Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-11-14

Search This Log:


EruditeHermit: nope
EruditeHermit: x 1.6
soreau: I guess I could try with dri1..
soreau: This just started last week and Im trying to find someone else with the same problem
soreau: or at least know why it happens
soreau: Its more an annoyance than anything else
soreau: EruditeHermit: Thanks for testing
EruditeHermit: i'll try with dri2 when I reboot
soreau: ok
Duskao: hello guys, I have a Radeon HD4850. Should I be using the Radeon driver or the RadeonHD driver?
Zajec2: Duskao: depends what you need
Zajec2: Duskao: radeon is... well,generally "better" whetever it means
Duskao: ok, well I'm generally trying the 3D capabilities of the video card with the drivers.
Zajec2: Duskao: radeonhd has soem nice features like HDMI and PM
Zajec2: Duskao: radeon has also PM, just not so good IMO
Duskao: So I should stick with the Radeon then?
Zajec2: Duskao: 3D does no depend on radeon/radeonhd choice
Duskao: ok
Zajec2: Duskao: any you want, really. i just preffer radeonhd for HDMI and PM and backlight, but that may mean nothing for you
Duskao: nope, I'm not using a laptop for the backlight, and I'm not using HDMI right now.
Duskao: can I install the catalyst driver over top of the Radeon Drivers even though I have the edgy PPA added?
Duskao: make that edgers PPA
Zajec2: Duskao: err, over top?
Zajec2: Duskao: you use radeon XOR radeonhd XOR fglrx
Duskao: I know that. But I mean when Ubuntu is installed it has the Radeon driver as the default does it not? Then you can install the fglrx Driver.
Zajec2: Duskao: sorry, i really know nothing about Ubuntu or fglrx
Duskao: ok
Duskao: Thanks
Duskao: I was just assuming I might be able to have both on my system (I know that could cause issues) and change the xorg.conf from "ati" to "fglrx" and vice versa.
Duskao: making the system use which ever I tell it to.
Duskao: perhaps it doesn't work that way.
soreau: That will not work for so many different reasons
Duskao: ok, thats why I asked before I tried it :D
soreau: General rule is if any part of fglrx is installed, the open drivers will be broken
Duskao: ok, thanks
Duskao: soreau: so is it possible to go back and forth at all then? install/uninstall fglrx to check how the Radeon drivers progress?
soreau: Not using whatever ppa stuff you have installed
Duskao: ok
soreau: fglrx is never compatible with latest kernel and x components
Duskao: ok
Duskao: yeah, they always seem to be a step or two behind.
Duskao: Thanks a lot, this is good info to know.
soreau: Not to mention for every step forward they take two steps back
soreau: Over the years the only thing fglrx has been good and consistent at is regressing
Duskao: Well, that might be entirely true, but their 3D is still a few steps ahead at this point with my vid card.
soreau: Yea we dont have GL2.1 quite yet
Duskao: You guys are doing great though.
soreau: and there is still plenty of optimizations to be done
Duskao: yep, thats true, HL2 is a slide show right now as is S.C.O.U.R.G.E.
soreau: whatever that means
soreau: I have a good feeling by time the next distro release cycle comes around next April, things will be looking pretty
Duskao: that means that about 1 frame per every 5 seconds or so with Half Life 2.
Duskao: I'll agree with you there.
Duskao: of course thats with wine.
soreau: why not use the native version? :o
Zajec2: does VRAM store it's content after powering down?
twnqx: for a split second...
Zajec2: :)
Zajec2: i've some weird locks up
Zajec2: 10 a day
Zajec2: started after I tried r600 for KDE4
Zajec2: but then it's unrelated
Zajec2: hope it's just my RAM that is broken
Zajec2: hm, removing one "bone" (?) of my RAM seems to resolve problem
twnqx: call it module.
Zajec2: twnqx: thanks
rhodan: Zajec2: I call them "bars"
hifi: umm, how do I force disable opengl vsync?
hifi: driconf didn't help
Zajec2: is this possible that single KDE4's widget can lock up my machine?
Zajec: maybe with some EXA operation?
gsedej: hi! Which is the best quick test for graphics? glxgeers is not good...
Zajec: gsedej: openarena maybe?
Zajec: google for "openarena demo benchmark"
gsedej: Zajec: Its 300MB...
Zajec: am i alive?
Zajec: ok, seems to :)
uzi18: :D
uzi18: Zajec: :)
Zajec: uzi18: s2ram test ;)
Zajec: Opera didn't break IRC connection somehow
uzi18: hmm
Zajec: gsedej: well, for real test you need something with advanced rendering
Zajec: gsedej: a lot of textures
uzi18: Zajec: not opera but it was standard lagg
uzi18: Zajec: this night i was launching almoust all tests form Mesa-master :P
Zajec: uzi18: passed?
Zajec: do we have dino shadow fixed?
Zajec: didn't test it for some time
uzi18: Zajec: yes but some of them are still unsupported
uzi18: Zajec: i'm using now xf86-driver-ati instread on radeonhd
Zajec: uzi18: i try to use KMS :P
Zajec: no much success... but still trying ;)
gsedej: I installed phoronix-test-suite (spent hour to enable gui, but failed :D) GLMark does not work (not even Vetrex Array)
uzi18: Zajec: looks like will be no kms for radeonhd ;/
Zajec: uzi18: yup, no reason to do that
uzi18: and it is porblem
Zajec: uzi18: it would be probably identical to radeon's KMS
uzi18: Xorg just hang on after launch ;/ and have insmoded radeon with mode_set=1
uzi18: (before)
Zajec: uzi18: ?
uzi18: this is why i'have moved to ati
Zajec: uzi18: do you mean hang when starting X on KMS?
Zajec: ah, some other issue
Zajec: err
uzi18: Zajec: yes
Zajec: but don't understand
Zajec: do you mean you've moved to ati non-kms?
uzi18: no from radeonhd to ati ;)
Zajec: because with KMS you have to use ati (radeon) anyway
Zajec: uzi18: ah, that's "expected" that starting X with radeonhd on KMS will give bad effects :)
uzi18: but radeonhd need workaround for this issue
uzi18: it will be pain for users
Zajec: uzi18: ?what pain? want KMS => use radeon
uzi18: Zajec: ok but user have insmoded radeon with kms (as it is almost mature)
Baivan: i know this may sound as a stupid question,but how to check wich graphic-driver im using?
uzi18: and launch X with radeonhd driver = hang
Zajec: kms mature... ? er
Zajec: Baivan: there are two kinds of drivers: 2D driver and 3D driver
uzi18: Zajec: works for me without issues
Zajec: Baivan: for 2D driver check your /var/log/Xorg.0.log (this can be radeon==ati or radeonhd for ATI cards)
Zajec: Baivan: for 3D drvier check glxinfo
uzi18: Zajec: i'm on all master gits kernel-zen, libdrm, Mesa, ddx ati
Zajec: glxinfo | grep OpenGL renderer string
Zajec: glxinfo | grep "OpenGL renderer string"
uzi18: driinfo ? :P
Baivan: so 2D is radeon and 3d is mesa
uzi18: Baivan: correct
uzi18: need restart :P
Baivan: is the Xfree86 better then mesa ?
uzi18: Zajec: direct rendering: No (If you want to find out why, try setting LIBGL_DEBUG=verbose)
uzi18: :P
Baivan: for an mobilyti radeon 9600
gsedej: Baivan: i have 9600m :P
Baivan: gsedej: wich driver u using for 3d?
gsedej: default on ubuntu 9.10 mesa 7.6
gsedej: xserver-xorg-video-ati 1:6.12.99
gimzo1: Baivan: Xfree86 is not a driver
Ingmar: Does the r600 driver need libdrm from agd5f/master, or is that only if I want experimental 3D for r600?
spreeuw: not sure at this point
spreeuw: about the lib
lagnus: help pls, http://pastebin.com/d621e0148
spreeuw: the kernel module must be the developement one
chithead: Ingmar: only use git master, do not use any branches or other repositories
spreeuw: or the 32 rc
lagnus: it is relation to kernel mode setting not working on my radeon 3200 HD using the git repositories
lagnus: i followed these installation instructions http://bbs.archlinux.org/viewtopic.php?id=79509
gsedej: is radeon driver using Gallium3D?
mjr: not the usual one, no
lagnus: so was just wondering if anyone can tell me if my card still isnt supported from the latest git
mjr: there's some ongoing work for r300g
lagnus: or if I made a mistake somewhere
chithead: gsedej: if you compile mesa with the right configure options, you will get gallium3d
adamk_: Baivan: Were you able to get things working yesterday after we talked?
Baivan: o yeah
Baivan: 2D improoved allot
gsedej: chithead: so I would get better performance in 3D?
chithead: gsedej: I don't know how you associate gallium3d with better performance
Baivan: adamk_:and i eaven updated my ports tree
Baivan: :)
adamk_: Woohoo :-)
gsedej: chithead: its newer :)
adamk_: Theoretically gallium3d could give better performance once it's complete, but that's quite a ways out.
gimzo1: gsedej: it's not yet finished
gsedej: ok :)
chithead: at present it will perform worse and support less 3d features
gimzo1: when we're at features, there's been some talk on phoronix about glsl working on r300, what's with that ?
chithead: see http://wiki.x.org/wiki/RadeonFeature for what works and what not
gimzo1: so that is the current status ?
lagnus: any help ?
lordheavy: lagnus: got it working with kms under archlinux with a 4200 HD (but without gallium driver) and it's working pretty well
lordheavy: with kernel-git too
lagnus: lordheavy, have you seen the two pages i linked to earlier ? What did you do differently?
lagnus: cool
gsedej: can someone explain me, why glxgeers is not good benchmark?
chithead: gsedej: it is the first hit if you google for "glxgears is not a benchmark"
gsedej: ok:D
lordheavy: not read, i've installed mesa-git (and git stuff) with this PKGBUILD http://pastebin.fr/5997
lordheavy: kernel-git (for 2.6.32-rc7) is useful to get proper DRM
osiris: agd5f: can you tell me if any of the TX_FILTER1_x reserved bits affect mipmap filtering quality?
gsedej: which benchmark do you use to test (R300)?
Ingmar: chithead: okay, thanks
adamk_: prefers openarena for benchmarking.
lordheavy: not tested Xreal.....
lagnus: lordheavy, ok compile kernel from git. i used drm-next before
adamk_: So when I tried 'git pull airlied_drm_remote drm-next' there were some conflicts in intel_display.c and intel_sdvo.c. Anyone know how to resolve them so I can get on drm-next?
lordheavy: not needed for me with the last kernel
lagnus: adamk_, posting those conflicts to pastbin and then posting the link here would help
adamk_: I'm not particularly familiar with git. All it said was that those files had conflicts. How do I get mor einformation on them?
unimatrix: hokay, so i've tested the UT2004 game for R700 and wanted to submit the results to your wiki, but it wont let me
unimatrix: could somebody submit it instead?
adamk_: None of the developers seem to be around at the moment.
unimatrix: well, whenever someone gets back, i'll be here all day :p
lagnus: lordheavy, thx for you help
lordheavy: i'm on #archlinux too if you need some help to build them
uzi18: adamk_: use kernel-zen :)
Zajec: think about discovering kernel bug :|
adamk_: Umm, yeah, I'd rather not.
adamk_: But, out of curiousity, how recent is drm-next in kernel-zen?
Zajec: can executing "dmesg" in loop cause any problem?
lordheavy: ttm segfault with mesa from git/kernel from git with openarena running fullscreen with kde4 and opengl desktop effects and 4200 HD chipset (no gallium) http://pastebin.fr/5998
gsedej: can someone tell me which if mesa-demos are good benchmarks and which of them?
adamk_: I doubt any of them are really good benchmars at performance. They are probably mostly used to demonstrate or test certain features and extensions.
Zajec: adamk: ++
adamk_: If you really want to benchmark, I suggest taking a look a this page and using one or more of the suggested items: http://dri.freedesktop.org/wiki/Benchmarking
dileX_: adamk_: unfortunately, there are files no longer downloadable and some URLs are dead.
adamk_: Yeah, anholt needs to replace his demos.
dileX: not only this
adamk_: But you can download the openarena files here: http://adam.npark.com/anholt.cfg http://adam.npark.com/anholt.dm_68
dileX: the radar demo for enemy-territory is not available and I didnt found it in the big WWW
dileX: I have that files already and did offer it here from my webspace
dileX: in general I would be pleased about a "standardized" benchmark which is "comparable" on the own machine and with others
adamk_: Then create one :-)
dileX: everyone is laughing at phoronix
dileX: they have a testsuite
dileX: but I didnt test
adamk_: Is it better than their reporting?
dileX: dunno?
gimzo1: dileX: I think phoronix testsuite is the only "standardized" benchmark
dileX: and its not only benchmarking, I lack also about informations on "this software/game is good for testing this/these feature(s)"
bjacques: Is the following code "legal"? :
bjacques: glEnable(GL_TEXTURE_2D);
bjacques: glReadBuffer(GL_COLOR_ATTACHMENT0_EXT);
bjacques: glReadPixels(0, 0, 100, 100, GL_RGB, GL_UNSIGNED_BYTE, buf);
spreeuw: dileX: I think open arena would be a good benchmark
spreeuw: for a general raw performance idea
bjacques: in particular, having GL_TEXTURE_2D enabled during this operation, which causes r300 to segfault.
spreeuw: replacing glxgears
spreeuw: but it requires cooperation of the openarena dev
dileX: spreeuw: I think OA is one favorite OSS game to test with
spreeuw: to build in a fixed config and demo
spreeuw: and a script to loop through 3 resolutions or so
dileX: and anholt has a benchmark-demo for that
gimzo1: how can you replace 20kb glxgears with 300mb oa ?
spreeuw: yeah except that that demo hasnt worked for years
spreeuw: its not compatible with the past years engine
spreeuw: 300MB is peanuts
gimzo1: not on liveCD
dileX: I dont know if you need all the data delivered with OA
jcristau: gimzo1: how can you expect a 20kb trivial app to be a useful/representative benchmark?
dileX: maybe it would be good to have a OA with anholts stuff
spreeuw: one of the glxgears clones are better though
gimzo1: I don't expect it to be representative, I just expect it to show me if the drivers are working :)
spreeuw: the one where you can pick the number of cogs and stuff
dileX: a "reduced" OA - but as I said I am not a gambler.
spreeuw: but i dot think it uses any textures
dileX: I only start these games for testing drm-radeon-kms
adamk_: gimzo1: It doesn't even do that well.
kdekorte: on my machine glxgears in software is 1000fps with hardware 2000fps
spreeuw: her 125 fps
spreeuw: with hw accel
spreeuw: depending on the window size
dileX: and parallelly radeon-gallium (with diverse state-trackers) with knowing gallium3d is very experimental technology
spreeuw: 250-125
twnqx: here 22000 in default size :>
twnqx: oh my, it's cpu bound?
spreeuw: which type of glxgears do you use?
