1. June 2010
2. May 2010
3. April 2010
4. March 2010
5. February 2010
6. January 2010
7. December 2009
8. November 2009
9. October 2009
10. September 2009
11. August 2009
12. July 2009
13. June 2009
14. May 2009
15. April 2009
16. March 2009
17. February 2009
18. January 2009
19. December 2008
20. November 2008
21. October 2008
22. September 2008
23. August 2008
24. July 2008
25. June 2008
26. May 2008
27. April 2008
28. March 2008
29. February 2008
30. January 2008
31. December 2007
32. November 2007

1. Radeon Wiki
2. xf86-video-ati
3. ATI Mailing List
4. Radeon @ Phoronix

1. RadeonHD Wiki
2. xf86-video-radeonhd
3. RadeonHD Mailing List
4. RadeonHD @ Phoronix

Radeon IRC Logs For 2009-11-15

cxo: drm-next is too outdated, its based on 2.6.29-rc2
Eddi|zuHause: hello, i just installed opensuse 11.2 and i suppose it installed the radeon driver for my 9700 pro, now i tried a 3d game (civ 4 under wine), and the 3D performance is insufficient, and i get graphics glitches (wrong textures). i was wondering what alternatives i have which drivers might be better suited
Eddi|zuHause: under 11.0 i used the fglrx drivers, but the support for 9700 pro was dropped
bridgman: I guess first step is to figure out whether the slow performance is really due to performance differences or if the lower level of gl support in the open drivers is causing the app or wine to use slower code paths
bridgman: level of gl support is being improved fairly quickly in the open drivers, general performance improvements will probably take longer
bridgman: but increasing level of gl support seems to help performance of many apps because they can use faster code
bridgman: if it turns out you really do need general performance improvements, my guess is that you would have to back date enough bits on 11.2 that it would be easier to go back to 11
bridgman: ie if you decide you need to run the fglrx driver
Eddi|zuHause: what i am missing (from googling) is a general overview about what actually happens, so i can figure out what i actually need
bridgman: yeah, I don't have that either; don't use wine much and not familiar with civ4 at all
Eddi|zuHause: for example i was reading the phoronix forum, and someone says " your solution is only to wait wait for kernel 2.6.32! and wait for the galium3D and OpenGL2 Opensource driver!"
Eddi|zuHause: and i can't find enough information about what that means or if this would help me at all
bridgman: might be interesting to search for how the game runs on intel open source drivers, since they have a higher level of gl support but not the performance ;)
bridgman: my guess is that you don't actually need gl2 - gl1.5 and glsl will probably be enough - and I expect that will get into the classic mesa code base so you don't have to wait for gallium3d to mature
Eddi|zuHause: well, it was not exactly fast on fglrx drivers either, but the graphics glitches...
bridgman: then again gallium3d has been moving pretty fast the last couple of weeks
bridgman: in the short term I guess your easy options are (a) work with wine and/or driver devs to understand the graphics glitches and see if they can be fixed (with current performance) or put 11.0 on a partition and use that with fglrx for gaming
Eddi|zuHause: well, i still have my 11.0 installation, so i could just go back, but that should be the very last option if everything else fails...
bridgman: I guess the question is whether you can live with the current open source performance for a bit if the graphics glitches can be fixed
bridgman: if yes, go that route, if no, then everything else has failed, go back to 11.0 ;)
bridgman: need to zzz, bbl
lagnus: KMS tries to load the firmware for my radeon 3200 RS780 but can't find it even though it is under /lib/firmware/rad
lagnus: any idea how to fix that?
cxo: Eddi|zuHause, wine doesnt play nice with the open source drivers
cxo: I've been messing with wine most of today
cxo: Wine does not know the capabilities of the open source radeon cards, so it chooses a very basic set of capabilities and continues
MostAwesomeDude: eosie: Sorry for not responding sooner, had to work.
MostAwesomeDude: eosie: Patch #1 looks pretty alright; I want to see if the emit case can be made a bit smoother, but it should be fine.
MostAwesomeDude: eosie: Patch #2 will need a bit more review; I'm not totally sure that it's doing the right thing, especially with the floating->fixed conversion.
Eddi|zuHause: hm... i just tried to go beyond the main menu, and upon trying to load a real game, it exits ("crash to desktop"), the last message on the console is "r300VertexProgUpdateParams:Params exhausted"
Eddi|zuHause: oh, and the main window seems to have an obiwan, the right and bottom row of pixels are not drawn to... (keeping white or blue lines instead of showing the animation)
cxo: MostAwesomeDude, what do you do for a living?
MostAwesomeDude: cxo: Play music.
MostAwesomeDude: Well, "living" is probably the wrong term; I don't make enough cash to live off of.
cxo: what instrument do you play?
MostAwesomeDude: Piano.
cxo: crazy. so you play at restaurants and stuff?
MostAwesomeDude: Yeah.
cxo: classical or jazz?
MostAwesomeDude: I played at a brewery tonight. Good crows.
MostAwesomeDude: *crowd, even.
MostAwesomeDude: Jazz. I can play other styles, but usually not for cash.
cxo: Not bad
cxo: If days had 50 hours in them, i'd play the piano too :)
cxo: spent many years in formal piano training (trinity college of music, cambridge), but plays the sax and guitar instead
MostAwesomeDude: Good times.
cxo: i really hated playing the piano at the time. playing the same damn scales 2 million times a day, and those 12 page ballads, god!
