Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-12-04

Search This Log:


siimo: by upgrading to kernel 2.6.32 am i likely to see performance improvement in 3D of X2300 RV550 using OSS driver?
siimo: the 3D performance is very poor, although i am using radeon driver not radeonhd driver, cause if i use radeonhd then i get flicking when watching movies :S
hifi: siimo: how poor?
hifi: like, does openarena run well
hifi: glxinfo | grep render
siimo: what is openarena?
siimo: glxinfo | grep render returns no output :( is that software rendering
hifi: umm, thats like broken GL alltogether
hifi: pastebin glxinfo
siimo: http://pastebin.com/m2f420125
hifi: what
hifi: what OS are you on?
siimo: slackware 13
hifi: the glxinfo output is weird if thats really all
siimo: what do you mean
hifi: http://pastebin.com/d76aac6f thats my glxinfo
siimo: hmmm
siimo: :S
hifi: anyway, it *seems* that you do have the correct driver working
hifi: how did you test it's slow?
siimo: i tried to run warcraft3 in wine
hifi: thats not a very good test
siimo: bascially does not work at all 4-5fps at best and random artifacts, it flies on my desktop with onboard nvidia 6150 and 100% stable and smooth
Leeds: glxgears!
siimo: glxgears is useless.. well im getting about 1100 fps http://pastebin.com/m101000c7
siimo: hifi: you seem to have a lot newer driver than me 2009 as opposed to 2006?? the date in the GL_Renderer
hifi: well, you *do* run slackware :))
siimo: its the newest version came out about 3 months ago
Leeds: I want to get superfast 3D running on my RV770 on slackware64 at some point soon - should be easy, I think
hifi: a lot happens in 3 months
siimo: hifi: any other test i can do that is useful? i guess i have no hope of playing warcraft 3 in wine with OSS ati driver lol
hifi: siimo: you could install openarena (a free FPS shooter) and jump around
hifi: though it seems that you do have accelerated rendering allright
siimo: it plays smooth as on nvidia with even 32mb graphics memory
siimo: no point using radeonhd driver right? for my card?
hifi: no, radeonhd is just the DDX (2D) part
hifi: also I think it was buried with all the ideologies it had...
siimo: openarena comes as zip archive, is that source code?
siimo: ah. btw i also tried this useless 3D simiulator thing called secondlife that my college uses its useless but yeah thats also unusable with this driver.. weird artifacts everywhere and 1-2fps
siimo: and vmware says i cant use 3d acceleration in VMs
hifi: isn't there a linux client for second life
siimo: yah thats the one i tried the linux one
siimo: very poor fps and random artifacts coming up.. fonts not showing up properly too
siimo: but works extremely smoothly on nvidia with 32mb shared ram on my desktop lol
siimo: i guess next time i dont buy ati laptop
rehabdoll: i dont get a text-console with modesetting, what am i missing in my .config ?
rehabdoll: [drm:drm_mode_getfb] *ERROR* invalid framebuffer id
hifi: fbcon probably?
siimo: sux how catalyst dropped support for me < 3 yr old card already 6 months ago
siimo: hifi: but huge 3D performance improvements for r300 driver likely or unlikely? even in current bleeding edge version
hifi: umm, you do have kind of old system, maybe
siimo: what do you mean?
siimo: software or hardware
siimo: i have this laptop http://www.asus.com/product.aspx?P_ID=Mmy7GVFZVPi8AJ58
siimo: ATI Mobility™ Radeon®X2300, External 128MB VRAM
Nightwulf|work: hi all
spstarr: hmmm
spstarr: http://cgit.freedesktop.org/mesa/mesa/commit/?id=c8cdce665790263bb2142d894a81c87abc4da9fb
spstarr: I wonder if this fixes VBO issues i have with radeon and assert
Telek: Hey radeondrmfb is still broken on RS100.
airlied: Telek: got a dmesg?
airlied: we don't have many rs100s around to play with
Telek: airlied: Nope, I'm gonna try setting up ssh though so I can :)
Telek: It's actually an IGP 345 from a P4 era HP Pavilion.
airlied: oh wierd I thought we got rs300 working though don't see many to know for sure
DanaG: Telek: if it's a desktop, serial console may be an option, too.
spstarr: sleep time, will test trunk everything later today :)
Telek: radeonfb works just peachy with it, but radeondrmfb has been blankscrening for a while.
Telek: DanaG: Nope, Laptop :)
Telek: And I don't have any serial cables handy anyhow.
Telek: Had a null modem cable and 25->9 connectors years ago, but lan was always less of a hassle :D
DanaG: P4 laptop? Bleh.
Telek: It's my second longest lasting laptop ironically enough :)
DanaG: wishes he could magically make Intel make up for all the power wasted worldwide through the use of P4s instead of AMD stuff.
Telek: And since I swapped a celeron into it it runs cool too :D
DanaG: Celeron is worse... no SpeedStep!
Telek: Uhmm... P4's didn't have speedstep.
DanaG: P4-M d id.
DanaG: did.
Telek: It was a *P4* not a *P4-M*
DanaG: aaaaaaaaah, even worse. =P
Telek: The first two years of P4 laptops were desktop chips shoved in a laptop.
DanaG: I have a similar-ish laptop around here: Toshiba Satellite 1415.
Telek: There wasn't even an option for P3's below the 1500 dollar pricepoint at that time, at least in store.
happycube: celeron-m's *still* have no speedstep
Telek: Was completely worthless as a laptop too, 1:15-1:45 minutes max, idling :D
DanaG: Has a GeForce *cough*FOUR*cough* MX in it ... horrible. And an EDID that makes the nvidia binaries (even for Windows) seem to be 966x768.
DanaG: STILL?
Telek: happycube: That's why you buy the cheapest laptop and shove a decent CPU in it :D
DanaG: Ugh, somebody should sue Intel for wasting power.
happycube: ;) and make sure it isn't soldered
DanaG: Even Semprons have PowerNow.
happycube: danag. still. even the 45nm celery-m's
DanaG: ugh.
Telek: I got a C700 and swapped a T7500 into it for less than a similiarly specced laptop would've cost :) Same Monitor up to the 1000 dollar range too.
happycube: ;)
DanaG: Time to print up signs: "Anthing with "CELERON" will waste tons of power."
happycube: desktop celerons *do* have speedstep, tho.
Telek: happycube: Don't buy SU series chips and you should be good... although by next year I expect sockets to be almost a thing of the past :(
happycube: the e3300 is actually good
Telek: And VTx :)
Telek: Also the 'new' E5xxx chips.
DanaG: I'd trade the P4-Celeron (ugh, barf!) for a P3-M any day.
happycube: telek - apple hasn't *ever* used socketed c2d's
Telek: happycube: Really?
happycube: at least in the non-pro's
Telek: Interesting, and good to know.
DanaG: Family friend has some Compaq laptop with a 1.2GHz P3-M, a "Mobility Radeon" of some sort, and TWO cd drive bays.
Telek: aurkued: still checking the disks :D
DanaG: Yes, TWO cd drive bays.
Telek: Nice.
happycube: i probably bought the only 2nd hand t60 to have been downgraded from c2d->c1d ;)
Telek: smirks.
happycube: whoever did the swap at least knew how to put thermal goop on right... it runs nicely
DanaG: My good, current laptop has a Mobility FireGL V5700.
DanaG: The last thing left that keeps me on fglrx, is lack of good power management on radeon.
DanaG: That is, for KMS.
DanaG: Non-KMS has some issues with some Compiz stuff.
DanaG: And KMS sucks watts like crazy.
happycube: danag - that's probably a rv100-m then
DanaG: I'll bet it'd work better than that NV17 junk. =P
Telek: Quite possibly.
Travis1: the changes to my comp failed again. Removed the xorg.cong file... got a blank screen on startup again
Travis1: Xorg.0.log = http://pastebin.com/m5d365351
Travis1: Xorg.0.log.old = http://pastebin.com/m6721b5c5
airlied: Travis1: can you edit radeon_driver.c search for aper_size == -1 and change the -1 to 0xffffffff
Travis1: uhh sure. Where do I go to do that?
airlied: Travis1: you got a favourite ediytor? edit xf86-video-ati/src/radeon_driver.c
airlied: then do the make; sudo make install again
Travis1: if (aper_size == -1)
Travis1: aper_size = 0x20000000;
Travis1: so change the first line to "0xffffffff"?
airlied: yup change the -1 to 0xffffffff
Travis1: ok saved the edited file
airlied: now make and sudo make instal
airlied: install even
Travis1: and I do the make and make install from: travis@travis-laptop:~/xf86-video-ati$
Travis1: like, that should be the folder?
airlied: yes
Travis1: ok done
airlied: okay try and start X again, cross fingers
Travis1: haha and toes
Travis1: brb
Travis1: remove xorg.conf when I restart right?
