Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-6-28

Search This Log:

yangman: well, look at that. I have KMS
yangman: didn't bring PANEL up in X at native res, though :\
airlied: yangman: its in native res outside X?
yangman: yes
airlied: got xorg log?
yangman: http://pastebin.ca/1477236
airlied: is your VGA connected?
yangman: looks like it is
yangman: it's on a dock. VGA's going to a 24" monitor
airlied: so it sees it and tries clone mode
yangman: forgot about that. it's actually sending a image to VGA
airlied: need to fix X to try RightOf by deafult
airlied: if it has a driver that can do resize
airlied: and the screen can fit inside max
yangman: there we go. native res :)
yangman: cool. I've been meaning to try KMS for months
yangman: doesn't seem to like being VT switched on suspend. came back with crazy signal timing
yangman: I'll poke around more later
glisse: nha: quick benchmark strongly encourage that we enable a fake texture in userspace otherwise i need to parse fragprog in kernel to catch kill instruction and checking cmd stream then takes way too much time
nha: k
airlied: so we need to enable TEX_0 for KIL to work?
nha: yup
nha: well, I think you can choose a different texture unit for the KIL, but it does need a texture unit enabled
nha: meh, my R400 is causing trouble with KMS :/
nha: how disappointing
nha: MostAwesomeDude: is gallium supposed to work on R300? At least glxgears?
nha: MostAwesomeDude: also, why is the gallium dri driver called radeon_dri.so instead of r300_dri.so?
OLiverr: hello, I have a radeon 9800 still some performance problems: it's enough for compiz and googlearth maybe but far behind the windows performance and games are not playable, so I read this: http://www.phoronix.com/forums/showthread.php?t=16554 compiled the stuff for myself (both the rewrite branch and the main branch, I think it's merged now?) but it doesn't seem to have any effect or it is even worse
OLiverr: and I think it is strange that it was compiled with the -g flag
adamk_: Well, until Gallium3d drivers are complete, it will likely remain noticeably slower than fglrx, particular with heavy duty games.
phoenix64: how will gallium btw help with this, and what is the problem anyways?
adamk_: Not sure I can give a really technical answer to that... As I understand it, though, gallium3d is much more focused on shaders, and will allow the open source drivers to support more extensions in hardware, thereby falling back to software less frequently.
nha: gallium doesn't help performance
nha: at least, not directly in any way
adamk_: It won't improve speed on extensions that are already supported, as I understand it, but it will allow more extensions to be supported in hardware.
OLiverr: well, the game (I need to play because I develop it) is a settlers 2 remake and therefore there is no special anything about it, no shaders or anything, I think it's mainly fillrate limited
nha: what it does is making driver writing easier, so that developers need not worry so much about stupid details and can instead focus on improving the actual code
adamk_: At least I think that's the explanation I've heard bridgeman give a few times.
nha: I don't think that's exactly true
nha: OLiverr: do they use OpenGL? If you're talking about widelands, I don't think it does (unless there have been some radical changes recently ;))
OLiverr: no, not widelands, it's an other project: www.siedler25.org and we use opengl indeed
nha: adamk: gallium isn't necessary to support more extensions, but it sure has the potential of making driver writing in general easier
nha: OLiverr: if it's fillrate limited, maybe you should keep your ears open for the texture and framebuffer tiling things that are hopefully coming
adamk_: Alright, so maybe I misunderstood... Or bridgeman misspoke :-)
nha: I don't think he did, but the difference is kind of subtle :)
nha: now GEM is the thing that doesn't directly affect performance but is absolutely necessary for many new extensions
adamk_: Hmm.. I'll have to give this game a try.
adamk_: Well, I know for a fact that he keeps saying that gallium3d will help narrow the performance gap between fglrx and the open source drivers. Not eliminate the gap, obviously, but shrink it down some.
adamk_: Hmmm... I wonder if I can find my disk somewhere.
OLiverr: hm maybe it isn't fillrate limited at all because I noticed there were no big fps changes after resizing the window .. :S but in glxgears I have very different values with different resoultions..
OLiverr: adamk, hehe, if you want to test it, you can download it for free http://www.abandonia.com/en/games/57/
adamk_: Hmmm... The game tells me "0 video drivers found! Video driver couldn't be loaded" and then it exits.
OLiverr: did you use the 0.6 binary?
adamk_: Nightly build.
adamk_: I can try 0.6, though.
NastX: did you follow the install instructions?^^
NastX: from the readme?^^
adamk_: Yep.
adamk_: It's installed to /opt/s25rttr. I have direct rendering, and I extracted the original game files to /opt/s25rttr/share/s25rttr/S2
adamk_: Hmmm.
adamk_: But, of course, I didn't start the game as it said.
adamk_: :-)
adamk_: Let's try that.
adamk_: It's downloading stuff, so this is a step in the right direction.
NastX: mhh use the sh (dont know where its located i am a win user because of some problems with the propitary ati drivers^^ and the slow rendering of the radeonhd driver (-;)
NastX: (-; )
adamk_: Yay... It started up.
adamk_: Well I haven't done much beyond start up a game, but performance seems pretty good here.
NastX: how much fps in the main menu?^^
adamk_: Hmmm..
adamk_: 34 FPS.
Airhardt: the main menu is a great benchmark :-p
adamk_: OLiverr, What were you getting?
NastX: but i ll get worse in the game =(
NastX: win xp - 710^^
adamk_: I did start up a game. I'm not in the main menu.
