection and Feedback huhno?
AstralStorm: it's the antique OpenGL 1.x feedback buffer...
AstralStorm: transform feedback is a different thing
AstralStorm: it works a step before end feedback buffer
AstralStorm: the funny thing is that I don't get any errors, just -1 elements written
AstralStorm: seems like it didn't intercept all those glMap2f etc. calls
AstralStorm: it does hijack basic drawing
airlied: in theory we should hit sw for feedback
airlied: same as selection
airlied: though not sure when I fixed GL_SELECT
AstralStorm: for now it hits either -1 or 0 counts :/
MrCooper: agd5f: would it be feasible to even trigger a modeset when the display priority is changed? :)
airlied: probably wouldn't require a full modeset either
agd5f: MrCooper: yeah. won't even require a full mdoeset, just an update of the watermark regs
MrCooper: that would be deluxe :)
agd5f: does changing the parameter generate some sort of event?
agd5f: MrCooper: ideally we make it a device attribute
agd5f: as I think that has a driver callback
agd5f: although, I'm not sure where we'd want to hang it in the sysfs tree
MrCooper: yeah, anyway this can always be improved later on, not having to reload the module is worth a lot
airlied: ah yes module params don't generate any feedback
DanaG: I keep thinking of watermark as in that "AMD Unsupported Hardware" mark.
agd5f: different kind of watermark :)
MrCooper: that's called an icon actually :)
MrCooper: (in the hardware)
DanaG: oh yeah, and thanks for that GPU reset stuff... now instead of getting my kernel dying, I just get my screen blinking.
DanaG: Though, the fact that I need a GPU reset worries me... is that just a driver issue wedging the gpu, or is it likely to be a hardware fault?
agd5f: danaG: likely a driver bug
DanaG: ah, good.
DanaG: I also make the comparison with fglrx, where running hl2dm in wine wedges the GPU, yet furmark native (and such) doesn't cause any issues.
MrCooper: wonders if we could make any good use of the hardware icon
agd5f: MrCooper: 2 hw cursors for mpx?
DanaG: Interesting place I'd seen a hardware overlay (perhaps?) used: on Gateway laptops, with the brightness OSD.
DanaG: It was actually done in the BIOS... and if I had just a CRT connected (and lvds connected), spamming brightness keys made my CRT lose signal for a moment.
MrCooper: maybe, or direct notifications from the DRM to the user :)
agd5f: or a overlay console
agd5f: although I think the icon is limited to 64x64 or some such
MrCooper: that would be pretty small though?
DanaG: hmm... maybe use it for the equivalent of a heartbeat LED?
DanaG: I've found a heartbeat LED tremendously useful on my HP laptop. /sys/class/leds/hp::hddprotect/
DanaG: ooh, resume from suspend seems faster, too.
DanaG: oh, and I fixed my gpu hang-reset-hang-reset-hang-reset I was getting with UEFI boot mode -- just had to entirely disable efifb.
airlied: DanaG: was efifb handover working?
DanaG: yeah, it seemed to be.... until Xorg tried to start.
MrCooper: using the icon for a beating heart might be interesting, though I suspect it might generate some bug reports :)
DanaG: Console was high-res as it should be, once I got gdm to stop respawning.
MrCooper: I guess the icon could be useful for GPU performance monitoring
DanaG: I do know that HP's UEFI implementation on the first-gen EliteBooks is screwy... it claims one video driver type (not sure which of GOP or UGA)... but then gives a framebuffer address of zero.
DanaG: ... and yet using the OTHER type works fine-ish.
airlied: DanaG: well efifb should give hires also
DanaG: Not on mine... it only gave 1024x768.
DanaG: lemme' dig up the pastebin with it not working.
DanaG: and I think this is the xorg log that goes with it: http://pastebin.com/ynYDJfAL
DanaG: Also weird on my system: adding ubuntu grub2 to efibootmgr gives working efifb (that is, 1024x768)... but hitting F9 and "boot from EFI file" gives broken text wrapping.
DanaG: Heh, and without a 1TB drive, what benefit is there to UEFI, anyway? I just mess with it as a curiosity.
TheBrayn: is xf86-video-ati-6.12.5 supposed to be able to turn down the video card's fan?
adamk_: TheBrayn: power management depends on the DRM, not the DDX. 2.6.33 supports some dynamic power control, but I believe the real improvements are slated for 2.6.34
TheBrayn: 2.6.34 doesn't seem to be in portage's x11 overlay
Tommeh: I don't know why it would be in an x11 overlay? Shouldn't it just be in the normal portage tree?
Tommeh: Granted I haven't used Gentoo for a while so I might be out of date :)
TheBrayn: I added an overlay to get more recent packages
adamk_: Well 2.6.34 hasn't been released yet, so I'm not surprised :-)
Tommeh: Sure, but I didn't think they'd include kernels in an overlay for X11 :)
MrCooper: it also depends on KMS, which 6.12.5 doesn't support
Tommeh: TheBrayn: are you sure that your flags aren't blocking 2.6.34-rc* for example?
TheBrayn: pretty sure
TheBrayn: yeah you were right
TheBrayn: x11-drivers/xf86-video-ati-6.12.192 <-
adamk_: Fair warning: I don't know that the newer PM patches have made it into the kernel tree yet.
