twiztid: anyone have any suggestion as to what the best driver for a 512mb AGP X1300 card would be?
Pallokala: radeon
Pallokala: i.e. xf86-video-ati or xf86-video-amd (should be the same)
MostAwesomeDude: No.
MostAwesomeDude: -amd is for Geodes.
MostAwesomeDude: -ati is right.
twiztid: Pallokala: ok, cool, my original delema is KMS + COMPIZ/GNOME = chop chop chop... so rather than disable KMS (which I need for GLXCairoDock to run smoothly), im using indirect-rendering, but the 'visual effects' in my apperance tab is greyed out and unselectable and no compiz effects seem to be applied
twiztid: ...d/l'ing and installing xf86-video-ati
Pallokala: which distribution you are using?
twiztid: Lucid
Pallokala: MostAwesomeDude: thanks for clarifying that. I had been wondering why the different names.
Pallokala: then you ought to have the best option from the start.
MostAwesomeDude: twiztid: Okay, Lucid offers the new drivers already.
Pallokala: does glxinfo report "Direct Rendering: Yes"?
MostAwesomeDude: So, if stuff already works, Don't Fuck With It. :3
edwin: drivers/gpu/drm/radeon/radeon_pm.c:430: warning: ignoring return value of ‘device_create_file’, declared with attribute warn_unused_result
Pallokala: twiztid: did you end up with lucid by upgrading your earlier ubuntu?
MostAwesomeDude: edwin: Patches welcome. It's not a big deal.
edwin: k
twiztid: thats what i thought too, but for some reason KMS and directrendering is unuseable, so rather than lose kms and cairodock ... disabling directrendering fixs the choppyness, but either way i dont have any effects
twiztid: its a fresh install of Ubuntu Lucid Lynx 10.04
MostAwesomeDude: Wait, it's unusable because of choppiness?
MostAwesomeDude: As in, horizontal tearing? Or as in, lag?
twiztid: lag, bad bad bad lag
MostAwesomeDude: You're using Compiz?
twiztid: Pallokala: http://pastebin.com/qJRPvABd ---direct rendering: yes---
twiztid: MostAwesomeDude: yes
Pallokala: that one seems correct.
MostAwesomeDude: twiztid: What kinds of numbers do you get with the Benchmark plugin? Is the lag constant, or only when doing certain tasks? Do you have certain plugins enabled that are known to be slow?
twiztid: MostAwesomeDude: no compiz effects are working... the lag is constant unless i disable KMS or use 'compiz --replace --indirect-rendering'
MostAwesomeDude: twiztid: Does the Benchmark plugin work?
twiztid: MostAwesomeDude: no, no compiz effects are working...
MostAwesomeDude: twiztid: Is Compiz actually running? :T
MostAwesomeDude: Your drivers appear to be all set up, you should go ask #compiz.
twiztid: MostAwesomeDude: and my visual effects options are greyed out, ...it doesnt seem like it is... like wobbly windows, cube, etc...
MostAwesomeDude: #compiz. :3
twiztid: MostAwesomeDude: k, thx for lookin into, that... ya im in compiz too... lol so i already have xf86-video-ati? cause this is a fresh install
MostAwesomeDude: Yes, you do.
MostAwesomeDude: Open source drivers work out of the box, imagine that.
twiztid: glxinfo reports this... http://pastebin.com/qJRPvABd
twiztid: bah, sry wrong tab... lolz...
benh: fixes glisse bugs
benh: glisse: jerome, ton code est pourri :-)
twiztid: im still unable to turn on the visual effects for compiz, what driver should i install?
benh: airlied: I suspect it's broken on non-cache coherent archs :-)
benh: at least it figures out things don't work and fallbacks to a dumb framebuffer
benh: which works
benh: so it's not completely hopeless :-)
twiztid: could someone recommend a driver to use for my ati x1300? compiz wont use the effects ive applied and the visual effects options are all greyed out...
MJEvans: http://en.wikipedia.org/wiki/Comparison_of_ATI_Graphics_Processing_Units#Radeon_R500_.28X1xxx.29_series
MJEvans: http://dri.freedesktop.org/wiki/Radeon
twiztid: MJEvans: wow! awesome! ive been googling forever, and this is greater than anything ive found... thx alot!
MostAwesomeDude: twiztid: You're already using that driver.
MJEvans: twiztid: ... it's not as awesome as you think
MostAwesomeDude: I'm not exactly sure what your malfunction is.
MJEvans: http://xorg.freedesktop.org/wiki/RadeonFeature << There we go, you have an R500 family graphics chip
MJEvans: The good news is if you use bleeding edge drivers you //might// get the features you want working with open-source code. The bad news is there's a __reason__ it's called bleeding edge.
twiztid: MostAwesomeDude: i know im so confused, this hasnt happend before... compiz normally works great without KMS ~or~ 'compiz --replace --indirect-rendering' but this time around, my visual effects are greyed out and so consequently compiz isnt utilizing any of the effects i have set
MJEvans: http://xorg.freedesktop.org/wiki/ATIProprietaryDriver
twiztid: MJEvans: funny you suggest that, i loaded and installed the bleeding edge xorg-edgers ppa, and NOW openGLcairo dock has black background (so i turned emulate transparency for it...) and the visual effects options are still greyed out...
MJEvans: Hum... does it require AA support?
MJEvans: I don't really care about compiz, I just want 2d and some video-decoding help.
twiztid: MJEvans: i believe i can disable AA, ive had compiz and glxCairo-dock running together on fresh install of Xubuntu, after commanding 'compiz --replace --indirect-rendering'; so to use the same workaround in Ubuntu, soreau suggested, 'mv /usr/bin/compiz /usr/bin/compiz.real {then make new file in /usr/bin/compiz with '#!/bin/bash [return, return] /usr/bin/compiz.real --indirect-rendering...
twiztid: ..."$@"' so it doesnt lag so bad with KMS/compiz..., but so after all that, compiz is like, "F&*K OFF!" >:(
MJEvans: http://wiki.compiz.org/Hardware/ATI
MJEvans: http://wiki.compiz.org/ATI%20with%20AIGLX
twiztid: ...thx im readin up...
MJEvans: http://www.google.com/search?q=compiz+ati&ie=UTF-8&oe=UTF-8 << more stuff if that doesn't help
twiztid: MJEvans: so should i use ati catalyst 9.3? or setup my ati with AIGLX?
MJEvans: twiztid: You're slightly closer to AIGLX, you may as well see what that does, as long as nothing else disqualifies you. Also make sure to google your symptoms in addition to my base search.
MJEvans: twiztid: then, of course, try other things if it still doesn't work.
adamk: twiztid, Pastebin your /var/log/Xorg.0.log file.
twiztid: MJEvans: ok cool ya i think ill try AIGLX; adamk: http://pastebin.com/mY8NN1W8
adamk: Well that's fine. What's the output of 'glxinfo | grep -i render'?
twiztid: adamk: http://pastebin.com/i2AwQBs1
adamk: So your drivers are fine.
adamk: Try starting compiz with 'compiz &' in a terminal and pastebin the results.
twiztid: adamk: hrm for some reason my visual effects options are greyed out, k im on that...
twiztid: adamk: http://pastebin.com/WXfP20qU
adamk: A) Never run it with sudo.
adamk: B) Your drivers are fine but your compiz installation is broken. I suggest going to #ubuntu.
adamk: Or, perhaps #compiz.
twiztid: adamk: ok, could it be because i did this? --> http://pastebin.com/zYbSGxBh
twiztid: i did this, as suggested by soreau, to allow indirect-rendering upon startup because for some reason i either HAVE to disable KMS ~or~ do compiz --replace --indirect-rendering to fix choppy window movement and slow gnome/compiz performance
adamk_: I'm not sure why he would have suggested that. It will simply break things in Lucid.
adamk_: So, yes, your problems are probably from doing that.
adamk_: Oh wait.
adamk_: Meh.
adamk_: If you had made the /usr/bin/compiz file executable, it should work just fine.
twiztid: adamk: how would i make in executable?
twiztid: it*
adamk_: Run 'sudo chmod a+x /usr/bin/compiz'
adamk_: Of course, since you ran compiz earlier with sudo, your compiz configuration might be owned by root now.
adamk_: But we'll deal with that later if it turns out to be a problem.
adamk_: I have to get ready for work.
twiztid: ouch, can i back off of root? ah ok,,, cool cool... i thank u, im runnin that last one...
twiztid: [2]+ Stopped sudo compiz is what it gave me
adamk_: No, that's what 'sudo compiz' gave you, not the last command.
adamk_: I'm out of here.
adamk_: bbl.
twiztid: k late
ferret_: does kms potentially break mainline kernel suspend-to-disk? :)
ferret_: There is no follow-up question; it's just I changed a bunch of things and kms is the main suspect
mjt: ferret_: Works For Me (tm) :)
mjt: actually, first kernel with kms was the first kernel which were able to perform suspend/resume cycle for me
mjt: before, they were all failing during resume
mjt: (even if X weren't running)
glisse: airlied, MostAwesomeDude avoiding state reemission seems to give near to no win on micro benchmark i did couple week ago
glisse: that's on r6xx
NTU: What's the difference between SConscript and Makefile and how do you know which new files to add to what file?
dileX: scons is a replacement for make
dileX: means use scons or make
NTU: oh ok
NTU: thanks a bunch!
dileX: changing depends on (new) source files
dileX: NTU: http://www.scons.org/wiki/FrequentlyAskedQuestions
PeterKraus: hello. I've got an interesting question raised in Xorg, wonder if anyone here could help me. i was wondering whether it's possible to:
PeterKraus: have 2 GPU's in a PC, say intel and ati. Have both outputting stuff onto their ports, each with its own driver.
themgdn: hello. just synced x11 overlay and seeing new "r600" useflaf (gentoo). what is it? btw, i'm using r700 hd4850
Tommeh: PeterKraus: As long as you're using open source drivers with RandR: I think that would work
Tommeh: It would be a nice test :)
themgdn: *in mesa ebuild
themgdn: :)
BioTube: themgdn: if the overlay contains mesa, that would be the 3D driver for r600/r700 chips
themgdn: looks like gallium support, no?