spreeuw: there are multiple programs called this way
spreeuw: with multiple versions
twnqx: pets his gtx285
dileX: spreeuw: thats a good point - with which software components (kernel, libdrm, mesa, ddx, xserver) do ppl test.
spreeuw: that too
spreeuw: you need those versions
spreeuw: the application version
dileX: what IMHO would be a fine thing is a "benchmark results" and a data-field/comment which describes the system of the tester
kdekorte: I use stock fedora 12, except mesa, I use that from git
spreeuw: and the exact CPU type
spreeuw: and actually also the chipset and ram
spreeuw: if you're going to gather info in some table
spreeuw: and the kernel cpu settings
twnqx: so a 285gtx is about 12times as fast as a mobility radeon hd 3650
spreeuw: many use powersaving scaling which severely hampers performance
twnqx: i don't see a difference between 1.6ghz and 2.6ghz
spreeuw: twnqx: no experience with nvidia sry
spreeuw: twnqx: what application did you use to compare?
spreeuw: and which drivers
twnqx: the pure glxgears from mesa
spreeuw: the closed drivers dont count
twnqx: uh, they do
spreeuw: no ;p
twnqx: fglrx doesn't count as it fails to enable my screen
spreeuw: might as well run windows if you dont care
spreeuw: its a superior desktop
twnqx: nah
spreeuw: in all but the license department
twnqx: windows as an absolutely unusable gui
dileX: the published "benchmark results" shall document the approach in the development
twnqx: has*
dileX: or can
twnqx: click to focus, autoraise on focus
spreeuw: yeah the default wm behaviour of windows sucks
twnqx: => unusable
twnqx: also, no win+r + typing to lauch apps
twnqx: or generally, to mouse-focused
twnqx: too*
twnqx: also, no decent media players
spreeuw: vlc?
twnqx: come on
twnqx: vlc is... junk
spreeuw: my moms w2k has mplayer
twnqx: mine has mpc-hc
spreeuw: junk? lol
twnqx: also i hate windows with a passion
twnqx: i have one where i'm clicking a 100 tiems a day "do not fucking reboot" for days now
spreeuw: at home I use floss only
twnqx: also i hate if a computer asks "do you want me to do what you just told me to do"
spreeuw: for apps and os, do play the occasional closed commercial game on it
twnqx: for me the advantage of foss is limited to the fact that i can fix the bugs myself once they annoy me too much
solf0r: does the radeon driver support kms yet?
chithead: there is no release that supports kms. you have to build git master
solf0r: how do i do that?
adamk: solf0r: Check the wiki in the topic.
chithead: either check if packages already exist for your distro (ubuntu ppa, fedora rawhide, gentoo x11 overlay, etc.) exist, or git clone and build manually
solf0r: xorg edgers ppa just reverts to software rast if you enable kms even with 2.6.32 rc7
chithead: probably because libdrm_radeon or somesuch is missing
chithead: the packages in ubuntu 9.10 release are kms capable already
chithead: (for <=r500)
solf0r: im using an x1250 which is r470 i think
hifi: chithead: yes there is a release that supports kms
solf0r: what is it?
hifi: ubuntu 9.10 does
solf0r: ?
hifi: does that count as a release?
hifi: x1250 sounds like rv5xx
uzi18: Zajec: have dri again :P
chithead: x1250 is rs690. the packages that ubuntu 9.10 ships are fine. just make sure that the radeon kernel module is loaded with modeset=1 parameter
solf0r: i have tried and it reverts to soft rast with kms enabled
chithead: solf0r: do you have libdrm_radeon installed?
solf0r: radeon.modeset=1 yeah?
chithead: also check Xorg.0.log
adamk: solf0r: We'd really need to see the log file to tell you what's going on.
solf0r: how doi get that?
solf0r: do i *
chithead: either radeon.modeset=1 kernel parameter or options radeon modeset=1 in /etc/modprobe.d
solf0r: tried that and thats when if falls back to soft rast like ai already said :P
solf0r: where is that log file stored?
adamk: /var/log/Xorg.0.log
chithead: /var/log
solf0r: thanks
uzi18: solf0r: yes
uzi18: solf0r: paste this file somethere
solf0r: kk
adamk: Just not in here :-)
hifi: btw. anyone else than osiris have patches for mesa that might improve some performance on rv570 and need testing?
chithead: put Xorg.0.log in a pastebin and give the link here
solf0r: pastebin?
solf0r: lol ok
solf0r: http://pastebin.com/m770ad3e6
hifi: and does r300g work at all yet?
solf0r: are you talking to me?
uzi18: solf0r: heh you have got vesa driver ;/ ?
adamk: That log file is with KMS enabled?
adamk: Because it shows that DRI is enabled.
chithead: hifi: it works but it is slower and has less features than classic mesa still
solf0r: huh
solf0r: im running on...
solf0r: hold on
hifi: chithead: worth testing out on kms?
uzi18: solf0r: You need kernel 2.6.32 rc6/7
adamk: rnoland: All the DRM changes for r6xx/r7xx are in -CURRENT only, correct?
uzi18: solf0r: libdrm latest
solf0r: OpenGL vendor string: DRI R300 Project
solf0r: OpenGL renderer string: Mesa DRI R300 (RS600 7942) 20090101 x86/MMX/SSE2 NO-TCL
solf0r: OpenGL version string: 1.5 Mesa 7.6
solf0r: OpenGL extensions:
solf0r: i have that kernel
solf0r: libdrm latest?
chithead: hifi: depends on what are your expectations
solf0r: i don't undertand xorg
chithead: solf0r: you have not kms enabled in that log
solf0r: so many pieces O____0
solf0r: i know
uzi18: solf0r: just 3 of these
hifi: chithead: run quake perhaps?
solf0r: i disabled it because it fell back to softrast for desktop effects
uzi18: solf0r: kernel,libdrm, xorg-driver-ati
adamk: solf0r: We need to see the log file from when you were using KMS.
solf0r: shall i enable it now and reboot?
adamk: solf0r: Otherwise we can't tell you why KMS causes you to use the software rasterizer :-)
uzi18: solf0r: just enable it and restart X ;D
solf0r: shall i install the latest xorg edgers packages again too?
chithead: hifi: even if it works properly (which I am not sure of) it will be slower than classic mesa
hifi: classic mesa you mean *classic* mesa or kms?
chithead: classic mesa works with kms or without
uzi18: solf0r: dont know what You have :P
solf0r: ill add the packages brb :D
hifi: chithead: I have bad performance with half-life and I'm hoping for any improvements, disabling multitexturing boots up fps though
chithead: before gallium reaches feature and perfomance parity some time will still pass
hifi: chithead: what do I need to test out r300g, just --with-gallium the current mesa tree?
chithead: enable gallium, gallium-radeon and dri state tracker at least
solf0r_: okkkkkkkkk
solf0r_: kms enabled with xserver 1.6.5
solf0r_: and
solf0r_: OpenGL vendor string: Mesa Project
solf0r_: OpenGL renderer string: Software Rasterizer
solf0r_: OpenGL version string: 2.1 Mesa 7.7-devel
solf0r_: OpenGL shading language version string: 1.20
solf0r_: OpenGL extensions:
solf0r_: lol
adamk: Please use pastebin for more than a few lines.
adamk: And now pastebin your /var/log/Xorg.0.log file.
solf0r_: sorry i copied more than i meant too
solf0r_: ok
lagnus: what kind of performance boost are we looking at here ?
lagnus: for using kms i mean
solf0r_: http://pastebin.com/m72393bae
hifi: lagnus: none yet
hifi: it's actually a bit slower than without
lagnus: hifi, you imply that there will be at some point
lagnus: hifi, or was that hope that they will improve it/
hifi: hope
lagnus: *
chithead: solf0r_: the drm failed to initialize. probably some error will be reported to dmesg. pastebin that too
lagnus: me too
adamk: Oddd...
adamk: (II) [KMS] drm report modesetting isn't supported.
solf0r_: wheres that ?
solf0r_: huh
solf0r_: it is enabled
lagnus: lol kernel-git has been compiling for about an hour now
solf0r_: when i ctrl+alt+f1
lagnus: thank god for dual core
solf0r_: i get bothing but black
hifi: the driver doesn't utilize the hardware fully yet nor it's optimized (yet), it'll get faster in the future I hope
solf0r_: that only ever happens with kms enabled
adamk: solf0r_: What's the output of 'cat /proc/fb' ?
solf0r_: that a terminal command?
adamk: You definitely should be getting more than a black screen if everything is working fine.
adamk: solf0r_: Yes, any sort of terminal.
solf0r_: ok i get colors now
solf0r_: not pure black
lagnus: hifi, well i am sure it will be fine until then
solf0r_: nothing
solf0r_: it just goes to prompt
chithead: could be that you don't have framebuffer console enabled
lagnus: that is the usual lifecycle, the same goes with evolution lol
lagnus: but hey those boneless fish with pressurized oil that are believed to be one of our earliest ancestors are still around and kicking
lagnus: well swimming
adamk: solf0r_: 'cat /proc/fb' didn't display anything? Yeah, then there's some quite wrong.
adamk: Unfortunately I have no experience with Ubuntu for the last year, so I have no idea what they are or aren't doing.
solf0r_: should it have the quotations?
chithead: if fbcon was built as module, make sure that it is loaded
solf0r_: whats the point in kms anyway?
rnoland: adamk: should be in -current, 7-stable and 8-stable now
rnoland: too many branches going now....
adamk: rnoland: Oh, excellent. Thanks. I told someone earlier that I thought -current was necessary, but that it might be in 8-stable. I'll let him know.
adamk: Oh, I should clarify... Does that include 3D support?
rnoland: adamk: yes, all of the -stable branches are mostly in sync now
rnoland: 3d is not in rc3
adamk: Ahhh
solf0r_: when is opengl 2.x gonna be supported by xf86-video-ati?
bridgman: gl 2.x support would be in mesa not -ati
BeerSerc: Hi there. Are there any chipsets with onboard graphics which work stable with the radeon driver right now?
solf0r_: well whatever lol
solf0r_: all the parts confuse me )___o
solf0r_: mine is pretty stable beer
solf0r_: do you mean 3d?
bridgman: I don't think there's a date but all the bits are being worked on
BeerSerc: solf0r_: which chipset do you have
solf0r_: x1250
BeerSerc: sorry, I mean onboard chipsets
solf0r_: that is an onboard chip
bridgman: x1250 is onboard, isn't it (graphics in the northbridge)
BeerSerc: need to buy a mb for my fathers pc and it should work with ubuntu or something
chithead: x1250 is rs690 igp
solf0r_: i thought is was r470?
chithead: which has rv410 3d core (minus hardware tcl), iirc
solf0r_: whats the difference?
solf0r_: no tcl? O___0
solf0r_: sh*t
solf0r_: i didnt know that
bridgman: rs400/480 (aka Xpress 1100/200) have rv3xx core
bridgman: rs600/690 (aka X1200/1250) have rc410 core
BeerSerc: hm, but they are not sold anymore right?
chithead: you don't need hardware tcl anyway
solf0r_: its equivalent to x700 if i remember
bridgman: rs780/790 have rv610 core
BeerSerc: so theres no mb out there which just works with radeon?
bridgman: rs880 aka 785 have rv620 core
solf0r_: how do you man works?
chithead: BeerSerc: the die-shrunk rs690 is called rs740 and is still on sale
solf0r_: mean*
bridgman: anything with a 690 aka X12xx should work completely out of box
solf0r_: it does
solf0r_: except 3d is pretty crap
bridgman: I would go with 780 aka HD3200 which will work pretty well out of box but which will be better going forward
chithead: tv out maybe not
solf0r_: no offence :P
chithead: rs690 is slow by today's standards. for compiz/kwin desktop effects it is sufficient though
bridgman: tvout is worth mentioning; still struggling with tvout on igp parts IIRC
solf0r_: it can still play some games on windows pretty well
solf0r_: spore wow etc...
BeerSerc: chithead: bridgman I dont need tv out nor crazy 3d stuff, basically office stuff
solf0r_: it will owrk great for that
solf0r_: work*
solf0r_: the 2d is great :)
chithead: BeerSerc: go with rs780 (radeon 3200 igp) or newer
solf0r_: why?
chithead: solf0r_: the wine support in the open source driver is still very bad
solf0r_: yes
solf0r_: playing wow is impossible
solf0r_: sadly
chithead: the rs780 is an order of magnitude faster than rs690
BeerSerc: so something like http://www.heise.de/preisvergleich/a353506.html
BeerSerc: that asus board?
solf0r_: if hes dosent want 3d then theres no point
BeerSerc: solf0r_: I dont need much 3d
solf0r_: 2d on an 9550 would be fine
BeerSerc: may be possible my father wants google earth sometimes
BeerSerc: dunno
chithead: BeerSerc: would work, but you will bite yourself later if you buy ddr2 mobo today
solf0r_: then go with what the other dude said :D
BeerSerc: chithead: it's not for me ... it's for my father, so the difference between ddr2 and ddr3 would only be the price
solf0r_: when will the opso driver work better with wine?
BeerSerc: chithead: and all the am3 boards have chipsets 785, is that equivalently working?
lordheavy: yes got an 785G (4200 HD) and it's working
chithead: BeerSerc: 785g is the successor to 780g, it is a bit faster and equally well supported
chithead: solf0r_: I think wine is pretty low priority still
BeerSerc: OK. then I'll go for one of those, and if ubuntu is not shiny with it, I will come and smack you with the board! ;)
solf0r_: ok then when will mesasupport opengl 2 for the opso driver?
solf0r_: mesa support
bridgman: when it's done, but people are working on it
solf0r_: sweet!
solf0r_: no more vista then :D
chithead: solf0r_: it will probably still take a few months
bridgman: richard got the rest of the flow control instructions working yesterday (ie subroutines) and we're hoping to push that some time around the end of next week
solf0r_: aww
lordheavy: BeerSec got this one http://fr.asus.com/product.aspx?P_ID=fcsXWSxnhzZE9rnR&templete=2
Ghworg: Is the plan to write glsl support for classic mesa, or is that only going in to gallium?
bridgman: I think it will go into classic mesa as well
Ghworg: Cool
solf0r_: is gallium ever gonna be released? the r300 3d driver is ready for that isnt it?
bridgman: I think people are working on getting it ready to replace the classic mesa driver but it's not quite there yet
bridgman: lots of activity in the last few weeks though
solf0r_: lol thi is why xorg confuses me mesa xf86-xxxxx gallium libdrm kms O.O
solf0r_: this is*
BeerSerc: lordheavy: I need ps2 mouse and keyboard support ... but thanks
bridgman: http://cgit.freedesktop.org/mesa/mesa/log/src/gallium/drivers/r300
bridgman: go Corbin
bridgman: solf0r; there are only a few parts; the tricky thing is that nearly all of them are being rearchitected at the same time, so there are a lot of different *combinations* being discussed right now
bridgman: 2 years ago it was simple, and a year from now it will be simple
solf0r_: why the confusion?
bridgman: I wouldn't call it confusion; just the challenge of shifting all the components to a completely new architecture while keeping things working for users
solf0r_: ??O.o
solf0r_: lol
solf0r_: i still dont get it
solf0r_: nvm
bridgman: so you end up with transitional changes, eg the radeon_rewrite effort which let mesa work with both old (DRI1) and new (GEM/TTM/KMS/DRI2) kernel drivers (drm)
solf0r_: lolol
bridgman: do you know what the main components do ? If not it will be confusing
solf0r_: i dont undertans that either
lordheavy: BeerSec, there is ps2/usb adaptators, i've got one everytime i buy a new mouse !
solf0r_: understand*
hifi: I'm profiling half-life slowdowns, what would be the library of interest where I should look for slow calls?