TheBrayn: what are pm patches?
adamk_: The power management patches.
adamk_: They might be in one of airlied trees
TheBrayn: http://codepad.org/7xKOx9Av http://codepad.org/DyBuSd7i
TheBrayn: compiling xf86-video-ati-6.12.192
MrCooper: probably need a newer libxext
NightNord: Greetings everyone. I've got a very strange problem with my RV710 card (Mobility HD 4570) and 3d accel
TheBrayn: I have 1.1.1
NightNord: Currently I have superior 2d acceleration, much better than proprietary drivers, but very strange bug with any kind of 3d
NightNord: It seems to lag badly with any GLX-enabled window, even if it playing some video, for example
adamk_: NightNord: What is the output of 'glxinfo | grep -i render'?
NightNord: I've tried some platinum apps from RadeonProgram, but noway
NightNord: adamk_: 3d is on.
NightNord: direct rendering: Yes
NightNord: OpenGL renderer string: Mesa DRI R600 (RV710 9553) 20090101 TCL DRI2
NightNord: KMS enabled, mesa 7.7, 33 kernel
pazof: firmware loaded ?
NightNord: live version of drivers (updated today)
NightNord: 2d and KMS is working
NightNord: Logs says that everything ok, I'll post Xorg.log now
adamk_: Mesa from git is at 7.9-devel, not 7.7, so how did you update?
MrCooper: TheBrayn: maybe xextproto or something then
suokko_: NightNord: Can you explain what you mean with lag?
NightNord: drivers - xf86-video-ati, I mean
AstralStorm: that's not enough
AstralStorm: old mesa doesn't have enough 3D
NightNord: Mesa version irrelevant, I've tested it on every possible version (git too)
AstralStorm: it is relevant
TheBrayn: should also be quite new
NightNord: I may update and report log, ok
AstralStorm: esp. ati dri driver version
NightNord: Lag - is very very very low fps. Most of 3d game/whatever will lag even while playing video (both wine and native)
NightNord: But wine says, that software rasterizer is in place, and this makes sense...
adamk_: Well that's not right.
NightNord: AstralStorm: ati-driver is live. Updated today, I've told this
adamk_: If wine is using the software rasterizer, it's gonna be slow.
AstralStorm: ati DRI, it's part of mesa
adamk_: As for native games, have you tried benchmarking openarena?
AstralStorm: also, drop us the log
AstralStorm: adamk_: if mplayer lags, it's software rendering
NightNord: I'm updating for git mesa now... I'd downgraded mesa to have same options as in RadeonProgram wiki page
AstralStorm: less = worse
AstralStorm: Xorg log would be more helpful
pazof: is it a Fresh xorg.conf ?
suokko_: NightNord: btw, you neeed 32bit mesa for wine
NightNord: Ok, I'm updating ;) adamk_: no, but nexuiz perfomance seems to badly depend on cpu frequency =)
MrCooper: NightNord: what suokko_ says
pazof: LoadModule: "exa"
NightNord: I'll test on openarena as soon as get newer mesa...
NightNord: pazof: it's 2d, 2d is working great
pazof: ok, sorry :)
NightNord: xorg.conf: http://paste.pocoo.org/show/195900/
NightNord: xorg-server is also live
adamk_: NightNord: Benchmark it with the anholt demo
suokko_: mumblesthat lag would be better word if fps was high but reaction to input was delayed
NightNord: adamk_: just to be clear: this one? http://dri.freedesktop.org/wiki/Benchmarking?
adamk_: NightNord: Yep.
NightNord: current mesa: 840 frames 32.9 seconds 25.6 fps 17.0/39.1/63.0/7.6 ms
NightNord: Will reload now for new...
AstralStorm: should've asked about the resolution
MrCooper: he's probably just missing a 32 bit r600_dri.so
adamk_: IF OA is 64-bit that shouldn't matter for this test. But would certainly explain wine issues.
AstralStorm: that's why I tested with ioquake
AstralStorm: it's available in both (after you build it, that is)
NightNord: 840 frames 32.4 seconds 26.0 fps 17.0/38.5/65.0/7.3 ms
suokko_: NightNord: No need to restart to upgrade mesa
NightNord: OpenGL renderer string: Mesa DRI R600 (RV710 9553) 20090101 TCL DRI2
NightNord: OpenGL version string: 2.0 Mesa 7.9-devel
adamk_: NightNord: Alright, so are you on a 64-bit or 32-bit distribution?
adamk_: NightNord: And is OA a 64-bit or 32-bit binary?
NightNord: /usr/games/bin/openarena: ELF 64-bit LSB executable
adamk_: OK, well there definitely is something odd going on there, then.
NightNord: But I'd no 32bit mesa libraries, that's strage, makes sense why wine is not working anyhow
AstralStorm: openarean at what resolution btw?
NightNord: adamk_: if it will help: hardware cursor is constantly dissapears, switch to console and back to X helps
AstralStorm: that's some bug, heh
AstralStorm: you said it's 2.6.33?