Travis1: as in remove it now and then restart
airlied: yes
Travis1: ok brb
Telek: airlied: http://pastebin.ca/1701555 <-- Dmesg off the laptop.
airlied: Telek: got fbcon loaded?
airlied: that looks fine that dmesg
Telek: Lemme double check
Telek: Should be built in the kernel.
Travis1: Xorg.0.log = http://pastebin.com/m51f74efe
Telek: And come to think of it, is. I was using the radeonfb driver previously, and just switched it over by enabling kms and disabling the radeonfb driver.
Travis1: Xorg.0.log.old = http://pastebin.com/m4b8262a7
Travis1: fail again
JohnDoe_71Rus: (WW) RADEON(0): Unknown vendor-specific block f
JohnDoe_71Rus: http://pastebin.com/m44c81c62
JohnDoe_71Rus: is normal?
airlied: Telek: good poing
Travis1: this is such a frustrating video card...
Telek: airlied: Yep fbcon is in, but broken. Also one of my other systems with either an rv740, or a r6-something had a similiar issue, only X windows would at least display on the screen when the fbcon was blank. This one doesn't display AT ALL.
Travis1: any next steps for my video card?
airlied: Travis1: yeah I'm a bit stuck at what is actually happening
Telek: airlied: There are some warnings while compiling about io_32.h and some memset __ functions?
airlied: Telek: shouldn't affect it I don't think
Travis1: right...
Travis1: so the log doen't hold any more clues?
Telek: nods.
airlied: Travis1: oh I can see its doing soemthing stpid just not quite how to fix it
Travis1: haha right
Telek: X2 550 black for 80 bucks with mobo worth it?
airlied: Travis1: can you wget http://people.freedesktop.org/~airlied/scratch/testdiff
airlied: in xf86-video-ati, do git checkout -f, then patch -p1 < testdiff
airlied: and get a new log, it won't fix it but I need some more info
airlied: dinner
Telek: waves.
Travis1: so the first wget I do that in xf86-video-ati as well yeah?
Travis1: ok I've done that.
Travis1: I assume restart now?
Travis1: be back in 30, dinner.
chithead: Telek: if the mobo supports acc so you can unlock to x4, yes
Telek: chithead: Nah, NV mobo with the combo :(
chithead: also if it is ddr2 you will bite yourself later that you didn't buy ddr3
Telek: chithead: Only if I didn't already have the DDR2 :)
chithead: I guess it is ok then
Telek: airlied: radeonfb is preferred over radeondrmfb?
chithead: radeonfb is obsolete, don't use it
Telek: You might want to have paid attention to my earlier conversation before you say uninformed decisions like that :)
Travis1: ok back.
Travis1: so I made those changes Airlied said. Remove xorg.conf file and restart time?
Travis1: let me know if there's anything else I need to do before restart/
Travis1: yes? restart now? I just don't want to do anything without approval :)
Travis1: ok I'm restarting. Hope someone will be here when I get back to look at the log...
Telek: Travis1: Nope, doesn't look like anyone is back yet.
Travis1: awesome. I'll post the log files anyway
Travis1: Xorg.0.log = http://pastebin.com/m49c2dde9
Travis1: Xorg.0.log.old = http://pastebin.com/m6189a664
Telek: Tried the regular -radeon/-ati driver, not the rhd one?
Travis1: not sure. I've tried a whole number of things...
Travis1: I just need people to talk me through what they think is a solution and I'm game to try anything
Telek: Your old conf was the radeon driver, not rhd.
Travis1: well neither are working
Telek: nods.
Travis1: this is a fresh install
Telek: I'm not really technical enough to give you any advice.
Travis1: install. No graphics.
Travis1: haha sure
Travis1: Airlied at dinner still?
Telek: AFAIK.
Ralesk: airlied: I have mesa 7.6
airlied: Travis1: wierd the second printf in my testdiff isn't printing
Travis1: so there's something not happening from the changes you just got me to make?
airlied: Travis1: agd5f was last person to look at something like this, but he won't be online until USA times
airlied: Travis1: yeah does git diff show something like http://people.freedesktop.org/~airlied/scratch/testdiff
Travis1: I just typed "git diff" into the xf86..etc folder and got this result: http://pastebin.com/m7d03d006
Travis1: is that what you were looking for?
airlied: then the make or make install failed
airlied: since that isn't the logfiles you lat posted
Travis1: I've been posting the two logfiles you asked me to post from each restart... do you need anything extra?
airlied: Travis1: the last restart didn't reflect the driver
airlied: Travis1: so either the make or make install didn't work
Travis1: hmmm shall I try again?
airlied: so redo those steps and look for errors
airlied: gotta run, hopefully with those logs, agdf might be able to give you a working patch
airlied: agd5f: ^^^
Travis1: Making all in src
Travis1: CC radeon_driver.o
Travis1: radeon_driver.c: In function ‘RADEONInitMemoryMap’:
Travis1: radeon_driver.c:1399: warning: format ‘%08x’ expects type ‘unsigned int’, but argument 2 has type ‘uint64_t’
Travis1: radeon_driver.c:1456: error: ‘aper0_base0’ undeclared (first use in this function)
Travis1: radeon_driver.c:1456: error: (Each undeclared identifier is reported only once
Travis1: radeon_driver.c:1456: error: for each function it appears in.)
Travis1: make[2]: *** [radeon_driver.lo] Error 1
Travis1: make[1]: *** [all-recursive] Error 1
Travis1: make: *** [all] Error 2
Travis1: sorry shall I just post it into pastebin?
airlied: oh remove the second 0 from aper0_base
airlied: just edit and search for aper0_base0 and remove the second 0
Travis1: how do I remove it?
airlied: just use your editor
Travis1: what file am I editing? There doesn't seem to be anything in "radeon_driver.c"
airlied: it shuold be in there
airlied: around line 1450 or so
Travis1: no as in there's nothing in that file. Nothing at all
Travis1: it's completely empty
Travis1: oh wait...
Travis1: I need to navigate to the /src or something don't I?
airlied: yup src/radeon_driver.c
Travis1: haha my bad
airlied: okay I gotta go, will be back on Monday for sure, on/off over weekend
Travis1: http://pastebin.com/m73227bcf
Telek: Cya!
Telek: Take care,
airlied: agd5f might see it and produce a working patch
Travis1: that look like that worked?
Travis1: yeah no worries.
airlied: Travis1: yup, boot again and get the logs
Travis1: thanks heaps for all of your help
Travis1: yeah will do
Travis1: ok another reboot fail with updated reboot logs
Travis1: can't wait for agd5f to come online and give me a hand. this has been 1 1/2 days of frustration...
Travis1: if anyone else is out there with some skills I would appreciate the help if you have time to tackle this beast?
TBBle: What sort of skills are you needing?
Travis1: so I've installed drivers, installed kernels, rewritten some code in the radeon_driver.c file, I keep on getting a black screen on boot if I remove my xorg.conf file
Travis1: want to see the logs?
Travis1: Xorg.0.log = http://pastebin.com/m5e2b35b1
Travis1: Xorg.0.log.old = http://pastebin.com/m380b430a
Travis1: obviously I needed to restart my computer and restore the xorg.conf file to be able to load ubuntu in safe graphics mode
maligor: boot in single mode?
Travis1: single mode?
maligor: aka 'recovery'
maligor: it gives you a root shell
Travis1: yeah so when I reboot my computer without the xorg.conf file, it goes to a black screen
TBBle: Immediately, or when it tries to load X?
Travis1: so I restart, enter recovery mode, get a root shell and then restore the xorg.conf file so I can load the OS again in safe graphics mode, bringing me back to here.
Travis1: umm when it tries to load x
Travis1: so I get the little white ubuntu symbol, then flickering, then black
TBBle: One of those pasted logs is radeonhd, and one's radeon.