BeerSerc: lordheavy: wrong way
BeerSerc: you get adaptors to connect usb mice to ps2
BeerSerc: but not to connect ps2 mice to usb
bridgman: ok, there are three main components - the X driver (aka DDX, ie -ati or -radeonhd)
bridgman: the 3D driver (aka mesa)
solf0r_: ok
bridgman: and the kernel driver (aka drm)
lordheavy: yes
solf0r_: no mnesa=no 3d
solf0r_: ?
bridgman: right
solf0r_: -n
BeerSerc: but I think I found something, asrock board with 785 chipset, hd4200
solf0r_: ok
lordheavy: http://www.ldlc.com/fiche/PB00037570.html in french sorry
BeerSerc: with ide and floppy and ps2 ...
bridgman: the X driver handles 2D and simple video playback
solf0r_: slightly clearer :D
solf0r_: yeah unlike the flgrz driver
BeerSerc: lordheavy: I know they exist
solf0r_: useless *****
bridgman: both X driver and mesa driver go through the kernel driver (drm) to access the acceleration hardware
solf0r_: ok
solf0r_: the x driver being xf86-video-ati?
bridgman: fglrx is actually structured almost exactly the same way, the only difference is that we wrap an installer around all the components to make it easier for users
bridgman: yes; -ati or -radeonhd
solf0r_: flgrx sucks
bridgman: so... 2 years ago the X driver accessed hardware directly for modesetting (ie programming the display portion of the chip), and owned all the memory on the video card
solf0r_: the dram?
bridgman: this made sense back when 2D was the main focus but was a problem for system security and for advanced 3d
bridgman: yep, although normally called vram (video ram) - it is dram though
solf0r_: is that why xserver will soon be rootless?
bridgman: just faster and more expensive ;)
bridgman: yep; kernel modesetting moves the display hardware control from the X driver into the kernel, so the X driver just makes API calls into the kernel
bridgman: so it doesn't need root privileges to access hw any more
solf0r_: isnt that more dangerous?
bridgman: ok, so that's KMS
bridgman: depends what you mean by dangerous; bad code with root privileges is always dangerous
bridgman: whether it's in the kernel or in an X driver
bridgman: moving it into the kernel makes it easier to write good code 'cause it's all in one place
solf0r_: yeah malicious code can go straight for the kernel?
solf0r_: anyway lol nvm that
bridgman: it also lets console and X use the same driver code
solf0r_: i dont undertans that
solf0r_: understand*
bridgman: right now X has a driver, console has a driver, directfb has a driver, and if you're very lucky they all get along
solf0r_: ok
bridgman: when you do a vt switch you're handing control from one driver to another, written by different people with different assumptions
solf0r_: so kms unifies them?
bridgman: yep
solf0r_: oh i see
Ghworg: One driver to rule them all, one driver to bind them
bridgman: next is memory management - GEM/TTM moves ownership of vram from the X driver into the kernel
bridgman: yep
solf0r_: thats good?
bridgman: makes it easier to share info efficiently between X and 3D efficiently, and provides the memory manager support required for things like VBOs, FBOs etc (the nice stuff people want in OpenGL)
solf0r_: ok!
bridgman: basically if you want OpenGL 2.x support you really need GEM/TTM
solf0r_: thank you
solf0r_: now it makes slightly more sense :_
bridgman: not finished yet ;)
solf0r_: :)*
solf0r_: lol
solf0r_: sorry
bridgman: Mesa has been around for a decade or more, and it has a standard API for hardware drivers
bridgman: inside the Mesa tree there are a bunch of drivers for different GPUs
solf0r_: mesa is 3d yes?
bridgman: yep
solf0r_: k
bridgman: so the Mesa driver is different from the X driver, even though they both do acceleration and both use the same graphics hardware these days (3D engine is used for 2D and video as well)
solf0r_: k
solf0r_: and gallium will repace it?
bridgman: Gallium3D is a new API which will hopefully allow the same driver to be used for 2D, 3D and video acceleration
solf0r_: replace*
solf0r_: uh?
solf0r_: so gallium ati will replace mesa and xf86-video?
bridgman: that means (a) gallium3D drivers need to be written, (b) the Mesa and Xorg drivers need to be modified to call Gallium3D rather than hitting the hardware directly
bridgman: parts of each, yes
solf0r_: lolol
solf0r_: i still dont really get that
bridgman: the gallium r300g driver will replace the current mesa r300 driver but most of mesa will still be there
solf0r_: why?
bridgman: remember that mesa is about a million lines of code and the r300 driver is only 20-30 thousand
solf0r_: isn't mesa outdated though?
bridgman: when people talk about it being outdated they're talking about the API between the upper levels of mesa and the hw drivers
bridgman: that API is really outdated since GPUs have changed so much
Pulsewidth: Does KMS automatically chose the EDID recommended mode? And if the EDID mode is wrong how do I specifiy my own modeline?
solf0r_: ok i think
adamk: Pulsewidth: Yes, it quieries the monitor.
bridgman: the Gallium3D driver API replaces the old mesa hw driver API, but most of mesa is still there
bridgman: OpenGL drivers are really big and complicated
solf0r_: for fallback?
adamk: Pulsewidth: If it's wrong, or you want to specify another mode, you can use video=DVI-0:mode when loading the radeon module.
bridgman: nope, most of the work is still done in the upper levels of mesa
Pulsewidth: adamk: thanks
bridgman: the HW drivers just handle the HW-specific bits
Ingmar: exit
solf0r_: ok
adamk: Pulsewidth: That apparently requires drm-next at the moment, though. I'd like to test it myself, but have been unable to do so.
soreau: has tested and knows it works
bridgman: so rather than mesa -> old hw driver -> drm we move to mesa -> gallium3d hw driver -> drm
Pulsewidth: What's the format for "mode"?
bridgman: now, take a deep breath
bridgman: KMS moved modesetting into the kernel
soreau: Alternatively you can put video=: as a kernel parameter, especially useful is you build the radeon module into the kernel
solf0r_: i get that
bridgman: Gallium3D provides a standard way to access acceleration
solf0r_: ok
bridgman: that makes it possible to write an X driver which doesn't know about hardware at all
bridgman: it uses KMS for modesetting and Gallium3D for acceleration
soreau: There was a good link detailing how to use it when it first came out
bridgman: that's the "xorg state tracker" you might have heard about
solf0r_: so the console in kms will be at the monitor res and not 640x480?
bridgman: a generic hw independent X driver built on KMS and Gallium3D
bridgman: I think so; KMS is supposed to run at native res AFAIK
bridgman: the defaults seem to work OK; what's weak right now is the ability to override
solf0r_: i think i sort of understand
solf0r_: is intel kms any better than ati?
bridgman: it's been around for 3-6 months longer so probably a bit more mature
bridgman: but they're all pretty new
Pulsewidth: I have a 120Hz LCD, and the EDID "native" mode is only 60Hz
Pulsewidth: I can override it in xorg.org with PreferredMode
solf0r_: ok one more question
Pulsewidth: Is the video=DVI-0:mode the same format as framebuffer modes?
soreau: Pulsewidth: For example, I have this as a kernel param 'video=VGA-1:800x600' even though xrandr shows my main monitor as VGA-0
solf0r_: how do i fix the artifacts in kubuntu when kwin is enabled?
Pulsewidth: soreau: The resolution is no problem, I want to give a modeline to set the refresh rate
solf0r_: it only does it when using the opso radeon drivers!
soreau: Pulsewidth: For X?
Pulsewidth: I already set mode for X, but I thought KMS let you have same mode for console and X
soreau: Pulsewidth: It does, not sure about refresh rates though. It probably automatically chooses what's best for your monitor by using whatever the bios does
Pulsewidth: soreau: That's no good because my monitor EDID recommends the wrong refresh rate
adamk: Yeah, hopefully it lets you specify the refresh rate... Why not try it and report back, Pulsewidth? :-)
soreau: Pulsewidth: Well you'd have to ask airlied or agd5f then. I can't seem to find the post that shows how to use it.
Pulsewidth: Which kernel should I use?
soreau: drm-next
adamk: You need drm-next.
adamk: Which I still can't get as of this morning :-)
Pulsewidth: I read the wiki, but it wanted me to clone the whole git repository
adamk: Yep.
soreau: indeed
Pulsewidth: And way to download just the latest snapshop?
soreau: no
adamk: Nope.
mjg59: You can pull the drm tree into an existing kernel git tree
adamk: And hope for the best :-)
solf0r_: no one knows about the kwin artifacts? :(
soreau: adamk: Now best my memory serves me, I have to do a fresh clone, do something with git merge before I tried to get drm-next pulled in
soreau: Is there some way I can grep the #radeon online logs? :P
mjg59: soreau: Shouldn't need to
soreau: mjg59: orly?
mjg59: soreau: Checkout a new local branch and reset it to the kernel that drm-next is based on, then just do a git pull
soreau: mjg59: Has merge conflicts
bridgman: Solf0r; which GPU and bus ?
mjg59: soreau: How?
soreau: mjg59: adamk can tell you
solf0r_: X1250
adamk: mjg59: At the present moment, these commands fail: http://pastebin.com/m46a1bf77
solf0r_: i dunno what bus
solf0r_: pci i think
adamk: mjg59: conflicts in intel_display.c and intel_sdvo.c
bridgman: that's OK, X1250 would be hypertransport with AMD CPU
bridgman: so it's not a bus issue
solf0r_: no intel cpu
soreau: I had the same problem and best my memory serves me, I had to end up doing some git merge magic
uzi18: Zajec: ok mesa tests works just just DRI_DRIVERS_DIR does not set before
mjg59: Yeah, drm-next isn't based on stable
bridgman: ahh, so Intel front side bus
solf0r_: i dunno lol
mjg59: You need Linus's tree
uzi18: solf0r_: what do you know ? :PPP
soreau: mjg59: Shouldn't that be changed in the wiki then?
solf0r_: random stuff lol
adamk: mjg59: Do you know the URL? And yeah, can someone update the wiki :-)
soreau: It's had this conflict for at least 3 weeks or more now
bridgman: not sure what current level of understanding is with rs600 (X12xx on intel), AFAIK we're still learning some of the details
adamk: Woohoo... The wiki has been wrong for a while :-)
mjg59: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
soreau: ;)
solf0r_: 01:05.0 VGA compatible controller: ATI Technologies Inc Radeon Xpress 1250
solf0r_: Subsystem: Samsung Electronics Co Ltd Device c034
solf0r_: Flags: bus master, fast devsel, latency 64, IRQ 28
solf0r_: Memory at c0000000 (64-bit, prefetchable) [size=256M]
solf0r_: Memory at d0000000 (64-bit, non-prefetchable) [size=64K]
solf0r_: I/O ports at 9000 [size=256]
solf0r_: Capabilities: [50] Power Management version 2
adamk: mjg59: Thanks.
soreau: good ghad
solf0r_: Capabilities: [80] Message Signalled Interrupts: Mask- 64bit+ Queue=0/0 Enable+
solf0r_: does that help?
adamk: solf0r_: Pastebinm!
soreau: solf0r_: Use a pastebin service, do not flood the channel
adamk: We've been over this.
solf0r_: sorry...
bridgman: it helps to annoy the other people on the channel ;)
solf0r_: lol
solf0r_: obviously :(
bridgman: are you using KWin effects via opengl or Xrender ? there's an option when you turn them on (a sub page IIRC)
bridgman: whatever you're using, try the other one
solf0r_: opengl
solf0r_: xrendr is too slow
bridgman: have you tried xrender ?
solf0r_: yep
bridgman: do you get artifacts ?
solf0r_: nope
solf0r_: but it's way too slow to us
solf0r_: too use*
dileX: solf0r_: normally, an lspci-output orgy starts - everyone wants to show my system is best
bridgman: the issue here is that the devs don't have your system, so the challenge is pinning down the issue enough that they can make a guess what might be wrong
solf0r_: lolol
solf0r_: an x1250 is not the best XD
bridgman: I'm serious; if the problem occurred on any of the dev systems it would probably be fixed by now
solf0r_: no
adamk: So who can update the wiki? :-)
solf0r_: it is actually a reported issue
bridgman: this is the same problem we have with fglrx; it works on the hundreds of systems we test with, just not with all the systems in the field
solf0r_: the ubuntu xorg devs are kinda sloppy though
soreau: adamk: Worked then I take it?
bridgman: there are ubuntu xorg devs ?
adamk: soreau: Still checking it out.
solf0r_: yea
adamk: Actually, I'll have to leave before it's done.
solf0r_: packagers/devs
bridgman: sorry, coffee hasn't kicked in yet
adamk: Sorry maybe I'll poke someone this afternoon if I get it working.
bridgman: goes to make another pot
solf0r_: do you have ant clue anyway?
solf0r_: any*
hbbs: Is there anyone else having problems with compiz and general slownesse using
hbbs: R500 hardware?
solf0r_: compiz no
solf0r_: kwin yes
hbbs: I'm using the latest kernel on Lucid Lynx
solf0r_: oh r500
solf0r_: nvm
hbbs: Which is compiled against the 2.6.32.rc7 kernel
hbbs: solf0r_, I'm having this issues on Lucid and Fedora 12
hbbs: Every single event lags
solf0r_: which isue?
solf0r_: issue*
solf0r_: i don';t get lag
solf0r_: i get artifacts
solf0r_: slight artifacts
bridgman: solf0r, you mentioned the problem was reported; is there a ticket # ?
solf0r_: no someone slse reported it ages ago
bridgman: this one may be specific to rs600, not sure
bridgman: reported where ?
solf0r_: thats why i say there sloppy
solf0r_: oh god
solf0r_: i last read that page about 4months ago
solf0r_: lol
bridgman: if there isn't a bug ticket active you can be assured nobody is looking at it
solf0r_: hold on lemme see if i can find it
solf0r_: lolol
bridgman: the devs are dealing with literally hundreds of issues at the same time
solf0r_: dont make me laugh!
solf0r_: they never look at bug tickets
solf0r_: "its an upstream issue"
solf0r_: ticket closed ¬¬
solf0r_: everytime
hbbs: A delay of a couple of seconds
hbbs: x1600pro APG over here.
hbbs: *AGP
bridgman: maybe it is; did you file an upstream ticket ?
hbbs: Which events, one may be wondering.. if I click on the gnome menu, or try to type on the gnome-terminal, all that takes a couple of seconds to actually happens.
solf0r_: no because it was ages ago and i went back to xp
solf0r_: i will see if ic an find it
solf0r_: ubuntu bug reporting is pretty bad
bridgman: all bug reporting is pretty bad; they call economics the dismal science but that has nothing on software development
solf0r_: no the ubuntu devs do it for reasons they can solve
solf0r_: several devs from other distros have said this
bridgman: if you pay big bugs for support you can sometimes find areas where there are enough people to actually deal with all the issues and keep everything up to date, but otherwise it's a small group of people trying to pick the biggest fires from a sea of tickets and deal with the big fires
bridgman: s/bugs/bucks/
hbbs: On Fedora 12 this delay is even worst.
bridgman: that's their job, isn't it ? dealing with ubuntu-specific issues ?
bridgman: they may also contribute to upstream but that's a separate task
solf0r_: i dunno but i lost faith in them ages ago and a lot of people say the same thing
bridgman: who do you think is better ?
solf0r_: are you refering to me?
bridgman: yep; I think you're really talking about the limits of what you can get for free rather than something specific to ubuntu
hbbs: airlied, agd5f, MostAwesomeDude So this general slowness (while compiz is on) is a known bug on R500 (AGP) hardware?
bridgman: when you buy stuff you pay for people to support it
solf0r_: i think you misunderstand what i mean
Zajec: bridgman: hi
bridgman: hi Zajec
Zajec: bridgman: could you say sth about releasing code for HDMI and IRQs?