BioTube: no
mjt: Tommeh: i tried that right before KMS, and been told that it does not work, kms is required.
themgdn: so, i need this flag or not? with r700
mjt: Tommeh: sure it were with randr
BioTube: themgdn: yes
BioTube: if you want 3D accelleration
Tommeh: mjt: KMS should be the default now.
Tommeh: As long as PeterKraus is using a recent distribution.
mjt: Tommeh: even more so for randr :)
Tommeh: Indeed.
themgdn: i've got 3d acceleration when i'm switched to git driver :)
PeterKraus: the problem for me with KMS is Xv
PeterKraus: maybe it's fixed now
mjt: what problem?
PeterKraus: Xv just doesn't work
mjt: WorksForMe(tm) :)
Tommeh: PeterKraus: it would help if you told us which adapters you are using...
mjt: that was my main requiriment, before 3d
PeterKraus: i'll restart and have a look
mjt: (xv thing that is)
Tommeh: No need
Tommeh: lspci
PeterKraus: Tommeh, whether it works
PeterKraus: didn't check in 2 months
mjt: 2 months ago it definitely should been working (what's the proper way to say it in english?)
BioTube: PeterKraus: if you've got an r600 or r700, you'll need extra firmware
PeterKraus: things have changed :P
Tommeh: Yes, lots of things
Tommeh: But if it's an R800 chip you might not have any changes yet ;)
Tommeh: So I could save you some time.
PeterKraus: it's r600
Tommeh: I was pretty sure Xv worked on it, so I'd give it ago.
PeterKraus: this machine needs restart after a week anyway
PeterKraus: brb
Tommeh: http://wiki.x.org/wiki/RadeonFeature
mjt: hm. Peter Kraus sounds familiar...
Tommeh: Funnily enough, the extra firmware is in Ubuntu 10.04, I think.
Tommeh: I didn't have to manually install that when I did a fresh install of it.
mjt: yay. HiJim :)
BioTube: it's in debian's firmware-linux-nonfree package, so I believe it
Tommeh: Oh good.
Tommeh: And with that, I wish my 2nd monitor wouldn't black-out/flicker during boot and/or loading Wine programs.
PeterKraus: meh, doesn't
Tommeh: The black-out literally just happened for a second. Good timing considering that I'm talking about it...
Tommeh: PeterKraus: which OS are you running and what versions do you have?
PeterKraus: Arch Linux 64b
Tommeh:
mjt: PeterKraus: xv should work, it works with radeon driver and r6xx for quite long time already
Tommeh: Also, I left you this link: http://wiki.x.org/wiki/RadeonFeature
PeterKraus: xfinfo says no adaptors present
PeterKraus: *xv
PeterKraus: right, i might just be doing something wrong
BioTube: dmesg | grep firmware
mjt: BioTube: it pauses during boot if not installed :)
PeterKraus: hmm failed to load
mjt: heh
PeterKraus: r600_rlc.bin
BioTube: http://people.freedesktop.org/~agd5f/radeon_ucode/R600_rlc.bin
Tommeh:
mjt: lol
mjt:
PeterKraus: strange
Tommeh: In fact, PeterKraus: failure to read /topic ;)
mjt: (or is that
PeterKraus: i'd think they'd provide it in the package
Ke: PeterKraus: which distro?
BioTube: it doesn't ship with the vanilla kernel
Tommeh:
PeterKraus: i've got the R600_me and R600_pfp
Tommeh: PeterKraus: read the link in /topic
Tommeh: All will be explained.
BioTube: got released after Linus decided to stop accepting blobs
mjt: which link - the build howto?
Ke: PeterKraus: pacman -Ss radeon and see, if there is some package named firmware os ucode oslt
PeterKraus: there's one on aur
PeterKraus: i'll have look at that
PeterKraus: that's the one!
PeterKraus: guess i'll need to restart again
PeterKraus: strange, the PC boots fine without the firmware
PeterKraus: someone needs some slapping in arch
PeterKraus: prolly me for not reading the wikis and being lazy to fix stuff
PeterKraus: hmm
PeterKraus: same error, but firmware is still in place
PeterKraus: *now in place
Ke: did you regenerate the initrd
PeterKraus: yeah right
PeterKraus: /facepalm
PeterKraus: i'm so stupid
PeterKraus: Following is probably not true since Linux 2.6.33, at least author didn't run into any problems: For newer ATI cards (R6xx and newer) extra microcode is currently needed. Grab radeon_ucode and radeon-initrd from AUR, build and install them and add radeon to HOOKS array in /etc/mkinitcpio.conf.
PeterKraus: can't bloody read!
Ke: pats PeterKraus
Ke: I couldn't read either
mjt: why it's in initrd?
mjt: can't it be loaded later?
mjt: i don't load it at all, it's loaded when X starts
Ke: well radeon probes for the initrd when it is loaded afaik
Ke: also KMS
mjt: "probes for the initrd" ?
mjt: it's a kernel module
Ke: gah
Ke: firmware
mjt: historically, initrd contained only stuff necessary to mount real root
PeterKraus: thats bettah
mjt: but radeon driver is not needed for that, obviously :)
mjt: apparently things changed and people put every shit into initrd :)
Ke: mjt: I think people ant faster boot times with initrd
mjt: heh.
mjt: it makes no difference in boot times
mjt: most time here is spent in kernel - probing and initializing hardware
jcristau: putting more stuff in initrd means you have to uncompress a bigger initrd.
mjt: maybe a few %%
Ke: kernel boot takes something like 2s for me
Ke: BIOS is very slow
mjt: jcristau: actually disk access might be slower than decompressing.
mjt: but the difference is very minor anyway
PeterKraus: wow
PeterKraus: stuff actually works with KMS
PeterKraus: good good.
mjt: news flash :)
PeterKraus: even game which would segfault before
PeterKraus: now works
PeterKraus: right so, what's so goood about gallium 3d? :)
yoshi314: PeterKraus: more opengl features, better performance?
Tommeh: Video acceleration via OpenGL
Tommeh: When someone continues working on it.
Tommeh: wishes he was a coder.
PeterKraus: right.
Tommeh: Right now, the OpenGL performance on anything up to and including R500 is better with Gallium than it is with Mesa
Tommeh: s/Mesa/'Mesa classic'/
yoshi314: yup
yoshi314: it's pretty reliable too
Tommeh: Though I don't think the r600/r700 code is up to the same scratch just yet.
Tommeh: But with Steam coming to Linux I hope it is soon :P
yoshi314: it's probably different under the hood
yoshi314: the >=r600 from r500
Tommeh: Yeah, quite different I'd imagine
scobby: hi
Tommeh: I'm not being impatient, but it's nice to know that Gallium has potential
yoshi314: i like how it improves day by day
yoshi314: i've started using git mesa with gallium enabled a few weeks ago
yoshi314: not planning to turn back for now ;)
yoshi314: unless it breaks xorg for some reason, that would require some hackery
scobby: i have a problem with tv-out on my radeon x1950 pro, i managed to get it working with 800x600, 1024x768 640x480. But the screen is too big, there are 2 cm cut off at top and side. with mesa driver (failsafe from ubuntu) it fit perfekt (sorry for my bad english)
Tommeh: scobby: it is probably related to 'overscan'
Tommeh: Which you need to compensate for.
Tommeh: But how to do this with RandR - I have no idea :)
scobby: and how?
Tommeh: ^ .. Worth a Google though to see if someone else has conqueued Overscan with X/RandR
scobby: i tried google 2h now .... this is my last option
Tommeh: I would imagine fglrx/nvidia binary driver control panels have some settings for this, but as for free drivers.. I don't know :)
Tommeh: scobby: but now you know it's overscan - it will make Google searches easier?
yoshi314: you could try some xrandr gui to make your life a bit easier
Tommeh: s/easier/'more effective'/
yoshi314: http://wilmer.gaast.net/blog/archives/34-Change-Intel-X.org-TV-Out-margins.html hmm i wonder if it works with ati
scobby: ok i try to search with overscan, yoshi314 tried arandr grandr xrandr and ubuntu
scobby: anyone know how to activate underscan on xorg.conf?
ferret_: mjt: Thanks for your response. Could you briefly tell me what suspend frontend you use (i.e. what program you run to suspend to disk), and any special configuration-stuff you might have done?
ferret_: If yours works and mine doesn't, I want to copy yours. I just want it to work at this stage.
mjt: for me it actually Just Works
mjt: even simple echo disk > /sys/power/state works as expected
mjt: (provided you've resume code hooked up)
mjt: pm-hybernate (sp) works too (that's the default on debian i think)
scobby: i tried this Option "TVOverScan" "0.5", nothing happend
adamk_: UNless I'm mistaken, the drivers don't support understand or overscan. I'm always seeing the developers tell people to turn it off on their TV.