NastX: ati radeon hd 3650^^
NastX: ah ok
OLiverr: adamk, i have windows at the moment but in the main menu i had around 100
NastX: ingame 200
adamk_: In the main menu, I get 105, and in the game, I get 34.
OLiverr: yes, i had the same..
OLiverr: around 40 in the game
adamk_: So not great, but certainly playable, at least so far.
OLiverr: hmm not really
NastX: if you play longer the
NastX: fps will suck^^
OLiverr: but as i said we don't use any extraordinary stuff, only modulate2x on the map
OLiverr: and with the same card on windows i have ~700 in the main menu and 100-200 ingame
adamk_: Hmmm... I wonder if libbugle would point out any software fallbacks.
NastX: libbugle?^^
Airhardt: http://www.opengl.org/sdk/tools/BuGLe/
adamk_: http://www.opengl.org/sdk/tools/BuGLe/
adamk_: lol
NastX: ah thx^^
adamk_: Of course I've had a really hard time getting recent versions of it to compile on my systems.
OLiverr: didn't know that, sounds interesting..
testor: Hello, I just switched to the latest OpenSource-Ati-drivers on gentoo (X1200-card), however, it keeps activating HDMI-0, though I have a dvi-monitor - how can I force it to use DVI-0?
edt: testor maybe xrandr -d DVI-0
edt: I would bet you need an xrandr command of some sort
testor: edt: It says it can't open Display DVI-0, or any display at all...
testor: edt: And in Xorg.0.log it says it did not detect any connected devices, and then falls to HDMI, which seems to be default - But a DVI-Monitor is connected!
edt: what xorg version?
edt: and what driver?
nha: OLiverr: please don't look at glxgears for performance discussions
nha: glxgears is essentially *clear performance* bound on almost all hardware accelerated setups
nha: and clear performance tells you nothing about anything else
edt: nha define *clear performance*
testor: edt: Xorg-1.5.3-r7 and xf86-video-ati-6.12.2 ^^
nha: edt: the speed at which glClear() is executed (and glXSwapBuffers() plays a role as well)
edt: testor (gentoo?) I have the same driver level but have unmasked the 1.6 xorg
edt: though 1.5.3 did work here
testor: edt: And the hardware worked with an old fglrx before, so nothing is broken - and framebuffer-terminal is still perfect... Jep, gentoo, but not yet unmasked 1.6, it pulls in so many unstables...
testor: edt: Well, I guess, it would work, if the driver detected the display... Is there no option for xorg.conf to force detection / set default display?
edt: fglrx is trouble - I have had a __much__ more stable setup with the open source driver
edt: testor do not know. Have you tried an empty xorg.conf? it sometimes helps...
testor: edt: Yep, that's why I wanted to switch (and because of the dropped support and tearing video ^^)
testor: edt: Ok, will try ^^
edt: testor btw 1.6 has worked fine here (amd64 with >4G) - the worst part of it is unmasking...
testor: edt: Amd64 here too, but just 2 GB, but might be worth a try... Though I am not sure whether this is really xorg or radeon-related...
testor: edt: Still no success with clean xorg :(
adamk_: testor, Can you force DVI-0 on by using xrandr with --newmode and -addmode?
edt: 1.6 gives xrandr 1.3 which might change somethings
testor: adamk_: xrandr always says it cannot open display...
testor: edt: Okay, I will give it a try ;-)
edt: luck
adamk_: testor, Well, you can specify a modeline for DVI-0 in your xorg.conf file and then use the PreferredMode option.
adamk_: I'm not sure that will work in cases where Xorg doesn't detect a monitor, but if you haven't tried it, it's worth a shot.
testor: adamk_: Okay, as it will take some time to compile, I might try that now ^^.
edt: btw you can use: gtf
edt: to generate modelines
adamk_: Yep.
adamk_: Or cvt.
testor: edt: Wow ^^. Did not know there was an easy tool for that, thx :)
edt: two tools. I learn something new every day
adamk_: testor, Here's an example of what I suggested: http://pastebin.com/m3487a1ef
testor: adamk_: Thanks :)
testor: Well, that's odd - it says there is no monitor-section for HDMI (which is correct), and then uses the DVI-0-Section-Modeline for HDMI-Output -.-
testor: Do you happen to see anything here: http://pastebin.com/f1487b1a2 that I don't?
edt: testor is DVI-0 screen 1?
adamk_: Honestly, it looks like a driver bug to me.
testor: edt: Yes, I only configured one screen, and the monitor section says dvi-0...
edt: looks like hdmi is screen 0 and dvi is screen 1. maybe you need to tell X to use screen 1???
edt: see line 490-489 in your pastebin
testor: edt: Hmm.. Yes, it detects two ports, but in xorg, there is only one screen configured.. so how can I tell it to use screen 1?
testor: That is my xorg.conf: http://pastebin.com/f1487b1a2
testor: Ops, sorry... http://pastebin.com/f29fdbcc1 That is ^^
daum: hey guys - can someone help me get my montiors setup to do dual monitor where when you maximize a window it only goes to the size of one of the monitors?
nha: daum: that's a decision of your window manager
daum: nha, do you know how to do it in kde?
nha: daum: in other words, the radeon driver has nothing to do with it
daum: (I asked in kDE and they said it was in how you setup the xorg
nha: no, I haven't needed it so far
nha: ...
edt: testor I am not at all sure about this but what I would try is adding a second Monitor section, and in my device section I would add a 'screen 1' line.