Zajec: bridgman: do you know state of it's reviewing?
soreau: \o/ OpenGL version string: 2.1 Mesa 7.7-devel
Zajec: bridgman: i talk about what agd5f told he has written locally
bridgman: yep; for IRQs we think we have made all of the necessary changes to pass review; agd5f pinged the review team late last week (I should have done it earlier in the week but was semi-off-sick) and we haven't heard back yet
bridgman: my guess is early next week we should get go ahead but we never know for sure until the end
bridgman: someone could come up with a problem at the last minute
Zajec: bridgman: sounds nice so far :) thanks for info
Zajec: bridgman: hope you're fine already :)
Zajec: bridgman: and anything about HDMI?
Zajec: agd5f told he has some imporvements that may make audio working on GPUs which Christian was not able to RE
bridgman: IIRC agd5f sent me a big list of registers to look at and my eyes glazed over when I looked
Zajec: huh
bridgman: will check ;)
Zajec: that one sounds bad :P
bridgman: there's some kind of flu bug going around, turns the brain to cheese
bridgman: hopefully not permanent
agd5f: bridgman: yeah, I've had a lingering sickness for the last week or so myself
agd5f: everyone's had it
gimzo1: soreau: what is that ?
bridgman: you seem to still be thinking though ;)
Zajec: bridgman: ;)
agd5f: it takes a lot of work :)
bridgman: "I keep trying to think but nothing happens" ;)
hnsr: takes a lot of work even when im not sick :(
soreau: Cool, et and oa at least run with the g ;)
hnsr: but with enough caffeine i can manage
soreau: gimzo1: The renderer string has been updated in gallium to report GL2.1 now
agd5f: I got hw i2c working last night and was able to probe the thermal chip on my card, read regs off it, etc.
gimzo1: is oa usable ?
soreau: nope
soreau: not here at least
gimzo1: is there a guide, I'd like to try ?
Zajec: agd5f: wow, nice
agd5f: just got to write an algo driver for the amd hw so we can re-use the hwmon drivers
gimzo1: speaking of i2c, will it be possible to have ddccontrol working ?
gimzo1: it was possible in distant past, having radeonfb module loaded
bridgman: agd5f; nice !
agd5f: gimzo1: with some work. you'd need to expose the i2c bus somehow
soreau: neverball is even rendering textures better than last time I tried it
logari81: I ve just experienced my 2nd Xserver crash today with
logari81: Fatal server error: failed to map pixmap -85
logari81: running KMS on a r300 card
soreau: using a DE?
osiris: yay! I just got
osiris: prey working
quintela: wonders what "firmware" means in topic :p
quintela: having some trouble with second monitor + radeon 4350
hnsr: cool
hnsr: thats a doom3 based game isnt it?
hnsr: or id tech 4 i guess i should call it
soreau: quicksilver: What trouble
agd5f: osiris: cool!
quintela: soreau: if you meaned me, second head is only able to get a resolution of 1280x1024, and monitor allows 1600x1200
osiris: hnsr: don't know
osiris: if I only knew how to properly calculate the damn rowstride alignement for compressed textures...
zhasha: quintela: firmware is code that's loaded onto the cards command processor in order to interpret the packets of data we send correctly. you'll have no acceleration at all if you don't have the firmware
soreau: quintela: Oh yes, auto-tab complete phail :)
soreau: quintela: You will need to add this mode with xrandr
quintela: zhasha: where/how I get the firmawer? Fedora 11 if it makes any difference
zhasha: fedora ships it
zhasha: you have it unless you intentionally deleted it
quintela: performance is crap on glxgear, but idea was to use two dell 2309 (2048x1152)
quintela: for specs, it looks like second DVI's isn't to the task :(
quintela: I am going to use it just with lots of gnome-terminals, i.e. speed is not so important
quintela: zhasha: I haven't removed it explicitely. But glxgears speed is way slower than previous FireGL V3100
quintela: (1300FPS vs 280FPS)
quintela: yes, I know that glxgears is not the best test, but I didn't have anything else handy :(
quintela: not a 3D guy for any means
cxo: try Nexuiz they have an example timedemo you can use
cxo: just launch the game and then run, "timedemo demos/demo1.dem" from the console
zhasha: quicksilver: are you on an HD2000 card or higher and NOT using AMDs own driver (catalyst/fglrx)?
zhasha: sorry quicksilver, my bad
zhasha: quintela: are you on an HD2000 card or higher and NOT using AMDs own driver (catalyst/fglrx)?
quintela: zhasha: yeap
quintela: I am using the opensource driver
zhasha: I'd switch to the proprietary driver if I were you
quintela: zhasha: we can start the other way around
quintela: zhasha: 1st I would throw the card through the window :)
quintela: zhasha: let me put it this way
quintela: I want to have 2 DELL SP2309 monitors (2040x1152)
zhasha: the reason glxgears is so slow, is because swapbuffers is extremely slow. The rendering itself is way faster
quintela: I work for redhat on fedora (i.e. propietary drivers will never work for me)
quintela: what card should I bought?
chithead: all radeon cards except for the hd 5000 series are supported with 2d, xv and 3d by the open source drivers
quintela: chithead: but this one (4350) is not able to handle the 2nd monitor at 1600x1200
quintela: only at 1280x1024
gimzo1: when is 2d for hd5000 expected ?
quintela: card has two DVI ouptus
quintela: outputs
quintela: I bought a 4350 because it was fanless and had two DVI's, I can change it for another one if this one is not going to work :(
quintela: I don't use compiz/etc just plain gnome with lots of gnome-terminals
chithead: quintela: I know for a fact that 4350 card supports 1 dvi and 1 hdmi monitors at 1680x1050 each, because my colleague has such a setup
bridgman: quintela, 280fps sounds like sw rendering
bridgman: is this with multiple x servers and DRI1 ?
quintela: bridgman: how can I know if it is using hardware/no?
quintela: bridgman: this is Fedora11, I think it should be DRI2
bridgman: glxinfo and look at the renderer string
bridgman: depends on the GPU; F11 was DRI2 for older GPUs, DRI1 for newer
quintela: bridgman: you wint: Software Rasterized
bridgman: OK, now which GPU ?
quintela: bridgman: moving to F12 would help
bridgman: sounds like it
quintela: bridgman: 4350
gimzo1: what does 3d have to do with display resolutions ?
quintela: 01:00.0 VGA compatible controller: ATI Technologies Inc RV710 [Radeon HD 4350]
bridgman: yeah, that'll be DRI1 on F11 so no acceleration for 2nd X server
bridgman: until the display works nobody can work on acceleration ;)
chithead: if you want a multiseat setup, use a nested X server such as Xephyr
quintela: bridgman: moving to F12 would help? or it is an "unknown"
quintela: chithead: no, I want a single user setup
quintela: just with 2048x1152 x2 monitors
bridgman: quintela, do you get acceleration on the first screen ?
bridgman: and do you have a Virtual line set in xorg.conf ?
quintela: that is the setup that I want, what card do I need to geth this?
chithead: quintela: what happens if you add 1600x1200 mode to second had using cvt && xrandr --newmode && xrandr --addmode
quintela: bridgman: no :(
chithead: second head*
bridgman: can you pastebin your xorg.log and xrandr output ? you may just need a virtual line
quintela: bridgman: no Virtual line
bridgman: what does xrandr output say your maximum res is ?
quintela: bridgman: 2048x1152 on DVI-0
quintela: 12801024 on DVI-1
bridgman: there should be minimu, current, maximum
quintela: I am not able to launch glxgears on screen 2
bridgman: ok, let's get your displays sorted out first
quintela: I run it on a terminal in the 2nd screen, but it launchs the window in the 1st screen
bridgman: can you pastebin xrandr output
quintela: bridgman: sorry, I am unable to run xrandr on the laptop to get the ouptut of the other machine
quintela: coping
quintela: it puts:
quintela: Screen 0: minimu 320 x 200, current 3328x1152, maximum 3968x1920
osiris: I've just pushed pushed radeon-texrewrite-clean branch to my private repo
osiris: please test and review
osiris: it's almost finished (the only lacking thing is proper TFP handling)
bridgman: ok, that's why you can't get 2x 2048 x 1152
osiris: this mipmap rewrite reduces necessary texture image migration between MTs to minimum
quintela: bridgman: do you know what I should do to get that resolution?
quintela: what card will work?
osiris: changing GL_TEXTURE_MAX_LEVEL,GL_TEXTURE_MIN_LOD, GL_TEXTURE_MAX_LOD params doesn't bring any overhead now
quintela: I want opensource drivers
bridgman: yeah, I think you need that Virtual line ;)
quintela: how do I get that virtualline?
osiris: also changing GL_TEXTURE_BASE_LEVEL (as long as the new base level is higher than current)
osiris: nha, airlied, glisse: ^^^
bridgman: in xorg.conf, hold on I'll find the details
osiris: also all the errors when migrating textures should be fixed now
bridgman: separate issue is why xrandr isn't showing 1600x1200 as an option, can you pastebin your xorg log ?
_entropy: Hi! I'm still seeing this http://imagebin.org/71137 with latest (master) libdrm, mesa and xf86-video-ati. Linux is 2.6.32-rc7 and/or Linux-next (I admitt, the last one I tried was from monday).
Ingmar: (WW) RADEON(0): Color tiling is not yet supported on R600/R700 <- Could someone explain what color tiling does?
_entropy: GPU is 4850, xorg-server is 1.7.1, KDE 4.3.3
osiris: piglit/lodclamp and prey game are now fixed
bridgman: 3D rendering normally uses two buffers, one to hold the pixels (color buffer or CB) and one to hold Z info (depth buffer or DB)
quintela: http://trasno.org/para/bridgman/Xorg.0.log
bridgman: until recently the driver used tiling on the depth buffer but not the color buffer
bridgman: I thought agd5f added color buffer tiling a month or so ago
bridgman: tiling involves shuffling address lines inside the chip so that "close together" memory addresses represent blocks on the screen rather than horizontal lines
bridgman: it lets the chip access video memory more efficiently
bridgman: but it's a pain when the CPU needs access to video memory because now it's mapped all funny
Ingmar: bridgman: interesting, thanks for explaining
bridgman: if you pick up the latest radeon code I think color tiling *is* supported
zhasha: when did that happen?
Ingmar: bridgman: even on RV770?
Ingmar: afaics, i'm using xf86-video-ati/libdrm/mesa all from git, on linux-2.6.32-rc7
hifi: osiris: doesn't build, do I need something extra?
osiris: hifi: what's the error?
bridgman: I thought so; going from memory though
bridgman: quintela; Virtual line goes in the Display subsection of Screen AFAICS
hifi: radeon_texstate.c:1024: error: ‘struct _radeon_mipmap_tree’ has no member named ‘firstLevel’
hifi: same for lastLevel
cxo: hello
hifi: also some warnings about incompatible pointer type arguments for radeon_miptree_unreference
cxo: I get this error when I try and run warsow http://pastebin.com/m23ab558b
cxo: it crashes and leaves my desktop at a different resolution
bridgman: quintela; try Virtual 4096 4096
osiris: hifi: uuh, darn. I'll need to align the r100 and r200 too. build swrast and r300 drivers only
hifi: I think I did
hifi: oh, radeon was for r100 and r200
bridgman: quintela; http://wiki.debian.org/XStrikeForce/HowToRandR12 is good reading
bridgman: search for "virtual" in the doc if you're in a hurry ;)
quintela: bridgman: looking at it, only using the virtual line don't help.
quintela: basically gnome-display-properties take over, and whatever is in xorg.conf is ignored
bridgman: hmmm... I can't really help with gnome stuff
bridgman: or KDE stuff, I'm equal-opportunity-ignorant there ;)
bridgman: you'll probably need that virtual line when the 2nd big screen shows up though
bridgman: anyways, for a 4350 you want F12 plus the "experimental" mesa driver from the repo
bridgman: there's a Phoronix article with details :
bridgman: http://www.phoronix.com/scan.php?page=article&item=fedora_r600_3d&num=1
hifi: osiris: +15fps for quake 2 in a high polygon area, gj
osiris: hifi: that was unexpected, but I'm glad to hear that :)
hifi: I need to retest that
quintela: found it
hifi: half-life maybe got a bit faster, probably not
quintela: bridgman: just to let you know
quintela: gnome display settings is very nice
quintela: and you can drag and drop where you put your displays relative one to the other
quintela: so far so good
quintela: if you put the 2nd monitor on the right, it gets a smaller resolution that the left one
quintela: you just move it to the left, push "detect monitors buttons" now res appears, and you can move it to the left now
quintela: bridgman: as you can see, completely obvious :p
hifi: osiris: my bad, no change
hifi: done so many tests today I forget what was the stock benchmark
quintela: bridgman: thanks very much
hifi: osiris: no boost for hl or q2, but they both at least run with your patch
bridgman: glad you figured it out; if UIs are made too intuitive then everyone stops thinking ;)
_entropy: http://imagebin.org/71137 any ideas, anyone?
bridgman: entropy; are you referring to the row of colored dots in top left hand corner ?
quintela: bridgman: ok, now it allows me to use the 2048x1152 monitor in both DVI outputs, strange
_entropy: hi! yes, and the black grandient things =)
mikkoc: _entropy: i have the same with kms enabled
quintela: bridgman: and the rest, thanks very much for your help!!!!!
mikkoc: it sux, im stuck to non-kms
_entropy: ah, ok - kms is enabled here aswell.
hifi: osiris: Mesa 7.7-devel implementation error: radeon_validate_texture failed to alloc miptree
hifi: when I tried to run lodclamp
_entropy: I thought this was considered to be fixed.