NightNord: And both openarena tests was with some dotted line on top edge
TheBrayn: xf86-video-ati compiled <3
AstralStorm: I'd try radeon-testing
AstralStorm: NightNord: dotted line I've seen in other apps too
NightNord: AstralStorm: 1600x900
AstralStorm: it's some kind of bug
AstralStorm: oh, it shouldn't be *that* slow then
AstralStorm: probably goes through software fallback on some shader
AstralStorm: note OpenGL 2.0
NightNord: openarena menu with flares enabled gives ~2 fps
AstralStorm: exactly, see above
AstralStorm: OpenGL 2.
AstralStorm: not OpenGL 2.1 or 3
suokko_: NightNord: Try to run with smaller resolution
AstralStorm: suokko_: nah, the resolution shouldn't matter
AstralStorm: it's a fairly powerful card
AstralStorm: (RV710 9553)
NightNord: AstralStorm: yes, I'm on 33 kernel, so called 'zen sources', KMS enabled, BFS shelduer (but it must be irrelevant)
AstralStorm: NightNord: zen-sources zen.git? that's not 2.6.33
AstralStorm: that has drm-next or drm-radeon-testing
AstralStorm: ah, stable
AstralStorm: BFS might be relevant, but is unlikely
AstralStorm: unless there's some lock contention that is
AstralStorm: priority inversion can happen
AstralStorm: try nicing the app
AstralStorm: like, nice -n 19 openarena
adamk_: AstralStorm: Going from 1152x864 to 1600x1200 in OA here (with a 4850) drops me by 20fps.
adamk_: AstralStorm: So resolution most certainly does matter.
suokko_: AstralStorm: That card is my option quite lower end card
AstralStorm: my RV670 manages better than this
suokko_: AstralStorm: And we are wasting the memory bandwidth quite badly
MrCooper: not sure 20-30 fps qualifies as 'very very low' anyway
AstralStorm: suokko_: ahh, that might be it
AstralStorm: MrCooper: it's 17-60 FPS spread
AstralStorm: bad really
AstralStorm: well, my ioquake deadly benchmark went ~30 FPS at 1920x1200
MrCooper: maybe it's hitting software fallbacks for some effects
AstralStorm: but that probably didn't excercise the card enough
AstralStorm: very likely
AstralStorm: the dotted line bug is real though
AstralStorm: needs a fix
AstralStorm: memory corruption?
suokko_: AstralStorm: looks like some rendering error
NightNord: AstralStorm: 840 frames 10.9 seconds 77.1 fps 7.0/13.0/42.0/3.3 ms
NightNord: niced on 640x480
suokko_: I think I reproduced it
AstralStorm: NightNord: not better
AstralStorm: in fact, a lot worse
AstralStorm: so it's likely not a priority inversion... but quake might be abusing sched_yield
AstralStorm: that will tank performance with BFS
suokko_: Now it is CPU limited when high resolutions are GPU limited ;)
NightNord: I may try to disable BFS to be sure
AstralStorm: yeah, do so
AstralStorm: really, people should learn that sched_yield is not a synchronization primitive
AstralStorm: no really.
AstralStorm: it's never been
AstralStorm: "The sched_yield() function shall force the running thread to relinquish
AstralStorm: the processor until it again becomes the head of its thread list."
NightNord: But doom3 was also lagging badly on initial video (id software logo) - sound works well and not lagging, but video... 1 fps or something like that. I've never managed to escape from menu by controls - it has much lower than 1 fps
AstralStorm: hint: it's not specified that they're run in round-robin order
suokko_: AstralStorm: openarena is single threaded. I doubt it is calling sche_yield
AstralStorm: suokko_: it is? hmmmmm
AstralStorm: now you've got me interested
AstralStorm: I bet it's running some shader via software fallback
suokko_: AstralStorm: Most likely 640x480 is runing 100% of sinle core while 1600x900 is running only about 50% CPU of sinlge core and waiting for GPU
AstralStorm: or more likely particle effect
NightNord: In videos too?
AstralStorm: yeah, scaling
AstralStorm: e.g. ARB_imaging - convolution?
AstralStorm: or it tries to do decoding in a shader and the shader fails with GLSL 1.1
AstralStorm: might be trying FBO too...
AstralStorm: is that accelerated now?
adamk_: NightNord: Out of curiousity, can you pastebin the full output of 'LIBGL_DEBUG=verbose glxinfo' ?
NightNord: One moment...
suokko_: AstralStorm: You don't do anything that complex to hit the limits
NightNord: adamk_: http://paste.pocoo.org/show/195910/
NightNord: No drirc? Or it's not needed?
adamk_: Not needed.
adamk_: Well that all looks normal.
suokko_: NightNord: real problems: Driver is not optimized and you don't have high end hardware
AstralStorm: suokko_: but you do something that complex to hit a software fallback ;p
NightNord: suokko_: eh?
NightNord: But it should not lag in videos, imo...
NightNord: At least =)
suokko_: NightNord: What videos are laging? HD?
AstralStorm: that's definitely hitting something really slow, like software
AstralStorm: it'd be great to get a dump of the calls
AstralStorm: can mesa do that?