MostAwesomeDude: I could have gone to conservatory, but I didn't. Dunno why.
cxo: i've been thinking of buying a cheap upright and putting it in the basement to mess around with, but i honestly wonder if i'll ever get anytime,
hifi: how do I know if I'm running gallium driver or not?
hifi: I compiled mesa with radeon gallium, installed it and rebooted X
soreau: hifi: glxinfo|grep nGL
hifi: i see nothing about Gallium
soreau: If you are using latest mesa and it reports 1.5 you are not using gallium
hifi: yeah, 1.5
hifi: ./configure --with-gallium --disable-gallium-intel --enable-gallium-radeon --with-demos= --with-dri-drivers=r300 --prefix=/usr
hifi: something missing?
soreau: hifi: What you probably want to do is install to a non-standard prefix and set LIBGL_DRIVERS_DIR to test the gallium driver
hifi: I can always reinstall my distributions package, right?
soreau: Of course, but that's irrelevant
hifi: is it?
soreau: To me it is
soreau: hifi: This is what I do http://pastebin.com/m4267df1d
soreau: After compiling in this manner, you can use LIBGL_DRIVERS_DIR=/opt/xorg/lib/dri to test r300g
hifi: but thats irrelevant if I just want to screw my system and install to /usr, right?
soreau: You make the call
hifi: it doesn't work even then, so I'm probably doing something wrong
soreau: "doesn't work" is meaningless
hifi: "doesn't show up in glxinfo"
soreau: Since r300g is not ready yet, you probably don't want to use it 'full time'
hifi: it doesn't matter now, I just want to test it out and install to /usr
soreau: It is very convenient to install to a non-standard prefix and test on a per-app basis
hifi: I'll do that later if first I can get someting out of installing it to /usr
soreau: But if you want to do that in your prefix of choice, no one is stopping you
hifi: last time I tried non-standard prefixes I had more problems getting everything to be looked up from the right place than not
hifi: *I* find it easier to make install /usr and aptitude reinstall libgl1-mesa-glx libgl1-mesa-dri after a test run
hifi: that way I can be quite sure I do have everything in place
simplexe1: hd5770 works with ddx from master?
da1l6: simplexe1: It does not even contain pci ids for anything that looks like an evergreen chip. So i'd say the answer is "no, not yet".
yupi: Hi. Id like to setup separate X server for second and third videocard (radeonhd). How can i disable first card from being used?
carldani: airlied: So you suggest to not load radeonfb, and try the suspend-to-ram/resume with KMS instead? Should X be running for the test?
carldani: and while I'm asking stupid questions: How does one call/name the console when KMS is used? Framebuffer console? KMS console? Text-with-KMS console?
dileX: carldani: dont use radeonfb (disable it kernel dot-config). and to answer your question: use/set CONFIG_FRAMEBUFFER_CONSOLE=y
carldani: dileX: I have radeonfb as module, so no recompilation is necessary. Thanks for the info.
carldani: dileX: do you have any speed comparisons between radeonfb and a KMS based console?
dileX: carldani: for radeonfb problem see -> "Troubleshooting KMS problems" (see also topic)
dileX: I dont use radeonfb
carldani: dileX: I need radeonfb to survive suspend-to-ram, and I love the speed I get on the console from using it (I frequently sift through huge log files and vesafb is slooooooooooooow).
dileX: so you use non-KMS often?
carldani: I've been using the radeonfb with X+radeon combination for the last 5 years.
dileX: carldani: then give KMS and fbcon+radeondrmfb a try :-). works fine here.
twnqx: halp
twnqx: my xchat doesn't show uppercase m any more
carldani: dileX: so KMS is only an option if it can reinitialize the graphics in my laptop on resume-from-ram without any video bios dependency (my video bios is trashed before or during suspend)
dileX: carldani: s/r is another topic
carldani: dileX: sorry, seems one of my lines was eaten by my client
carldani: dileX: I want console speed, and I want my laptop not to crash on resume. These are he two criteria.
carldani: dileX: I'll try KMS and look if it survives suspend/resue
glisse: carldani: gpu ?
glisse: kms should survive
glisse: it's one of the thing we take great care to achieve
carldani: ATI Technologies Inc RV350 [Mobility Radeon 9600 M10]
carldani: glisse: my laptop has special resume handling code in radeonfb (Samsung P35)
glisse: carldani: well if kms doesn't work at least we know we can look at radeonfb
carldani: glisse: True. I'll try and report back.
glisse: carldani: easiest way to try is to use fedora 12 livecd
carldani: glisse: Thanks for the tip.
da1l6: Did setup R600 KMS and 3D on my M76 with drm-next and mesa git code yesterday. It works stable with no crashes. :D Only minor glitches with KDE4 Compositing. Really amazing work!
olaf: hello , i'v a lot of tearing playing video with totem using xv:textured video or mplayer , i'm using last gits drm/xf86 ati/and mesa
olaf: is it normal ?
Zajec: uh, great, conn lost
DeX77: hi
DeX77: so.. now I'm on 2.6.32.rc7.g156171c and udev 147 and still: bogus firmware :(
DeX77: if someone here got a working rv770 with kernel 2.6.32 I would like to have its config :)
olaf: i'v only a 730
DeX77: olaf: but it works?
lagnus: anyone getting this issue for KMS, [drm:radeon_driver_load_kms] *ERROR* Fatal error while trying to initialize radeon
olaf: DeX77, no pb with firmware , only tearing when playing video
DeX77: lagnus: bogus firmware length? me too
olaf: DeX77, i'm noob , but have you the right firmware in / lib/firmware ?
lagnus: DeX77, can you explain ?
DeX77: olaf: yepp.. agd5f already checked the sha1sums of them..
olaf: ok
DeX77: lagnus: do you see something like "bogus firmware length 3392" in your dmesg?
lagnus: i the firmware file is installed under /lib/firmware so it has to be finding it
DeX77: olaf: can you perhaps upload your config somewhere? pastebin.ca?
lagnus: DeX77, no let me get you a pastebin of the whole thing
lagnus: DeX77, http://pastebin.com/d26f20315
DeX77: lagnus: erm.. rv780!? this a evergreen chip?
biotube: DeX77: evergreen is r800
DeX77: rs780 is 4890?