_entropy: Same setup with r300 and KMS works perfectly fine.
hifi: back to tv ->
osiris: hifi: strange, it works for me on rv530
osiris: hifi: what file/line does this assertion fails?
bridgman: Ingmar; agd5f just corrected me, color tiling for 6xx/7xx has *not* been enabled yet
Ingmar: ah :)
bridgman: apparently being able to access the fb from sw *is* important ;)
cxo: is going to see if 3dmark 2000 works with the radeon driver
Vinky: do I need to have drm_kms_helper loaded to get kms working, got a lot of unknown symbols when trying to load it, for example drm_mode_vrefresh
cxo: shouldn't it just pull that in if it needed it
Vinky: it loaded it before I recompiled stuff when I just got a blank screen
cxo: depmd?
cxo: ^depmod
Vinky: anything special to look for, cant find the symbol it looks for when doing depmod -n drm_kms_helper.ko
Vinky: http://dpaste.com/120480/
DanaG: Interesting... nexuiz renders pretty well, though slowly, and you can't see any of the other players. Weird.
DanaG: oh, and where was that clock info supposed to go?
bridgman: DanaG; last time I was playing Nexuiz I could see the players, are you running latest code ?
bridgman: didn't help though, players still killed me
bridgman: this was DRI1, not DRI2
DanaG: I'm on DRI2.
DanaG: Not sure... I also have visual settings cranked up (aside from now turning off GLSL, of course).
DanaG: 2.6.32-rc7, and ... oh wait, my DDX is confused. ah.
DanaG: I'm mixing Karmic and Lucid repos, that's why.
Vinky: reinstalled drm.ko and drm_kms_helper loads but not radeon.ko
DanaG: Oh, and it still takes tons of power. :(
Vinky: missing drm_calloc and drm_cleanup_pci
bridgman: yeah, the power saving code is only in the DRI1 paths AFAIK
bridgman: that'll change though
cxo: You wanna know something funny? Wine shows my r700 as an Nvidia FX5600
DanaG: I'm fine with just static "keep in low-power mode" for now.
cxo: wonders if it'll work faster now
DanaG: wishes fglrx used alternatives, instead of diversions, for libgl.
DanaG: s/gl/GL/
cxo: 3dMark2001 seems to work, but its horrendously slow
cxo: 0fps
Vinky: brb
bridgman: DanaG; don't think KMS has static power saving either, that's only in DRI1 paths
bridgman: with KMS you can have dynamic power management, but that's more work to implement so right now you don't have anything ;)
gimzo1: I think ideal power management would be an applet with "low power" "full power" selection :)
DanaG: Right now, the power usage is a showstopper for me, even when on AC... It makes the laptop noisy.
DanaG: oh, updated libdrm, restarted... nexuiz is faster, but still can't see opponents.
agd5f: DanaG: the missing stuff in nexuiz is a app bug
bridgman: gizmo1; the problem is the code that actually hits the hardware; it needs to be in the kernel so it can quiesce 2d and 3d acceleration operations first, so the current UMS code can only set it at startup and not make on the fly changes
DanaG: http://pastebin.com/f48e19536
DanaG: I'd be fine with "at startup" for now. =þ
bridgman: now that KMS is running it's possible to write more capable code in the drm but AFAIK priority has been getting KMS stable before adding features
DanaG: there's my glxinfo.
bridgman: yeah, I think the issue there is just lack of an easy way to get config options into the KMS driver, not sure if that has been addressed yet
gimzo1: couldn't it be done like cpu frequency switching, ie kernel provides interface and userspace app just interacts with the user ?
DanaG: hmm, and why is there a do-nothing dynclks parameter on the radeon module?
agd5f: DanaG: it does stuff for r1xx-r5xx
DanaG: tries fixing his libgl drivers path...
DanaG: zap:
bridgman: gizmo1; I think that's the plan, but again that's more work than a kernel only solution and so far the devs have higher priority stuff to work on
DanaG: ah, now I do have DRI.
DanaG: And, now I can see the other players.
bridgman: has already forgotten what problem DanaG was trying to solve, decides to go back to bed
DanaG: Couldn't see opponents in Nexuiz.
DanaG: Turns out it was being indirect, because I had messed with libgl_drivers_path.
gimzo1: bridgman: I'm not asking for it to come faster, I'm just wondering what it will be like
DanaG: Wow, the performance is pretty damn good on nexuiz. Now, the power really IS the last reason I have to use fglrx.
bridgman: gizmo1, it will probably be "all of the above" - first something only in the kernel, then some kind of API added so apps can control power, then once the limits of apps are reached more decision making will go back into the kernel, requiring new apps, and the...
DanaG: Hmm, how long would it take to add a minimal "set this speed, before we do anything else" statement? I really don't care about API and such.. I just want to manually make the thing not be running full bore -- even if it means it's slow.
bridgman: dunno... probably a few hours to make it work on the first system, then a few hours to make it work on the second and third systems, then a few hours to make it work on the 4th through 8th systems, then a few hours to...
DanaG: ah.
bridgman: If the current KMS code actually sets the engine clock then it would probably be pretty easy to hack that code and force a lower clock speed
bridgman: if it uses bios defaults then it would be a bunch more work
amarks: how much power is kms really burning? my aeroplane runlevel with kms on my thinkpad is under 9W (intel gpu)
bridgman: I think it depends on the size of the GPU; small chips draw a lot less power even running at full speed
DanaG: With fglrx, idle is like 20 watts... but with radeon, it seems to be like 40 watts.
DanaG: That's +/- 3 in either direction, though.
bridgman: does anyone remember if there's a cgit interface to latest kernel source ?
DanaG: ah, 36W.
amarks: linus-tree?
agd5f: bridgman: git.kernel.org
bridgman: got it
bridgman: thanks
DanaG: Anyway, for the sake of my sanity (and removing fan noise), I'm rebooting to get fglrx back, for now. At least with karmic+lucid repo mix, I get 2.6.32 kernel and KMS-enabled libdri, yet keep fglrx-compatible X server.
amarks: i guess a nerd knob for power control would be nice
amarks: even if it's crude
amarks: maybe a /sys/devices/ thing
DanaG: heh, maybe I should run an irc proxy somewhere.
amarks: or /proc/acpi/video
DanaG: So I don't keep missing things.
bridgman: honestly, once the code is written there are a bazillion ways to use it
bridgman: someone has to write the initial kernel code first
glisse: /win 1
elbeardmorez: hi. i'm having dvi->hdmi issues. i've added the modelines which are detected.when the vga port is connected to the plasmas vga, and can then get xrandr to give sensible output. but the screen remains blank, whatever mode i change to. could this be a hdcp problem? is there anyone here who has a working dvi-hdmi setup with a pioneer plasma? thanks.
cxo: 40 3Dmark 2001 points, haha
_entropy: bye!
bridgman: realizes he doesn't know exactly where the KMS code lives in the current kernel tree
glisse: drivers/gpu/drm/radeon
MostAwesomeDude: drivers/gpu/drm
bridgman: yeah, that's what I thought, but there are only about 8 files and none of them seem to be modesetting stuff
bridgman: looks again on his pitifully slow dialup line
amarks: sends more bytes down bridgeman's line
bridgman: noooo !
Ghworg: Lets DDoS bridgman by all typing at once ;-)
bridgman: noooo !
amarks: fix a bug or be ddos'd! lol
MostAwesomeDude: No, let us not do that.
bridgman: I think I was looking in the wrong project
bridgman: 2.6.29 stuff
amarks: has ordered a NEC 3090WQXI, looking forward to 30 inch niceness
bridgman: ok, this is better...
amarks: what product do people here use for monitor calibration?
bridgman: mark 1 eyeball ?
amarks: lol
yangman: amarks: I use a Pantone Huey with ArgyllCMS
amarks: i read NEC stores the profile in the monitor and that it is not required to use the video card to do the tweaking
DanaG: wonders when the heck somebody will make a >=120dpi desktop LCD.
DanaG: The highest I've ever seen on desktop LCDs, is 109 or so.
amarks: yangman: good experience?
yangman: fancy. sounds like that would need proprietary software
amarks: yangman: apparently part of their spectraview s/w
yangman: amarks: as gooas it gets on linux, I guess. by which I mean I first had to fix bugs in the driver by rewriting the LUT loading code ;)
amarks: did you submit the fixes upstream?
yangman: yeah, that was ages ago
amarks: ok so should just work you reckon
yangman: indeed
amarks: cool thanks
bridgman: DanaG; my cursory glance at the code makes me think that KMS doesn't set engine clocks at all yet, ie that bios defaults are used
DanaG: ah, and HP's BIOS defaults to full-power. bleh.
amarks: what is the bios default? max?
DanaG: I think so.
DanaG: I've viewed the vbios in RaBit (radeon bios editor), and that seemed to be true.
bridgman: there's no standard; every hw vendor does something different
bridgman: I keep seeing references to radeon_static_clocks_init being called during 6xx/7xx setup but haven't found it in the code yet
amarks: what is req'd at the hw level? flip some bits?
bridgman: atombios call to set the plls
Ghworg: bridgman: Its in ./drivers/gpu/drm/radeon/radeon_clocks.c at the end of the file (on drm-next anyway)
bridgman: looks *again*, thinking more sleep would have been a better idea
bridgman: wishes that Linus's last name started with A rather than T so he didn't have to wait for the whole page to load
bridgman: Ghworg; it's there now, wonder how that happened ;)
bridgman: still looks like 6xx/7xx use bios defaults and don't have engine clock setting code
Ghworg: bridgman: The code fairy must have visited just after you looked the first time
bridgman: that's my theory too
bridgman: happens mostly on the weekend
amarks: so bridgman, want to save the environment with some green code, or would you rather sleep :)
bridgman: sleep
DanaG: Or sleep now, and fix it later -- after writing on a sticky-note so you don't forget?
DanaG: I tend to often think of stuff right before I go to bed... and then never ever remember it when I wake up.
bridgman: then I'm going to cut down trees and drive around in an SUV ;)
amarks: but it's OK right, you have or will have an evergreen chip
bridgman: I don't have one, but that doesn't matter
bridgman: agd5f has one, and that *does* matter ;)
bridgman: he gets the new toys first
MostAwesomeDude: I also don't have an Evergreen and TBH I don't want one.
amarks: indeed
MostAwesomeDude: I've already got enough toys.
amarks: supposedly low idle power!
MostAwesomeDude: Or, as my mom used to say, "I'm not buying you more Legos! You keep leaving them out, and then we step on them, and it hurts!"
bridgman: if the driver sets it that way
bridgman: ;)
amarks: bridgman: ssshhh ;)
DanaG: oopsie, note to self: don't run window managers over ssh.
DanaG: =t
DanaG: =þ
amarks: i read lego is making a comeback
bridgman: especially in video games
bridgman: I think there's a toolkit that lego-izes your drawing commands
agd5f: the atom set engine/memory clock functions are there in kms. you could just add a call to set a different clock
agd5f: there for pre-atom as well
amarks: and all we need is hitchhiker's toolkit and we're complete
amarks: is there a way to read graphic thermals/clock values etc?
amarks: for lm_sensors etc
hnsr: that would be neat to have something like that
hnsr: because right now I manage to get my 4850 to spontaneously reboot my PC unless i limit the FPS so I suspect it might be some heating issue
airlied: osiris: look good to me, from a quick review
bridgman: agd5f; that's what I thought, just couldn't find it in the code ;(
agd5f: bridgman: it's defined in radeon_atombios.c
agd5f: asic specific functions are set in radeon_asic.c
bridgman: thanks
airlied: osiris: r600 uses some of that code also
osiris: airlied: hopefully it's not much to update there
maleadt: is there any news yet on the bug that prevents r600 users from clicking anything in clutter-based apps (FDO 24408)? Kinda ruins the gnome-shell experience
bridgman: agd5f; just so I understand, the atom calls are defined but not actually used today, is that right ?
bridgman: for setting engine clock in kms
agd5f: bridgman: yeah
bridgman: got it... I was looking for where they were used and losing big time ;(
airlied: maleadt: I can clock on things, I'm running g-s on my desktop
airlied: click even
airlied: if you are using Fedora the mesa in koji is what I'm using
airlied: its not perfect thouh sometimes it seems to lose clicks
maleadt: airlied: ah ok, I just enabled updates-testing but no luck, I'll have a look at mesa in koji
kdekorte: airlied, I still get a lot of warnings from clutter even using mesa from git
kdekorte: They look like this: (fosfor:9394): Clutter-WARNING **: The required ID of 4013373 does not refer to an existing actor; this usually implies that the pick() of an actor is not correctly implemented or that there is an error in the glReadPixels() implementation of the GL driver.
airlied: kdekorte: ah so we must still have some bug
airlied: probably some flushing issue
kdekorte: Works fine on r5xx, so it is only r6xx and higher
maleadt: installed it, gonna test
kdekorte: And clutter + gstreamer is dog slow on r5xx... hoping the work from osiris helps with that
maleadt: airlied: seems to work now, though not always. As you said many clicks get lost
airlied: need to geta piglit test for clutter picking I suspect
maleadt: arlied: another thing, friend of mine got severe plymouth corruption (no animation at all, just a framebuffer dump of the POST screen), though the system is fully functional when it hits GDM. Also on fedora 12, RS780 chip
maleadt: lspci @ http://pastebin.com/m6abd43d7, xorg.log @ http://pastebin.com/m237786cc
EruditeHermit: bridgman, HI
EruditeHermit: err
EruditeHermit: hi
EruditeHermit: airlied, hey, the last patch I tried didn't work
nha: osiris: just saw your mipmap work, cool :)
osiris: nha: thanks
soreau: Hi nha :)
osiris: nha: there's just one thing I can't figure out how to fix. in mesa/progs/tests/mipmap_limits you can see that blending between two mipmap levels is pretty ugly
soreau: nha: I am assuming it would be a mesa bug that caused compiz magnifier fisheye mode to break. Would this be a correct assumption?
soreau: If so, I will attempt a bisect
osiris: nha: and I believe this problem is also affecting piglit's lodclamp-between
nha: soreau: yes, that's a pretty safe assumption
nha: osiris: is this a regression of your work or did it happen before?
nha: lodclamp-between is a fairly new test iirc
osiris: nha: no, it happend before too
nha: oh okay
nha: then it's quite possibly some problem in the way the hardware is programmed
soreau: nha: ok I will try to track it down then, thanks
nha: if you can get an old version of fglrx to run it would also be useful to compare that
osiris: nha: no way. my distro is too new
nha: i recall that at least with cubemaps there is some mipmaps strangeness in the hardware itself
osiris: nha: yeah, I've been trying to fix it too without a success
nha: as I feared :/
osiris: also properly calculating rowstride alignement for compressed textures is giving me a headache
eosie_: osiris: I've tested fglrx, it doesn't pass lodclamp-between
osiris: eosie: could you put somewhere a screenshot of mesa/progs/tests/mipmap_limits?