NightNord: suokko_: any video in glx window (gameone)
NightNord: I'm not watching any HD videos, so I dunno about mplayer perfomance
AstralStorm: if that's hitting a fallback, it is really in a sorry state
AstralStorm: or non_power_of_two
suokko_: NightNord: XV is only well supported for video palyback currently
AstralStorm: suokko_: ... it shouldn't work that slow
AstralStorm: it works fine and fast here
AstralStorm: with RV670
AstralStorm: this includes funny scaling fragment programs
NightNord: Hm, wine seems to work ok with non-lib32 mesa libs
NightNord: So, I may post wine log for cnc4 saying some strange thigs about 'formats'
AstralStorm: NightNord: does mplayer -vo gl work slow too?
suokko_: AstralStorm: 3850?
suokko_: or which card?
NightNord: It's working just ok
AstralStorm: suokko_: here it works all fine duh
NightNord: At least, I don't see any problems
AstralStorm: NightNord: ok, so it's some weird pixel format
suokko_: AstralStorm: Because if you have that one your card is about 5x more powerfull ;)
AstralStorm: suokko_: not *that* much more powerful
NightNord: VIDEO: [WMV3] 1280x720 24bpp 1000.000 fps 3000.0 kbps (366.2 kbyte/s)
NightNord: I suppose it's quality enough to make problems, if any...
AstralStorm: so that -vo gl would work fast and clean here and not on rv710 ;p
AstralStorm: NightNord: weird pixel formats are likely unsupported
AstralStorm: or suppored in software ;p (= slow)
AstralStorm: NightNord: doesn't matter really, just going fullscreen would cause problems
AstralStorm: try then -vo gl:yuv=4
AstralStorm: if that works fast, fragment *programs* are fine
NightNord: Small lags are here
NightNord: Let's try something more heavy...
AstralStorm: if it gives a black window, they're broken
AstralStorm: small lags are fine
AstralStorm: that fragment program wants 4 texture units ;p
AstralStorm: probably loading your card to the brim
AstralStorm: try yuv=2 then
NightNord: Nothing bad
NightNord: Everything ok. some dvd quality video from mkv also working fine with this options
AstralStorm: this uses 3 texture units
NightNord: I've got kernel recompiled without bfs and i/o accounting
AstralStorm: so 3D works
NightNord: Should I also switch off timers stats? They may make problems in some cases
AstralStorm: it's definitely whatever openarena does
AstralStorm: oh right
AstralStorm: check CPU usage while you're at it
AstralStorm: here it's about 30% of a core on some 720p
AstralStorm: nah, shouldn't matter much
ferr_: i have problems with radeon and dri, i have a new 2.6.34 kernel with next patch, have now working KMS and dmesg says dri2 is working, but in Xorg.0.log i see: [dri] radeon kernel module version is 2.1.0 but version 1.17.0 or newer is needed. and glx info says OpenGL renderer string: Software Rasterizer. Can anyone help me please?
AstralStorm: ferr_: you need a newer version of xf86-video-ati
AstralStorm: or libdrm
AstralStorm: or both :)
adamk_: Or you need to disable KMS.
AstralStorm: is it debian?
ferr_: yes it's debian (sid)
ferr_: i built xf86-video-ati, mesa, and libdrm from git
NightNord: "err:winediag:X11DRV_WineGL_InitOpenglInfo The Mesa OpenGL driver is using software rendering, most likely your OpenGL drivers haven't been installed correctly"
suokko_: ferr_: echo radeon >> /etc/modules
NightNord: Same thing still, strange...
ferr_: it's already there
AstralStorm: suokko_: uh, he's running with KMS d'oh
AstralStorm: ferr_: you're missing --enable-radeon-experimental-api for your libdrm
AstralStorm: or something like it
AstralStorm: might've messed up the name of the configure option
jcristau: AstralStorm: no, suokko_'s right.
NightNord: AstralStorm: X + mplayer has ~40% of CPU
NightNord: Video off and load fades out
AstralStorm: rebuild libdrm, then rebuild xf86-video-ati again
AstralStorm: alternatively, your mesa is really old
AstralStorm: but you said git, so unlikely
AstralStorm: or your mesa is built w/o radeon, r500 or r600 driver
AstralStorm: NightNord: that's fine, it has to decode video somewhere :)
AstralStorm: jcristau: except KMS loaded, so radeon is loaded ;p
spreeuw: NightNord: with HD video?
spreeuw: thats decoding , not display
suokko_: ferr_: Can you paste your full log?
spreeuw: its normal
ferr_: xorg log?
ferr_: if i try libdrm with experimental api i get:
ferr_: configure: WARNING: unrecognized options: --enable-maintainer-mode, --enable-radeon-experimental-api
ferr_: suokko_: jut a sec
AstralStorm: at least should be
NightNord: spreeuw: nope, some trailer laying around... But resolution is fine enough
mokoloko: Does HoN work with r600 card with radeon drivers? :) http://www.heroesofnewerth.com/ open beta starts soon :P
NightNord: I've found some nice thing in dmesg
NightNord: It might be a problem
NightNord: radeon 0000:01:00.0: LVDS-1: EDID invalid.
NightNord: And before that few messages like [drm:edid_is_valid] *ERROR* Raw EDID: (all 00 00 ...)
NightNord: That's ok?
AstralStorm: mokoloko: does it require OpenGL 2.1 or newere?