DeX77: got a 4850 that is (I think) correctly detected as RV770
DeX77: my error is this: http://pastebin.com/m187df82b
taiu: lagnus: you have radeon built as module or builtin?
olaf: just an idea , but could you have rights pb ? -rw-r--r-- 1 root root 3392 nov. 14 19:40 RV770_pfp.bin
lagnus: taiu, built in , kernel-git
taiu: lagnus: you probably need to build in firmware also then
lagnus: DeX77, my card is an onboard Radeon HD 3200
lagnus: taiu, yeah i add a firmware package that install the pfp and me file under lib/firmware/radeon
lagnus: taiu, so I really dont know why it isnt loading it up without know what the coding is trying to achieve at the exact point before it fails
DeX77: -rw-r--r-- 1 root root 3392 14. Nov 15:44 /lib/firmware/radeon/RV770_pfp.bin
taiu: lagnus: CONFIG_FIRMWARE_IN_KENRL
taiu: KERNEL
lagnus: compile flag?
taiu: in kernel config
olaf: lagnus, cat /usr/src/linux/.config | grep CONFIG_FIRMWARE
DeX77: zgrep CONFIG_FIRMWARE /proc/config.gz ;)
lagnus: CONFIG_FIRMWARE_IN_KERNEL=y
lagnus: CONFIG_FIRMWARE_EDID=y
lagnus: CONFIG_FIRMWARE_MEMMAP=y
lagnus: all enabled it seems
olaf: looks ok , i'v the same and no firmware pb
glisse: DeX77: did you test fedora 12 livecd ?
DeX77: # CONFIG_FIRMWARE_IN_KERNEL is not set
DeX77: I'll try with that enabled..
taiu: lagnus: donno this part, maybe then deselect PREVENT_FIRMWARE_BUILD
lagnus: taiu, CONFIG_PREVENT_FIRMWARE_BUILD=y
lagnus: deselect as in n ?
taiu: yes
lagnus: taiu, how does that affect me? it does seem self explanatory but I am just looking for a bit of techincal insight
lagnus: actually , i will just look it up lol
lagnus: boy is a recompile going to take a while
lagnus: i think i will have to try that later
legume: olaf: Do you have more than one screen connected (even if one is TV and they are off)
olaf: legume, nop , it's a laptop , i'v just hdmi and svga port (not used)
lagnus: spent all day yesterday compiling the kernel
legume: olaf: OK, maybe not that then - though it may be worth doing xrandr --verbose and checking only one output shows as connected.
lordheavy: lagnus got this with a radeon HD 4200 (onboard card)
lordheavy: [drm] Loading RS780 CP Microcode
lordheavy: platform radeon_cp.0: firmware: requesting radeon/RS780_pfp.bin
olaf: compiling is good for my room temperature (i'm in the north of france and it's winter time)
DeX77: using prescott for room heating? ;)
lordheavy: under /lib/firmware/radeon/
lagnus: lordheavy, no a radeon 3200
olaf: legume, nop , others are disconnected , i'v only LVDS connected 1366x768+0+0 (0x54) normal (normal left inverted right x axis y axis) 344mm x 193mm
lagnus: lordheavy, at least that is what my specs tell me and which als repeats in my xorg config file after using proprietary drivers
lagnus: lordheavy, i am just the open ati driver these days
legume: olaf: Hmm not that then - I assume xvinfo looks OK and things are not falling back to x11.
olaf: legume, xvinfo say that i'v have a textured video port
legume: olaf: OK, does XV_VSYNC say 1
olaf: legume, "XV_VSYNC" (current value is 1)
olaf: legume, glxinfo say IRQ's not enabled, falling back to busy waits: 2 0 , but i'm not sure it's related (sorry for my bad english ;) )
lordheavy: lagnus, i use the default kernel-git config without any changes
lagnus: lordheavy, Please correct me if I am wrong and the 3200 is actually a 4200 in degise with a chip detection of RS780/R780
legume: olaf: That doesn't matter and is normal. If the tearing you see is no vsync type tearing then I don't know why it isn't working.
lordheavy: 4200 is an "updated" 3200
lagnus: lordheavy, ok. Well i didnt change anything related to the video card. just opted for my particular cpu and removed some drivers i didnt need
lagnus: lordheavy, is that only software based?
olaf: legume, i'v sometime two "tearing line" in the same frame , not sure it's vsync related
lagnus: lordheavy, or is a slightly different gpu?
lordheavy: gpu is slightly different (4200 got uvd2 for example)
legume: olaf: Yea you can see two lines when vsunc is off. Maybe it's something specific to your chip.
lordheavy: how did you boot to the git kernel ?
taiu: lagnus: dunno exactly how this builtin works, maybe it tries to include it from /lib/firmware
taiu: lagnus: so you have to install it once at least: make install_firmware
olaf: legume, ok , thanks for your help , i'll wait and hope :)
lagnus: taiu, i have the firmware in lib/firmware already
lagnus: under radeon dir
lagnus: -rw-r--r-- 1 root root 21504 2009-11-14 23:31 RS780_me.bin
lagnus: -rw-r--r-- 1 root root 2304 2009-11-14 23:31 RS780_pfp.bin
lordheavy: lagnus, how did you boot the kernel ?
lagnus: lordheavy,how do you mean? Live cd, hard disk
taiu: lagnus: then I'm at loss, maybe it fails because of something else also, SYSFS, FW_LOADER ...
lagnus: as far i know i just use vmlinuz and kernel26.img from boot after using mkinitcpio
lordheavy: is it with your regular archlinux installation ? what is your grub entry ?