eosie: osiris: with fglrx?
osiris: yes
eosie: I need to reboot then...
eosie: osiris: http://storm.unas.cz/mipmap1.png http://storm.unas.cz/mipmap2.png
osiris: eosie: thanks
osiris: looks like it's broken there as well
osiris: eosie: btw, what card is it?
eosie: osiris: RV530, i.e. Mobility X1700
eosie: octe: it might be worth noting that mipmap_limits uses GL_NEAREST_MIPMAP_NEAREST
eosie: octe: sorry
eosie: osiris: it might be worth noting that mipmap_limits uses GL_NEAREST_MIPMAP_NEAREST
osiris: eosie: switching it to linear doesn't help
osiris: eosie: see how the software renderer does this
eosie: osiris: replace "if (NearestFilter) {" by "if (0) {" and see the magic
eosie: I see, the InitValues function re-initializes NearestFilter
_Groo_: hi/2 all
_Groo_: agd5f: ping
_Groo_: airlied: ping
eosie: osiris: how exactly did you change the filter to linear? ;)
osiris: eosie: press f
cxo: Is there any good 3d stuff in the pipeline for radeon, coming out soon?
osiris: cxo: yes
eosie: didn't know that
cxo: great
eosie: osiris: yes, it works with fglrx
osiris: eosie: can you create a screenshot for linear?
eosie: sure
osiris: cxo: harware blit support which will give huge speed boost for apps using glCopyTex[Sub]Image, and glReadPixels, glCopyPixels if I will find more time
cxo: Is that whats needed for Alien Arena?
eosie: osiris: I thought glReadPixels cannot be sped up without PBOs
eosie: osiris: http://storm.unas.cz/mipmap3.png
osiris: eosie: it certainly would be much faster with PBOs, but even without PBOs we will gain some fps because current (software) implementation is very slow
eosie: I see
_Groo_: hey osiris
osiris: _Groo_: hi
cxo: I get 42 points on 3dmark 2000 and 40 points on 3dmark 2001
cxo: on a r700, i wonder if it would be faster on an r300
carldani: hi
_Groo_: osiris: could you take a look at bug 23715?
carldani: How do radeonfb and KMS interact?
carldani: Right now I need radeonfb to survive suspend-to-RAM on my Samsung P35 laptop (ATI Technologies Inc RV350 [Mobility Radeon 9600 M10])
eosie: MostAwesomeDude: http://storm.unas.cz/eosie-11-14.tar.gz - please review/push these, the first one needs testing because I have only r500 here
orzel: hello. I'm trying latest kernel/ati-drivers/mesa (all from git) on my xpress-200+gentoo+xorg-server 1.7.1, using KMS. It has really improved since i first tried it few months ago. 3D rendering is buggy and the console size is weird at boot time, but otherwise it's pretty much working. If i can be of any help testing whatever, i'd be happy to do so.
Ghworg: cxo: Strange, I get 2800 on 3dmark 2001 using an r600. Would have thought the 700 would be similar
_Groo_: orzel: i have a 200m also, can you resume from suspend to ram fine?
cxo: Ghworg, hmm interesting, maybe one of the devs would know.
osiris: _Groo_: what gpu? rs480?
_Groo_: osiris: rs482
Ghworg: cxo: I do have to set the offscreen rendering mode to backbuffer or the tests crash though
orzel: _Groo_: i never really tried any kind of suspend.. I haven't done this test and don't really know what to try/test. I'm testing boot, X working, KDE effects/composition, and 3d using googleearth. Which is weird.
cxo: Ghworg, i didnt change any of the defaults, i'm using wine-1.33
soreau: Ok. I'm on mesa master and I want to bisect it so I want to checkout a commit from about a month ago. What would be the best way to do this?
soreau: I'm looking on cgit, but I have horrible trouble navigating there
_Groo_: orzel: wait? GE is wroking with your rs48x? can you run driconf and check something?
_Groo_: orzel: the GE zooming works fine? or is garbled? like flying over a city
Ghworg: cxo: Try setting backbuffer as shown in http://wiki.winehq.org/UsefulRegistryKeys it might make a difference. Be interesting if it did
orzel: ge is working more or less (first time ever for the last 4 years) : i managed to take a screenshot of 3d bugs : http://img69.imageshack.us/i/bugrendering1.png/
orzel: _Groo_: i guess this is 'garbled' ??
orzel: _Groo_: i'm installing driconf
orzel: can i run driconf through ssh or do i need to connect. I dont always get kdm running, most of the time i get a black screen with blue line and i cant do something. I still dont know of a reliable way to get kdm/xdm running fine
_Groo_: orzel: thats exactly what i reported, see bug 23715 in freedesktop.org, you can circumvent it by disabling some tstuff in driconf but the performance goes down the dain, since cpu takes over gpu
cxo: Ghworg, I dont even have a Direct3d submenu in HKEY_CURRENT_USER
orzel: _Groo_: i see
orzel: i'm going to subscribe to the bug
Ghworg: cxo: Just create one, it's only used by wine
_Groo_: orzel: add a note to it so devs can see im not yje only one suffering from this bug
_Groo_: osiris: still there? :)
osiris: _Groo_: yes, I'm checking it
_Groo_: osiris: thanks osiris youre the best, maybe its something TCL related since rs482 is no-tcl
orzel: _Groo_: done :)
cxo: Ghworg, nope, no difference
cxo: I'm running in a window if that makes a difference
Ghworg: cxo: I do run fullscreen, I'll have a go in a window
_Groo_: orzel: there are 2 mroe bugs you can test, 24670 and 16537, the last one is a bitch to find
orzel: _Groo_: ok, i'll see if i can reproduce them. Though currently i don't seem to manage to make it work again. Gr. And there does not seem to be any error message.
osiris: _Groo_: I can't reproduce it on my rv530 with forced swtcl so it maybe be IGP or RS480 specific bug
osiris: _Groo_: ping me tomorrow about, I've got to go now
osiris: _Groo_: and reassign the bug to r300
orzel: _Groo_: mm..... it seems that 'kdm restart' never works BUT if i did not touch my computer for a long time. For example right now it has just worked. So it might be related to some kind of resume. But i have the inverse problem of you : it works better then....
_Groo_: osiris: if i use the softwrae fallback it works fine like i wrote in the bug report
_Groo_: osiris: if you make any patch add it to the the bug report so i can test it and report back
osiris: _Groo_: sure
_Groo_: osiris: since orzel with the same card has the same problem, is probably specific to rs48x
osiris: _Groo_: or IGP specific
_Groo_: osiris: :) these cars are evil i say!
_Groo_: cards
osiris: I agree
orzel: _Groo_: osiris : i try to describe my tests on http://www.freehackers.org/Orzel:Asus_W3Z#Video.2FXorg
orzel: i'm actually quite confused concerning card naming.. 480, xpress 200m, x300, r300 ...
osiris: orzel: check out http://en.wikipedia.org/wiki/Comparison_of_AMD_graphics_processing_units
Ghworg: cxo: Nope. 2731 marks in a 1024x768 window
_Groo_: orzel: 200m = rs482/rs485
_Groo_: orzel: its actually the most evil of them all, its a r200 2d chip slapped with a r300 3d chip and a agp acting as a pci-e...
_Groo_: orzel: radeon engineers where drunk when designed our card
happycube: that's just wrong
_Groo_: orzel: and the thism this shitty card doesnt have and vertex shaders!!! only pixel... ence the non-tcl when you see the glxinfo
orzel: i bought it in 2006 and quickly found out that it was one of the worse supported card in linux... now i take care before buying :-)
happycube: nowadays at least they put the new 2d+uvd in before upgrading the 3d
happycube: welcome to the amd notebook ghetto :P they're cleaning it up, just not quickly enough ;)
cxo: Ghworg, as a sanity test, what do you get on glxgears, default res
osiris: _Groo_: write some more info in the bug report (like mesa/ddx version, whether you are using kms)
_Groo_: it was the first mobile card from ati... they just slapped something together, and at the time was actually quitte fast compared to the market
orzel: _Groo_: how can i know if i have a rs482 or rs485 ? it's not obvious in lspci -v.. ?
Ghworg: cxo: About 1000 fps
_Groo_: osiris: latest drm/ddx/mesa git master with kernel 2.6.32 rc7
_Groo_: orzel: actyually it is just do a lspci without switches
orzel: _Groo_: i did say what i'm using (if we speak about #23715)
cxo: Ghworg, ok i get above 9000fps, so it must be some missing r700 specific acceleration
orzel: _Groo_: i only have this then : "01:05.0 VGA compatible controller: ATI Technologies Inc Radeon XPRESS 200M 5955 (PCIE)"
osiris: _Groo_: did you check UMS?
Ghworg: cxo: Very strange
cxo: indeed
orzel: osiris: was for your the "i did say what i'm using (if we speak about #23715)"
_Groo_: osiris: and i DID put that, its the comment #1!
osiris: _Groo_: sorry, I've missed it
_Groo_: osiris: :)
_Groo_: osiris: UMS?
osiris: _Groo_: try also some older mesa versions (like 7.6 and 7.5)
orzel: _Groo_: by the way, if i use CONFIG_FB_RADEON in the kernel i have no more /dev/dri/* and X of course is not happy with that.
osiris: _Groo_: user-mode mode setting (the opposite to KMS)
osiris: _Groo_: I've got to go now
_Groo_: osiris: same behaviour with older mesa.. only change is the use of the fallback ?
_Groo_: osiris: sure, thanks :) hope you can pinpoint this one
_Groo_: gotta go ppl, thanks for the help rozel
_Groo_: orzel
AstralStorm: interesting
AstralStorm: I still haven't solved the -vo gl problem
AstralStorm: why did it work and doesn't now? despite rolling back kernel, libdrm, mesa...
AstralStorm: but not the driver, hmm.
AstralStorm: can DDX affect fragment programs?
orzel: _Groo_: see ya
cxo: Anyone know if there are any differences between the status of r600 and r700 driver?
bridgman: not much; a better distinction would be between "cards the developers have" and "cards the developers don't have" ;)
bridgman: but should be same status, same features AFAIK
cxo: Ghworg and I found some disparities in performance
cxo: he has an r600 and I have an r700
cxo: He gets 2800 points on 3dmark 2000, i get about 40
cxo: He gets 1000fps on glxgears, i get about 9000
cxo: We were wondering if this was meant to be happening because of something WINE was using that was unavailable yet on r700 but was available on r600
cxo: And if that was the case, why it wasn't reflected in the status page http://www.x.org/wiki/RadeonFeature
MostAwesomeDude: cxo: Occam suggests that you're on 64-bit and you don't have 32-bit DRI libs built. :3
cxo: hey you could be right
cxo: So I should build mesa for 32bit and try?
MostAwesomeDude: Yeah, you'll need to build your multilibs. libdrm and mesa.
MostAwesomeDude: I have *zero* idea how your distro does multilib, so you'll have to ask them.
cxo: can handle multilib
carldani: so how does KMS interact with radeonfb?
bridgman: it obsoletes it AFAIK
carldani: oh, that's bad
MostAwesomeDude: Not really.
bridgman: not really, it was obsolete already ;)
carldani: I need radeonfb to survive suspend-to-RAM
MostAwesomeDude: There's a couple bits of funky code in there that need to be ported...
MostAwesomeDude: ...but it should go die in a fire.
MostAwesomeDude: Also it doesn't do a lot of newer radeons.
MostAwesomeDude: carldani: Mac?
carldani: No, Samsung P35 laptop
carldani: RV350
carldani: If I use vesafb, it hangs hard during resume (no interrupts whatsoever)
carldani: text mode works if I make sure to manually POST the card before switching back to X, but text mode screen contents will be corrupted
carldani: radeonfb combined with radeon X driver worked absolutely fine
carldani: well, worked fine since I added code to radeonfb which supported my card
carldani: linux-2.6/drivers/video/aty/radeon_pm.c has static struct radeon_device_id radeon_workaround_list[] = {
carldani: and my laptop is mentioned there, as well as others
cxo: MostAwesomeDude, can you just have a look at my config summary for mesa to see if its good
cxo: http://pastebin.com/m7db2d0a1
cxo: just to see if everything that needs to be enabled, is enabled
MostAwesomeDude: cxo: Looks fine, I suppose. Again, I don't know how to do multilib for you.
cxo: just noticed that mesa's automagic is not setup to build outside of the srcdir
MostAwesomeDude: Mesa doesn't use automake, just autoconf.
cxo: oh no wonder
carldani: auto* are evil
cxo: likes automagic
cxo: Automagic gets a bad name for itself because people try to hotwire it too often. When you start messing with computer generate code you can turn magic into hell really quick
carldani: So it is like AOL? Not inherently bad, but the majority of users is giving it a bad name?
cxo: I prefer the 70s 911 turbo. Most of them ended up round a tree, not because they were bad handling cars, but because it was mostly rich 50+yr olds in them trying to impress girls
cxo: Not inherently bad at all :)
carldani: Hm yes.
cxo: Whats this glut thing and why does mesa want it?
bridgman: I didn't think mesa would want it, other than for demo programs etc...
stikonas: gl utility toolkit -
bridgman: then again demos get built by default I think
cxo: says it wants glut and GLEW
cxo: ok looks like my CFLAGS didnt propagate all the way
cxo: you see, what happens when people try to mess with automagic
soreau: cxo: I think you might want --with-demos=""
cxo: tries
cxo: Does mesa have arch specific /usr/includes ?
cxo: wonders if he should just shot circuit make install from replacing his /usr/include/GL/* that came from his native (64bit) mesa build
cxo: Is there anything other than opengl and directx for graphics? Like what happened to Glade and all that stuff from the 90s?
bridgman: openvg, I guess
Dr_Jakob: openrt :)
cxo: Is there a non "open"GL anymore?
Dr_Jakob: I think that nintendo has homebrewed GL like api's for Wii and DS.
MostAwesomeDude: GX?
Dr_Jakob: yeah
Dr_Jakob: something completely different on the DS
Dr_Jakob: but no th only public ones are GL and DX
Dr_Jakob: the rest of them died out.
cxo: Does SGI still do their own GL?
Dr_Jakob: not that I know of.
eosie: MostAwesomeDude: have you noticed the patches I sent you?