Tommeh: 'cause every TV supports that :/
NTU: Xserver controls which version of GLX is in use, right? 1.7.X = 1.2 and 1.8.X = 1.4? If so, why isn't this handled in Mesa?
scobby: i found a tool sudo tvo_set which could solve my problem, anyone know how to get it on ubuntu karmic?
adamk_: It's not in the Ubuntu repos, so unless you can find a 3rd party repository with it, you'll have to download the source and compile it yourself.
NTU: The reason I ask is because unless the user compiles Xserver from source, compiz --replace complains about not having 1.4 support
adamk_: It's just a harmless warning :-)
scobby: is there a complete list of tv-out options?
NTU: brb. my window manager is hosed
Tommeh: scobby: it's worth updating to 10.04 if you can.
Tommeh: Given the added radeon development that has gone into it.
Tommeh: Or rather, there have been millions of changes that were rolled into 10.04
soreau: I saw glx version jump from 1.2 to 1.4 and I'm on rv350. What's the deal with that?
soreau: That is to say, it used to be 1.2 in the past and now it reports 1.4
scobby: Tommeh: hmmm i use vdr repo from vdr-team, i think currently only karmic supported
twnqx: the better question, when will we have open gl 4.0 support :P
soreau: slaps twnqx with a large trout
soreau: twnqx: "When it's ready"
Tommeh: scobby: really? That's a shame.
twnqx: :D
twnqx: not that i'd care
Tommeh: Lucid has been out for weeks now.
jcristau: soreau: says 1.2 on server <= 1.7, and 1.4 on server 1.8 + dri2.
twnqx: i'm more whining after the loss of decent powersaving
Ke: opengl4 is a lot closer when gallium3d support improves
twnqx: the bigger problem is this damn PC
soreau: jcristau: Right but why? I thought that we would be fortunate to see glx 1.3
twnqx: and the lack of open stores to buy a dvdrom
twnqx: i wonder if windows can boot of firewire connected cdroms...
soreau: jcristau: I guess what I'm asking is if this is 'really' glx 1.4 or some 'quirk' that just reports this version
NTU: Did AMD have the docs to GL 4 before they were published on the net? If not, how in the world in like... 3 days or so, they got GLSL 4 + GL 4 in the AMD 10.3 opengl beta driver?
jcristau: soreau: i don't think it's a quirk
soreau: jcristau: The thing that confuses me is the mesa warning. It says 'this is an app bug' but I think it's actually a driver bug for reporting pixmap create/destroy without glx 1.3
soreau: *shrug*
Ke: NTU: they probably have some of people listed that were compiling the standard
soreau: jcristau: ie. nothing changed in the application side but the bug (claiming an app bug) is fixed by newer glx version reported in X server
soreau: which doesn't seem to make much sense if you ask me :P
soreau: Further, I don't think that (harmless) warning should have been added in the first place because there's really no need for it (other than to confuse users because they think it's an error and it does look like an intimidating message)
NTU: well it intimidates me because compiz doesnt work for me
soreau: NTU: Do you need help getting compiz to work?
jcristau: soreau it's a bit of a mess. aiui we supported enough of new glx to make compiz work, but not enough to claim 1.3, and there's no way to express that in the protocol.
NTU: well it works in KDE 4 since its built in
soreau: NTU: Compiz is *not* built in to kde
NTU: in gnome, it doesnt
NTU: no i mean compositing
soreau: kde uses it's own kwin kompositing effects
jcristau: soreau: thing is, the warning may not be harmless. it is for compiz, not necessarily for other stuff
soreau: jcristau: it *is* for compiz? doesn't this warning show on other apps like kwin and cairo-dock?
jcristau: soreau: i meant afaik it's harmless for stuff like compiz
soreau: jcristau: ah ok. fairnough then :)
NTU: http://pastebin.com/6EPFhD3m
NTU: my toolbars disappear and then i need to restart GDM
NTU: nevermind. installed KDE 4 :)
adamk_: What distribution is this?
NTU: highly modified archlinux
NTU: its closer to slackware :)
PeterKraus: :)
adamk_: Hmmm...
adamk_: I'd be more interested in seeing the output of 'compiz-manager &' if you have that installed.
NTU: http://fpaste.org/E3Ug/
mjt: it gives internal server error :)
adamk_: Yes, yes it does.
adamk_: That's because fpaste.org sucks and does not work behind certain proxies.
NTU: sorry ill change it to pastebin
NTU: http://pastebin.com/6AKjLARR
adamk_: Does X crash?
NTU: no my toolbars though
adamk_: That is odd. I thought it might be due to being run in lxde, something I have no experience with, but I just tried it and it worked fine here.
NTU: yeah ive seen youtub videos of people with arch running compiz + lxde just fine
NTU: lucky me :)
ickle: krh: is/window 22
soreau: NTU: Can you start a gui program from your terminal like a text editor? (and see if it gives similar messages)
NTU: seems fine
soreau: tried rebooting/restarting X?
soreau: Not sure where that message is coming from but it's not a common one
NTU: a lot
soreau: Which wm are you currently using?
NTU: right now, openbox
soreau: Can you replace openbox with any other wm besides compiz?
NTU: let me try
NTU: yes with desktop effects!
NTU: kwin! :D
soreau: ugh :P
soreau: Did you check xsession-errors to see if there's anything of interest there?
NTU: how do i find that out? xorg log?
soreau: ~/.xsession-errors
NTU: just authentication stuff
soreau: Can you try killall kwin; compiz-manager ?
soreau: or whatever wm you have running
NTU: and broken wm again
soreau: What glx version is reported by glxinfo?
NTU: 1.2
NTU: i can make it 1.4
soreau: I'm just curious as to what the problem actually is
NTU: im not in a mood to recompile anything though.
soreau: I'm pretty much done guessing right now so let us know when you figure it out ;)
NeKit: GPU: ERROR: Need GL_EXT_framebufer_object for multiple render targets
NeKit: GPU: *********Does your video card support OpenGL 2.x?
MostAwesomeDude: You either have an ancient Mesa, no KMS, or both. Also, you're going to be very broken-hearted when you upgrade and find that MRTs still don't work because GL_ARB_draw_buffers is missing.
NeKit: so, nothing can be done?
Dr_Jakob: MostAwesomeDude: I thought roland added MRTs?
MostAwesomeDude: Dr_Jakob: To classic? I think it only works on Gallium.
Dr_Jakob: right
NeKit: HD3870x2
MostAwesomeDude: NeKit: Ah, that stuff doesn't work on r600 yet anyway. It'll be a bit.
Ronis_BR: hi all
NTU: hello!
Ronis_BR: is there any improvements in kernel 2.6.34_rc7?
Ronis_BR: NTU: :)
adamk_: So what would cause "failed to validate texture for unit 0. unsupported texture format in setup_hardware_state" ?
adamk_: I'm guessing that's a DRM thing and not a mesa thing? :-)
NTU: if you get: ntransform_fuction 168 unsupported ERROR(r600_pipe_shader_create 114) is that supposed to help the developer what to turn on and off in gallium? And no, I'm not trying to just add random stuff in the r600g files by making it show different stuff in glxinfo. I'm trying to make it work with 7.8 branch merged.
NTU: I was just wondering if that output is supposed to mean something...
adamk_: Alright, gonna try upgrading to 2.6.34-rc7 to see if I still get that validate texture error.
adamk_: Alright, if I 2.6.34-rc7 I get the same error.
adamk_: Maybe this really isn't DRM related after all, then.
NTU: what exact validate texture error and on what hardware? like... maybe I can try whatever you are doing?
adamk_: r600, it happens with ut2004 and RTCW.
adamk_: Let me see if it happens with anything else.
NTU: oh i have r600!
adamk_: Latest Mesa from git and 2.6.34-rc7
NTU: latest DDX and libdrm too?
adamk_: No, DDX is at 6.13.0 and libdrm is at 2.4.18
evil_core: hi all
NTU: oh ok. Are you playing it through wine, vbox or what?
NTU: evil_core hey how are you?
adamk_: NTU: They are native.
evil_core: NTU: dunno, I wanna glisse there!
NTU: you're not using the pay verion of ut 2004 are you? otherwise i cant test.
adamk_: NTU: I am.
NTU: oh ok nvm then :)
adamk_: Well, for what it's worth, libdrm from git has the same issue.
NTU: i have gl-117 installed. can you see if it acts weird on that?
adamk_: That seems to be fine.
NTU: its kind of funny.. about 3 years ago when I first heard about gl-117 i was getting around 30 FPS at its lowest settings. now im getting flat out 60 FPS 3 years later.