nha: sorry about this back and forth, but as long as you have both monitors setup properly, can move windows from one monitor to the next and so on, you have configured xorg correctly
edt: a good read of man xorg.conf is probably needed to get it exactly right
nha: how far your windows are maximized is entirely up to the window manager, and I assume it defaults to maximizing to the current monitor's size because that's what the majority of people prefer
daum: well i'm using xrandr to make it span across both currently
nha: (as it happens, that's also how I prefer it, which is why I don't know how to change it)
edt: bbl
nha: oh wait
nha: I entirely misread your statement
nha: so when you maximize a window, it spans the *entire* desktop, across both monitors?
daum: yeah
daum: so its as if it made one super wide montor
daum: monitor
testor: edt: I tried it, did not work... as you can see in my xorg, I already tried a second monitor section before...
nha: I thought they'd fixed that in KDE :/
nha: but still, maximizing the windows is the window manager's job
daum: nha, that's why it hink it might be my Xorg? am i suppose to have to use xrandr to do it?
adamk_: Yeah, that sounds like a KDE issue to me.
daum: nha, it is also just spanning the entire menu bar across both screens, it literally believes that there is just one big monitor
nha: daum: xrandr should tell you that you have two monitors, and what their resolutions are and so on; beyond that, xrandr is not responsible
nha: daum: this sounds like multi-monitor support in KDE is sub-par
adamk_: daum, What's the output of 'xdpyinfo -ext XINERAMA | grep head'
adamk_: nha, It works as expected here :-)
daum: adamk, sec gotta get my old xorg config back was trying another one, one min
MostAwesomeDude: nha: Morninig.
MostAwesomeDude: *Morning, even.
nha: MostAwesomeDude: Afternoon ;)
MostAwesomeDude: Um, glxgears should work on r500 chipsets. r300 hasn't gotten much love, and the fragment shaders there are probably still broken.
MostAwesomeDude: For some reason, the clear color broke a couple weeks ago, and I haven't bisected it quite yet.
nha: k
nha: the trouble is, my r500-based card refuses to setup KMS correctly
MostAwesomeDude: Hm, that's strange.
nha: [ 13.087471] [drm:radeon_ring_test] *ERROR* radeon: ring test failed (sracth(0x15E4)=0xCAFEDEAD)
nha: [ 13.087477] [drm:r100_cp_init] *ERROR* radeon: cp isn't working (-22).
daum: adamk, hm it appears i don't have xdpyinfo installed sec
nha: this is with an uptodate kernel from linus' git
nha: it used to work on this very same hardware before things got merged
MostAwesomeDude: Yeah, r300 fragment shaders on gallium just haven't been fleshed out much. The emit for them is also not working.
nha: k
adamk_: daum, I really recommend installing it. That will tell us if it's a configuration issue with Xorg or a problem with KDE>
nha: well, I would like to get something at least somewhat working on both r300 / r500 and then work on the shader compiler
daum: yep installing
MostAwesomeDude: The stuff I committed yesterday was nearly all related to getting depth buffers unbroken; packed_depth_stencil broke stuff, and then I decided to track down and eliminate those stupid warnings for the unsupported textures.
daum: adamk, nothing in the output
adamk_: Sounds like you don't have the xinerama extension loaded then.
adamk_: daum, Pastebin your /var/log/Xorg.0.log file please.
nha: and I'm stuck with a completely unexpected and incredible kernel bug :/
daum: my xorg.conf is just bogus now the one i had tried is bad, now i 'm just getting duplicate monitors, pasting in sec
nha: right now, I'm compiling a kernel from just after the big commit that added KMS into the kernel master, and see if it's already broken there or if something later down the line caused the breakage
MostAwesomeDude: nha: Is this an IGP?
nha: nope, it's a discrete chip, via AGP
daum: http://pastebin.com/d13455316
nha: the R400-based card also fails, though slightly differently
nha: only the R300-based one works, although it also shows a somewhat "interesting" problem: https://bugs.freedesktop.org/show_bug.cgi?id=22519
adamk_: daum, Does 'xpdyinfo | grep -i xinerama' show anything?
daum: adamk, my xorg.conf is : http://pastebin.com/d128c559d
daum: adamk, the only output ont hat is XINERAMA
adamk_: Well that is good... But for some reason 'xdpyinfo -ext XINERAMA' isn't showing any head information.
adamk_: Pastebin the entire output of 'xpdyinfo -ext XINERAMA'
daum: http://pastebin.com/d1b4e9f8f
testor: edt: I just upgraded to xorg-server, but xrandr still says it can't open display... shall I do a revdep-rebuild?
adamk_: testor, Are you running xrandr from within X?
testor: adamk_: I can't, because it is on hdmi-0, and I can't see that ^^
adamk_: Try running 'DISPLAY=:0 xrandr'
MostAwesomeDude: nha: Those are just weird.
nha: ya think? ;)
testor: adamk_: Of course, you are right, works.. says HDMI is turned on, DVI off, and both disconnected.. will play around a bit ^^
daum: adamk, is my xorg conf right?
testor: adamk_ , edt: IT WORKS ^^. Just activated the mode for DVI-0, and though it still shows disconnected, output is there... Now I just have to find a way to xorg.conf it...
daum: http://pastebin.com/dc99e18d is my new xorg.conf it has it so that the desktop literally spans across both monitors(as in it thinks it is just on monitor) after i use xrandr -output DVI-0 --mode 1280x1024 --left-of DVI-1 --mode 1280x1024
testor: edt: Any idea how to do it? As soon as I do xrandr --output DVI-0 --mode "1440x900_60.00" it works, but xorg.conf just uses the modeline for the first port (hdmi-0)...