lagnus: for*
lagnus: mkinitpcio to build the new kernel image*
lagnus: lordheavy, one sev
lagnus: sec*
lordheavy: lagnus : revelent parts from my menu.lst http://pastebin.fr/6006
lordheavy: you should remove nomodeset to get kms working :-p
lagnus: kernel /boot/vmlinuz26-git root="rootdrive" ro
lagnus: initrd /boot/kernel26-git.img
lagnus: those are the lines i use to boot
lagnus: lordheavy, but i am using lzma compression for the img instead of the standard gzip
lagnus: lordheavy, though I cant see how that may be a problem
lordheavy: It would be good to know why the firmware fail to load ......
olaf: have a good day , bye
adamk: Alright, so I'm using drm-next and video=DVI-0:1280x1024 doesn't seem to change the resolution on boot. I also tried video=DVI-1:1280x1024 as I understand that the Xorg output name may not be the same as the KMS one.
adamk: soreau, ping? Do I have the syntax correct?
fredl: hi folks. I'm using the radeon driver on with a HD4350 card on Debian Sid.
fredl: I'm having a bit of a problem with it that I don't know how to fix. There's a Sony AV between the HDMI output of my system and my TV
fredl: EDID vendor "SNY", prod id 63232
fredl: the TV itself can do Full HD but it would seem that it reports 1280x720 as the maximum resolution to the system.
fredl: I have not yet tried to hook up the TV directly to the system.
fredl: but is there a way that I can tell the system that there's really a 1920x1080 screen attached?
fredl: I'm going to walk downstairs and hook up the TV directly to see if that makes a difference.
Xeno: @fredl: Check your TV specs, my sis got full hd one but it offers something like 1300x768 as PC display
twnqx: fredl: is there a non.working 1920x540, by any means?
fredl: ok back again.
fredl: the TV itself reports 1280x720 as well and a lot more than the SNY does
fredl: so the SNY probably just reports the highest res.
fredl: I have never had the TV itself on Full HD over the HDMI input with the computer, this is a new setup for me.
fredl: I have however forced the playstation 3 which goes through another port of the SNY into 1920x1080
fredl: so I know it's possible through HDMI, just not reported, so perhaps it drives the circuitry to the max.
fredl: on the DVI port of the TV it does support 1920x1080
fredl: I guess... perhaps I could hook up DVI, copy the Modeline and put that in xorg.conf and force it to use that Modeline?
fredl: twnqx, no there's no such mode, neither for the direct connection as the connection through the AV
twnqx: how do you connect?
twnqx: you'll need HDMI for HD
fredl: the SNY has three HDMI in ports, one HDMI out. the HDMI out is connected to the TV's HDMI in.
twnqx: i c
fredl: it almost seems as if the HDMI input to the TV is limited compared to the DVI port.
fredl: it's possible correct, to override the Modeline's that are probed?
twnqx: yeah
twnqx: if you ask if the assumption is correct
twnqx: if the result will be correct... there's only one way to find out!
fredl: well read up, I've used the PS3 on the 1920x1080 resolution before through HDMI
fredl: so I know it *can* do it, it's just not reported.
PsyMan`: just add the resolution manually
fredl: I will try that yes. Is that the only thing I need to do, just add the Modeline to the Screen section and DPMS won't override that again?
PsyMan`: not quite sure
fredl: lemme walk downstairs and hook up the DVI cable to see if I can figure out the modeline that way before I have to go Google for it :-)
PsyMan`: don't use linux here :p
fredl: to be continued :)
adamk: Alright, so just using video=1280x1024 on the kernel command line changes it on both monitors.
adamk: However, 1280x1024 still flickers, so I guess it's running at 85 Hz, which KMS seems to have problems with on this morning.
adamk: And then when I start X, it goes to 1600x1200 if I don't specify a PreferredMode, which seems odd. If I'm specifying a mode to the kernel module, and xf86-video-ati is built with KMS support, then one would think that xf86-video-ati would select 1280x1024, too.
adamk: So time to try video=1280x1024@75 on the kernel command line to see if that works, because 85 is unusable.
adamk: Hey, it looks like it worked!
fredl: is there a way to indicate xorg should use that Monitor on the HDMI-0 output?
fredl: coz right now in the log I see: Output VGA-0 using monitor section Tatung V42
fredl: while it also says: Output HDMI-0 has no monitor section
adamk: Finally, KMS is usable on both my at home. Unfortunately, of course my LCD is now running at 1280x1024@75 in the console and is selecting 1280x1024@60 for X.
adamk: So I still need to use a PreferredMode's for both monitors, but that's fine.
chithead: heh, there goes flicker free vtswitch
adamk: Right, but if I get the PreferredMode right, it should work fine.
adamk: I just need to figure out the modeline that is being used in the console.
adamk: Well, this one clearly isn't it.
adamk: Heh.
lagnus: lordheavy, indeed it would be
adamk: Shoot.. Is there someway to find out what modeline the console framebuffer is using under KMS?
Zambezi: With fglrx I have full resolution, but a very unstable system. With radeon(hd) I have a super stable system, but just a tiny window in fullscreen: http://tinypic.com/m/6dssbb/1 Any idea how to fix?
chithead: adamk: maybe fbset
adamk: No, it just shows the resolution. The refresh rate even comes up as 0.
soreau: Zambezi: I'm not sure what I'm looking at there. and that is a horrible image hosting site btw
gimzo1: Zambezi: looks like xv is not working, what does mplayer say ?