AstralStorm: if so, it won't work really
NightNord: Hm, it seems to post this whatever video is called
ferr_: suokko_: http://pastebin.com/0AbMaiJ4
adamk_: ferr_: KMS loaded and your console resolution changed *before* X started?
suokko_: ferr_: you need video-ati from master branch
suokko_: 6.12 doesn't support KMS
ferr_: yes resolution changed to native while booting
suokko_: ferr_: But sid has already new enough package
suokko_: + sid has 1.7.6 xserver if you just upgrade
NightNord: I'll try to update for git-kernel than, it seems to be some kms bug...
AstralStorm: mokoloko: the movie doesn't look like it really *needs* advanced shaders, but...
AstralStorm: devs sure get lazy
AstralStorm: NightNord: ?
NightNord: AstralStorm: invalid EDID on LVDS
AstralStorm: which kernel?
NightNord: I dunno about consecuences, but kms seems to be unhappy
mokoloko: AstralStorm: It needs at least ogl2.0 support
NightNord: 33 zen currently, i'll try to upgrade to 34_rc2.2
suokko_: NightNord: EDID is information that your monitor has for driver to detect supported modes
NightNord: suokko_: hm... And Xorg reads it just ok... Anyway, i'll give it a try =)
AstralStorm: mokoloko: so *might* work
ferr_: suokko_: i'm upgrading
AstralStorm: as long as they don't lie about 2.0 when they mean 2.1
ferr_: suokko_: i have now xorg 1.7.6 and radeon 6.12, anithing else to do? or just restart?
mokoloko: I wonder if there's something big coming up for r600 cards because there haven't been any updates in mesa-git for a few weeks :P
suokko_: ferr_: 6.12.902?
suokko_: hmm. That should be right
mokoloko: AstralStorm: From min req: Video Card - 128MB fully OpenGL 2.0 / GLSL 1.20 compliant
ferr_: ok i try to restart:
suokko_: ferr_: + you probably want also the latest mesa if you didn't upgrade it
ferr_: i did a full upgrade
adamk_: mokoloko: GLSL is at 1.10 in the r600 driver.
adamk_: And I'm not even sure if that's at 100%
mokoloko: Darn well I'll try anyway :) Hopefully catalyst 10.4 comes soon so that it'll finally "work" with fedora 12
NightNord: Heh, as for me, proprietary drivers has good 3d perfomance (but buggy) and unasable 2d perfomance state...
Pallokala: mokoloko: afaik, r600g is in the making
Thunderbird: the 2d acceleration part will be fixed soon
Pallokala: dunne when it becomes usable though
Thunderbird: what would help most drivers if more tests were added to piglit and if more drivers (including proprietary ones) would use it for testing
madduck: any idea why my X 7.5 with two radeon cards in a triple-head setup lost Xvideo in the last two weeks?
madduck: nothing in the logs. :(
mokoloko: Yeah I know I guess people are focused on porting the old one to the gallium arch. Hopefully it won't take too long and will be usable by HoN release :)
TheBrayn: http://paste.pocoo.org/show/195938/ <- anything that should be in here?
xming: there is a prerelease of catalyst 10.4 which works with xorg 1.7.x
TheBrayn: catalyst == fglrx?
suokko_: madduck: Did you upgrade the video-ati?
madduck: suokko_: 1:6.12.192-2 is the latest available one and yes, it's installed
ferr_: suokko_: it's working now, thank you
suokko_: madduck: I was jsut thinknig did you upgrade it in the period when you lsot xv
madduck: suokko_: lemme check
madduck: yes, i was running 1:6.12.99+git20100218.e68d3a3-1~dileX+1 before
Ke: I tried the prerelease on gentoo and for some reason it didn't seem to install any kernel module
Ke: (also claimed there is no proper hardware)
suokko_: madduck: Not many commit between there. Is there chance that you could bisect it?
madduck: suokko_: yeah, maybe on the weekend
suokko_: madduck: You could now try to downgrade as a quick workaround
madduck: yeah, if I find the package again. ;)
Tommeh: Hmm... Looks like airlied's pull request didn't get honoured for -rc3 :(
jcristau: Tommeh: he retracted it, so not surprising
Tommeh: jcristau: I didn't realise! Nevermind then. :)
obscurant1st: my GUI wont work after installing ati catalyst driver for radeon x2300, how can i remove and restor the GUI?
adamk_: Questions on the catalyst drivers should go to #ati
adamk_: Oh, you decided to ask there and here. Well there is the correct place :-)
obscurant1st: adamk_, thx,
ferr_: hmm, i have now texture problems, if i type xrandr --output LVDS --auto (horizontal lines blinking up)
zhasha: ferr_: I have the same issue
ferr_: that's bad:\
zhasha: well, I only occasionally notice a 1px thin line blinking
adamk_: Can anyone confirm that the X2300 is an r500 card?
Tommeh: Wikipedia has a list of AMD GPUs and their respective details. :)
Tommeh: (I'd only look there to find out.)
zhasha: adamk_: just from the fact that it doesn't have HD in the name, you should be able to tell that
adamk_: Hmmm... Mobility Radeon X2300 is RV535.
adamk_: zhasha: Except that the HD2300 shows up as RV550 in wikipedia.
adamk_: zhasha: So either that rule is wrong, or wikipedia is.