adamk_: daum, And windows still maximize across both monitors?
daum: yeah
daum: i found a poste saying i need to have kde and everything compiled with the xinerama use flag to get it so that it doesn't do that?
adamk_: daum, I'm not completely convinced that's the problem...
daum: so recompiling
adamk_: daum, Your xinerama extension isn't showing any information about your monitor placement.
daum: hm let me see after it recompiles with xinerama enabled what it outputs?
adamk_: Sure
daum: adamk, that did it!
adamk_: What did you recompile this time?
daum: kde and a couple other packages iwth xinerama enabled
adamk_: Very cool..
testor: Mhhhm... I now fixed my problem by calling some xrandr-commands when X ist started automatically.. did not get it to work with xorg.conf :(
adamk_: testor, You should search for a bug report on freedesktop for that problem and, if you can't find one, open one up.
testor: adamk_: Yep, will do, as soon as I have some more time ^^. too much work atm, but at least, I can work a bit now ^^
testor: adamk_, edt: Thanks for everything up to now :) Have a nice sunday!
testor: bye all ^^
okias: Hello, is already avaliable any testing version of KMS on R600 (exactly HD 2400)?
nanonyme: If you mean one that you could use X with, not afaik...
muep: What's the difference between the framebuffer stuff that has been in Linux for ages, and KMS?
nanonyme: Well, one of the differences anyway is that X and fbcon (framebuffer) are both just clients on KMS, I guess? :)
nanonyme: Before you had modesetting both in X and framebuffer consoles, now it's centralized and they cooperate nicely.
masa-: darn.. what am i missing this time..
mjr: and kms should play nice with the kernel graphics memory management...
masa-: hw 3D is not working with X600 and xf86-video-ati
masa-: i have 2.6.30-gentoo-r1 with drm as module, xf86-video-ati-6.12.2 from testing and xorg-server 1.5.3-r6
muep: ok, thanks
muep: masa-: I think 3D acceleration for >= R600 isn't yet available
masa-: X600 is RC380...
masa-: RV380 even
okias: agd5f: Hey, is already avaliable KMS for r600?
masa-: 19:21 < okias> Hello, is already avaliable any testing version of KMS on R600 (exactly HD 2400)?
masa-: 19:32 < nanonyme> If you mean one that you could use X with, not afaik...
agd5f: okias: not yet
okias: okey, thanks
muep: oops, confused R600 vs. X600 :-p
nanonyme: Happens.
Neo_The_Xserver: R600 is better than X600
nanonyme: If you don't count the driver support. ,9
nanonyme: s/,9/;)/
Airhardt: and the energy consumption
Neo_The_Xserver: im all about the preformance of the card itself
Neo_The_Xserver: *per
muep: R600 doesn't necessarily take more power
Neo_The_Xserver: :)
muep: there are low end models of them, too
Neo_The_Xserver: muep: sure are
Airhardt: ah, right
Neo_The_Xserver: my dad has a low end R600 card
muep: I have a mid-range HD3650
muep: passively cooled
Neo_The_Xserver: dang!
Airhardt: i always associate r600 with that HD 2900 monster :-)
Neo_The_Xserver: 2900 monster?
Neo_The_Xserver: 2900 is that good?
muep: wouldn't have even wanted to have the 2900
Neo_The_Xserver: heh! the 4770 is what im gettin
nanonyme: Well, not counting driver support has classically been the way to waste lots of money on hardware useless in the project timeframe. *shrug*
Airhardt: Neo_The_Xserver: yeah, it's that good - for your electric power company :-p
Neo_The_Xserver: Airhardt: hahaha! i always thought those power saving cards were junk]
nanonyme: Playin_PS2_games: For God's sake, use away messages for that like decent people.
_Groo_: agd5f: hi alex
_Groo_: airlied: greetings
_Groo_: hi/2 all
Playin_PS2_games: nanonyme: i like registering names
agd5f: _Groo_: hi
_Groo_: agd5f: google earth isnt working anymore :D probably because my card drivers are screwed up...
_Groo_: agd5f: it crashes after a few seconds
_Groo_: agd5f: with
_Groo_: [ 2873.287179] [drm:r300_packet0_check] *ERROR* Invalid texture format 17
_Groo_: [ 2873.287185] [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !
agd5f: _Groo_: probably need new mesa bits
_Groo_: agd5f: and ddx...
_Groo_: agd5f: ddx kms branch is broken for rs485 (cute little squared green and white boxes)
_Groo_: agd5f: and since yesterday, mesa breaks kwin 3d and a lot of working 3d games
_Groo_: probably is related
agd5f: _Groo_: it you updated your drm, you'll probably have to update your ddx and mesa
_Groo_: agd5f: i cant.. rs485 doesnt work with latest mesa/ddx
_Groo_: agd5f: its broken
_Groo_: agd5f: im using latest drm, latest kms, mesa from friday and your ddx tree
agd5f: _Groo_: my ddx tree?
_Groo_: agd5f: sorry its kms-branch but an older one before the rs485 break
_Groo_: agd5f: from 19 jun
agd5f: _Groo_: you probably need to use an older drm then
_Groo_: agd5f: im waiting for the rs485 fixes...
masa-: when trying to run glxgears i get:
masa-: Error: couldn't get an RGB, Double-buffered visual
agd5f: _Groo_: I've been on vacation for the last couple weeks, so I'm still catching up on what'e been happening
masa-: what is missing..?
nanonyme: agd5f: Had a nice vacation?