TBBle: The radeonhd one appears to have X dying at the end.
Travis1: umm well I have no idea what's going on I'm just doing what people tell me to do. I'm a massive noob
Travis1: first time on ubuntu
TBBle: Which of the two logs is the broken one?
Travis1: first time typing into a terminal... hehe
Travis1: yeah I have no idea. So the http://pastebin.com/m5e2b35b1 is my Xorg.0.log file
adamk_: Well you aren't like to get help with radeonhd on this channel :-)
Travis1: haha airlied told me to jump on this channel for him to help me out
Travis1: like I said, happy to do whatever people tell me to do... I just really want to get this video card fixed...
maligor: ubuntu karmic ships with a snapshot?
TBBle: Well, assuming you're in X now, Xorg.0.log is probably working. Which is radeonHD, but appears to have crashed. The Radeon driver being loaded by the Xorg.0.log.old appears to load fine though.
TBBle: Final test. What's the driver line in your xorg.conf?
Travis1: http://pastebin.com/m368fe7fa
Travis1: that's my entire xorg.conf file contents
Travis1: so yeah I've been renaming my xorg.conf file and rebooting, bringing me to the black screen when Ubuntu tries to load. And then I have to restart in recovery mode, restore my xorg.conf file and restart again so I can come back here to find a fix.
TBBle: If you change "radeonhd" to "radeon" in your xorg.conf, does it also cause a black screen?
Wizzup: I was wondering... I have been playing a bit with 3d support on my r700, and some stuff works quite neat, while some other stuff is still lacking. Is it already the time to mention bugs when using, for example, wine for an other 3d game?
Travis1: not sure. I can change it and reboot and let you know what happens
adamk_: Wizzup: I don't see why not. They are certainly accepting bugs for 3D issues with r6xx/r7xx.
TBBle: Sure.
Wizzup: ok... /me boots up laptop
Travis1: alrighty changed. Be back in a tick with any logs if it fails
crimsonflame123: I've been getting the following messages in /var/log/messages whenever I happen to boot on my dual monitor setup
crimsonflame123: [TTM] Failed moving buffer. Proposed placement 0x00060004
TBBle: I'm pleased with how easily the whole kms thing comes together, but at the same time I wish I had more motivation to do new git pulls every day. As it is, everything's just working. Nice job everyone.
crimsonflame123: [drm:radeon_object_create] *ERROR* Failed to allocate TTM object
crimsonflame123: and [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object
crimsonflame123: X restarts randomly whenever I happen to be opening a window for some app
crimsonflame123: I'd really appreciate if someone takes a look at this problem
maligor: TBBle, heh, I just get a black screen on X load if I enable it (rv670), but haven't poked around it much yet
Wizzup: I am using wine 1.1.33 to play battlefield. it does the intro music, and the main interface, but actually playing doesn't work. Wine keeps throwing this error message:
adamk_: crimsonflame123: Have you checked to see if there is a bug report for this problem on freedesktop?
Wizzup: fixme:d3d:context_check_fbo_status FBO status unrecognized (0)
Wizzup: err:d3d:context_check_fbo_status FBO 0 is incomplete, driver bug?
TBBle: I am curious to know if there's any goal for stabilising/stamping the libdrm_radeon ABI?
adamk_: If not, you definitely should open one.
maligor: Wizzup, I wouldn't bother with d3d stuff yet
Wizzup: I guess it is more something that is incomplete rather than a bug
adamk_: Didn't osiris write the FBO code recently? You may want to check with him and see if he knows what's going on our has any ideas.
maligor: yeah, and then proceed to crash on something else :P
Wizzup: ok
crimsonflame123: adamk: I did open a bug on redhat bugzilla
crimsonflame123: I use fedora 12
TBBle: I'm on an rv770 (I think) and I haven't had a single hard failure since I switched over to KMS about a month ago. I had some weird colours on the xdm login box early on, but nothing once I was in my X session.
maligor: I'll have to get KMS working sometime tho, it's just fantastic for playing nethack in
adamk: crimsonflame123, Then be patient :-) Hopefully someone will get to it. You might need to poke the developers here or on the mailing list periodically.
TBBle: The only thing I've had to patch in is HDMI audio support. I don't even really need KMS, I only got it because I couldn't use fglrx with Xorg server 1.7 and I was going to see about doing some more Xinput 2 work on Wine.
Travis1: fail
crimsonflame123: adamk_: and there is something weird that I am seeing. I have a laptop and a screen which I use as my second monitor. I've set both to resolution 1280x1024. Still I see messages like these in /var/log/messages "kernel: [drm] LVDS-10: set mode 1400x1050 36" ... Why in world would anybody try setting 1400x1050 when I want to to be 1280x1024. Typically I've noticed these messages to happen *moments* before the X crash.
TBBle: OK, good. So it's not the lack of Xorg.conf, it's just that the radeon driver doesn't work for you.
Travis1: was that to me TBBle?
TBBle: Yeah. Sorry.
Travis1: http://pastebin.com/m3f40921e
Travis1: that's my latest Xorg.0.log
Travis1: yeah my video card is a radeon mobility hd 3650
mikkoc: yo
mikkoc: wrong chat :/
Travis1: so I would think I need the radeonhd driver
maligor: the radeon driver in karmic seems to be 1:6.12.99+git20090929.7968e1fb-0ubuntu1, so it's a old git snapshot...
TBBle: That Xorg.0.log is using the vesa driver. I assume that's you in failsafe mode? What's your Xorg.0.log.old like?
TBBle: Yeah, the other log with radeon claimed to be 6.12.99
Travis1: umm I guess
Travis1: http://pastebin.com/m1a6053af
TBBle: Travis1: What version of the radeon driver are you running? I assume you haven't built your own, so dpkg -l xserver-xorg-video-radeon should tell you.
Travis1: that's the Xorg.0.log.old file
crimsonflame1231: well I got disconnected... did I miss anything said on my problem
Travis1: http://pastebin.com/m5f775292
TBBle: OK, so presumably you're on karmic? Or was that said before?
TBBle: Uh, wait. Jaunty's current, isn't it?
Travis1: yeah on Karmic
Travis1: Karmic is the latest. 9.10
TBBle: OK, good. Temporary alphabet fail there, sorry.
Travis1: haha no probs
Ronis_BR: Travis1: :D
Travis1: if you have an idea of what to do next, let me know. I'm seriously all ears
Travis1: even if I have to start from scratch, if you think you've got the time and you can help... that'd be amazing
Travis1: hey Ronis
Ronis_BR: Travis1: hi
Ronis_BR: it seems kernel 2.6.32 is stable now!
Ronis_BR: does it contain the most recent DRM for radeon?
Travis1: so I went through the online ubuntu community guide for getting the driver working which didn't seem to help much...
TBBle: Travis1: You could try https://edge.launchpad.net/~xorg-edgers/+archive/ppa to see if the
TBBle: Ooops, premature enterjaculation.
TBBle: to see if the newer ATI driver works for you. But it's called bleeding-edge for a reason.
TBBle: Oh, neat. There's a revert-to-official script. Nice one, Ubuntu.
maligor: yes, unlike the non-bleeding edge driver karmic has included
TBBle: Hehe.
maligor: because it's a old bleeding edge
maligor: a bloodied edge?
TBBle: I was thinking more than the xorg-edgers PPA pulls in new Xserver, libdrm, mesa etc, as well as a new xserver-xorg-video-ati
TBBle: _And_ a new radeonhd, so it's possible that we might lose even that much of the working setup.
TBBle: Man, I had another bad brain fail. I thought the radeon driver was up to 6.13.99 for some reason.
Travis1: so what are you suggesting I do?
TBBle: Go to that page I linked. Follow the instructions. Particularly read the instructions on how to undo the damage, _before_ it breaks. Then see if the updated radeon driver works, and the updated radeonhd driver works.
Travis1: I've already added these PPA's to my 'Other Software' list:
Travis1: http://ppa.launchpad.net/xorg-edgers/drivers-only/ubuntu
Travis1: http://ppa.launchpad.net/xorg-edgers/ppa/ubuntu
Travis1: I added them last night I think
TBBle: OK. dpkg -l xserver-xorg-video-radeon | more
TBBle: Sounds like you're ahead of me.