NTU: that was back with my 9800 PRO. :)
adamk_: figured it out. I have to explicitly disable s3tc support in driconf
adamk_: Not even sure why I ever enabled it.
NTU: what does driconf talk to? or is it just a stand-alone config that mesa or whatever looks for?
adamk_: It creates a ~/.drirc file that the mesa drivers read.
adamk_: Or an /etc/drirc file if you're root.
evil_core: NTU: in internet everything is free as in beer :)
Thunderbird: BTW does anyone know the maximum texture size of a Rage 128? I'm getting some wine bug report and glTexImage2D fails with GL_INVALID_OPERATION for no clear reason
Thunderbird: nothing special is done except the texture size is for instance 640x480
chithead: Thunderbird: does rage128 support npot textures?
Thunderbird: perhaps it is related to that
Thunderbird: or perhaps not the right dimensions are printed in the log; we allocate the nearest power of two size
otaylor: Thunderbird: from the age of the card, it's pretty likely that it only supports 512x512
otaylor: Thunderbird: but I dont' have specific infomratoin
Thunderbird: I was thinking the same but I'm not entirely sure since if my memory is right a RivaTNT (which is of the same age) supported 2048x2048 I think
otaylor: Comment in th esources says
otaylor: 10, /* max 2D texture size is 1024x1024 */
Thunderbird: err, it gave a GL_INVALID_VALUE
Thunderbird: but this is nothing special:
Thunderbird: trace:d3d_surface:surface_allocate_surface (0x14a960) : Creating surface
Thunderbird: (target 0xde1) level 0, d3d format WINED3DFMT_P8_UINT, internal format 0x1908,
Thunderbird: width 1024, height 512, gl format 0x1908, gl type=0x1401
Thunderbird: Mesa: User error: GL_INVALID_VALUE in glTexImage2D(level=0, width=1024,
Thunderbird: height=512, depth=1)
Thunderbird: just GL_RGBA + GL_UNSIGNED_BYTE
MostAwesomeDude: Hm.
evil_core: halflife 1.x and 2.x games run for anyone in d-r-t, master ddx/mesa and especially xorg 1.8?
MostAwesomeDude: Thunderbird: I have here a pastebin from you from a few days ago http://wine.pastebin.com/KWZTUmw6, what did you want me to do with it?
evil_core: MostAwesomeDude: you wouldnt be a cowboy, 12seconds too late ;)
Thunderbird: hi
MostAwesomeDude: Hey.
Thunderbird: btw did you have time to try that s3tc test yet?
MostAwesomeDude: That's what I was asking about. What did you want me to try?
Thunderbird: let me look it up again
Thunderbird: MostAwesomeDude, http://wine.pastebin.com/KWZTUmw6
Thunderbird: it is an issue I encountered on fglrx but according to some appdb entries it is also around on other dri drivers
soreau: MostAwesomeDude: FWIW, the bug I asked you about the other day ended up being a strange metadata issue (for ordering or gl and composite plugins) but only happened with open drivers using libdri
soreau: s/or/of
MostAwesomeDude: errorB is 0x501 and errorC is 0x500.
MostAwesomeDude: This is on Gallium, r300g without dxtn.
MostAwesomeDude: soreau: Huh, alright.
MostAwesomeDude: I'm on a single-card setup right now, so I can do the whole Compiz thing if you need any more tests.
soreau: MostAwesomeDude: There's only one more plugin that has a similar issue but I'm pretty sure it's somehow related (to the bugs we've already fixed)
Thunderbird: what happens if you have dxtn around? (or just enable the decompression part using driconf) I'm interested in the difference with the #if on and off
zhasha: MostAwesomeDude: soon your baby will take its rightful place among the FOSS drivers :)
soreau: MostAwesomeDude: If you have some spare time, we need testers for 0.9 (or you can just try it for yourself). We are trying to squash as many bugs as we can before 0.9 official release
MostAwesomeDude: Thunderbird: The other path gives errorA as 0x500.
MostAwesomeDude: I don't have dxtn, and Gallium's DRI doesn't allow forcing it on (yet).
soreau: MostAwesomeDude: Here's some build instructions, there are more on the wiki build++ page and there is also a script I wrote to automate the compilation/install process (d/l's code to ~/src/compiz/compiz++ and installs to /opt/compiz++ by default) http://forum.compiz.org/viewtopic.php?f=112&t=12565
Thunderbird: ok, too bad
soreau: and also anyone else here that would like to help us test the development version of compiz ^^
soreau: mainly we're just trying to get all the crashes (and 'noncrash' crashes) worked out right now
MostAwesomeDude: Thunderbird: What were you testing for, exactly?
Thunderbird: fglrx doesn't seem to allow npot compressed textures using glTexImage2D
soreau: MostAwesomeDude: Oh yea, from the other day: Focus Blur crashes in the driver
soreau: I think both c and g
Thunderbird: but it works fine using glCompressedTexImage2DARB
MostAwesomeDude: Thunderbird: That sounds hilarious.
MostAwesomeDude: soreau: I have to grab from git?
soreau: MostAwesomeDude: nope
Thunderbird: but we need to use glTexImage2D since glCompressedTexImage2DARB can't be used reliably for allocating a texture (the behavior on NULL is undefined)
soreau: MostAwesomeDude: You should be able to reproduce with 0.8 on rv350 (and probably all r3-5xx)
Thunderbird: MostAwesomeDude, I saw the same error in a WoW appdb report
Thunderbird: so that's why I think it occurs on this driver as well
MostAwesomeDude: Thunderbird: Hmm, I'll have to grab dxtn and see.
soreau: The only thing libdxtn has done for me is cause rendering errors. One time it sped up doom3 a lot but with a ton of rendering errors
soreau: with et, it causes bright pink and purple spots on textures that are at a steep angle in relation to the camera position
soreau: and the errors painfully obvious looking at the planet in the menus of doom3 and etqw
MostAwesomeDude: soreau: Haha, yeah, it segfaults. How can I gdb this?
soreau: MostAwesomeDude: fedora?
MostAwesomeDude: soreau: Yep.
soreau: MostAwesomeDude: Guess you don't need debug symbols for compiz.. just enable crash handler plugin
soreau: trace should be dumped in output and to file in /tmp
MostAwesomeDude: I don't have that plugin. Time to go search the repos!
soreau: heh
soreau: moeSizlak: compiz-fusion-plugins-extra
soreau: meh
soreau: MostAwesomeDude: btw, what chipset are you testing with?
MostAwesomeDude: soreau: r580.
soreau: ok
MostAwesomeDude: And so far, blur hasn't caused a crash.
MostAwesomeDude: The crash was from something else, apparently.
soreau: MostAwesomeDude: So focus blur works there?
MostAwesomeDude: soreau: No, it doesn't *work*, but it doesn't crash.
soreau: ah..
MostAwesomeDude: If I set the window key to "any", it should match e.g. my Fx window, right?
soreau: MostAwesomeDude: make sure you have focus blur enabled in the blur plugin and window type match set to 'any'
soreau: zactly
MostAwesomeDude: How should I tune the effect to be easily visible?
soreau: MostAwesomeDude: Basically it's supposed to blur every window that does *not* have focus
NTU: if you get: ntransform_fuction 168 unsupported ERROR(r600_pipe_shader_create 144) is that supposed to help the developer what to turn on and off in gallium? And no, I'm **_NOT_** trying to just add random stuff in the r600g files by making it show different stuff in glxinfo. I'm trying to make it work with 7.8 branch merged. I was wondering if the error would pinpoint the exact location of what isn't right so I can fix it.
MostAwesomeDude: No idea, haven't ever seen an error like that before.
soreau: MostAwesomeDude: It definitely crashes in r300_dri.so here but I don't have -g
MostAwesomeDude: soreau: Heh, I just got some hilarious crashes. The crash handler plugin completely and totally failed.
soreau: I think for experimental driver repos like xorg-edgers they should enable symbols by default
soreau: MostAwesomeDude: Well that's not cool.. did you remember to restart compiz after installing -extra?
MostAwesomeDude: Bug tormod about that.
MostAwesomeDude: soreau: Oh yeah, and I reconfig'd things in ccsm.
soreau: figured you would..
MostAwesomeDude: I'll just pull the git version.
soreau: :P
soreau: Typically the crash handler plugin only fails when the code path never reaches crash handler
soreau: like some hard segfault
soreau: In which case I guess you could valgrind it
soreau: MostAwesomeDude: is tormod ever on irc/freenode?
MostAwesomeDude: soreau: Yes, and that's his nick when he is.
soreau: ok
soreau: MostAwesomeDude: Hm. With gallium it is 'working' but only makes the non-focused windows black
soreau: I was trying to get a useful bt..
soreau: IIRC, focus blur was much worse with gallium about 4-6 months ago :)
soreau: MostAwesomeDude: So I guess the bug is 'doesntwork with gallium and crashes with classic mesa'
MostAwesomeDude: m.
MostAwesomeDude: Alright.
soreau: oh good, icanhas output
soreau: MostAwesomeDude: http://pastebin.org/231504
soreau: that is from gallium and seems to be caused by focus blur effect
MostAwesomeDude: soreau: Uf, yes, that would be.