_Groo_: agd5f: i know, relax, airlied knows my grieves..
agd5f: nanonyme: yes. I went to visit some friends in au
_Groo_: agd5f: im probably the most annoying user of this channel
nanonyme: _Groo_: I'm sure developers won't consider you annoying just for telling you have issues as long as you do your best to produce enough useful information on what caused it so it can be fixed. ;)
_Groo_: nanonyme: i fill bugs, i pinpoint the troublesome commits to the best of my klowledge... but since rs485 is so broken as of late, i cant even keep up with the development anymore
agd5f: _Groo_: no problem. just don't really have context for most stuff at the moment
_Groo_: agd5f: yeah like i said, i just need to be patient and wait for you guys start slapping out the bugs for 485
daum: hm anyone know why it seems like my mouse cursor at some poitns bugs and disappears(or gets really distored)
agd5f: daum: should be fixed in git master and 6.12-branch
_Groo_: ouch,drm/kms just crashed.. fortunatelly linux was able to proceed
_Groo_: [ 5257.101557] [drm:radeon_cs_ioctl] *ERROR* Failed to initialize parser !
_Groo_: [ 5308.041719] Xorg: page allocation failure. order:4, mode:0xc0d0
_Groo_: [ 5308.041742] Pid: 3883, comm: Xorg Not tainted 2.6.29-custom #1
_Groo_: wish he could help writing the radeon drivers.. but its skills are useless for driver development
Veerappan: agd5f: Would you happen to be non-AFK?
Zajec: _Groo_: increase them :)
Zajec: _Groo_: start with something really ease, read code, hack :)
Zajec: don't start digging into 3D for sure ;)
_Groo_: Zajec: i might do just that
agd5f: Veerappan: I'm back from vacation. just cathing up on stuff now
Veerappan: yes, I realize that. Hope you had a fun time
Veerappan: There was a patch you had posted on phoronix forums a while back that fixed Xv for my laptop's x300 so that it wasn't just showing the upper-left quadrant of the video (something to do with an offset on r370 or something)..
Veerappan: Do you happen to know if a similar fix will be making it into the radeon git tree sometime, or were there technical issues with it?
agd5f: Veerappan: http://www.botchco.com/alex/xorg/fix_xv.diff
Veerappan: I've been applying the patch by hand on my laptop, and it still does the trick, but it's probably something to fix before a major release
agd5f: I need to sync up with dave and jerome, but I'll commit it to the kms branch soon
Veerappan: k. Just wanted to make sure it was still on the radar, and it hadn't been forgotten about.
agd5f: Veerappan: don't worry, there won't be a release without a fix
agd5f: the kms code really need to to be synced with master. it's pretty out of date for a lot of stuff
Veerappan: I'm not too worried, just wanted to check. If it was released that way (which I doubted you would do), I'd just keep applying the fix manually.
Veerappan: You're a much better judge of that than I... I just spend my time lurking in the background (at least until I learn the code well enough).
Veerappan: Only other issue I've got currently is suspend/resume in Ubuntu 9.04, but I'm still trying to figure out if that's a distribution issue (re-posting the video bios or something on resume, maybe) before I dig too deep.
Veerappan: I've found documentation in the past suggesting it's an Ubuntu problem.
daum: hm my second screen just got a ton of green color added to it randomly any ideas why?
airlied: agd5f: plans for syncing kms/master, but cleaner than git pull :)
airlied: I've started adding some macros to the kms tree that make the interfaces from the accel code cleaner
airlied: and I want to collapse the 4-5 conditionals into one if I can.
soreau: wb agd5f. I took the regdumps I got from fglrx and plugged the obvious values into radeon_tv.c and 1024x768 tvout works, but there are still timing issues. I was wondering if there was more to the regdump I don't know about or how to get the ntsc/pal timing values. I can post any information, but still want to get this working correctly
darksun_: not necessarly a driver question, but im looking for a new vid card for an older pc i have here, and was wondering if anyone knew the diffrences between a fireGL 9700 and the radeon 9700?
m03sizlak: well i just installed kernel -198 and xorg-x11-drv-ati-19 and now running fusion icon kills X instantl
soreau: agd5f: and by working, I mean it is in color but flickering and scrolling rapidly with some portions of the screen.. well http://www.speedyshare.com/839752753.html
agd5f: soreau: can you pastebin the patch you are using?
soreau: how should I create the patch, with git?
agd5f: git diff
soreau: agd5f: http://pastebin.com/m1c4f57ce
agd5f: soreau: you'll have to change the crtc pll settings as well
soreau: agd5f: And I need to know the code for that. I was tinkering with them blindly and no luck. Would you like to see the regdump?
agd5f: soreau: sure. email it if you could
soreau: I assume it's agd5f@freedesktop.org or..(?)
agd5f: soreau: that should work too, but alexdeucher AT gmail
soreau: Ok, thanks
Zambezi: I have a Radeon HD2400 Pro 256 (or 128) MB PCI. But today I saw same card with 256 MB, but with AGP 8x. I assume the performance will be slightly better, but do you have any ideas?
agd5f: Zambezi: AGP is asking for trouble
Zambezi: agd5f: How come?
agd5f: Zambezi: it's a nightmare hardware-wise
agd5f: you're in for driver problems
Zambezi: agd5f: Well, good I said earlier today I wouldn't swap it until tomorrow cause I wouldn't go to bed a problem is solved...