MostAwesomeDude: eosie: How did you send them to me? I don't think I saw them.
eosie: MostAwesomeDude: http://storm.unas.cz/eosie-11-14.tar.gz - please review/push these, the first one needs testing because I have only r500 here
elbeardmorez: hi all. still struggling.. ..can anyone throw me a bone please? i've tried standard and many non-standard timings (that are working in windows ..hate saying that) ..but applying these as modelines in linux just results in a blank screen. the EDID info is giving only 1920x like resolutions, not the list of 6 or so that Powerstrip can read, including the standard 1280x720. any ideas on how to progress please? , .
eosie: MostAwesomeDude: it doesn't matter, I'll e-mail you next time
cxo: I built 32bit drm, 32bit mesa and wine again and now thinks are little weirder. 3dmark2000 doesnt work at all, and 3dmark2001 runs the first "car chase" test fine in low detail, but when it switches to high detail it crashes (While it didnt crash when using software rendering)
cxo: s/thinks/things
Ghworg: cxo: That's what I get when using default wine settings. There is a problem with framebuffer objects somewhere
cxo: xe: ../../../../drm/libdrm/radeon/radeon_cs_gem.c:121: cs_gem_write_reloc: Assertion `bo->space_accounted' failed.
bridgman: elbeardmorez, are you sure the timings you apply are being accepted by the driver (by checking logs or looking at your display's OSD) ?
cxo: I switched backed to backbuffer and 3dmark2001 works, i get 2508 points
cxo: 1024x768, default settings
cxo: but 3dmark2000 doesnt work at all even with backbuffer
cxo: i think wine has a lot of driver dependent code
cxo: probably because of how bad fglrx is
cxo: fixme:d3d_caps:wined3d_guess_vendor Received unrecognized GL_VENDOR "Advanced Micro Devices, Inc.". Returning VENDOR_WINE.
cxo: fixme:win:EnumDisplayDevicesW ((null),0,0x32d5f8,0x00000000), stub!
Ghworg: wine is really heavily nvidia focused. There are tonnes of bugs for other platforms that work fine for nvidia cards because that's what most of the devs have
Ghworg: I'm hoping that the open drivers will start to change that
foo: http://x03.ath.cx/temp/gnome-bug.png - gnome screenshot. I have dual monitors, 20" and 22"... the second half of my 22" monitor is a tad whack... as you can see from the screenshot. Any tips on a fix/workaround?
cxo: Ghworg, but i sincerely think thats due to just because nvidia drivers are so much better than anything else out there, so the ifdef'd code has now consequently a more mature and functioning code path than the generic path
adamk: foo, What model GPU?
foo: adamk: How can I find that out? Hm
adamk: lspci | grep -i vga
foo: 05:00.0 VGA compatible controller: ATI Technologies Inc R423 UI [Radeon X800PRO (PCIE)]
adamk: That card has a maximum texture size of 2048.
DanaG: nice: "200m = rs482/rs485 -- it's actually the most evil of them all, its a r200 2d chip slapped with a r300 3d chip and a agp acting as a pci-e..."
DanaG: AGP acting as PCIe? I can't quite figure out what's meant by that.
adamk: foo, I assume you are using compiz? Any textures larger than 2048 in compiz will be screwed up.
adamk: foo, And that includes the wallpaper.
cxo: Ghworg, and thus having perfect open source drivers wont fix that unless wine devs actively start using such drivers and rewriting wine accordingly, which thanks to inertia is likely to happen soon
adamk: foo, If you disable nautilus from drawing it's own wallpaper and instead use compiz' wallpaper plugin, it may work better.
cxo: Ghworg, s/likely/unlikely :)
cxo: compiz wallpaper plugin?
adamk: cxo, ?
cxo: hasnt had a nautilus drawn desktop for ages,
foo: adamk: interesting, let me try that
chithead: cxo: it would be a good start if the attitude of wine developers changed
cxo: adamk, whats the advantage of the compiz wallpaper plugin
Ghworg: cxo: I'm hoping that open drivers will equal good quality drivers, so that more wine developers will buy ati cards and hence the code for same will improve. Will be a slow process, and there are no guarantees it will happen, but I can dream :-)
adamk: cxo, Well, it treats each monitor as a separate texture, for one.
adamk: cxo, It also lets you have a different wallpaper on each viewport.
cxo: chithead, yeah, but if i put myself in their shoes, i can see why they dont want to change
adamk: Alright, I gotta run.
adamk: bbl.
foo: adamk: how would I go about adding a compiz background? I only see regular wallpapers in ubuntu, hmph. What's the compiz manager name again?
Ghworg: The ability to track a bug all the way down from wine code into the driver stack would be very attractive to me at least.
DanaG: hmm, they could at least give you an option to export TREAT_ALL_VIDEO_CARDS_LIKE_NVIDIA, or such.
chithead: cxo: yes. the open source community has to deliver better product (not necessarily faster)
DanaG: Or more neutral naming: DISABLE_CARD_SPECIFIC_WORKAROUNDS
elbeardmorez: bridgman: hi. thanks. i'm sure that they are applied, just not accepted. i get the furthest (towards usable) if i boot with dvi only connected, then i get a 720x400 mode that works, the rest of the available modes (applied via xrandr) result in the black screen, or ..for the likes of 640x480, a green garbled mess. . i can then.
elbeardmorez: ..no.. ..erm, i'm using KMS, and note that modes on the DVI don't get set at boot at all in a dual-screen config like i see with other peoples dmesg outputs..does the R480 have any known problems regarding reading EDIDs i wonder?
cxo: DanaG, doesnt look too bad, there is a d3d_caps function that takes in the vendor string and then assigns capabilities, patching it looks like child's play
eosie: MostAwesomeDude: so what do you think?
adamk: Hey guys... Can someone here confirm that r300 has a maximum 3D context size of 2560?
airlied: carldani: you should try with out radeonfb loaded and kms at console
airlied: kms handles s/r itself, but it needs testing
airlied: you can't load radeon kms and radeonfb at the same time anyways
MightyMu: I'm having problems with a number of apps crashing, and I may have narrowed it down to these two errors on my Xorg.log:
MightyMu: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed to open the DRM
MightyMu: (EE) GLX error: Can not get required symbols.
MightyMu: And modprobe drm tells me that "Error inserting drm: Operation not permitted."
[Enrico]: MightyMu: check that fglrx is not loaded
[Enrico]: likely you are still using fglrx kernel module
MightyMu: huh
[Enrico]: MightyMu: you can check with lspci -k and lsmod | grep fglrx
MightyMu: "Kernel modules: radeon, radeonfb"
[Enrico]: MightyMu: these are the avaible module, what about the module in use ?
MightyMu: [Enrico]: Wasn't sure which channel was the correct one. :)
MightyMu: fglrx is not loaded, according to lsmod
[Enrico]: MightyMu: if it says Kernel driver in use: fglrx_pci then this is not good
MightyMu: which makes sense, as I've never installed
MightyMu: never installed it, that is.
[Enrico]: MightyMu: mhm and lsmod | grep radeon ?
MightyMu: no result
MightyMu: wtf?
[Enrico]: MightyMu: what OS are you using -?
MightyMu: Ubuntu, 9.10
soreau: What about 'modinfo radeon'
MightyMu: my xorg.conf specifies the ati driver
MightyMu: Well, that spews a bunch of info
[Enrico]: MightyMu: just mv xorg.conf xorg.cong.mv you don't need it :D
MightyMu: [Enrico]: Why not?
[Enrico]: MightyMu: well if you get "Error inserting drm: Operation not permitted." this mean that some other driver is using the video card and that radeon (and drm) cannot be loaded. or you are not root :D
MightyMu: Every time I mess with that file, I end up with a black screen, so I am hesitant to. :)
soreau: MightyMu: That spewing a bunch of info means that radeon was compiled and as a module
MightyMu: I am not root. :)
MightyMu: soreau: thanks.
soreau: -_-
[Enrico]: MightyMu: can you paste lspci -k pls to some nopaste service?
MightyMu: sure
[Enrico]: MightyMu: but you are using sudo modprobe radeon isn't it ?
soreau: You always want to load and unload kernel modules as root or at least with sudo
MightyMu: oh. Duh. stupid me.
[Enrico]: MightyMu: by "you must be root" i mean you must use sudo (or sudo su -)
MightyMu: yeah
MightyMu: anyway, lspci output is at http://pastebin.mozilla.org/683733
[Enrico]: MightyMu: no module in use, so you have to modprobe one :D
[Enrico]: and the correct one is radeon
MightyMu: sudo modprobe worked
[Enrico]: :D
MightyMu: can't believe I didn't try that
[Enrico]: MightyMu: don't worry it happens
[Enrico]: just keep it in mind :D
MightyMu: Now... let's see if that fixed anything. :)
MightyMu: Also - how to get that driver loaded automatically?
[Enrico]: MightyMu: it should if the correct driver is specified in xorg.conf or without any xorg.conf
MightyMu: Hm. Same crash with Boxee
[Enrico]: boxee ?
MightyMu: Should I change 'ati' to 'radeon' in xorg.conf, then?
[Enrico]: MightyMu: pls try with something safe before :D
[Enrico]: MightyMu: well ati is a radeon alias afaik
[Enrico]: so it is the same
MightyMu: hm
[Enrico]: MightyMu: but indeed you can
soreau: ati is a wrapper that will almost always choose radeon
MightyMu: the crash is happening with any GL application, I believe - including glxinfo
[Enrico]: soreau: thanks for the correct explanation :D
[Enrico]: MightyMu: can we see the glxinfo or glxgears log pls ?
MightyMu: glxinfo crashes, no output.
[Enrico]: MightyMu: wow
MightyMu: do I need to restart X after having modprobed?
[Enrico]: MightyMu: then can i see you /var/log/Xorg.0.log may be there is some clue there
[Enrico]: MightyMu: indeed you need!
MightyMu: apparently needs a bit of hand-holding. :)
MightyMu: Back in a moment.
[Enrico]: MightyMu: we will make ubuntu do it automatically don't worry
MightyMu: hahaha
[Enrico]: (i wonder why it doesn't)
[Enrico]: it should just do it!
[Enrico]: mine does
MightyMu: wait - what should do what? X should restart itself?
[Enrico]: MightyMu: well the best way it to press ctrl + atl + backspace but afaik it doesn't work anymore on recent X
MightyMu: yeah, that's disabled
[Enrico]: MightyMu: are you using gnome ?
MightyMu: I haven't gotten aroudn to re-enabling
MightyMu: Gnome, yes
[Enrico]: MightyMu: so i guess you are using gdm right ?
MightyMu: Yup, Ubuntu's messed-up new version. BOOOOOO.
[Enrico]: MightyMu: ok then it is easy
biotube: MightyMu: it was disabled upstream
MightyMu: was just going to log out/back in
[Enrico]: MightyMu: close all you app and in a terminal write sudo /etc/init.d/gdm restart this will restart X
[Enrico]: MightyMu: logout and login back is not sufficent
MightyMu: Ah.
MightyMu: brb
Masanori: is there a way to totaly get ride of video tear with xv on rv770? I didn't set up any xorg.conf
soreau: Masanori: Using kms?
Masanori: soreau: no
Masanori: soreau: im running 6.12.99.git20091014-1
soreau: You should probably try drm-next, and libdrm, mesa and ddx from git
Masanori: ok i'll give it a try
MightyMu: WEll. That was fun. X will not start with the radeon driver loaded.
MightyMu: had to ssh in and rmmod radeon
hagabaka: do you happen to know how to use drm-next on ubuntu?
MightyMu: Nope.
MightyMu: can learn :)
MightyMu: step 1: Google drm-next
[Enrico]: MightyMu: don't do that
[Enrico]: MightyMu: you don't need drm-next
[Enrico]: MightyMu: you need drm-next only with r600 or 700 cards (like mine)
MightyMu: good
MightyMu: it looks terrifying
cxo: what does that have that 2.6.32rc7 doesnt have?
[Enrico]: MightyMu: it is, you have to recompile the kernel
[Enrico]: MightyMu: anyway can you paste /var/log/Xorg.0.log.old so we can see why with radeon X doesn't start may be ?
MightyMu: Yet another linux task I've never successfully done.
MightyMu: I've been looking at the X logs... they're remarkably error-free
MightyMu: which is weird as hell
cxo: you probably have some sort of conflicting setup
[Enrico]: MightyMu: just paste them pls
cxo: I dont think X log will have anything useful at this stage
MightyMu: [Enrico]: sure
cxo: you need to look at kernel logs
MightyMu: dmesg?
[Enrico]: MightyMu: be sure to use the .old one (it should be the one generated by the non working X )
hagabaka: when I use 2.6.32rc6, I can get compositing and KMS with radeon 9800Pro, but there are flickering lines on the screen both in the console and X. I have to change the refresh rate from the default 60 to 75 to remove the flicker. with rc7, as well as one of the earlier rcs, X just crashes when I change refresh rate
cxo: hagabaka, i think i had a similar problem, it seems to have gone now with the latest git
hagabaka: oh
cxo: but it wasnt a continuous flicker, it was kind of a static distorted interlace
hagabaka: my flickering lines only get obvious near the right edge of the screen, and they're actually shifting images of the normal display. when I move the mouse cursor to the right of the screen it appears to have a flickering shadow
cxo: mine were at the bottom edges
cxo: How friggin big is this git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git
cxo: 13% - 74mb?
hagabaka: I'll wait for the next ubuntu kernel ppa package I guess...
MightyMu: MightyMu's Xorg.0.log.old: http://pastebin.mozilla.org/683738
MightyMu: the fun stuff isn't marked with (EE), which is why I didn't see it when I grepped
[Enrico]: MightyMu: strange that /proc/ati/major doesn't exist
[Enrico]: MightyMu: can you check it manualli with cat /proc/ati/major pls ?
[Enrico]: manually*
MightyMu: there isn't even a /proc/ati
[Enrico]: ok
MightyMu: Something is broken in a bizarre and terrible way, huh.
[Enrico]: MightyMu: double check that radeon is loaded pls with lsmod | grep radeon
MightyMu: it shouldn't be - I rmmod'ed it to get x started
MightyMu: checking anyway
[Enrico]: oh
[Enrico]: right
MightyMu: yeah, not loaded
[Enrico]: MightyMu: well indeed you removed it
[Enrico]: MightyMu: ok listen you need to be able to get on irc from the tty. so get another pc or learn to use irrsi
MightyMu: Oh, that's where it gets WAY better
MightyMu: on account of TTYs don't display on this monitor at all
[Enrico]: uh?
MightyMu: ctrl+alt+Fx results in lovely colored lines. :)
[Enrico]: MightyMu: oh well it happens change between f1 and f7 a couple of times
[Enrico]: at some attempt you should get it
[Enrico]: rofl ubuntuers
MobileMu: yeah... they *used* to result in colored lines... now just black
MobileMu: f6 is X, right?
[Enrico]: MobileMu: f7 yes
[Enrico]: MobileMu: well you are now on another pc i guess
MobileMu: that's not getting X back. drat.
cxo: you can otherwise just hit ALT+LEFT/RIGHT arrow and it'll switch tty
MobileMu: well.. iPhone
[Enrico]: MobileMu: rofl
MobileMu: heh
[Enrico]: MobileMu: well we need a tty now
MobileMu: I'll see if I can boot into one
[Enrico]: just wonders why radeon on ubuntu doesn't "just work"
[Enrico]: it always just work
MobileMu: it did work, until I upgraded to 9.10
[Enrico]: something gone bad i guess
MobileMu: compiz, GL, everything was great
MobileMu: yup
soreau: [Enrico]: It does justwork. until you start messing with it
[Enrico]: soreau: rofl
soreau: MobileMu: Did you have ppa stuff installed before you upgraded?