Zambezi: soreau: That a video in full screen.
adamk: Yeah, I see some strange guy with a mask and a snuggle or snuggie, or whatever that is.
soreau: adamk: Last I asked airlied if there was a way to set fb res in real time, he said 'not yet' (I assume you'd be able to do something similar to xrandr and view the current res too)
soreau: Zambezi: How are you starting the video? what player? what command?
adamk: Zambezi, Oh , you didn't mentioned that you were talking about video... Kind of an important piece of information.
Zambezi: adamk: It's not me btw. :-P
Zambezi: soreau: Just mplayer filename
Zambezi: gimzo1: A moment.
soreau: Zambezi: You need mplayer -zoom /path/to/file.ext
adamk: Zambezi, mplayer is probably using the x11 output device if it's not scaling. Xv should scale automatically.
soreau: adamk: I don't think it does
adamk: If mplayer isn't detecting an xv device, the you probably don't have direct rendering.
adamk: It always has here.
soreau: tests
gimzo1: mplayer always defaults to xv, and zoom works, so I guess there's something wrong with xv here
adamk: Zambezi, Does 'xvinfo' show any Xv ports?
soreau: gimzo1: Is this your setup?
Zambezi: Output from console: http://www.pastebin.ca/1672225
adamk: Alright, this is ridiculous. Both monitors are using the same PreferredMode, and I'm passing video=1280x1024 on the kernel command line. Only the CRT switches to the console flicker free, though.
adamk: Zambezi, Right, no Xv.
adamk: Zambezi, So you don't have direct rendering setup and working.
Zambezi: soreau: Zoom worked, but god it became slowmotion.
adamk: Zambezi, Right, you want to use Xv.
soreau: Zambezi: Try adding -vo xv
adamk: soreau, Did you see the mplayer output? It's not seeing any Xv ports.
soreau: adamk: Ok.
soreau: adamk: Yea, it does auto detect here. But in the past (possibly with dri1) I've had to use -zoom otherwise it would look like this in the middle of the screen
Zambezi: soreau: That said: no video.
adamk: -zoom should only have ever been necessary with the x11 output.
adamk: Zambezi, Pastebin your /var/log/Xorg.0.log file.
adamk: And now back to my problem... If the console on both monitors is using the same resolution, and both monitors are using the same PreferredMode in X, how is it that only one monitor is flicker free when switching between X and console? Argh.
Zambezi: adamk: I think there's a way to upload the file to pastebin.
adamk: Zambezi, Sure, just open it in an editor, select everything, copy ansd paste.
Zambezi: adamk: That only works the same size as the window I'm in but I switch desktop.
adamk: Not sure I understand. If you open it in gedit, for example, you can select the entire file.
soreau: Zambezi: If you have curl installed, you can use: cat /var/log/Xorg.0.log | curl -F 'sprunge=<-' http://sprunge.us
soreau: which should give you back a link containing the file in it's entirety
soreau: adamk: Are those both the same monitors btw?
adamk: soreau, No, one's an LCD and one's a CRT.
adamk: I'm still trying to find the right combination of video= arguments and PreferredMode to get this working.
soreau: Well that might be why. When I have different monitors, they behave differently, one showing output while the other does not for example during a second when the kernel first boots or switching to tty, one will show tty first while the other takes a bit longer but not showing what happens during the transition
Zambezi: Pastebin with Xorg.0.log and xorg.conf. Running Debian Lenny, Ion3. No Gnome, KDE etc: http://pastebin.ca/1672231
soreau: adamk: Would it happen to be the crt that does the flickering?
adamk: soreau, No, the CRT is fine when switching between X and the cosnole.
Zambezi: soreau: Of course. Now I remember. The packages name is pastebinit.
soreau: Zambezi: Without even looking at that log I will take a stab and say because it's lenny, the packages it's using are too old
adamk: With radeonhd you have to explicitly enable DRI with:
adamk: Option "DRI" "on"
soreau: ewww
adamk: But, yes, it is also quite possible that your kernel and DDX are too old to support Xv.
gimzo1: (EE) RADEONHD(0): The fglrx kernel module is loaded.
soreau: lol
adamk: Oh, that's bad too.
adamk: :-)
adamk: So remove fglrx, enable DRI with that option, reboot, and try again.
adamk: If it fails, either stick with fglrx or upgrade :-)
Zambezi: soreau: Now you made my smile. I know. Sometime they are just too old. Latest Fx (Iceweasel) in Lenny is 3.0.6. I run 3.5.3 so I'm fine.
Zambezi: adamk: 2.6.26-2-686
adamk: Right....
adamk: I don't know when DRI support was added to the kernel.
soreau: Zambezi :)
soreau: adamk: Shoulda been 'mmmright...' hahaha
Zambezi: I had like to hard freeze. Then I just switch to radeonhd and got happy again. I can always try Squeeze. That's a bit better. Then I might even use my two monitors at the same time. ;-)
Zambezi: soreau: You were spot on. I like Debian cause it stable, but some packages are a bit too old. But it's at least better than Ubuntu.
soreau: Zambezi: How is it better than ubuntu? They at least try to keep up with the latest package versions
adamk: soreau, Do you specify different resolutions for both monitors on the kernel command line?
Zambezi: soreau: I don't like sudo, don't like Gnome. Maybe alternate then, remove sudo.
adamk: Or just don't use it.