Zambezi: agd5f: It's 02.30 here now, but I said that about an hour ago. Too warm to sleep now.
virtuald: Results 1 - 10 of about 67,100 for agp nightmare. (0.12 seconds)
virtuald: o.o
daum: hey guys does anyone have any ideas why it seems that the color on both my montiros are off?
bridgman: for everything, or just video, or...
agd5f: bridgman: how goes? when are you heading out?
bridgman: looks like I'm ont; one by one people dropped out
bridgman: and I really should be working on the old house, so...
bridgman: s/ont/not/
agd5f: bridgman: bummer. sounded like a good trip
bridgman: guess I'm going to "veranda beach" this year ;)
bridgman: sounds like you made it home ok ?
bridgman: yeah, it would have been fun
agd5f: yup. flight home was a bit rough, but good trip otherwise
bridgman: then again I think I already have a years supply of insect bites
bridgman: it's long, isn't it
bridgman: or rough like bumpy ?
agd5f: long and tiring
daum: bridgman, everything well one monitor is really green the other it seems like certain apps jsut have messed up colros
bridgman: vga or dvi connection ?
agd5f: trans-pacific part was ok, but the last leg from SFO to IAD was brutal. seemed longer than the trans pacific leg
agd5f: daum: what card?
daum: one is VGA one DVI, in firefox grays are rendering like a light blue. x1550 pcie x1
soreau: agd5f: You couldn't sleep at all?
agd5f: soreau: not on that leg. I was on the aisle and it seemed like people went out of there way to bump me and wake me up as soon as I fell asleep
mjg59: agd5f: Yeah, transcon is almost always more miserable than transoceanic IME
soreau: agd5f: I've never ridden for very long trips, but I never would sleep unless it was an unusually long trip. Even then, I guess a layover would scratch that plan mostly to pieces
agd5f: soreau: I slept ok on the way over. I slept about half of the flight over (~14 hours)
yangman: daum: check your calbe connection for VGA first. having a couple of the pins not connected properly can cause a single channel to dissapear
daum: yangman, you are very smart
soreau: sure to grab beer, eat, fly! :)
soreau: agd5f: Glad your travels went well
agd5f: soreau: thanks
daum: yangman, any ideas on tis one: http://img194.imageshack.us/img194/2262/testscreene.jpg try to visit the actual url yourself you'll notice that the blue is actually a gray
agd5f: daum: looks fine in the screen shot
yangman: daum: gamma curve might be messed up. try running xgamma -gamma 1.0
daum: hah yeah just looked at the screenshot on my laptop it does look fine
daum: yangman, hm no luck
yangman: well, the grey looks tinted blue to me, so it's probably normal
spstarr: agd5f: welcome back from vacation, hope you had a good time
agd5f: spstarr: thanks!
daum: hm i got it much closer to the actual color, had to set my green to 0 and blue and red on monitor to 100
spstarr: :)
spstarr: hello bridgman
bridgman: hi spstarr
spstarr: hey
King_InuYasha: hey bridgman
bridgman: howdy\
King_InuYasha: right now, im doing one of the most dangerous things you can do to a Linux machine
King_InuYasha: Live upgrading the distribution
bridgman: Intrepid to Jaunty worked for me
bridgman: no blood
bridgman: fglrx or radeon/drm/mesa ?
yangman: I do that all the time
yangman: :p
King_InuYasha: well, Intrepid to Jaunty basically killed graphics on Skuld
King_InuYasha: fglrx
bridgman: yeah, not sure how well dkms is working there
King_InuYasha: i dont think it is...
bridgman: in theory the kernel driver automatically gets rebuilt
King_InuYasha: in practice, that is generally hit or miss
bridgman: yeah, that's what I've seen
King_InuYasha: unfortunately, in my experience, its more miss than hit
King_InuYasha: somehow Ubuntu botched dkms
bridgman: there's a lot to be said for uninstalling any driver that doesn't come with the new kernel
bridgman: before you upgrade
bridgman: that's what I do
King_InuYasha: that's what I'm doing now
King_InuYasha: im also removing proprietary software packages that didn't come in a repository
King_InuYasha: since the repo manager screams FAIL every time I try to update when they are installed
King_InuYasha: so, opera is gone
King_InuYasha: and so is skype and flash plugin
bridgman: yep, sounds right; I don't know if any package managers have a "watch what I do and pretend it was a package" option
King_InuYasha: i wish they did
bridgman: but that would be mighty handy
King_InuYasha: yum is one of the few repo managers that lets you install and remove locally installed packages
King_InuYasha: its one thing I loved over APT
King_InuYasha: being able to do solid dependency resolution for a local rpm over the command line is really useful
King_InuYasha: it puzzles me how GDebi can do it, and APT itself cannot
King_InuYasha: the one thing that worries me is how Radeon performance is going to be after this is done on Yggdrasil, my main machine
King_InuYasha: it has a Radeon X1300 (I know, quite old)
bridgman: which distro are you moving to ?