MostAwesomeDude: That should be "fixed" with a Mesa from the past few hours.
MostAwesomeDude: But that's a card limitation; that shader will never work on pre-r500.
soreau: Well it worked with fglrx in the past :P
MostAwesomeDude: Hm.
MostAwesomeDude: fglrx cheated somehow.
soreau: In a way I'm kinda glad we have had fglrx. Otherwise you guys would just say anything and we would have no other choice but to take your word for it ;)
DanaG1: focus blur makes compiz quit on gma950.
DanaG1: bleh, DEcelerator.
MostAwesomeDude: Well, we seriously cannot do more than a handful of texture indirections on the hardware.
MostAwesomeDude: I have a feeling that the shader limits weren't advertised prior to a commit made to Mesa earlier in the day.
soreau: MostAwesomeDude: IIRC, focus blur caused a small slowdown similar to alpha blur (on fglrx)
DanaG1: Bleh, compiz deals badly with lack of shaders.
DanaG1: Instead of unloading the plugin, it entirely aborts.
soreau: MostAwesomeDude: Also when you say 'fixed in mesa a few hours ago', is that for c, g or both?
MostAwesomeDude: soreau: For the shader giving you the indirection error.
soreau: DanaG1: It typically fails with rendering errors or crashes in dri if it's trying to do something the driver can't
soreau: MostAwesomeDude: No I mean is it fixed in classic mesa, the gallium driver or both?
MostAwesomeDude: Oh. Gallium only; classic should already do the right thing.
soreau: ah ok
soreau: well classic from xorg-edgers crashes and gallium just makes the window black that compiz is trying to blur
DanaG1: Weird... it shouldn't TRY in the first place! =þ
DanaG1: It should check and say "I can't do that". =þ
soreau: DanaG1: If the driver reports that it can do it, it should be able to
soreau: can't really do much in compiz to fix driver crashes (should be pretty obvious 'dont do that' when a certain effect causes a crash)
soreau: MostAwesomeDude: but yea, this is latest mesa from about ~30 mins ago and it's giving these with gallium
soreau: I'll try to get a bt with symbols from classic
Curan: just a quick question: how much does "mostly" for the Gallium Mesa support for r300 mean? Same OpenGL level as the defaul mesa driver for r300, more or less (especially: which OpenGL level is implemented with hw acceleration)?
MostAwesomeDude: 2.1, mostly.
Curan: On software or with (almost full) hardware acceleration? Or put differently: is it viable to use Gallium for acceleration instead of the "standard Mesa"?
Curan: (which only has 1.5 and makes some applications really slow)
Curan: s/On/In/ (for the message at 22:29:14 (CEST))
MostAwesomeDude: We don't do hard fallbacks. It might eat your babies, but I think at this point that's not much of a problem.
soreau: we could use less whining kids anyway
mokoloko: they taste good too
Twiztid: adamk_: hey jus an update, from yesterday about compiz and the --indirect-rendering file that soreau, reccomended; you we're right on the money to make it executable; so to conclude; compiz and GLcairo are in harmony on my Ubuntu Lucid install! xD
Curan: oh I sacrifice a few kids gladly if the 3D throughput gets up :)
NTU: ahh finally I'm getting closer. http://pastebin.com/aSnhRSfv The errors I'm getting is the output of %s %d (Integer and String) in line 144 of r600_shader.c. I'm so close! ...i think
Curan: MostAwesomeDude: thx, then I'll try to build Gallium and give it a try
soreau: Twiztid: --indirect-rendering is a compiz arg, not a file.. what did you have to make executable?
Twiztid: soreau: hey man i made that compiz.real file for Ubuntu's compiz to not complain, but yesterday, i hit a snag when i couldnt enable compiz, so with some googling, and ultimatly the help of adamk i, yes, had to make it executable... lolz
soreau: Twiztid: Was it giving permission denied or similar message?
Twiztid: soreau: well, kinda, the visual effects options we're greyed out...
soreau: That points to automatic plugin sorting not being enabled
soreau: ccsm should work even if compiz isn't running
soreau: oh wait..
soreau: you mean in ubuntu visual effects preferences
soreau: suddenly remembers
Twiztid: soreau: really? ya ccsm works, but none of the effects we're 'on'; yeah.. the prefs...
soreau: Twiztid: Ah, when I told you to make that script
soreau: I assumed you knew to make it executable
Twiztid: soreau: lol no worries, ya i did it to a tee, but jus didnt make it executable...
soreau: no wonder
Twiztid: soreau: i know i know, no blame no shame...
Twiztid: soreau: im still making the leap from windows to linux so i overlooked a relitivly simple step
soreau: I figured it was just some ubuntu black magic. I was thinking g-a-p was md5 sum checking the compiz binary so it makes sure the user isn't trying to fool it or something xD
Twiztid: soreau: wow thats too intense for me, lol im dyin to get into a linux or ubuntu course... but ya, opensource driver is handling the eyecandy without a hitch! yay!
soreau: I always tend to end up over or under thinking things. Every once in a while I get one right ;)
Twiztid: hey, you we're spot on with xubuntu and named a perfect workaround for whatever weird problem compiz/openGL/dirict-rendering/KMS was having, im still kinda curious but i dare not break it.. lol
soreau: Twiztid: Hang on to what you have. --indirect-rendering makes my screen go black and input stop working :P
soreau: MostAwesomeDude: Oh yea btw, --indirect-rendering makes my screen black and input stop responding
soreau: MostAwesomeDude: Ok, here is classic mesa with symbols and focus blur crash http://sprunge.us/OiKI
MostAwesomeDude: soreau: Hmm. No idea why.
MostAwesomeDude: Hmm, sneaky. Looks like we fallback to software and then something (else) goes wrong.
soreau: gah, looks like no symbols?
soreau: ok building with symbols now (hopefully)
NTU: does this look right to anybody? struct r600_pipe_shader *r600_pipe_shader_create(struct pipe_context *ctx, unsigned type, const struct tgsi_token *tokens)
Curan: can sombody with an X.org Wiki account please update
MostAwesomeDude: NTU: Right now an r600g merge is getting planned. If you wait a day or so...
NTU: oh i had no idea!
adamk_: jcristau: 3D works on r600/r700 in Lenny, right?
jcristau: hmm, no
jcristau: the r600 driver didn't exist in mesa 7.0.x
adamk_: Oh, right it was released originally back in February of 2009, wasn't it?
jcristau: yeah
evil_core: !help
NTU: evil_core i dont think bots live here
evil_core: NTU: did you heard glisse there in last week?
cTn: test
cTn: ah registartion works, wonderfull :)
cTn: could i trouble someone for 2 minutes regarding powersaving issue ?
cTn: i already posted a question on launchpad few days back, but still without a reply, so i am stuck :(
svenstaro: cTn: just ask the question
cTn: svenstaro: i own an MSI laptop which have hd4670m inside, i am running latest fully updated distro of ubuntu
cTn: but my "laptop" is heating badly when i am on AC power, also battery time is heavily reduced
cTn: more info could be found here https://answers.launchpad.net/ubuntu/+question/110525
cTn: i guess there is some sort of powersaving problem but i have no clue how to debug it and so on i cant fire an bug ticket to get it fixed
cTn: any ideas?
svenstaro: cTn: kms/ums?
cTn: kms
svenstaro: tried ums?
cTn: not yet, i am not really an "experienced" linux user
svenstaro: use grub?
cTn: yeah
svenstaro: put "radeon.modeset=0" to the end of your primary kernel line in menu.lst
svenstaro: reboot, try again
NTU: or grub.cfg
cTn: gonna fire up my laptop and try it, be right back
soreau: cTn: Yea, if you disable kms, you can use forcelowpowermode option in x conf
soreau: or whatever it's called
cTn: svenstaro: is it the same, when i edit my boot option via "e" in grub and add radeon.modeset=0 to the end of the "string" ?
svenstaro: cTn: yes
svenstaro: but it wont be permanent
cTn: okey thats what i wanted, booting now
cTn: okey, what now, should i check the "battery time" ?
cTn: or do we have some nice debug commands that i could use :)
svenstaro: check the heat, the noise and the performance
cTn: okido
egns: so
egns: any tutorial about compiling r300g?
egns: i'm lost, compiling something that i don't remotely now what is
egns: xD
svenstaro: get git source, type make? it's what I would try
egns: that's what i did
svenstaro: If you get stuck, shout profanties at MostAwesomeDude
egns: after some fight with dependences
egns: -_-
egns: so now i have r300_dri.so
svenstaro: oh, so it worked
egns: that's r300g?
svenstaro: If you got the right branch, yeah
egns: how do i know if that's the right branch ?
svenstaro: did you compile with --enable-debug?
egns: no
svenstaro: oh well
cTn: svenstaro: laptop battery time is the same, fan speed is little higher, and for the heat we will need few minutes, but judging from the battery time, heat problem will be present because "normal" battery time is cca 2hours 30minutes on battery, currently only 1hour 10 minutes (like before)
egns: i did ./autogen.sh --enable-gallium-radeon --with-state-trackers=dri,glx
svenstaro: cTn: how's 3d performance?