Travis1: hehe system apparently is up-to-date from about 14 hours ago...
TBBle: I can see why airlied'd want you here to debug the driver, but I think right now is an unlikely time to hear from him.
Travis1: so do I setill need to write that dpkg etc. thing you posted?
TBBle: Yeah, best to be sure.
Travis1: http://pastebin.com/mb455148
Travis1: that's the result
TBBle: OK, cool. So you are running the absolute latest driver in the PPA. And you had this problem before you added xorg-edgers, right?
Travis1: yeah I had this problem from when I installed Karmic
Travis1: the video card has never been working
Travis1: I'm a previous windows user.
Travis1: I would fall in love with Ubuntu if I can just get this video card working...
Travis1: guess I'm learning about commands and stuff quickly since I kinda have to
TBBle: To just make it work, sounds like you want to remove the radeon driver, and just stick with radeonhd. Without an Xorg.conf, I guess it's picking radeon first because it's alphabetically before radeonhd, assuming both claim the device's PCI ID.
TBBle: Of course, making it work in radeon would be better, but sounds like it might be hard to organise. Certainly I don't see anything in the logs that leaps out at me, but that doesn't mean it's not there.
Travis1: hmmm
maligor: might be something strange about the vbios
Travis1: like I said, I'm the noob here. Preach it to me, and I will follow orders
maligor: which is sometimes bundled with systembios updates
Travis1: so I've gone through: https://help.ubuntu.com/community/RadeonHD
TBBle: Well, unless you're up for a remote debugging session, with a side of hands-off coding, and can actually tie a dev down to do it, I'd just plunk for radeonHD, and try radeon when it gets updated in case this problem gets fixed.
Travis1: yeah sure I don't really care if it's radeon or radeonHD... I just want a resolution and screen size that's not so massive and stretched...
Travis1: at this point I'm not focused on gaming. I can worry about that later.
TBBle: OK. Then prolly stick with the 'has an xorg.conf' option. Maybe see if you can uninstall xserver-xorg-video-ati, although that's not really necessary if you're using an xorg.conf to pick the driver.
maligor: Travis1, actuallyl, on a laptop I wouldn't use the opensource driver yet
maligor: the power management features are a bit limited
maligor: you'll probably get much better battery life with fglrx
TBBle: Although you'll have to downgrade your X back to 1.6.5, probably using that ppa-purge script.
Travis1: like I said, I'm a REAL noob, so I'm not sure what the difference is with an opensource driver or not. If you guys can just tell me what you think will work 0 I'm happy to do it.
Travis1: I've already tried installing the System - Administration - Hardware Drivers - ATI/AMD proprietary FGLRX graphics driver
TBBle: Oh, xorg-edgers is still 1.6.5, so that's OK.
Travis1: that was my first thing since I thought the system would give me the right option and driver...
Travis1: but alas, black screen on loading Ubuntu
Travis1: I had to re-install Ubuntu from scratch at that stage cause I didn't know what else to do
TBBle: Interesting. So fglrx doesn't work, radeon doesn't work, but radeonhd does.
Travis1: well... not that I know of. As in I've never been able to load my computer and change the screen size...
spreeuw: xrandr
Travis1: if radeonhd works, I've not seen it allowing me to change my screen size and getting away from this large and stretched screen... hehe
TBBle: Oh? Didn't know about that.
TBBle: Hmm. I wonder if Ubuntu's X does something funky when the radeonhd driver crashes, like autoswitch to vesa or something.
Travis1: yeah I'm not sure. I've been looking in a lot of forums and such
TBBle: Is the screen stretching you have now the same as it was when I first started "helping" you?
Travis1: one person had listed in the bios they changed some setting for their card... not sure if I should try that
Travis1: yeah
Travis1: always stretched
Travis1: the screen is always set to "monitor: unknown" and Resolution: 1024x768 (4:3)
Travis1: nothing is able to be changed
Travis1: no dual screen possibilities.
Travis1: no working video card.
Travis1: always some dodgy default/safe graphics mode
TBBle: Does xrandr -q work?
Travis1: type that into a terminal?
Travis1: Screen 0: minimum 1024 x 768, current 1024 x 768, maximum 1024 x 768
Travis1: default connected 1024x768+0+0 0mm x 0mm
Travis1: 1024x768 0.0*
Travis1: haha that answer your question? hehe
Travis1: the card isn't working. I just want it to work.... :(
TBBle: Yeah. If you put radeonhd back in your xorg.conf and reboot, does it do the same thing?
Travis1: yeah pretty much.
Travis1: so I've just changed my xorg.conf to be like this: http://pastebin.com/m61806106
Travis1: can you confirm that all looks as it should and I'll reboot my comp and then show you the log?
TBBle: You've got some jive-ass stuff in that config file. I assume it's fglrx leftovers... But it should be OK. reboot, and see what xrandr -q says.
TBBle: Oh, and what's your laptop's actual native resolution?
Travis1: uhhh not sure. I can't remember what it was. I just know it's defo not meant to be this one... hehe
Travis1: ok I'll rebook and then run xrandr -q and let you know what it says
Travis1: I really appreciate this help mate
Travis1: travis@travis-laptop:~$ xrandr -q
Travis1: Screen 0: minimum 1024 x 768, current 1024 x 768, maximum 1024 x 768
Travis1: default connected 1024x768+0+0 0mm x 0mm
Travis1: 1024x768 0.0*
Travis1: fail
TBBle: OK, post your Xorg.0.conf, just to make sure it's doing what I think it is.
Travis1: http://pastebin.com/m56deba5a
Travis1: there's the log
Travis1: I think you mean Xorg.0.log tho right?
TBBle: Yeah
Travis1: yep
Travis1: hopefully something jumps out at you...
TBBle: OK, so new theory. Radeonhd isn't working either, and Ubuntu's kicked in some kind of failsafe to get you into X with vesa or something running, and is logging it somewhere else.
Travis1: right...
TBBle: So, your best option might be to go 'round to AMD's offices, and leave a burning envelope full of feces on their doorstep.
TBBle: The fact that fglrx doesn't work suggests your card or it's BIOS, more likely the latter, is doing something majorly screwy and out-of-spec, and the driver will need a hack of some kind to compensate.
Travis1: right,,,
Travis1: I might look at what the bios is doing since I haven't had a look at that yet
TBBle: Lines 473 and 474 of http://pastebin.com/m5e2b35b1 might be relevant. Or might be irrelevant. I'm pretty sure they're not grapefruit, at least.
Travis1: but everything considered, the xorg.conf file should be working yeah?
TBBle: Might want to check for a bios update
Travis1: ok sure
Travis1: back soon. Checking out the bios settings now
TBBle: Ah, rebooting for video games. TTFN.
marvin24: I'm getting "mesa/lib/gallium/r300_dri.so: undefined symbol: align" with todays gallium build
marvin24: something changed?
marvin24: maybe this one: http://lists.freedesktop.org/archives/mesa-commit/2009-December/014699.html
phercek: Did anybody get TA Spring running fine with radeon driver? I tried with git version from last weekend and some models are missing textures. The game works fine with nvidia card and proprietary drivers.
phercek: ach; this attempt was with RV670PRO
chithead: phercek: spring does not work properly, there are additional issues with shadows and ARB_fragment_program_shadow
brian_: Does the radeon driver also work for radeon xpress cards in laptops?
chithead: radeon driver works for all radeon chipsets, igp and discrete, desktop and mobile
chithead: except for the newest 5000 series
brian_: ok, thanks
phercek: chithead: thanks
hatseflats: hmmm, getting rather low fps on my r670, around 74 fps in fullscreen glxgears (1920x1200), is this normal?
marvin24: wonders, cause align is defined as INLINE in u_math.h
hatseflats: ah, it's software rendered, that'd explain it
agd5f: airlied, Travis1: http://www.botchco.com/alex/xorg/limit_aper_256.diff
agd5f: that patch should do the trick
roysjosh: airlied, I think I had the same issue as Telek with the IGP345. see https://bugzilla.redhat.com/show_bug.cgi?id=537140
roysjosh: agd5f, can you please take a look at https://bugzilla.redhat.com/show_bug.cgi?id=537140#c15 and see what you think? I know that's now the right place to put the fix, but does it seem logical?
agd5f: roysjosh: I'll whip up a patch. thanks for tracking that down
roysjosh: agd5f, thanks!