Tommeh: I'd imagine given that it's a low-tier part in the HD2xxx range, that it's just a re-badged GPU from an older, more profitable, tech.
Tommeh: Nvidia do it all the time, so I wouldn't be surprised if ATI/AMD did/do it.
adamk_: I'm sure, which is why it's not always easy to figure out what chip is used based on the GPU name :-)
adamk_: But, yeah, I'm gonna go with wikipedia on this one.
Tommeh: In fact, the 880G/GX IGP is just a re-badged 785G IGP, albeit with higher clocks. They actually brand it as a HD4xxx instead of HD3xxx though.
ferr_: adamk: i have x2300, and OpenGL renderer string: Mesa DRI R300 (RV515 7188) 20090101 x86/MMX/SSE2 TCL DRI2
zhasha: adamk_: that does seem to be an exception to the rule, but AFAIK X2300 is an R500
zhasha: I didn't thoroughly check but that was the overall impression I got from my old classmates then new laptop
MrCooper: IGPs tend to be bastard children with different components from different generations
Tommeh: Whatever provides the most features with the lowest power and highest yields, right? :)
jcristau: and lowest cost i guess
Tommeh: highest yields ~= lowest cost.
Tommeh: But yeah.
madduck: suokko_: problems persists after downgrade, so this is likely a change in X.org
madduck: I also had the same problem on an intel machine the other day
madduck: there it was fixed by upgrading the intel driver
AstralStorm: mokoloko: yeah, you're screwed
AstralStorm: as I said, 1.10 doesn't really cut it
AstralStorm: lack of arrays is a real pain in the rear end
AstralStorm: saying "OpenGL 2.0 compliant" while requiring GLSL 1.2 is a bit contradictory
AstralStorm: as GLSL 1.2 is OpenGL 2.1
AstralStorm: not sure if there's a card that has the one and not the other
AstralStorm: 2.1 has GLSL 1.2 and PBOs
AstralStorm: just about all the difference from 2.0
agd5f: yes, x2300 is an r5xx card. rv550 actually which is an rv515 with UVD
chithead: I thought the difference between x2300 (m64) and hd 2300 (m71) was the uvd
hnsr: i wonder, what is "11:03.0 VGA compatible controller: ATI Technologies Inc ES1000 (rev 02)" ?
hnsr: i have that in a server i rent, but ive never heard of it
agd5f: hnsr: it's a rv100 based server chip
agd5f: no 3d engine, limited bandwidth
tmkt: I have Radeon 2100 and get a 2 pixel wide purple line on the left side of my monitor at all times, any ideas?
hnsr: agd5f, ah okay
Obscene_CNN: tmkt, (excuse the stupid question but someone needs to ask it) is it there in windows?
tmkt: don't know
tmkt: don't use windows
tmkt: it wasn't there before when i had the nvidia card there
chithead: does the tv also have a vga connector?
tmkt: might..i'll have to check that tonight,
Obscene_CNN: tmkt, can you try the vesa driver and see if it is still there
tmkt: ok so driver 'vesa' will do
Obscene_CNN: it will "work" but will be slow
tmkt: yeah..i don't do any 3d
tmkt: just HD
tmkt: its my mythtv box
adamk_: I'm guessing that vesa isn't being offered up as a permanent solution, just a test :-)
Obscene_CNN: you'll want some acceleration for that to avoid tearing
tmkt: ok..so 2 tasks tonight...try vesa and radeonhd
Obscene_CNN: well, radeonhd is on life support (unfortunately)
tmkt: ah ok i won't bother with that then
Obscene_CNN: just see if the vesa driver eliminates the purple line. If it does its probably not a hardware issue.
tmkt: i'll give that a shot now
tmkt: ok..waiting for feedback if that helped or not
AstralStorm: vesa won't do with HD
AstralStorm: it can't even enter the necessary resolutions
AstralStorm: not to mention that scaling is very cpu intensive, add H.264 decoding to that and your cpu is not enough
tmkt: my screen is 1360x768
tmkt: so not asking for 1080
Obscene_CNN: AstralStorm, the vesa suggestion is to just test the hardware
tmkt: eitehr way..rebooting back to radeon once i figure out if the purple line is there or not
AstralStorm: why not? downscaled is excellent quality
AstralStorm: better than 720p
AstralStorm: assuming your filter doesn't suck
AstralStorm: cubic is fine
Obscene_CNN: ponders a picaso filter that does nothing
AstralStorm: and that's available either via XV if you have r500 or lower, or via gl, lscale/cscale=1 or 3 (talking mplayer)
AstralStorm: you mean nearest neighbor scaling, try sws=1
AstralStorm: uh, -sws 4 actually
AstralStorm: that's really sucky
AstralStorm: the correct one is -sws 5 - area filter, "fine" for downscaling, just as sucky for upscaling
AstralStorm: of course those are software
AstralStorm: the default being bicubic (not bicubic spline like available in gl)
AstralStorm: bicubic spline unfortunately is fairly blurry as far as filters go
Obscene_CNN: notes that for that Picaso joke to fail people need more culture
AstralStorm: hey, mplayer has all kinds of filters
AstralStorm: probably a "cubism" transform too
tmkt: I have Radeon 2100
Obscene_CNN: okay you know who Picaso is and have more culture than just between your toes :)
AstralStorm: Obscene_CNN: for now, you get to live with libcaca
AstralStorm: asciiart is very picasso-like
AstralStorm: tmkt: oh, an old one, is that r300?