svenstaro: egns: recompile with --enable-debug
egns: k
egns: omg
egns: 30mins more
svenstaro: then launch glxgears like LIBGL_DEBUG=verbose LIBGL_DRIVERS_PATH=path-to-your-new-so's RADEON_DEBUG=all
egns: ah no
svenstaro: er, and put glxgears at the end
egns: ok
svenstaro: egns: 30mins man? what machine is that on?
egns: that's what mesa took
egns: p4 @ 2ghz
svenstaro: oh dear
egns: nah it just relinked it
evil_core: svenstaro:rebuild ddx and libdrm with --enable-debug also?
svenstaro: evil_core: not sure to be honest
cTn: svenstaro: to tell the truth i have nothing to compare it to, glxgears reports 1650 frames in 5 seconds, that seems to be little low to me
svenstaro: cTn: what card?
cTn: hd4670m
svenstaro: omg
egns: it works
svenstaro: cTn: glxinfo |grep -i rend
egns: but i'm getting TOO much debug info at the console :p
svenstaro: egns: er thats alright, hows the looks?
svenstaro: corruption etc?
egns: no, perfectly
MostAwesomeDude: No such thing as too much debug. :3
svenstaro: egns: you sure you put my line there correctly and it loads the new libs?
egns: MostAwesomeDude: konsole's crying
cTn: direct rendering: Yes
cTn: OpenGl rendered string: Software Rasterizer
svenstaro: egns: sounds good then, try somethign complex
egns: svenstaro:
egns: LIBGL_DEBUG=verbose LIBGL_DRIVERS_PATH=/home/neiko/mesa/lib RADEON_DEBUG=all glxgears
cTn: heat problems starting to show
svenstaro: cTn: congrats, you are not using your radeon at all!
svenstaro: egns: try crysis now, baby steps
egns: crysis, lol
egns: that would make my computer explode
egns: i'll try openarena
egns: xD
svenstaro: er fine :(
svenstaro: You just *HAVE* to jump the gun
svenstaro: it will probably explode
egns: works, veeeeeeeeeeewy slow
egns: probably due to the huge amount of debug info being spitted to the console
egns: i'll have to shut that thing up
svenstaro: remove the debuginfo thingie
svenstaro: does it look alright?
cTn: svenstaro: so if i understand this right, everything is running in software mode ?
svenstaro: if it looks good, it's MAD's fault, so go shout profanities at him
svenstaro: cTn: yes
svenstaro: cTn: probably the firmware issue
svenstaro: see /topic
egns: argh
cTn: i c there are some ubuntu updates pending, i give them a shot
egns: it's slow
cTn: interesting i cant restart
svenstaro: cTn: I bet you can
egns: so it's slower than the radeon driver, werent it supposed to be faster? ._.
cTn: terminal ftw
svenstaro: egns: Dunno, ask MAD
MostAwesomeDude: What's slow?
egns: r300g compared to radeon
egns: BUT it was my fault
MostAwesomeDude: With which program?
evil_core: -DNDEBUG collide with --enable-debug?
egns: disabled the debug info
egns: and now it's faster
egns: openarena, MostAwesomeDude
MostAwesomeDude: How much debug info were you dumping?
MostAwesomeDude: Oh, RADEON_DEBUG=all? Don't do that.
MostAwesomeDude: Seriously.
egns: points at svenstaro
evil_core: enable-debug causes much slowdown?
MostAwesomeDude: Just don't enable RADEON_DEBUG and it'll be fine.
egns: so how do i make Xorg to use this driver permanently ?
MostAwesomeDude: Debug builds aren't that slow. RADEON_DEBUG is very noisy though.
evil_core: I know
MostAwesomeDude: You can install it like any other DRI driver, although you really shouldn't.
evil_core: but will I loose visible amunt of FPS?
cTn: svenstaro: ok i rebooted witout the radeon.modeset=0 opengl rendered is mesa dri r600... should i add the radeon.modeset=0 and reboot again ?
egns: why ?
MostAwesomeDude: I don't recommend it, that's all.
MostAwesomeDude: Distros don't do it -> you shouldn't do it either.
egns: if the worst thing is my machine freezing, i can live with it
egns: :p
cTn: svenstaro: 10 000 fps / 5sec, battery time is the same (1 hour 5 min) and laptop is heating up
svenstaro: cTn: thats without KMS now?
cTn: thats with the radeon.modeset=0
cTn: i am using latest "stable" ubuntu kernel so i guess its KMS only ( i am really not pro in this)
cTn: i am currently tring radeon.dynpm=1 and then i try radeon.modeset=0 if necessary
svenstaro: cTn: dont do that, you are running ums.
svenstaro: cTn: so if the card still goes full throttle perhaps you should ask other folks in this channel since I only have r300 cards.
evil_core: KMS classic sucks, but gallium with pagteflip is getting usable
svenstaro: is using trusty ums
evil_core: but UMS lack pageflip
cTn: svenstaro: are you sure i am running ums? the cat /sys/kernel/debug/dri/0/radeon_pm_info
cTn: is giving me results
cTn: but message state: PM_STATE_ENABLED is missing
cTn: still running on full clock
agd5f: cTn: You need drm-radeon-testing branch of the drm for the latest kms power mangement code
evil_core: and after yesterday updates pageflip not longer hangs while playing q3 based games
evil_core: I cannot run radeong, it segfaults Xorg on start
cTn: agd5f: i guess i should also grab v2.6.34-rc7-lucid ?
cTn: current ubuntu is running on 2.6.32 i f i am correct
agd5f: cTn: I don't know if that has the patches or not
evil_core: relfextions dont slow down neverball in gallium recently
svenstaro: OOhh really? So render to texture now works?
MostAwesomeDude: RTT has always worked.
svenstaro: It works FAST now? :D
MostAwesomeDude: Fast enough for people to not bitch about it, apparently.
svenstaro: Then make it run faster :P
NTU: heh
soreau: cTn: If you're going to build a kernel, use drm-radeon-testing branch from airlies kernel tree
svenstaro: I will send you a bunch of cookies if you can get hardware mip-mapping running again.
evil_core: I am not saying it doesnt slow downs
svenstaro: :(
MostAwesomeDude: We're already doing hardware mipmaps...
cTn: soreau: first i am going to try latest rc kernel for ubuntu, compiling kernels isnt really my strong side
evil_core: but last time I checked it did intro in UMS as slideshow in UXGA res
svenstaro: Last time I checked not too long ago it was still mesa in KMS
evil_core: libdxtn sucks hardly
Agrajag-: g'day, something is causing my 2nd monitor (dvi) brightness to change depending on what's displayed on the screen. it seems that if there's lots of graphics on the screen it goes very bright, if there's not it goes darker. i have no idea what might be causing this, it's very annoying because it often goes very bright. any ideas?
soreau: Agrajag-: what card model is it?
svenstaro: Agrajag-: it is not the screen configued for dynamic contrast is it?
Agrajag-: ATI Technologies Inc RV620 LE [Radeon HD 3450]
Agrajag-: svenstaro: i don't think so. this happened when i upgraded to ubuntu 10.04, wasn't happening in 9.10
Agrajag-: where is dynamic contrast configured?
egns: wow, thank you guys - i though i would never be able to run openarena at 800x600 xD
cTn: soreau: well with the RC kernel i am atleast getting an message state: PM_STATE_DISABLED, which is interesting whn i added radeon.dynpm=1
soreau: cTn: I don't know how that pm stuff works. I just know they've been working on it lately and the latest bits are in the d-r-t kernel/drm
agd5f: cTn: the dynpm option is gone in the latest code. you configure pm on the fly via sysfs
cTn: agd5f: coud you be a little more specific if thats possible? :)
agd5f: cTn: http://lists.freedesktop.org/archives/dri-devel/2010-May/000492.html
egns: errr, i want to run something that needs GL_ARB_texture_non_power_of_two, any way to get this with a r300 ?
agd5f: egns: should already be supported
egns: not in the driver im using
Droste: it's only available with r300g
egns: thats why is upposed
egns: i did cp *_dri.so /usr/lib/xorg/modules/dri/
egns: then restarted
egns: but it does not appear in glxinfo yet
evil_core: # sysctl -n /sys/class/backlight/acpi_video0/actual_brightness
evil_core: error: ".sys.class.backlight.acpi_video0.actual_brightness" is an unknown key
evil_core: how to get value trough sysctl?
Droste: egns: with LIBGL_DEBUG=verbose glxinfo you can see which driver is used (to check if /usr/lib/xorg/modules/dri is right)
cTn: agd5f: hmm ubuntu implementation lokks little different but i guess thats expected
egns: libGL: OpenDriver: trying /usr/lib/xorg/modules/dri/r300_dri.so
egns: that's it?
agd5f: cTn: that information only applies to drm-radeon-testing branch
evil_core: using any compositing manager on laptop is bad idea?
agd5f: evil_core: why?