brian_: What's a good result for glxgears on a radeon xpress 1250? I'm getting anywhere from 1600-1700
fabio123: hi
BioTube: brian_: glxgears is not a benchmark
fabio123: a question
fabio123: why exa is so slow tith the flash in firefox?
fabio123: with*
BioTube: IIRC, Flash uses an opengl rendering path
fabio123: i would like to use 16 bit color depth bu i can't
fabio123: due to the flash in firefox
fabio123: Flash should use Xvideo or something like that
[Enrico]: fabio123: flash is slow by desing
[Enrico]: it doesn't use xvideo
[Enrico]: flash uses cpu only afaik (at least the oddobe version)
[Enrico]: gnash uses some video accel
[Enrico]: adobe*
fabio123: radeon is broken in Ubuntu 9.10
fabio123: exa acceleration is very slow, xaa is fine, at least with my igp 345 (rs200)
brian_: biotube: then what's it for?
maligor: flash on linux is abysmal anyway
fabio123: i'm running gnash now
maligor: albeit it's better than it used to be (flash 7)
fabio123: it's faster
BioTube: brian_: it's a simple "is it working?" test
maligor: yeah, gnash is pretty nice but doesn't work with everything
fabio123: it show no video on youtube though :(
[Enrico]: unluck gnash works with almost nothing :'(
maligor: there's swfdec also, it's better in some things
[Enrico]: maligor: swfdec is no more developed :'(
fabio123: see later
[Enrico]: maligor: it was in a better state of gnash but gnash devs instead of helping the swfdec development started a new project from scratch :'(
maligor: well, it was certainly more stable than gnash when I tried it
maligor: can't say I'm having performance issues with adobe flash 10 with radeon on rv670 tho
maligor: (in chromium)
[Enrico]: it works good here on firefox (rv620 with radeon from git)
maligor: yeah, I'm on git master also
marvin24: http://pastebin.com/m3997d4bb fixes a missing header in gallium radeon build
agd5f: roysjosh: http://www.botchco.com/alex/xorg/0001-drm-radeon-kms-legacy-set-overscan-regs-on-modeset.patch and http://www.botchco.com/alex/xorg/0002-drm-radeon-kms-legacy-set-common-regs-to-sane-value.patch
roysjosh: agd5f, great, thanks, I'll try those out over lunch
uyf: Hello, i have had the driver working for me in kms mode perfectly with compiz for a while. but now the fps goes down to like 5. i'm using drm next with latest git pulled and patched for irqs vram and clip vram to agp ap size
uyf: i'm wondering i should leave out the vram patches i'm going to try the 2.6.32 with patches instead
kdekorte: uyf, I'm using drm-radeon-testing with mesa, libdrm and xf86-drv-ati all from git an performance is excellent on my rv635 card
kdekorte: I'm using compiz as well
uyf: kan i do a make clean then make mrproper and pull radeon testing to my drm-next dir?
uyf: nm i'll try it
kdekorte: drm-radeon-testing is a branch of drm-next
[Enrico]: mhm someone said me that drm-next contains some powersave patches for kms..... is he lieing ?
[Enrico]: lying*
kdekorte: Ask Zajec, he is doing the patches
[Enrico]: kdekorte: thanks, does he joins here sometimes ?
kdekorte: Yes quite often
[Enrico]: kdekorte: thanks again i will wait for him :D
kdekorte: [Enrico], however, I don't see them in drm-radeon-testing, so they might be an external patch you have to add in
uyf: thanks for the help kdekorte going to compile radeon-testing, since i have a rv635 too
kdekorte: uyf, no problem
[Enrico]: kdekorte: well someone said me it was already merged in drm-next but well i never heard about it, that's why i asked. i heard too there was a *separate* patch
[Enrico]: but i will just wait for the drm-next merge in the latter case
[Enrico]: you know i'm on a rv620 and for now i don't use kms and i force lowpowermode (laptop here) but i would be happier with a bit of automatic powersave
[Enrico]: (using kms disable lowpowermode so the pc becomes hotter)
kdekorte: [Enrico], it is definitely coming as a lot of people have been asking for it
[Enrico]: eheheh good :D
PuffTheMagic: do i need to use libdrm from git with latest kernel drm now
PuffTheMagic: i updated something on my system this week and I cant get into X any longer when kms is enabled
kdekorte: PuffTheMagic, you need libdrm and then to compile xf86-video-ati after you install it
PuffTheMagic: yeah i do that
PuffTheMagic: and i know i need libdrm
PuffTheMagic: my question was which libdrm
kdekorte: git
PuffTheMagic: well thats awesome, libdrm from git does not build
kdekorte: PuffTheMagic, yup your right, seems someone forgot to commit config.h.in
PuffTheMagic: kdekorte: do you have the missing file or have the ability to push it?
kdekorte: After running autoreconf for libdrm git, it fails in modetest.c
kdekorte: PuffTheMagic, run autoreconf
kdekorte: but it still fails
kdekorte: but seems to fail in the tests
PuffTheMagic: yeah thats where its failing for me
[Enrico]: if fails for all there
[Enrico]: a simple syntax error imho
PuffTheMagic: [Enrico]: if its simple give send me a patch
[Enrico]: PuffTheMagic: i'm not an expert of libdrm, i don't want to make some disaster :D
[Enrico]: i just know the c syntax
BioTube: [Enrico]: I managed to fix a null pointer dereference in an emulator without even knowing what exactly was wrong
[Enrico]: BioTube: well i do it too normally, but well libdrm is to....... critical for a blind patch
BioTube: true
[Enrico]: it can couse damage
[Enrico]: so i just wait
[Enrico]: or well i can hack libdrm to know how it works, but well it requires time :D
PuffTheMagic: [Enrico]: i takes nearly no skill to read the error message from gcc, its obvious what is wrong
[Enrico]: PuffTheMagic: indeed, but the fix is not trivial
BioTube: PuffTheMagic: sometimes those messages can get the cause completely wrong
[Enrico]: PuffTheMagic: i mean let's say i delete/add the missing member, what about if this is not the right solution?
PuffTheMagic: well someone pushed a fix
PuffTheMagic: thanks who ever did that
MostAwesomeDude: eosie: I don't agree with patch (1); I think that r300_winsys.h needs to get nuked.
MostAwesomeDude: But I'm applying and pushing them all anyway.
MostAwesomeDude: I'll get to it later, when I've a bit more spare time.
PuffTheMagic: ahh its good to have kms and X working together again
ernstp: wow, I get like 3 fps on http://alteredqualia.com/visualization/evolve/
ernstp: I wonder if this isn't related to the radeon driver somehow
ernstp: have to try it without exa
hifi: so, can ut2k4 be made to work without s3tc?
MostAwesomeDude: ernstp: No, it just takes a while. It's up to your CPU.
roysjosh: agd5f, the patches work, I can see my screen. thanks! if you have a minute, one more thing: suspend/resume comes back with a black screen. backlight is possibly on (radeontool with no args says: on) here's the before and after diff in regs: http://fpaste.org/4hVp/
hifi: hmm, cool
hifi: drmRadeonCmdBuffer: -22. Kernel failed to parse or rejected command stream. See dmesg for more info.
agd5f: roysjosh: does changing MEM_STR_CNTL help?
agd5f: hifi: either too old drm or too old mesa
agd5f: most likely
hifi: I tried to run a wine app that requires s3tc
hifi: http://pastey.net/129789
hifi: thats the dmesg bitr
roysjosh: agd5f, doesn't appear to
agd5f: hifi: tex compession is broken with kms at the moment on r100-r500 cards
hifi: agd5f: can I do something to get something to work? :)
hifi: or help
MostAwesomeDude: I got a request to look at the S3TC tex check; I might tackle that tomorrow
agd5f: hifi: fix the size calculation in the kernel cs checker
hifi: noo, no kernel building
agd5f: compressed textures are block based rather than pixel based, so the calcs need to take that into consideration
hifi: why everything I try to do today ends up badly :p
hifi: first I wore a gray hat and got scolded for that, now with a white hat the driver itself didn't work, not the wine app
roysjosh: agd5f, I'm in a debug kernel with a fair amount of debugging turned on, if it helps: http://fpaste.org/Hpy3/
roysjosh: ^ dmesg
agd5f: roysjosh: it's going to be a matter to sorting out the reg differences most likely. you might try this patch: http://people.freedesktop.org/~airlied/scratch/combios-readback.patch
roysjosh: agd5f, ok, thanks
agd5f: or this one: http://people.freedesktop.org/~airlied/scratch/combios-readback-v3.patch
ossman: agd5f, ping
agd5f: ossman: pong
eosie: MostAwesomeDude: yeah... the important thing in that patch was to fix implicit declarations of functions
ossman: agd5f, I was gonna ask what that patch changed, but I figured it out
ossman: agd5f, still, I don't think it fixes it as CRTC2 hangs before any attempt is made to disable it
agd5f: ossman: ok
ossman: agd5f, that patch does raise the question why my patch worked at all though
soreau: MostAwesomeDude: (or anyone) http://pastebin.com/m6236333e Care to take a look at why Driver "modesetting" crashes?