AstralStorm: pci or agp?
tmkt: its onboard
AstralStorm: you can check with lspci -v
tmkt: which field?
AstralStorm: duh... fairly obvious?
tmkt: Flags: bus master, fast devsel, latency 32, IRQ 27
tmkt: doesn't say if it is pcie i mean
tmkt: its pci
tmkt: not agp
AstralStorm: actually -vv
tmkt: I'd assume its pcie since it plays everything smoothly
AstralStorm: probably yeah
AstralStorm: the problem sounds like a misclocked output
tmkt: when i had a pci video card in there it was too slow to handle the HD, and this handles it no problem
tmkt: 00:06.0 PCI bridge: ATI Technologies Inc RS690 PCI to PCI Bridge (PCI Express Port 2)
AstralStorm: huh? card too slow to handle... 2D graphics, wtf?
tmkt: pci is to slow to handle hd
AstralStorm: no really, that is PCI to PCI bridge
AstralStorm: of course it's not
AstralStorm: PCI can do 133 MB/s
Obscene_CNN: in theory
tmkt: Hmmm...not sure then..the answer i got from the mythtv people was that my video card was plenty powerful enough
AstralStorm: and that is not enough for full HD, yeah
tmkt: until i told them it was pci version of the card and not pice
AstralStorm: because it takes 237 MB/s
Thunderbird: note that a PCI card can be fast enough for HD playback if the GPU does all decoding
AstralStorm: because then only compressed data is sent
tmkt: so..purple line still there as vesa
AstralStorm: tmkt: hardware problem then
Thunderbird: I have seen reports of people who use geforce 8 cards using vdpau on a p3
Thunderbird: yes thanks to the compression
AstralStorm: now get another card and check the screen
AstralStorm: I assume you have one
Obscene_CNN: check cables first
tmkt: AstralStorm: yeah..I have a NVIDIA 9600
AstralStorm: if it's analog, digital should fail completely
tmkt: that worked great
tmkt: just the fan was so loud
AstralStorm: fans are fixable ;p
tmkt: so i decided to go to the onboard card
AstralStorm: just drop some WD-40 from time to time
tmkt: and purple line is less annoying then the fan
AstralStorm: will extend the lifetime 5x or so
tmkt: nah..the fan brand new...was loud to begin with
AstralStorm: then replace with a better model, maybe :)
adamk_: tmkt: I'm inclined to go with agd5f on this one and suggest that you enable or disable overscan, if it's an option.
tmkt: guess i could dish out the cash and get a passive cooling card
AstralStorm: check the screen setup
tmkt: ok...just have to figure that out
AstralStorm: sounds like EDID might've failed
AstralStorm: care to check syslog/dmesg for such messageS?
tmkt: nothing in dmesg about edid
AstralStorm: should be some
tmkt: dmesg|grep -i edid
AstralStorm: oh no, it won't be spelled like that
AstralStorm: | grep drm
AstralStorm: see Connectors and Encoders
tmkt: one thing to note...i set drive to vesa last reboot
tmkt: but dmesg is talking about radeon all over the place
AstralStorm: tried to load it nevertheless, yes
AstralStorm: check previous boot with /var/log/messages
tmkt: anything useful?
tmkt: lsmod has drm and radeon modules loaded
mjt: Kernel Modesetting"... "mostly"
mjt: wtf is "mostly" ? :)
stikonas: mjt: "MOSTLY" means that it is implemented but has some known bugs.
mjt: well, i see the definition (which you repeated)
mjt: just.. funny... ;)
mjt: ignore me :)
spreeuw: mostly pregnant
hnsr: mostly harmless
makkalot: hi all can someone help me with my radeon hd 3650 ,i'm using radeon driver and have black screen when startx,i'm able to use my system with vesa, xorg.conf : http://paste.pocoo.org/show/196074/ xorg.log : http://paste.pocoo.org/show/196073/
makkalot: any ideas ?
makkalot: is channel dead ?
Obscene_CNN: be patiant
tmkt: anyone get a chance to look at my pastebin and see if edid was my issue?
dk_mag: is there a bug filed for this [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object while running nexuiz? it seems like it appeared with latest mesa git
makkalot: (my machine shut-down suddenly) i have black screen with no actions i use radeon driver and xorg.conf is : http://paste.pocoo.org/show/196074/ xorg.log : http://paste.pocoo.org/show/196073/ any ideas ?
suokko_: makkalot: Does it work if you move the xorg.conf and try without it?
makkalot: (sorry for polluting channel again) i have black screen with no actions i use radeon driver and xorg.conf is : http://paste.pocoo.org/show/196074/ xorg.log : http://paste.pocoo.org/show/196073/ any ideas ?
makkalot: also when startx with no conf the same again
NightNord: Greetings again, everyone
NightNord: May someone told me, why wine's renderer detection may differ from glxinfo's?
DanaG: hmm, I turned off UEFI boot mode, again.