Zambezi: adamk: Also an option. I have an extra harddrive connected. I can always try it.
soreau: adamk: No. In fact with S-video plugged in and it being hard coded to 800x600, upon boot the output (with no video line) is 800x600 on tv-out, and VGA-0 at 1280x1024 but all of the output in the upper left 800x600 corner. So, I set VGA-1:800x600 so it fills the screen and this happens to look very nice.
adamk: Gotcha.
adamk: I'm curious what the correct syntax is for specifying different modes for each monitor... I'm guessing something like 'video=DVI-0:1280x1024@75,DVI-1:1280x1024@60'
soreau: Zambezi: That is no reason not to use ubuntu. You can run a root terminal, you can run a standalone session.. there's nothing you can do with debian that you can't (or couldn't a couple years ago) with ubuntu
adamk: But when I try that, DVI-0 is stays at 1600x1200
soreau: adamk: Start out with making sure you get the correct. With xrandr, it reports VGA-0 while the kernel param must be VGA-1
soreau: adamk: and make sure you are using drm-next ><
soreau: I don't know if this works in any other kernel yet
adamk: Oh, I'm using drm-next.
soreau: ok cool
Zambezi: soreau: I'm not going to install a DE, but I think it's possible to skip with alternate. At least with server.
adamk: But with that video= option above, I'd expect the CRT to be 1280x1024 @ some frequency, instead of 1600x1200 :-)
soreau: adamk: So does stderr get spammed with output when running compiz there?
adamk: Oh, right, I forgot about that. Give me a few minutes and I'll test.
soreau: adamk: Well obviously if the

is wrong, it's not going to affect it at all
adamk: Right, but it's hooked up to a DVI port. I'd expect the output to be either DVI-0 or DVI-1 since I have two monitors, each hooked up to a separate DVI port :-)
adamk: Do you know some way of finding out what KMS says the name is?
soreau: nope. I just guessed and got lucky, but I was reading something agd5f gave me telling how to use this new feature and I should have saved the link
soreau: Zambezi: With the packages you're using, you're probably stuck back in the ages when fglrx was the thing to do :P
Zambezi: soreau: I have rtorrent 0.7.9. A dev called it ancient. I'm living in a world with no news.
adamk: soreau, No, I do not get those same errors you had, just the GLX 1.3 is not supported message.
adamk: soreau, But I haven't updated Mesa in a bit. I'll try that in a few.
adamk: But first I want to get both monitors flicker free when switching to the console :-)
soreau: adamk: Somehow I think that's wishful thinking ;)
adamk: Bah... I will win :-)
Zambezi: soreau: Proper drivers and dualmonitors I need randr for example. 1.2.2 in Lenny, 1.3 in Squeeze. I think that's a big difference, but xserver-xorg-video-radeonhd is 1.2.1 compare to 1.2.5. And that's the version used by latest Ubuntu. Might be a whole new experience.
adamk: soreau, I think you're right.. It is wishful thinking. Even if I pass video=1280x1024@60 on the kernel command line, and setup a modeline for both monitors to use, it works fine on the LCD now, but not the CRT.
adamk: So I'm pretty sure I need separate video= options for each monitor, but I just don't the correct syntax for doing separate modes for each monitor.
adamk: reboot
adamk: D'oh
adamk: Well, separate video= options doesn't do it either.
soreau: adamk: Is 1280x1024 listed as a mode by default (no x conf and not adding it with xrandr)?
adamk: soreau, That may be part of the problem. Both monitors list 1280x1024, but the CRT only lists is @ 75 and 85. The LCD lists it only @ 60
soreau: adamk: If you just try video=DVI-1:1280x1024, does have any effect?
soreau: man, I really wish I had that link
soreau: It tells you how to setup even two monitors and if my memory serves me correctly, it's all with one video='s line
adamk: Argh.
adamk: Well video=DVI-0:1280x1024 doesn't change anything on the CRT. video=DVI-1:1280x1024 resulted in a locked box... Let me try that again.
adamk: Maybe I just mistyped.
adamk: Yeah, I must have mistyped. I just tried again, and it boots up, but the CRT is still at 1600x1200.
adamk: The only thing that seems to impact the console resolution of the CRT is video=1280x1024@75 or video=1280x1024@60.
adamk: Which would be fine except that when I then use the same PreferredMode for both monitors, only one of them is flicker free.
adamk: Which, I guess, is not that big of a deal at the moment.
adamk: I might just revisit this when airlied or agd5f are around during the week.
adamk: If you give me a few, I'll update libdrm and mesa and try compiz.
adamk: Heh... Now the vmware kernel modules won't compile against linus' tree.
adamk: soreau, I do not get those same warnings/errors you received.
soreau: adamk: While you have the latest code there, can you test mags fisheye mode please?
soreau: I'm pretty sure it's an r3xx specific thing
adamk: It does something... But what it does just ain't right.
soreau: Cool
soreau: That's what it does here
adamk: Have you tried adjusting the radius?
adamk: Let me see what happens here.
soreau: I tried bisecting mesa yesterday since I found a good commit but I still haven't narrowed down the faulty commit 100% (even though the bisect seemed decisive, reverting one commit from the first bad commit it found didn't fix it)
soreau: adamk: I have a good mesa commit (for the fisheye bug) if you are interested
adamk: Yeah, that didn't make a difference.
spreeuw: http://freegamer.blogspot.com/
soreau: I am trying to figure out exactly how to make the bug report though since it broke in two different ways
adamk: soreau, I can't remember the last time I used magnifier, actually.
adamk: Heh.
soreau: adamk: Thanks for testing.
adamk: No problem. If you'd like me to test something else, let me know.
soreau: adamk: Well if you have time, you could test what I've found to be a good mesa commit here
adamk: I wonder how it works on r600.
adamk: Sure.
soreau: This is the latest good commit I have found so far 84c7afd9e0f2df72d90dd82d38384c4f2f45173e
soreau: something happened when they were doing some 7.5/7.6 merging stuff afaict
soreau: But the first bad commits I found, one was for 'i965: Add support for KIL_NV in brw_wm_emit.c' by anholt and the other was 'mesa: use new _mesa_meta_CopyTex[Sub]Image() functions' by brianp ;)
soreau: The first one just causes a 'ring' around fisheye so to speak and the second breaks it very badly in the way it's current state is now
adamk: That's exactly how I would report it. :-)
soreau: Well the thing that I question is checking out the first bad commit and reverting one commit (in the case of the former) with git checkout HEAD~1 does not fix it
soreau: I discovered this when I went to take screenshots for the bug report
soreau: The latter I have not went into depth yet
adamk: I can confirm that 84c7afd9e0f2df72d90dd82d38384c4f2f45173e works here.