Droste: agns: radeong_dri.so is the gallium driver. make a symlink to the driver in a different directory (cd /usr/lib/xorg/modules/dri && mkdir gallium && ln -s ../radeong_dri.so r300_dri.so) and run glxinfo again with LIBGL_DRIVERS_DIR=/usr/lib/xorg/modules/dri/gallium glxinfo
evil_core: agd5f: because of battery/interrupts
evil_core: I dunno, I heard that about compiz many times
Droste: egns: you can set the LIBGL_DRIVERS_DIR for every application you want the gallium driver for.
evil_core: egns: http://carme.pld-linux.org/~evil/radeon/dirhack/useful_scripts/
egns: thanks
evil_core: egns: or simply copy that scripts to /usr/local/bin, chmod +x, and run mesa_exec and gallium_exec
cTn: agd5f: looks like i really have to compile my own kernel
cTn: damn
egns: what would Jesus do ?
egns: heh
egns: dl'ing
agd5f: evil_core: any app that uses vsync interrupts would probably keep the CPU from entering higher C states
egns: evil_core: ./mesa_exec: line 10: : command not found
evil_core: agd5f: so using vsync is not recommeneded on laptop?
Droste: egns: the paths are also wrong for your system. you have to edit it
agd5f: evil_core: It's fine. Just might not last quite as long on battery
evil_core: egns: mesa_exec glxgear
agd5f: but I'm not sure how much difference it makes
egns: argh
egns: Droste: ok
evil_core: agd5f: I want longer battery life, because radeon sucks at it, got biggest battery in T60p and 3h is whatI usually got
evil_core: buyt using pageflio/vsync and not using compositing is not cutting battery life, right?
evil_core: I am not saying about plyaing games, but using vim, etc
evil_core: and I guess using pageflip/vsync on laptop and playing 3d game would helped because of frame cutting helping save much power than 60 vblank interrupts, am I right?
agd5f: evil_core: composting doesn't affect battery life. just things like vsync interrupts that fire a lot
evil_core: I heard in the past that generally OpenGL cuts battery life, so compiz does too..hm..
evil_core: I was wondering if should I choose xcompmgr or compiz
agd5f: evil_core: benchmark it
evil_core: I switched from wmaker to fluxbox, because everything in wmaker in separate process == many interrupts for docks
agd5f: evil_core: I doubt the window manager makes much difference
evil_core: I am not even sure If I need compositing manager, but I asterted playing with conky again, and would be nice to see it under xterm or urxvt ;)
cTn: agd5f: looks like latest ubuntu kernel has some implementation of this but all i can see is /sys/class/drm/card0/device/power/control
cTn: when i gedit control there is "on"
cTn: all i can find about this :-/
agd5f: cTn: not related to power management
egns: ah.
egns: seems that i wasn't using gallium or something.
egns: 50fps at glxgears.
egns: WTF.
Droste: egns: with gallium?
egns: yes
egns: at least now i have GL_ARB_texture_non_power_of_two
MostAwesomeDude: GINAB
egns: wat
MostAwesomeDude: Gears Is Not A Benchmark
egns: so i played openarena
Droste: oh is this the official abbreviation? :-D
egns: well, 'play' is a wrong way of calling it
egns: because staring at a frozen menu
egns: is not playing
MostAwesomeDude: Hm. Works for me. Are you sure you're using an accelerated driver?
MostAwesomeDude: Pastebin the output of this for me: $ LIBGL_DEBUG=verbose glxinfo
egns: k
egns: without gallium_driver, no?
MostAwesomeDude: With whatever isn't working for you.
Agrajag-: i have fixed my brightness issue by upgrading to from 2.6.32 to 2.6.33, i believe this https://bugs.freedesktop.org/attachment.cgi?id=34718 fixed it. i still have an issue with not all the modes of my monitor being detected though, i'm seeing "[drm:radeon_vga_detect] *ERROR* VGA-1: probed a monitor but no|invalid EDID"
egns: so with it
egns: wait
Agrajag-: to get my mode to work i have to do "xrandr --addmode VGA-0 "1280x1024; xrandr --output VGA-0 --mode 1280x1024"
egns: this is pissing me off
Agrajag-: again, this used to work, edid wasn't invalid before
egns: OpenGL vendor string: DRI R300 Project
egns: OpenGL renderer string: Mesa DRI R300 (RV350 4153) 20090101 x86/MMX/SSE2 TCL DRI2
egns: if i see that, is because i'm NOT using gallium, right?
Droste: right
egns: ._.
egns: so what the hell did i compile
Droste: did you make the symlink as I told you and ran glxinfo this way: LIBGL_DRIVERS_DIR=/usr/lib/xorg/modules/dri/gallium glxinfo ?
egns:
egns: the problem here is that i'm not agns, so i didnt read that
egns: :p
Droste: sorry :-)
egns: heh, no problem
egns: anyway seems that you are missing a 'cd gallium' there
egns: ooooh boy
egns: OpenGL vendor string: X.Org R300 Project
egns: OpenGL renderer string: Gallium 0.4 on RV350
Droste: now you can test glxgears: LIBGL_DRIVERS_DIR=/usr/lib/xorg/modules/dri/gallium glxgears
egns: 130 fps
egns: uhm
Droste: twice as much as before ;-)
egns: 850fps without compositing
Droste: sounds reasonable
svenstaro: er no :/
egns: no? uh
svenstaro: or well, for what glxgears measurement is worth, it probably is reasonable
Droste: go and check a game or something (start it with LIBGL_DRIVERS_DIR=/usr/lib/xorg/modules/dri/gallium
svenstaro: I get 2400 something on a shitty r300 card in UMS
egns: ah wait
egns: Running synchronized to the vertical refresh. The framerate should be
egns: approximately 1/1835103081 the monitor refresh rate.
svenstaro: wat
Droste: yeah you can't comapre KMS to UMS with glxgears
svenstaro: Droste: figured
svenstaro: egns: try crysis
egns: ...
svenstaro: :(
egns: that wouldn't work
egns: even with windows
MostAwesomeDude: agd5f: I just realized how the COL_CN bits in RS work; write backface B000 -> RGBA is pretty clear, what's write fbuffer XY00 -> RGBA ?
evil_core: so vsync sucks power, compositing is for transparency and to avoid tearing(which needs vsync), so generally avoiding tearing on laptop sucks?
MostAwesomeDude: Compositing doesn't figure into it, but yes.
cTn: agd5f: so i swtiched to http://kernel.ubuntu.com/~kernel-ppa/mainline/drm-next/current/ (which should have all i need)
cTn: dmesg |grep drm returns 8 power modes
evil_core: radeon vblank can be "extra_timer_interrupt" in powertop?
evil_core: vblank==vsync? after enabling PageFlip and not using opengl, am I loosing power in X11 too?
egns: urr
dgbaley27: I've disabled a monitor with video=VGA-1:d at the kernel command line. Now when I try to enable it with xrandr, nothing happens: xrandr --output DVI-0 --primary --output VGA-0 --auto --right-of DVI-0
dgbaley27: After running that, my Xorg log spits out some stuff about DVI-0's modelines
dgbaley27: Here's some output: http://pastebin.com/zWmf7Wf4
Droste: dgbaley27: try just: xrandr --output VGA-0 --auto --right-of DVI-0
dgbaley27: Nope
Droste: ah VGA-0 is disconnected
dgbaley27: It's physically connected though.
dgbaley27: I had an intel card where the same functionality worked fine. I could do video=
cTn: laptop is burning
cTn: weeeeee
cTn: :D
Droste: hm ok... there's no chance to enable it with xrandr if xrandr thinks there's nothing connected. if you don't add video=VGA-1:D to kernel command line does xrandr report VGA-0 as connected?
dgbaley27: Yes, but then it's also activated and the monitor on DVI is limited to the size of VGA
Droste: you can change the resolution for both screens (they can be different) with xrandr or xorg.conf
dgbaley27: Oh, yeah, the resolutions are different, but there's a lot of unused space on DVI. Get what I'm saying?
dgbaley27: fyi - tried to manually add a mode to VGA and enable it - didn't work
Droste: no really... are we talking about the console or X?
adamk_: Droste: His issue wasn't with the monitor resolutions in Xorg, but in KMS.
adamk_: In console.
Droste: ah
Droste: ok :-)
adamk_: The console uses the largest resolution available to both monitors.
adamk_: So he disables the VGA port via the kernel command line but I guess it won't reactivate later.
egns: how do i force the tty resolution to 1024x768? :S
dgbaley27: Yeah that's my problem.
adamk_: Sounds like something that just hasn't been implemented yet.
adamk_: egns: video=1024x768 on the kernel command line.
egns: k
dgbaley27: And I assumed it's a driver specific ability because it worked with intel
adamk_: I'm sure it is.
Droste: yeah... if that's the case, I'm out of ideas. File a bug? :-)
adamk_: Are you using the latest code in test?
adamk_: It might already work there. I really don't know.
dgbaley27: adamk_: I read a lot that you shouldn't use video= with KMS, is that an outdated suggestion?
adamk_: Otherwise, yeah, open up a bug report.
adamk_: Not sure where you read that.
adamk_: But I don't know why that would be the case.
adamk_: It's a perfectly valid way to set the monitor resolutions.
dgbaley27: Don't use vga= and video= is what i always have read
adamk_: As I said, I don't know where you read that, but it sounds like stupid advise to me :-)
dgbaley27: Can I do something like video=VGA-1:0x0,DVI-0:1680x1050
adamk_: You're asking the wrong guy :-)
dgbaley27: I've had a hard time finding the syntax of the video option
adamk_: At this point you know more about it than I do.
dgbaley27: Ok, well, I'll try using the latest driver
dgbaley27: I have to build the kernel module and the xorg driver together?
adamk_: No.