MobileMu: yes, but I don't remember precisely what
redex: Hi im trying to get PSX emulator running and im getting this: "Application calling GLX 1.3 function "glXCreatePbuffer" when GLX 1.3 is not supported!" Is it driver related?
soreau: and you expected an upgrade to go smoothly, why?
[Enrico]: MobileMu: well you can check your source.list the previous ppa are commented
soreau: redex: Harmless mesa warning
MobileMu: note to self: ppas screw upgrades
[Enrico]: MobileMu: at least can
soreau: redex: Actually, it's glxCreatePixmap and DestroyPixmap warnings are harmless when running compiz, not sure about a psx emulator, they might not have appropriate checks in place
MobileMu: soreau: I actually didn't, it worked way better than I thought it would
[Enrico]: MobileMu: got a tty?
redex: thanks
MobileMu: [Enrico]: working - I'll let you know when I get it
soreau: MobileMu: So why not waste your time and install 9.10 fresh?
[Enrico]: MobileMu: well we can also do it in gnome indeed, but when we need to restart it (and xorg will freeze) we will need a tty anyway
soreau: Aside from being afraid of grub2..
[Enrico]: oh well grub2 works
[Enrico]: nothing to be afraid about normally
MobileMu: soreau: that is indeed a last resort
MobileMu: and the only PPA was banshee
MobileMu: [Enrico]: let's try it in gnome, I can ssh in from here when needed
[Enrico]: MobileMu: it should not hurt then :D
[Enrico]: MobileMu: oh well ssh is a tty :|
MobileMu: yeah
[Enrico]: MobileMu: ok then sudo modprobe radeon, paste any error displaied if any
MobileMu: lemme get real IRC RUNNING
MobileMu: also turn off my caps
MightyMu: I was getting too used to IRCing from the iPhone
MightyMu: This is faster.
MightyMu: Okay, modprobing radeon
MightyMu: No errors.
[Enrico]: MightyMu: do /proc/ati exist ?
[Enrico]: MightyMu: and pls paste the output of dmesg | grep -i radeon
MightyMu: /proc/ati does not exist
MightyMu: the output is short pasting here:
MightyMu: [ 890.041202] [drm] radeon default to kernel modesetting DISABLED.
MightyMu: [ 890.041724] [drm] Initialized radeon 1.31.0 20080528 for 0000:01:00.0 on minor 0
[Enrico]: MightyMu: mhm i don't remeber if for r300 cards firmware is needed
[Enrico]: MightyMu: dmesg | grep -i firmware ?
cxo: grep -i drm
[Enrico]: MightyMu: try what cxo suggested: dmesg | grep -i drm
soreau: He probably needs to boot with radeon.modeset=1 since ubuntu kernels have it disabled by default
MightyMu: grepping for firmware returned nothing
MightyMu: jsm@curtis:/proc$ dmesg | grep -i drm
MightyMu: [ 889.950604] [drm] Initialized drm 1.1.0 20060810
MightyMu: [ 890.041202] [drm] radeon default to kernel modesetting DISABLED.
MightyMu: [ 890.041724] [drm] Initialized radeon 1.31.0 20080528 for 0000:01:00.0 on minor 0
soreau: I'd like to think the one in the ppa has it enabled by default but i wouldn't be surprised
[Enrico]: well it should work without kms
[Enrico]: i don't use kms couse i want to forse the card to low powermode, but it works great
MightyMu: opens a beer, passes the six-pack around
MightyMu: okay, so we're talking about kernel modesetting.
MightyMu: learns about that
[Enrico]: MightyMu: i'm sorry but i have no clue, seems all ok to me...... i really have no idea why xorg is looking for /proc/ati here it is not mentioned in my log....
[Enrico]: i go to sleep
[Enrico]: gn guys :D
MobileMu: night
MobileMu: too slow
bridgman: am I the only one who thinks radeonfb is likely to be a problem ?
cxo: doesnt use radeonfb
cxo: wtf is radeonfb anyways?
bridgman: its an old kernel driver for r1xx-r4xx that gave you a graphics console at native resolution
bridgman: nice to have but AFAIK it breaks the current ddx/mesa/dri stack
cxo: MobileMu, lsmod | grep radeon
MightyMu: radeon 636576 0
MightyMu: ttm 36308 1 radeon
MightyMu: drm 160224 2 radeon,ttm
MightyMu: i2c_algo_bit 5760 1 radeon
bridgman: hmmm... I might be confusing MightyMu with another user
bridgman: thought I saw radeonfb mentioned
cxo: how come he doesnt have drm_kms_helper
MightyMu: Don't worry, I'm confused, too. :)
bridgman: < MightyMu> "Kernel modules: radeon, radeonfb"
bridgman: so that's what is available not what is loaded I guess
cxo: MightyMu, which kernel are you using?
MightyMu: 2.6.31-14-generic-pae #48-Ubuntu
bridgman: so can someone refresh me with history here ? 9.04 worked ok, updated to 9.10 and doesn't work, are we trying kms as well ?
MightyMu: which is, afaik, whatever ubuntu provided with the 9.10 upgrade
cxo: MightyMu, i think you need the latest kernel and mesa/drm/ati toolchain
hagabaka: would drm-next not help at all for radeon 9800 Pro?
bridgman: there were some interim versions of 9.10 that were kinda broken for graphics; did you update after 9.10 was released or before ?
MightyMu: ok, getting newer kernel I can do, what s involved in getting the rest?
MightyMu: bridgman: after
cxo: why cant drm-next just be a patch to the latest git instead of an entire tree,
bridgman: drm-next is where the work gets done, it can't be a patch
cxo: MightyMu, well the rest can be taken from freedesktop git
cxo: so drm-next will always be used, or is it a temporary stop gap?
bridgman: IIRC drm-next is where patches are accumulated to go into the "next" Linux kernel, ie its where things pile up before the merge window opens
cxo: Does it stay in sync with the rest of linux?
cxo: ie the upstream vanilla kernel tree
bridgman: not sure, airlied would know
bridgman: my guess is "not quite but close"
cxo: this is why i'd prefer a patch for just the updated drm/kms/radeon/whatever bits
bridgman: I imagine there's another repo where patches for the currently active kernel version are accumulated before sending to Linux
bridgman: Linus
bridgman: once you sync down the tree you can just get incremental updates, you don't have to keep pulling a bazillion gigawatts of code down every time
bridgman: if I were MightyMu I would :
bridgman: 1. fresh install 9.04 and see if it works
bridgman: 2. fresh install 9.10 and see if it works
bridgman: 3. decide what to do based on results from 1. and 2. ;)
cxo: obviously, but i'd like to track the rest of upstream linux as well
MightyMu: bridgman: Fresh 9.04 worked fine
bridgman: oh good
cxo: if I were MightyMu I would:
MightyMu: and option 2 is getting more and more likely. :(
cxo: 1. Scrap n00b00n2 kernel
MightyMu: lol
cxo: 2. Vanilla 2.6.32rc7
cxo: 3. Build mesa/drm/radeon et friends
cxo: 4. Hit everything with a hammer until it all sticks together
bridgman: mine is easier and you can drink beer while the installation is happening
cxo: 5. reboot
cxo: ubuntu installs really fast actually on new systems, you could do 2 installations in under 20mins
bridgman: could be... old habits die hard, so I usually start the installation and go home ;)
bridgman: it probably finishes while I'm still in the parking lot
cxo: probably
MightyMu: Yeah, I recall the 9.04 install being really fast
MightyMu: But... If that's the course, I think I'll wait to do it.
cxo: The last few releases of Ubuntu have just been amazing, 9.10, 9.04, 8.04
MightyMu: Got an early day tomorrow, and then a long-ass week, and I really just want to sit back and watch some tv. :)
MightyMu: With beer, of course
MightyMu: I certainly plan to use cxo's step 4 as often as possible.
cxo: thats basically all i do really, i just get into this zombie mode (like those koreans that play starcraft) and just keep hitting everything until it works
MightyMu: goddamnit, now I want to play starcraft
bridgman: I write down the last thing I did, make a note not to do it again, wipe the drive and start over ;)
bridgman: life is too short
bridgman: and we are all far too creative in finding ways to muck up a system
cxo: wine really doesnt like the open source driver
cxo: its worse than fglrx actually in terms of wine compatibility
cxo: i think the wine devs just go so used to fglrx they probably built an entire driver just dancing around fglrx's bugs and got everything working finally
cxo: I was thinking of starting a thread on their mailing list about writing more driver neutral code, but i realise, they'll just end up flaming me, so i didnt
MightyMu: is out - thanks for everything, I really appreciate your help.
bridgman: cxo; I don't think you need to start that thread; originally wine was written around nvidias bugs (which was a problem for fglrx because it had *different* bugs) but over the last year or so I think the wine devs have gone a long way to making it driver neutral
bridgman: we fixed a bunch of bugs as well so everyone's happy
damentz: i would think that wine attempts to code their opengl code for the strict standards of opengl
damentz: since it has to transform direct3d to opengl stuff
happycube: the prob is what they need is extensions
happycube: shader assembly, etc
damentz: who?
happycube: wine devs, dx9+
bridgman: I'm sure it's coded to the strict standards of opengl... and then tweaked until it works, like every other piece of code on the planet ;)
bridgman: it's the tweaking that makes it driver specific
damentz: lol
damentz: right
happycube: forget extreme programming, in the real world spegetti programming rules
happycube: also forget about cleaning the walls.
bridgman: the only fundamental problem I saw was that Wine sucked up all of the available memory
bridgman: that was ok for NVidia drivers (which seemed to preallocate everything they needed) but was screaming death for fglrx
damentz: hmm, i don't get what you're saying
bridgman: something like that anyways
damentz: you mean video ram or something?
bridgman: tries to remember something from 2 years ago that he didn't pay much attention to at the time
bridgman: fails
happycube: eh, as long as it works now
damentz: i'm really interested in that new mesa work that's happening
bridgman: I did know at one time, honest ;)
damentz: claiming 100x speed enhancement is quite peculiar
happycube: it's a local/setup enhancement
bridgman: we got 100x speed improvement by hw-accelerating GLSwapBuffers in the 6xx/7xx 3D driver ;)
happycube: ooo. nice :)
damentz: btw, does anyone know if xserver 1.7 / xorg 7.5 increase 3d performance?
bridgman: unfortunately glxgears ran at about 20 fps before
damentz: there was a benchmark on phoronix against fedora 12 beta and ubuntu
bridgman: don't know but I wouldn't expect any real changes
damentz: and fedora was pushing more frames per second on every opengl game
damentz: right
bridgman: the X drivers can make a big difference but not so much the server
damentz: maybe there was bottleneck we were unaware of?
bridgman: maybe tiling or something
damentz: i'm waiting for it to propogate into sid so i can do some testing myself
damentz: but sid takes forever to get new X packages
bridgman: fedora does tend to have the newest code so it'll either be faster, crash, or swap itself to death ;)
damentz: lol
damentz: bridgman, i made a bad buying decision lately
damentz: there was an nvidia 6800 gt agp model on sale at frys for $70, so i got that instead of the a radeonhd card for about the same
damentz: i just realized that nvidia may not support this card in the future
damentz: plus it generates heat like no ones business
damentz: not sure if i can return it in exchange
damentz: i should though, i had to already replace it after 20 days for display corruption
bridgman: the 6800gt was a pretty powerful chip for its time IIRC
bridgman: it was nv's response to the r420 I think
bridgman: but IIRC all the chips of that generation were big power hogs
bridgman: "winter cards" ;)
bridgman: didn't they sell for ~500 or so originally ?
cxo: I cant remember, i usually just buy whatever has the best bang for buck each 2nd year
cxo: So far its been Nvidia, Nvidia, ATi, Nvidia, Ati (2008)
cxo: next year i'll likely buy a new card, whether its ati or nvidia just depends on how much value i get from it
factor: Where is the proprietary driver for the radeon 9200
factor: the only one I am finding is from 2006
EruditeHermit: factor, there isn't one
factor: gotta be a newer one than that
factor: not even a legacy one
EruditeHermit: Catalyst 9.3 supports r300 and newer
cxo: the open source driver for a 9200 should really rock
factor: wel it crashes
EruditeHermit: 9200 is r200
factor: so I have to go back to the closed source one.
factor: wil keep trying both
factor: but need GL stuff for now
EruditeHermit: factor, the open source driver does all that the card can do for the 9200
factor: except work
factor: yes it has the GL but when i use it it crashes
EruditeHermit: it shouldn't
bridgman: versions, messages etc ?
factor: well I would hope not
hifi: I think there isn't a closed source driver for r200
EruditeHermit: if it has issues, file bugs
bridgman: are you talking windows or linux ?
factor: I know, just need to get GL gping for now.
factor: linux
bridgman: what distro, and which driver versions are you using ?
factor: trying to write a HID code for second life.mod
factor: I need GL to work I dont have time right now to debug the minor issue to get to the major
bridgman: ok
bridgman: were you using the closed source driver before ?
factor: could not get the HID to compile in 64 bit so had to install 32 bit but have nvidia on 64 bit and ati on 32 .
factor: I think
hifi: lspci | grep VGA
hifi: what does it say
factor: on which sys ati one
hifi: the one you think is ATI
happycube: 6800gt? where'd they find that?
factor: RV280 [Radeon 9200] rev 01
hifi: what distribution do you have
factor: debian 64 bit ubuntu 9.10 32 bit wiht ati
hifi: ok, what does "glxinfo | grep renderer" say?
bridgman: the closed source driver won't work with ubuntu 9.10
factor: ok I will just keep trying with the oss version
factor: what is the command to check for gl activeation
bridgman: so 2d runs ok but 3d crashes ? any messages ?
factor: yes
factor: hold on
hifi: factor: try the command above
factor: rebooting into init 3
bridgman: glxinfo | grep renderer will tell you what driver you have
factor: ok glxinfo that is right
factor: I give up
factor: crashed booting into X
bridgman: did you update to 9.10 or fresh install ?
factor: will switch it to a working card.
damentz: happycube, i got it at fry's electronics
damentz: looks like they found some old 6800 gt cards
factor: will be back to debug ati when I get my original rpoblem resolved
factor: thanks
bridgman: a few people are reporting problems with r2xx family after updating to 9.10, not sure why yet
damentz: i got it since i knew it could play essentially anything i threw at it
bridgman: I was going to suggest he try 9.04, that seemed to work for *everyone* with r2xx
damentz: but then i realized what a power hog it was
bridgman: yeah, electricity used to be cheaper ;)
damentz: hey bridgman, did you watch the wayland server video?
damentz: from the plumbers conference apparently
bridgman: is still on dial-up