DanaG: NightNord: are you on a 64-bit Debian-based distro (such as ubuntu)?
NightNord: I'm on 64bit gentoo
DanaG: If so, the ia32-libs package may have an older version of Mesa.
DanaG: May still be true of Gentoo.
NightNord: Hm... Ah, I get it, probably you are right...
DanaG: Wine uses the 32-bit libraries, since it is 32-bit -- at least on my system.
mjr: it's 32-bit generically; well, there's a 64-bit version in development, but it'll only be for win64 stuff
NightNord: Right you are, I've checked that normal (non live) mesa also don't have 32bit and decide that that's ok =)
NightNord: But now I see, that there is emul libraries in place
NightNord: Now need to compile newer version somehow
laumonier: hi everyone i have a problem sometimes when im watching a video or when im playing a video games my screen becomes just line of color which appears and disapears fastly someone have the same problem and is there a way to fix it????
xming: NightNord: you need 32bit libs for wine, the way on gentoo to have that is by using the multilib portage
NightNord: xming: yes, yes... I'm already trying to automate eapi=2 use flags handling (don't want to have lib32 in global...)
NightNord: Ah, I've forgoten to say - 2.6.34 kernel has no more hardware cursor dissapear bug
Hazelesque: out of interest, I currently have a "Radeon Mobility X1400" in my laptop, and a "RV516 [Radeon X1300 Pro]" (according to lspci) in my work desktop... I'm thinking of grabbing a new ATI card for a box at home... would a HD4350 work OK with the radeon driver? I'd like to be able to play 1080p video on it
BioTube: Hazelesque: maybe even better at this point
BioTube: just remember to install the R700_rlc.bin firmware
Hazelesque: mostly all I do is use IRC, browse the web, and play video
BioTube: if you use KMS
Hazelesque: I don't think I'm using KMS - should I?
Hazelesque: (this will be for a different box, so I could upgrade the kernel/etc if I need to)
BioTube: it's the future, but the main drawback now is that you're capped at OGL 1.4
BioTube: and, of course, only one screen can use DRI at a time
Hazelesque: the HD4350s I'm looking at have one VGA connector, one DVI connector and one HDMI connector
[Enrico]: Hazelesque: well if you will use kms you will get the dynapic power management with kernel .34 (or .35 i don't remeber)
Hazelesque: can I use the HDMI output and the DVI-D output simultaneously?
BioTube: if you've got two users logged in, only one can have acceleration with UMS
Hazelesque: BioTube: oh, I see, only one X instance?
BioTube: Hazelesque: not quite. You can have as many as you need, but you need KMS to have acceleration on them all
[Enrico]: dyn PM is in .34 rc ok i found it
Hazelesque: I never run more than one X instance, although I may have more than one monitor attached, and enabled (dynamically, using xrandr(1)) at once
Hazelesque: I can still do the latter?
Hazelesque: just wanted to check, when you said "one screen"
BioTube: I only use one monitor
BioTube: the main difference with multimonitor is that KMS doesn't need a virtual line
Hazelesque: ah, because it can dynamically resize the framebuffer?
NightNord: [Enrico]: .34
NightNord: I'm using it, powersave is in place, at least dmesg says this
[Enrico]: NightNord: you are in late, just said it. but thanks :)
BioTube: Hazelesque: yes; in fact, a lot of the new features come from the ability to dynamicly manage VRAM
NightNord: Er, sorry, missed this =)
[Enrico]: NightNord: yeah i tested it too (both the one in drm-next and drm-radeon-testing.... a bit choppy with composite, but not bad at all)
NightNord: Strange thing - I've been unable to get any kind of accel working without KMS
[Enrico]: NightNord: you know my r200 has a similar problem. with UMS glxinfo reports dri working and xorg log too but when i try to play nexuiz i get less than 1fps i can't even move the mouse in the start screen. with kms i get a non-playable framerate, but well it works at least
[Enrico]: this on fedora 12
NightNord: Hm, for me UMS reports that dri is failing due to some strange error (don't remember what it was)
[Enrico]: mhm here i didn't understood if the problem was something that was forcing nexuiz to fallback to swr or something AGP related
[Enrico]: but i guess swr
NightNord: Let's just hope, that gallium will solve it all one day =)
MostAwesomeDude: We won't do Gallium for r200.
[Enrico]: and there is no reason to do it :)(
[Enrico]: :) *
Hazelesque: right, well, I added a HD 4350 to my order
Hazelesque: will let you guys know if it works! heh :)
Hazelesque: saves me having to nick the graphics card back out of my work box, heh
DanaG: argh, xorg must do something weird when unminimizing windows... because it seems to cause gpu hangs sometimes.
airlied: DanaG: try latest -ati
DanaG: I'm actually guessing it's that same "readback from video RAM" that plagues fglrx and sometimes nvidia. =þ
DanaG: ... because it also gpu-hung on resize sometimes, while resizing was giving a white border.
Vash63: Mmmm, new patches from git make kde composited feel faster.
Vash63: Not quite Nvidia binary fast, but pretty good.
Vash63: I like how youtube works like 10x better if I fullscreen Firefox and use the Cairo zoom
Vash63: Than if I just run it fullscreen.