King_InuYasha: Fedora
King_InuYasha: I have long since been sick of Ubuntu
King_InuYasha: I still can't get 1024x768 on S-video on a Radeon 9200SE, which is arguably one of the best supported ATI cards by FOSS drivers
King_InuYasha: the only reason I install Ubuntu nowadays is for people that can't quite manage their machines well
bridgman: you're talking about 1024x768 scaled down somehow ?
bridgman: the video standard doesn't allow that kind of res otherwise
King_InuYasha: well, my TV supports it
airlied: no it doesn't.
airlied: the card scales it down to NTSC
bridgman: it might support it through another connection
King_InuYasha: probably
mjg59: Though if a TV supports 1024x768 natively, you're probably in the wrong aspect ratio
King_InuYasha: it has an HDMI, Component Video, S-Video, and Composite video inputs
mjr: you probably want the hdmi
bridgman: yep
mjr: though your 9200se can't output it, of course...
mjg59: It'll probably lie to you via EDID, though
King_InuYasha: my dad is too cheap to buy an HDMI cable
mjg59: mjr: HDMI->DVI is an option
mjr: mjg59, yes, of course, though I (perhaps hastedly) assumed 9200se not to have that either
King_InuYasha: and I'm trying to save money to get Windows 7 and a few upgrades for my computer by the time I have to head to college
mjg59: Most 720p TVs will claim to be 1360x765 or something
mjg59: Because EDID sadness
King_InuYasha: i believe thats what my TV's highest res is
King_InuYasha: something like that
King_InuYasha: oddly enough
mjg59: It'll almost certainly actually be 1366x768
bridgman: yep, those are the standard lcd panels
King_InuYasha: I have a smaller one that has a VGA input and supports "real" 16:10 resolutions like 1440x900
mjg59: But you can't express 1366x768 in the normal EDID parameters
bridgman: or is that only for projection ?
mjr: mjg59, ...
mjr: "okay"
mjg59: Horizontal has to be a multiple of 8, vertical has to be an aspect ratio
mjr: edid sadness indeed
mjg59: There's a hack in X to turn 1360x765 into 1366x768
King_InuYasha: when I upgraded that machine that is connected to to TV from Gutsy to Intrepid, graphics was automatically switched to vesa
King_InuYasha: from radeon
King_InuYasha: the resolution didn't change, but OpenGL support suffered a lot
mjg59: Which will doubtless blow up when we find someone who actually has a panel that's sufficiently fucked up to *actually* be 1360x765
King_InuYasha: so the computer got real slow
King_InuYasha: mjg59, I have a friend that does have a TV at that res
mjg59: No you don't
King_InuYasha: its an older projection style TV, but it outputs to it
King_InuYasha: it doesnt work at 1366x768
King_InuYasha: but at 1360x765 it works
mjg59: You need to define "work" very carefully here
bridgman: and that's over S-video ?
King_InuYasha: well, it doesn't look like stretched taffy with strange lines running through it
King_InuYasha: yep
King_InuYasha: that was a messed up TV
mjg59: Well, you're already doing it wrong
King_InuYasha: of course, when component video was used, the lines werent there, but it still looked like taffy
mjg59: Projection TVs are unlikely to have "native" resolutions
mjg59: But given that you're not going to get EDID over component or s-video, it doesn't really matter
mjg59: We won't do any fixup
|niko|: I have an r500 card with the latest stable radeon driver in debian sid
|niko|: and I just installed KDE4 but compositiing and video playback cannot work together
|niko|: Is there something I can do to VLC to make it play back my videos while using a compositing window manager?
airlied: make it use texture video adapter
|niko|: Or does the radeon driver not yet support compositing + video playback for r500 cards?
|niko|: Whoops, sorry bout the caps
airlied: it needs to use a different xv port
airlied: I'm not sure how kde/vlc does it
|niko|: Is this something that needs to be fixed in radeon driver or in VLC?
|niko|: airlied: So in the different ouput module choices in VLC's config I should be using Xvideo?
airlied: it should offer xvideo port choices
airlied: you'd need to talk to vlc ppl
|niko|: XZvideo adapter number?
airlied: yes
|niko|: It is set to -1 now
|niko|: What should it be?
airlied: run xvinfo
airlied: and see what Texture Video is
|niko|: 0
|niko|: Adaptor #0: "Radeon Textured Video"
airlied: oh wait r500 doesn't have overlay
airlied: no idea whats going on then
airlied: sounds like KDE doing it wrong
|niko|: yeah it did not work
|niko|: hrmm, so radeon driver is fully capable of playing video and compositing window managers at this point?
|niko|: If so, I'll big #kde
|niko|: *bug
airlied: make sure you aren't using a GL backend for VLC or something
airlied: as long as its using Xv it should work fine
|niko|: airlied: XVideo is selected, but the checbox "Accelerated video output" is checked. Should I uncheck that? Can radeon support that?
airlied: no idea what that does
|niko|: airlied: Any idea why I don't have an /etc/X11/xorg.conf?
|niko|: Or how I can generate one that is friendly with the radeon driver?
King_InuYasha: |niko|, most distros no longer require it
|niko|: Whoa, that's a weird thought.
King_InuYasha: |niko| in fact, Ubuntu purged theirs I believe
|niko|: I'm running Debian
King_InuYasha: you can still generate one though
|niko|: unstable
|niko|: Looks like it is purged here
King_InuYasha: Debian unstable is the repository tha Ubuntu is based of off, so yes it would be
|niko|: #kde told me to add
|niko|: Option "AccelMethod" "EXA"
|niko|: To fix my issue
King_InuYasha: you will have to generate one and write it up from scratch then
King_InuYasha: I believe there is a dpkg-reconfigure or something like that
crdlb: EXA is likely the default; you can grep /var/log/Xorg.0.log to be sure
King_InuYasha: that will let you generate a basic xorg.conf that you can use
|niko|: (II) RADEON(0): XAA Render acceleration unsupported on Radeon 9500/9700 and newer. Please use EXA instead.
|niko|: crdlb: It looks like... EXA is not default?
crdlb: apparently not for you :)
|niko|: Where in Debian would that be set?
|niko|: Somewhere within X11
|niko|: You set EXA as default...