Droste: no
adamk_: dgbaley27: I really would be interested in knowing where you read not to use video= . I can understand not using vga= because that controls the vesa framebuffer, which you don't want to mess with if radeondrmfb is enabled, but the developers here still recommend video= for various things.
dgbaley27: http://wiki.archlinux.org/index.php/ATI#Early_KMS_start http://wiki.archlinux.org/index.php/Nouveau#Late_start
adamk_: So I'd like to correct whoever is saying not to use video=
dgbaley27: I am pretty sure i've seen it in places other than Arch though
adamk_: Bah.
adamk_: That wiki is crap, frankly.
dgbaley27: I like my arch wiki =)
adamk_: I know previously it was incorrect in it's discussion of radeon KMS, talking about it being enabled by default, and then disabled by default...
svenstaro: adamk_: name a better wiki
adamk_: So I wouldn't put much credence in it.
adamk_: I don't monitor wikis :-) Sorry.
adamk_: The one in the topic here was pretty good at one point, but I haven't checked it out in a while.
dgbaley27: Oh, back to building radeon, so I only need to build the kernel driver?
adamk_: Yeah, you would need to build a newer kernel.
adamk_: airlied has various branches with newer DRM code for testing purposes.
adamk_: But that's also the extent of my knowledge on the subject.
Droste: the advise to remove vga= and video= was/is because video= wasn't implement in the early versions of the radeon kernel driver.
dgbaley27: I have to rebuild the whole kernel? I can't just compile it against my current headers?
Droste: the advise to remove vga= is still true :-)
dgbaley27: Droste: thanks, I'll update the arch wiki then
dgbaley27: Droste: but you don't know if the video= removal necessarily applies to intel and nouveau however, right?
Droste: no, I don't know if it's true for intel or nouveau. sorry
dgbaley27: That's fine, the nvidia card in my laptop died, pretty much forcing me to get a whole new computer - i'm 100% ati now.
evil_core: it does too, because of KMS
evil_core: vga= is vesafb, am I right?
Droste: yes
evil_core: so it will bvreak KMS
chithead: in recent kernels, vesafb handover to kms is implemented
evil_core: didnt knowed
dgbaley27: .34?
dgbaley27: chithead: Is there any specific reason to do that, or is it 'just in case..'
evil_core: yay, power managmenent and pageflip is finally fixed for r500 :D
evil_core: except flashesh and xorg crashesh for classic window apps
evil_core: but gallium really rulez now
cTn: so looks like there is no power_state in ubuntu?
cTn: :x
sneakers: Do I still need to put "radeon.modeset=0" at the end of my kernel line?
sneakers: Using Fedora 12.
cTn: agd5f: still here?
MostAwesomeDude: sneakers: Only if it's necessary for things to work.
cTn: any idea how to compile latest ubuntu kernel with drm-radeon-testing ?
DanaG: yeah, I wish there was a ppa with d-r-t.
cTn: i am currently downloading latest ubuntu kernel sources via git but i have no clue what to do after :(
DanaG: You'd have to do some merging of git trees... I don't know how to do that.
DanaG: It's easier to just build the git radeon kernel as-is.
DanaG: er, all together.
cTn: me neither but i guess i am going to find out how to do so very soon
NTU: git clone
NTU: well... git pull
NTU: to compile kernel in ubuntu just use fakeroot make-kpkg
NTU: sudo apt-get build-dep linux && sudo apt-get install build-essential && fakeroot make-kpkg --initrd --append-to-version=-custom kernel-image kernel-headers kernel-source
woden: bout half the time when I startup Fedora, the system crawls very slowly. The screen redraws very slowly, menus open very slowly, typing goes slowly because the text shows up delayed, etc. Right now I just reproduced it. How do I debug it to find out what is the problem?
woden: I am using Fedora 12 and my video card is ATI Technologies Inc RV770 [Radeon HD 4850]
woden: Should I try the radeonhd drivers instead?
NTU: Have you installed mesa-experimental-dri or something?
NTU: It might speed things up. let me find exact package name. 1 sec
NTU: mesa-dri-drivers-experimental
NTU: also install the radeon firmware. xorg-x11-drv-ati-firmware
woden: NTU,
woden: rpm -qa |grep mesa
woden: mesa-dri-drivers-experimental-7.7-4.fc12.x86_64
woden: mesa-libGL-7.7-4.fc12.x86_64
woden: mesa-libGLU-7.7-4.fc12.x86_64
woden: mesa-dri-drivers-7.7-4.fc12.x86_64
NTU: mesa-dri-drivers-experimental. google lied to me sorry
woden: Should I be using the radeonHD instead, since I have an HD video card?
NTU: no
NTU: radeonhd has no KMS support IIRC and KMS is good :)
woden: Does that list look right?
NTU: uhm.. yes.
NTU: also install the firmware package :)
woden: NTU, I am using radeon.modeset=0 at the end of my kernel line.
NTU: oh KMS doesnt work for you?
NTU: why do you have that set, is what im actually asking
woden: rpm -qa |grep xorg-x11-drv-ati
woden: xorg-x11-drv-ati-6.13.0-0.21.20100219gite68d3a389.fc12.x86_64
woden: xorg-x11-drv-ati-firmware-6.13.0-0.21.20100219gite68d3a389.fc12.x86_64
NTU: xorg-x11-drv-ati-firmware that package
woden: NTU, Does ^ look right?
NTU: yes it looks right
woden: NTU, I am not using KMS because in my game (Urban Terror) I get really low FPS. Once I did that modeset thing (at the suggestion of a member on this channel) my frame rates went way up.
NTU: hmm... interesting
woden: But I still get this very slow system problem about half the time.
woden: (when I restart Fedora)
woden: Right now it is very slow, so I was wondering if there is a way to debug it.
NTU: so what does rpm -qa do? list currently installed packages or does it just scan for rpm's that you _can_ install ?
NTU: uhm.. debug X in fedora... yes its very simple since there is dbg packages for Xserver but wheather or not debugging X itself via ssh in gdb will help you, is another story
woden: NTU, Currently installed packages.
NTU: ohh.
NTU: so you already have the goodies :-)
woden: I have no idea why this problem is happening, or if it is even due to radeon driver.
NTU: well pastebin /var/log/Xorg.0.log and complete output of LIBGL_DEBUG=verbose glxinfo to pastebin
woden: #fedora suggested to use the radeonHD driver since I have an HD video card.
NTU: lets not... for now
woden: http://fpaste.org/WaX8/
NTU: xf86-video-ati has the PCI IDs for the 4850 and the card is perfectly supported in the -ati driver
woden: http://fpaste.org/K6hs/
NTU: does the screen or whatever still run slow with KMS when _not_ playing Urban Terror
NTU: ?
woden: Yes
woden: the problem is noticeable at the login screen even.
woden: when typing, the text takes a long time to show up. Menus take a long time to click open. Moving windows and the screen redraw is painfully slow, etc.
NTU: thats very strange
woden: when the problem is not happening, the system is very fast and responsive.
woden: it happens about 5 out of 10 boot ups.
NTU: so its random???
woden: yes
woden: rrandom but easily reproducable because it happens so often.
woden: to fix it i just keep rebooting until i get a good boot up
NTU: the only thing i know in KMS that is random is ppl that DanaG told me about
DanaG: I've seen radeon KMS lag horribly with certain pages in Firefox.
DanaG: Those pages must be doing something weird.
woden: this lag is everywhere on the system
woden: not just internet or web browser.
NTU: Do you think new_pll=0 would solve it?
NTU: or is there other random things besides the pll code?
NTU: *algo.. i should say
woden: NTU, If I removed radeon and installed radeonHD and the problem went away entirely, would that be helpful for debugging purposes?
NTU: I don't know the answer to that. I'd like to say yes but uhh..
NTU: wait.. before you do that, can you post dmesg really quick?
woden: sure
woden: http://fpaste.org/XMAA/
woden: I asked on #radeonhd and they can't tell me how to remove radeon and installed radeonhd on Fedora 12 :/
NTU: oh you dont need to remove radeon
NTU: just set "radeonhd" in xorg.conf
DanaG: Isn't radeonhd obsolete, anyway?
woden: NTU, I dont have an xorg.conf
NTU: DanaG: i think so, yes
woden: hrm.
woden: im stuck. i have no clue why this is happening.
NTU: neither do i
NTU: everything looks good...
NTU: maybe some of the people who work for redhat here can help you out
NTU: but i dont think any of them are actually here.
woden: am i supposed to have mesa-dri-drivers AND mesa-dri-drivers-experimental at the same time?
NTU: experimental overwrites classic mesa i think
woden: if i try to remove mesa-dri-drivers there are lots and lots of dependencies that want to be removed also
NTU: ok dont do that
woden: ok
NTU: is GL_EXT_texture_array the 1D/2D Texture arrays that are implemented in core mesa and swrast in the GL3.txt documentation?
MostAwesomeDude: I think so, yes.
NTU: thanks :D