soreau: adamk: Cool. Which chipset is that again?
adamk: x850, an R480.
soreau: adamk: btw, what did you end up having to do to get drm-next pulled?
soreau: linus-tree worked?
adamk: Yep.
soreau: Well I would update the wiki but I feel like I need permission first ;)
DeX77: so I have firmware built in now: still bogus length
DeX77: I'm out of ideas.. I guess I have to downgrade to 1.6x XServer
ossman: is there any way to disable kms so that you can reload the module?
ossman: having to reboot is a bit annoying
ossman: I'm guessing fbcon is what needs to be killed
DeX77: I can't even relad it when its not loaded correctly
taiu: DeX77: so how does your RV770_me.bin look like on disk
TCW: compiz 0.8.2, xserver-xorg-video-radeon 6.12.3, kernel 2.6.31, rv530chip (x1600 card), does one come random (more or less) x-session crashes in mind? Especially when I rotate the cube with the mouse (compiz feature) it will crash after a few seconds and drop me back to gdm login
dutchfish: can someone explain me if kernel 2.6.32 when it goes stable will have full support for RV610 and what are my changes that /usr/lib/dri/r600_dri.so will be ready? i am a bit confised because i thought support will be complete for this card in 2.6.32. Thank you for a replay.
spreeuw: nice, the old vegastrike engine runs ok
TCW: I have this problem since a few months and every now and then I ask about it if others had a similar issue :)
spreeuw: gave privateer gemini gold a run
spreeuw: dutchfish: for 3d you mean?
dutchfish: spreeuw, 3d/opengl
spreeuw: it will have support for r600 yes
spreeuw: but its not done in the kernel really
spreeuw: but the has supporting code in 32 yes
dutchfish: spreeuw, i am running now 2.6.32-rc7 with latest mesa from git, but either i did something wrong or i dont understand. I have excellent 2D and video support tho.
spreeuw: mesa does the real work
spreeuw: maybe you need an up to date libdrm too
dutchfish: and gallium is not ready. That i understand.
taiu: TCW: 3d is done by mesa, might try to update it
TCW: taiu, oh, I forgot to mention mesa stuff has version 7.6 here
dutchfish: spreeuw, so if i do it right i have hardware opengl 1.5/3.0 atm?
spreeuw: 1.5 ish yes
spreeuw: its not feature ocmplete yet
spreeuw: not even for 1.5
spreeuw: it lacks s3tc for instance
dutchfish: spreeuw, and we not talking over softpipes right?
spreeuw: no idea
dutchfish: ;)
spreeuw: I'm just a user
dutchfish: ah ok, np
spreeuw: I'm running it at the moment on rv620
TCW: spreeuw, btw. how does privateer gemini gold feel? Worth playing? Enjoyable? At least in good shape / complete?
spreeuw: quake and stuff is accelerated
dutchfish: spreeuw, glxgears?
spreeuw: TCW: think it uses the old privateer data
spreeuw: should be polished
spreeuw: dutchfish: of ocurse
TCW: spreeuw, oops? was that privateer code / data released under a free / open license?
dutchfish: spreeuw, atm i come to 600 FPS rv610 (HD2400XT)
spreeuw: TCW: I dunno maybe they redrew the pixelart
dutchfish: spreeuw, but i uses softpipes
spreeuw: glxgears is no benchmark
spreeuw: and theres no good alternative
spreeuw: for comparison atm
spreeuw: lets just say quake3 engine games are playable on 1280x960
dutchfish: spreeuw, http://pastebin.com/f2d144183
dutchfish: dont mind the quirck in EDID detectection (known bug)
spreeuw: dutchfish: check glxinfo
spreeuw: its a command
spreeuw: direct rendering: Yes
spreeuw: OpenGL vendor string: Advanced Micro Devices, Inc.
spreeuw: OpenGL renderer string: Mesa DRI R600 (RV620 95C5) 20090101 x86/MMX+/3DNow!+/SSE2 TCL
spreeuw: OpenGL version string: 1.5 Mesa 7.7-devel
spreeuw: this is present mesa git
dutchfish: spreeuw, http://pastebin.com/m17aeb555
TCW: bbl
taiu: dutchfish: you don't seem to have r600_dri.so
dutchfish: ;)
spreeuw: dutchfish: yep you use soft rendering
dutchfish: yep
spreeuw: it should be using r600_dri.so
spreeuw: see if the file is instlaled
dutchfish: spreeuw, nope
taiu: dutchfish: mesa from git should say mesa 7.7-dev
taiu: dutchfish: you installed it somewhere else
spreeuw: then something went wrong when buiklding mesa
dutchfish: taiu, wich git repo you used?
dutchfish: taiu, expirimental?
spreeuw: indeed 7.6 is too old
spreeuw: maybe you have 2 copies instlaled or so
taiu: there's only one git://anongit.freedesktop.org/mesa/mesa
dutchfish: spreeuw, actually 4, this was my latest build because >7 crashed my xorg
dutchfish: taiu, thank you
kilna: Greetings all. I am attempting to get an ATI video card working in osx86. Am I in the correct place?
taiu: when you run configure specify at least --prefix=/usr and with-dri-drivers=swrast,r600
DeX77: taiu: the sha1sums are the same as agd5f, blisse and others.. so I think they are fine (on disk)
dutchfish: taiu, ok, thank you, and you running KMS?
dutchfish: taiu, or userspace?
taiu: DeX77: well, show me
taiu: dutchfish: recently kms
spreeuw: I'm using dri1
dutchfish: taiu, with curent version?
dutchfish: sorry for all the q ;)
taiu: pretty much, havent updated mesa in couple of days though
dutchfish: taiu, ok, fair enough, thank you so much !
dutchfish: spreeuw, thank you too !