MostAwesomeDude: soreau: I'll need to see for myself, unless you care to post a disassembly. :3
soreau: MostAwesomeDude: If you care to tell me how to do it, I might be able to
lordheavy: sometime xorg (kde) stall many second when doing various things (opening a new tab/switching windows/...) i've try with/without kms, with older kernel (2.6.31) but no changes. How can i trace these problems ? (it's with a RS880 and all is build from git, except xorg-server)
maligor: lordheavy, is the system responsive in between?
lordheavy: really slowly, i saw that hdd led is light on
lordheavy: the system isn't swapping
lordheavy: (4go mem)
lordheavy: nothing in Xorg.0.log or dmesg
soreau: MostAwesomeDude: Well anyway, it happens here on my rv350 as shown in the log. I assume modesetting works elsewhere..
lordheavy: will try oprofile ....
Ronis_BR: does kernel 2.6.32 have the lasted drm for radeon?
Ronis_BR: lastest*
spreeuw: a working one yes
Ronis_BR: I'll see
dandel: 2.6.32 drm is bad, hypernate/resume fails ><;
dandel: I have a bug report i'll be updating soon with the tests.
spreeuw: dandel: oh I'm on a normal pc
dandel: well it works, but i am on laptop tho ><;
dandel: so hybernate/suspend/resume are high priority.
dandel: pretty much with hybernate/resume i get a drm:radeon_cs_ioctl and drm:radeon_ib_schedule errors.
Pallokala: was CONFIG_FB_RADEON required to get KMS working?
Pallokala: or unnecessary?
jcristau: unnecessary
BioTube: Pallokala: IIRC, venomous to KMS
Pallokala: ok, unselectinh
Pallokala: so I have agpgart and CONFIG_DRM_RADEON and that is enough?
Wizzup: What times is osiris on?
kdekorte: Pallokala, pm'd you a bunch of settings
gmartyn: i'm using the experimental kms driver from fedora 12. Should I expect screen corruption, or report it? thanks.
roysjosh: gmartyn, you should probably ask if something has been fixed recently for your card, and then report it if the answer is "no"
gmartyn: ok, Mobility Radeon HD 2600 XT here. The worst is that the task bar buttons are blacked out. Screenshot: http://imgur.com/RR7CM.png
gmartyn: I also see random horizintal lines across all my applications that go away when app under them is repainted. can't screenshot that.
gmartyn: plasma applets have extraneous squares, see http://imgur.com/0xOYi.png and http://imgur.com/gq15J.png
gmartyn: those squares also appear on task bar buttons when i mouse over them
airlied: gmartyn: using desktop effects?
gmartyn: yeah
airlied: turn them off
airlied: see if it helps, though I think we have an X server bug we need to fix
gmartyn: problem remains
gmartyn: i don't see the random horizontal lines
gmartyn: but the other problems (task bar buttons, plasma applets) remain
amarks: gmartyn: the extraneous squares seems fixed in recent git
amarks: for my card anyway
airlied: gmartyn: yeah I think its a bug in the xserver in F12 hopefully get some time next week to look for it
gmartyn: airlied: cool, thanks
dandel: http://pastebin.com/d6cbaa478 (Nice little hang on radeon driver, first error at: [ 115.296036] [drm] Clocks initialized ! )
glisse: dandel: agp gpu ?
airlied: glisse: IGP
airlied: probably sideport agani
glisse: yeah maybe
glisse: i thought sideport memory never hit the market, i was completely wrong with this assumption
airlied: we just need to detect it and flush VRAM object I suppose
dandel: no sideport.
dandel: it's radeon 3100 (rs780mc)
dandel: the card pulls a minimum of 256mb from system memory.
glisse: airlied: btw did you push ajax's patches to f12 ?
glisse: i need to relearn cvs command to check history
airlied: I thought he had, but he probably hasn't yet, I'll rebase on Monday
dandel: I have a lot of test results, also, it appears upon resume i get only a black screen with mouse (ubuntu 9.10 with latest kernel git)
dandel: resume from suspend results in black screen that is.
airlied: agd5f: btw LVDS fix helped the s/r but broke normal display https://bugzilla.redhat.com/show_bug.cgi?id=541562
airlied: think we might need to copy some of the bios ref divs around or somethnig
agd5f: airlied: yeah, lvds is pretty picky on those rv4xx cards
airlied: I'll break out raedintool nexxt week with him and track it down
Ralesk: hi again airlied :)
agd5f: airlied: got hpd working well
agd5f: jsut about to send out the patches
evocallaghan: rnoland: Around bud, long time no see?
Ralesk: gmartyn: hmm, random horizontal lines... I got such last night when I was moving around a window with just kwin running
airlied: agd5f: excellent
gmartyn: Ralesk: same here. Also, switching tabs in firefox is all it takes for them to appear
evocallaghan: any know have more a clue about automake then me?
Ralesk: on the other hand, running a full plasma desktop makes things crash as soon as a window is on the screen
evocallaghan: Trying to build DRM on FreeBSD, wondering if this looks ok? http://pastie.org/728140
gmartyn: Ralesk: i haven't had any crashes
evocallaghan: that's ageist the latest git trunk for drm
Ralesk: I have drm 2.4.15, should I get a newer one for my yesterday's radeon driver?
soreau: Ralesk: You probably want kernel, libdrm, mesa and ddx all from the same daay ;)
turmlos: airlied: Is there an open bug regarding the sideport issue? I'm wondering if that's what's causing my problems (X1270).
evocallaghan: why is no one talking to me on any channel today? :(
evocallaghan: feels ignored
glisse: evocallaghan: we don't know much about freebsd
soreau: doesn't know much about freebsd either
evocallaghan: Its more a automake question glisse
glisse: last time i check freebsd drm driver were directly included in the freebsd kernel tree
airlied: turmlos: not sure, if you can disable sideport then its worth testing
evocallaghan: ahh! I was starting to think my irc connection was not working for some reason
soreau: evocallaghan: It isn't, we still can't see you.
evocallaghan: hehe
evocallaghan: glisse: prob has been imported, but should the trunk not build also?
evocallaghan: looks like it just needs some automake tweaks
glisse: dunno i am not automake expert but i think the freebsd fella did have it compiling
turmlos: airlied: Probably not, my BIOS is pretty barebones. :/
Ralesk: soreau: well, hmm, kernel is 2.6.32~rc8, could try to build a .32 release, can't get any newer than that though.
soreau: Ralesk: 32-rc should be fine, just make sure to build libdrm with --enable-radeon-experimental-api and build mesa and ddx against that libdrm
Ralesk: dig, that's how it's been already, but for some reason things are crashing here left and right when I use composite
evocallaghan: glisse: because I need to get drm building to build the latest mesa trunk for which I am interested in G3D you see.
evocallaghan: thus I need to get DRM building.
airlied: evocallaghan: do you have kernel drivers btw?
airlied: evocallaghan: I'm not sure how G3D can help without porting KMS to your OS kernel
glisse: yeah you likely need kms at least for radeon hw you need kms
glisse: freebsd don't has this
evocallaghan: well, first I just want to get it to 'compile' rather then even work
glisse: won't compile
evocallaghan: I know
evocallaghan: I plan to work on it............
glisse: gallium will ask for things the freebsd kernel header can't provide
evocallaghan: but this automake stuff is in my way :p
evocallaghan: glisse: I know, I know.