|niko|: I mean if I were to file a bug report on which package would I do it? (in order to make EXA default)
|niko|: If Debian/Ubuntu has purged xorg.connf... how can I configure two X11 screens for each of my two DVI ports?
mjg59: You can write an xorg.conf
|niko|: mjg59: I'm doing that now...
|niko|: But I mean, how are "normal" users expected to configure that kind of stuff?
|niko|: If they don't even have the xorg.conf to begin with?
|niko|: It seems to me that xorg.conf is not yet obsolete (although ideally it should be)
|niko|: Also... should I have Option "AddARGBGLXVisuals" "On" or get rid of it?
crdlb: that's an nvidia option
muep: the ideal is probably to configure with xrandr
|niko|: And should I have a DRI section with a line Mode 0666 ?
|niko|: Is that an nvidia option as well?
crdlb: |niko|: on ubuntu, you can run 'sudo dexconf' to get a tiny xorg.conf (only Device, Monitor, and Screen, all empty)
|niko|: crdlb: Handy info, thanks.
crdlb: from the 'd', I'm guessing it's on debian too
|niko|: Running dexconf as root didn't generate it though on Debian
|niko|: It's okay... I made my own.
crdlb: the dri mode won't hurt, but you don't need it anymore
|niko|: okay good.
crdlb: if you have 3d acceleration, you are already set there
|niko|: Okay, going to reboot to test EXA
|niko|: Whoa
|niko|: Now I can;'t even turn on compositing
|niko|: and Xorg is taking 99% of my CPU
|niko|: every is extremely sluggish
muep: without compositing?
|niko|: yup
|niko|: COmpositing won't even turn on
|niko|: It's like my CPU is rendering the graphics or something
muep: on my computers, kde 4 is perfectly usable even with the vesa driver
|niko|: I don't think this is a KDE issue
muep: me neither
|niko|: I used to have no xorg.conf driver
|niko|: err, file
|niko|: Now I ceated one and it feels like direct rendering is disabled or something
muep: you probably misconfigured something, or it's just a Sid issue
soreau: |niko|: What does 'glxinfo|grep renderer' say?
|niko|: soreau: one moment, install mesa-utils
|niko|: soreau:
|niko|: direct rendering: Yes
|niko|: OpenGL renderer string: Software Rasterizer
soreau: Your assumptions are correct
soreau: Can you pastebin your current X log?
|niko|: not my xorg.conf?
|niko|: sure
tpocra: Bah
tpocra: It's unsably slow
tpocra: Had to connect via my laptop
tpocra: working on pastebin from here...
tpocra: soreau:
tpocra: http://pastebin.ca/1478012
tpocra: is my log
tpocra: and my xorg.conf is: http://pastebin.ca/1478014
soreau: (WW) RADEON(0): Direct rendering disabled
tpocra: soreau: How should I enable it?
soreau: The lines preceding that might have something to do with it but not sure
tpocra: Did I do something wrong?
tpocra: (in my xorg.conf I mean)
soreau: tpocra: You're |niko|, yes?
tpocra: yes
soreau: You shouldn't even need x conf for x 1.6.x
tpocra: soreau: but EXA is disabled
tpocra: or rather it is not default or something
tpocra: So I needed to add Option "AccelMethod" "EXA"
soreau: It's probably not default yet. (it needs to be or will be eventually though)
tpocra: So in the meantime I would like to 1) set EXA to default in the xorg.conf without losing direct rendering
tpocra: and 2) know the proper Debian package to file a bug report on to set EXa as default
ghoti: looks around
tpocra: soreau: #1 is more important... do I need to add or remove something from xorg.conf to get my direct rendering back?
tpocra: Like, should I be loading some of the modules I commented out?
tpocra: maybe dri or glx..
yangman: tpocra: EXA is already the default in git. it'll be part of the next release if not already
tpocra: yangman: next release of what?
yangman: tpocra: radeon
tpocra: Oh, okay, so this is something in the radeon driver...
tpocra: maybe if I got a newer radeon driver package this would be resolved...
tpocra: There aren't per chance any precompiled amd64 snapshots in deb format... are there?
soreau: tpocra: You could use a very simple config like this if none at all is using xaa for some reason http://pastebin.com/m2cb01bc0
tpocra: Okay thanks, rebooting
tpocra: soreau: still software rasterizer
soreau: It's something to do with drm 'closing' might be an indication your kernel is incompatible
tpocra: soreau: I am running the 2.6.30 kernel in sid
soreau: good luck with that, it looks like a drm kernel module issue
tpocra: soreau: Why would it work without an xorg.conf at alll, though?
tpocra: minus the EXA bit
soreau: Because it might be using xaa ? not really sure
tpocra: oh no
tpocra: I deleted the xorg.conf and direct rendering is still disabled
tpocra: performance is much, much better though
tpocra: Xorg process is no longer chewing up my CPU
tpocra: I don't think I've changed anything else besides generating xorg.conf
tpocra: Oh, hmm, I switched kernels
tpocra: That is something =P
tpocra: 2.6.26 gives me direct rendering again... I'm going to try using 2.6.26 with EXA in a minimal xorg.conf now
tpocra: oohh it works
tpocra: EXA + 2.6.26
tpocra: okay so 2 issues i found were:
tpocra: #1: EXA is not default in Debian's radeon driver, which is annoying because an xorg.conf must be generated where there was none before
tpocra: #2: Debian's current 2.6.30 kernel breaks radeon's direct rendering somehow.
tpocra: Are these radeon issues or Debian issues?
yangman: #1 is radeon. it's already "fixed" in git
yangman: unsure about #2