glisse: well i am not automake expert but i think Kristian got it sorted out
glisse: evocallaghan: ask him on dri-devel if he thinks this change are sane
evocallaghan: well I was just looking for some feedback on the above patch
airlied: I though rnoland had latest libdrm buildign fine
evocallaghan: ok
evocallaghan: I thought rnoland was *the* only freebsd guy working on X 'stuff'
glisse: yeah i did think too that rnoland did build lastest thing properly
evocallaghan: I don't think he is about at this time, if I remember correctly
evocallaghan: glisse: I don't know what Kristian nick is, sorry.
glisse: krh
evocallaghan: ah ok
airlied: evocallaghan: I assume you are using FreeBSD 8.0 or so
evocallaghan: yessy
airlied: not sure what versions of automake/autoconf are needed
airlied: but you might have too old ones
glisse: anyway time for zzzZZZzzz
glisse: &
evocallaghan: err, well ./autogen did not warn me its too old and it _does_ have a check
evocallaghan: night glisse
airlied: okay then you need to fix that
airlied: probably need something from ports
evocallaghan: autoconf (GNU Autoconf) 2.62
evocallaghan: sounds fairly new
airlied: automake?
evocallaghan: automake (GNU automake) 1.4-p6
evocallaghan: looks old
airlied: yup I think 1.10 is needed
airlied: later &
evocallaghan: hmm, should warn me!
evocallaghan: installed later version
ossman: I'm having some performance issues and this function accounts for 88% of the time in oprofile: radeonReadRGBASpan_ARGB8888
ossman: anything that's obvious to you guys?
Wizzup: osiris_: I heard you recently did a part of the FBO implementation? Can you tell me how far is it done?
osiris_: ossman: if you have r300,r400 or r500 card try my r300-blit branch
ossman: osiris_, r600 unfortunately
osiris_: Wizzup: nope, I implemented accelerated glCopyTexImage
ossman: it is projectM that's having issues.
ossman: I think it used to run a lot smoother a while back though....
Wizzup: Ok, so fbo isn't done yet? Then I won't file a `bug' report :)
ossman: RADEON_DEBUG=fall to see if the driver does software fallbacks, right?
osiris_: Wizzup: FBOs were implemented a while ago by airlied
osiris_: ossman: yes
ossman: and output should come to stderr?
osiris_: yes
Wizzup: Ok. Cause I'm trying to run bf1942 with wine (d3d), and I am getting this warning: (and the game won't actually run games either)
Wizzup: fixme:d3d:context_check_fbo_status FBO status unrecognized (0)
Wizzup: err:d3d:context_check_fbo_status FBO 0 is incomplete, driver bug?
ossman: osiris_, I'm not getting anything. got any good flag that will definitely print something so that I can test that I get the debug output?
osiris_: Wizzup: it looks to me like a wine bug. driver is allowed to set framebuffer as incomplete if an image of unsupported format has been attached to it
Wizzup: I don't have this issue on other computer, which has nvidia binary driver though
Wizzup: Same wine
osiris_: ossman: RADEON_DEBUG=cs
ossman: thanks
osiris_: Wizzup: nvidia driver may support more formats than radeon driver
ossman: gah!
ossman: ok, so nothing wrong with the output
Wizzup: osiris_: Ok, I'll wait for a bit then
osiris_: Wizzup: talk to stefand on wine channel , maybe he'll be able to provide more info
Wizzup: Ok, thanks
ossman: osiris_, I have swizzle_copy as the second dri function in oprofile's list. any clues from that?
osiris_: ossman: not all fallbacks are logged currently.
ossman: bummer
osiris_: ossman: that's probably part of software implementation of glCopyTexImage (just like radeonReadRGBASpan...)
ossman: osiris_, my OpenGL knowledge is not that complete... that function copies the framebuffer to a texture?
osiris_: ossman: yes
ossman: osiris_, I'd imagine projectM makes heavy use of that as it constantly morphs the previous frame
ossman: osiris_, I take it that function is not yet accelerated. Anyone working on it though?
osiris_: ossman: I don't think so
agd5f: airlied: hpd patches sent
ossman: osiris_, would it be difficult to do?
agd5f: ossman: accelerated blit?
ossman: I'm not sure what the full extent of the function is
ossman: but if it's not too difficult then maybe I'll be able to sort it out :)
agd5f: I was planning to do it for r600 once osiris_ merged r300 support. basically a copy of the exa blit code
agd5f: but feel free to beat me to it
agd5f: I probably won't have time to look at it for a while
ossman: the idea is basically to tell the gpu to blit from the framebuffer to the texture?
osiris_: agd5f: I'll push r300 support this weekend
ossman: osiris_, got a tree I can start looking at?
agd5f: ossman: right. basically just a copy using the 3d engine
ossman: what about format conversions?
ossman: or are they guaranteed to be the same format?
osiris_: ossman: nope, I didn't work ok r600 at all yet
ossman: osiris_, I meant for the r300 work. I suspect I can grasp the general idea from looking at that code and trying to do something similar :)
agd5f: so load the shaders from the exa copy code, and set the state the same way
osiris_: ossman: http://cgit.freedesktop.org/~osiris/mesa/log/?h=r300-blit&=switch
ossman: agd5f, any specific parts of the exa code I should look at?
ossman: osiris_, thanks
agd5f: ossman: exacopy. actually the Xv code might be easier to understand
agd5f: r600_texturedvideofuncs.c
ossman: and the Xv code is something I've already poked around in :)
agd5f: just use the non-csc pixel shader
ossman: agd5f, non-csc?
agd5f: ossman: colorspace conversion
ossman: ah
dandel: just spotted another interesting thing... using radeon driver instead of radeonhd (i'll double check suspend against that)
ossman: agd5f, osiris_, I really should have been getting some output for this case with RADEON_DEBUG=fall, right?
agd5f: ossman: r600 fallback tracking is pretty broken
ossman: k
agd5f: there are a lot of cases where we should fallback but don't
flyback: prepares to call mastercard and threaten the shit out of them
spreeuw: flyback: uhh
spreeuw: number taken?
flyback: no decline
spstarr: is in trunk drm-next, ddx, libdrm, mesa
spstarr: tests for stalls
spstarr: glxgears shows stalls
spstarr: when full screen
spreeuw: try the correct build order
spreeuw: drm libdrm ddx mesa
spstarr: libdrm, ddx, mesa, kernel
spreeuw: did you build it for the first time?
spstarr: yeah stalls still happening
spstarr: no
spstarr: testing the new irq support
spreeuw: does it look like rotating backwards?
spreeuw: oh ok
spreeuw: hey will irqs improve performance?
spstarr: very much so
spreeuw: I'd like some more beef in nexuiz
spstarr: ive already seen a 40% improvement with the latest changes with KMS + DRI2
spreeuw: very brutal effects in that game :D
spstarr: but the gears 'stutter' sometimes when spinning
spreeuw: nice
spstarr: same in secondlife
spreeuw: is thta a missed irq?
spstarr: so they are still not done the optimization/paths
spreeuw: where are irqs implemented?
spstarr: back to intel GPU, i continue to test as they progress
spreeuw: drmnext?
spstarr: the stalling may be due to some irq issues i dont know
Lebenskuenstler: Hello. I have a ATI Radeon HD 3200 onboard gpu connect via DVI and F12, using F12, latest git build from koji. The screen is flickering quite often. Is this a problem with KMS? If I use the a VGA-connection the flicker does not occur, but the the image quality on my 28" LCD is too poor.
spreeuw: 1920x1200?
agd5f: Lebenskuenstler: try this patch: http://www.botchco.com/alex/xorg/0001-drm-radeon-kms-pll-fixes.patch
gimzo: I had that flickering, F12, hd4200, DVI, 1920x1200 and I put back my old x700 so I don't know how to fix it
agd5f: gimzo: try this patch: http://www.botchco.com/alex/xorg/0001-drm-radeon-kms-pll-fixes.patch
gimzo: agd5f: I'll take a look, but I'll probably just stay on r400 for now
gimzo: x700 sholud be faster in 3D than hd4200 ?
agd5f: gimzo: it's not likely to get fixed unless I get some feedback
agd5f: gimzo: depends on the app
spreeuw: no a 4200 is 3x faster
spreeuw: if its not an onboard chip
gimzo: it is
spreeuw: then I dunno
spreeuw: but I had a 3450 with its own ram and that blew my old x700 away
spreeuw: now on 4670
spreeuw: which in turn blows the 3450 away
flyback: whew they just wanted to verify that the paypal use was real :)
spreeuw: even now it can do 1280x720 nexuiz more or less playable
flyback: had a feeling mabye it was paypal
spreeuw: with unoptimized drivers
gimzo: agd5f: I patch kernel with this ?
agd5f: gimzo: yes
gimzo: ok
gimzo: I'll try it tomorrow
Lebenskuenstler: spreeuw: yes. 1920x1200
gimzo: and wil then benchmark hd4200 against x700 :)
spreeuw: gimzo: on higher res you notice they have more power
spreeuw: but pitting a real videocard against an onboard may lessen the gap
gimzo: I'll take a look at it, I notice now on 1920x1200 in openarena there is input lag, but it's not there with gallium
flyback: turns out at least with some nvidia cards you can replace the default fonts
flyback: is that possible on some radeon card bios's?
flyback: cause that would be really awesome to replace the fonts with ones that are compatible with 480x234 resolution portable dvd player lcd attached to tv out port
maligor: spreeuw, heh, I get around 15-50fps on 1680x1050 in nexuiz
MostAwesomeDude: flyback: Don't flash VBIOS if you want support from us.
flyback: well I might do it anyways
flyback: it's for a project box
spreeuw: maligor: it looks so purty on normal detail
spreeuw: but I saw some screenies of even purtier texture
spreeuw: son the weapons
MostAwesomeDude: flyback: Okay, but we can't support flashed cards.
flyback: I understand that :P
flyback: card is most likely going to be for diagnostic purposes and not even get a driver loaded
flyback: I was just curious if it's possible to change fonts on ati's vbios
Lebenskuenstler: ag5df: is there anything I can do until i can apply the patch?
maligor: spreeuw, yeah, I was impressed how well it works, even if it isn't really quite playable :P
spreeuw: some maps are very heavy
spreeuw: I played a bit with that fireball BFG against bots
spreeuw: card chokes on thta ;p
spreeuw: also tried opengl2 shaders
spreeuw: gave me black walls
spreeuw: but occlusion alone speeds things up
Lebenskuenstler: test
dandel: invalid ioctl with kms radeon_cp_setparam_kms on radeon 3100 onboard.
dandel: also, it has error just after that about radeon_cp_init_kms
dandel: i can't get xorg running on latest git of radeonhd ><; kms errors.
Ralesk: airlied: well, with a newer libdrm (.16) and mesa and xserver 1.7.3 rebuilt against it, my GL compositing seems to be stable
Ralesk: X takes about a constant 35% CPU in idle though, and plasma 14%. without compositing, plasma does 43, X 24 and dbus-daemon another 14 :P Well, old machine and old card, but this is a bit ridiculous :)
Ralesk: also soreau ^
soreau: If it compiles, ship it!
soreau: hides
Ralesk: haha :)
Ralesk: well... that's kinda what we do usually XD seems to work!
Ralesk: nah, there's more to it, especially with core stuff :P
Ralesk: instantaneous VT switching is just way too hot
Ralesk: (not to mention 210x65 consoles :) )
sonne: hello :)
sonne: i'm using a dual-head setup with both monitors at 1280x1024, in a xinerama-like setup
sonne: and everything is working smoothly
sonne: now, i was thinking of replacing my two monitors with some new ones which have a 16:10 (not 16:9) ratio
sonne: for what you know, is such a resolution supported?
gimzo: sonne: I have 1920x1200, that's 16:10 and works ok
sonne: great!
sonne: think it could give any kind of problems on 16:9?
sonne: erm
sonne: think it could give any kind of problems on dual head?
gimzo: most widescreen monitors are 16:10 AFAIK
Ralesk: soreau: hmmmmmm, this is odd :) there are some white, horizontal lines popping up for a fraction of a second... just a few pixels long, I'd guess 16 because it's a nice round number :)
soreau: Ralesk: What card is it?
Ralesk: radeon 9600
gimzo: sonne: I don't know, but I don't think there should be problems
soreau: Ralesk: Are the lines happening on the right side of the screen mostly? oO
Ralesk: yeah, I'd say so, yes.
soreau: Ralesk: While 3D is happening?
sonne: well thanks for the info :)
soreau: like compositing WM?
Ralesk: there's a conky window up, and I think it might be when it gets updated, but the lines flick up on the desktop. and yes, compositing/gl is on
soreau: Ralesk: Well I could tell you the fix but then I'd have to kill you
Ralesk: hmmm, nope, the last two didn't appear when conky updated, my bad
Ralesk: heheh :3
Ralesk: is it something like "buy a newer ati card already"? :D
soreau: In the kernel, drivers/gpu/drm/radeon/r100.c:2176 or so, under where it says /* get modes */, change the if statement to read if (ASIC_IS_R300(rdev)) {
soreau: basically, there's something that needs to be ported to KMS that hasn't been yet (or at least afaik)
soreau: this is somewhat of a hack to work around it
Ralesk: and I guess that modification is very specific to my card so we shouldn't put it in our stock kernel as a patch? :)
soreau: I have no idea. All's I know is I have the same problem on my rv350 and that is what fixes it
Ralesk: ah, dig :)
Ralesk: wonders what r100.c has to do with r300 series cards if there's r300.c :D
soreau: Because r300 uses r300, r200 and r100 code
Ralesk: now to see if ccache makes my kernel building less than 3 hours :P
Wizzup: Ralesk: Wow, are you like compiling every kernel feature as module?
HuiJun: hello
HuiJun: has anyone gotten the 4850x2 to work?
HuiJun: @_@
Ralesk: Wizzup: it's our distro's stock kernel :)
Wizzup: ah
Ralesk: it's kinda upsetting that the kernel rebuilds everything even at the slightest poking (even outside our build system...) :/ believe me, it was "fun" to figure out why the bugger didn't want to boot on that one machine.
HuiJun: the xf86-video-ati is supposed to support the 4850x2 with R700 but apparently KDE doesn't think so
HuiJun: @_@
Wizzup: HuiJun: How recent is the driver?
HuiJun: it's from git
HuiJun: so about 2 hours
HuiJun: lol
Wizzup: o :)
Wizzup: Ask one of the devs, I'm not one, so I don't know too much about it
HuiJun: hmm
HuiJun: ok
HuiJun: let me fudge with it some more
HuiJun: @_@
HuiJun: i knew i should have bought a nvidia
Ralesk: HuiJun: it might be a good idea to get a very recent libdrm and compile mesa and xserver again using it
HuiJun: yeah
HuiJun: i got libdrm and mesa from git too
HuiJun: lol
HuiJun: it looks like it's the radeon driver though
HuiJun: looking at my xorg log
edt: drm-next + 2.6.32 + KMS gives an unbootable kernel. It hangs loading the r600 firmware. It does load two firmware files for rs780 then fails loading R600-rlc.bin. With the non KMS kernel just the first two rs780 files are loaded...
edt: this is with drm-next as of a few mins ago
edt: the last two messages are:
edt: [ 1.821686] platform radeon_cp.0: firmware: using built-in firmware radeon/RS780_me.bin
edt: [ 1.830461] platform radeon_cp.0: firmware: requesting radeon/R600_rlc.bin
edt: here is no opp or other message on the serial console. It just stops. dead.
BioTube: edt: it's trying to load the IRQ firmware
edt: and dieing. how do we find out on what? I do have a serial console attached...
BioTube: edt: you can download the firmware here: http://marc.info/?l=dri-devel&m=125960757404659&w=2
edt: were do I put it?
BioTube: firmware/radeon and then tell the kernel to build it in
roy_hobbs: 2.6.32 has fixed my suspend issues with KMS enabled
dandel: roy_hobbs, which card?
roy_hobbs: 01:00.1 Display controller: ATI Technologies Inc RV516 [Radeon X1300 Pro] (Secondary)
roy_hobbs: 02:00.0 Multimedia controller: ATI Technologies Inc Theater 550 PRO PCIe
edt: BioTube it boots once you get the kernel builtin... Wonder why Alex posted .bin files instead of .ihex. The later would have made (manually) adding them to the kernel cleaner.
edt: Thanks for the pointers/help.
edt: no more glx complaints about IRQ either...