DeX77: taiu: 097a76ed71a2e53f240ad83d1e0d9709f7f6d17c /lib/firmware/radeon/RV770_me.bin
DeX77: c8ef7bf917d3af091da14b80670762f79ee5e3e4 /lib/firmware/radeon/RV770_pfp.bin
taiu: DeX77: do you have initrd of some kind also?
DeX77: nope
taiu: intresting... and what's ls -l firmware/radeon/RV77* in your kernel tree?
DeX77: taiu: ls -l firmware/radeon/RV770_*
DeX77: -rw-r--r-- 1 dex users 5440 15. Nov 16:01 firmware/radeon/RV770_me.bin
DeX77: -rw-r--r-- 1 dex users 14972 15. Nov 14:56 firmware/radeon/RV770_me.bin.ihex
DeX77: -rw-r--r-- 1 dex users 3392 15. Nov 16:01 firmware/radeon/RV770_pfp.bin
DeX77: -rw-r--r-- 1 dex users 9340 15. Nov 14:56 firmware/radeon/RV770_pfp.bin.ihex
DeX77: taiu1: you got the last post?
taiu1: mhmh, intresting that it get's the first file ok, then gets the second the same as first
DeX77: taiu1: it think thats a race
cxo: airlied, hey
cxo: Are all the changes you're doing in drm-next under drivers/gpu ? I'd like to patch those against current mainline linux
soreau: cxo: If you use the instructions for drm-next in the wiki, you can clone linus-tree instead of what it has there..
cxo: Is linus'-tree stable? there have been some nasty regressions in ext4 lately
dutchfish: hi, i just compiled over form the latest git but there is no tag of a 7.7 branch, i only see a merge of 7.6 and a few commits after that. nevertheless i recompiled with the same result (only software pipes for mesa) the i saw this: http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?h=6.12-branch My q: Is RV610 really supported?
MostAwesomeDude: Yup.
cxo: soreau, where is the drm-next wiki?
soreau: cxo: Checkout the /topic here
spreeuw: dutchfish: just the current mesa
soreau: ;)
spreeuw: git
dutchfish: ok, i delve into it further
spreeuw: abefd7dcdf28c90454b59faaf9401fa6e6c6f526
dutchfish: thanks guys
spreeuw: top git log result
dutchfish: spreeuw, i have another tag ed9c4933a francisco jerez
dutchfish: as master
spreeuw: thats the next one
dutchfish: oki
spreeuw: should be fine for r600
spreeuw: you should only build swrast,r600
dutchfish: yep, i did
spreeuw: the rest only increases the chance of buildbugs
spreeuw: k
dutchfish: it build fine, but does not work
spreeuw: something seems in the way then
spreeuw: if it considers r600 at all it should be in the xorg log
cxo: doesnt like that method of pulling changes
spreeuw: did you upgrade your libdrm yet?
dutchfish: spreeuw, wich version?
spreeuw: not sure which one is needed now
dutchfish: spreeuw, and do you mean libdrm2?
dutchfish: i have used 2.4.14-1
spreeuw: not sure if linux 6.32rc7 is good enough either
dutchfish: oh?
hnsr: what kernel should I be running if i want to test the latest and greatest 3d stuff? i was using drm-next so far but I see drm-radeon-testing has more recent commits in it
spreeuw: are those modules loaded?
spreeuw: drm.ko and radeon.ko
dutchfish: yep
cxo: What are you getting with drm-next?
cxo: (other than all those things listed on airlied page)
cxo: What end-user functional benefits
dutchfish: me?
cxo: Is there some game or app that didnt work before, going to start working?
dutchfish: i posted that a bit earlier
cxo: (no, open question)
cxo: i just logged in 20mins ago
DeX77: taiu1: any ideas what could be wrong? I seem to be the only one with that problem :(
dutchfish: spreeuw, thanks again, i delve into it further, you sure that this is not getting in my way? http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?h=6.12-branch
spreeuw: dutchfish: thta should be updated too
spreeuw: to something recent
cxo: i suppose i'll just wait for 2.6.32rc8, too lazy to stitch the changes against 2.6.32rc7
soreau: cxo: drm-next is where all the latest code goes which will most likely end up in linus tree eventually. You will likely see fixes for kernel related stuff there first, such as s/r fixes and fb mode caps to name a couple
dutchfish: and last but not least: kudos for all the hard work the developers did so far !
dutchfish: spreeuw, ok, i will thanks
cxo: soreau, yeah i got that, i just wanted to know what was currently on drm-next that 2.6.32rc7 didnt have in terms of visible end user benefits
soreau: cxo: no idea.
cxo: waits for 2.6.32rc8 :)
soreau: you'd have to ask airlied
cxo: Like all that stuff on here sounds great: http://airlied.livejournal.com/68097.html
cxo: But what does it mean? Is Warsow suddenly going to start working? Is wine going to start working? etc..
dutchfish: cxo, when your distro catch up yes
Ronis_BR: can anyone help me to set 3D in a R7** card with xf86-video-ati driver?
cxo: dutchfish, What do you mean?
dutchfish: cxo, mesa/X/kernel/dri etc
cxo: Ronis_BR, just build kernel,mesa,drm,xf86-video-ati from source and it should just all work
Ronis_BR: cxo: I'm using gentoo, I'm wanting something more specific :)
cxo: dutchfish, I don't understand what you said "when your distro catch up yes" in relation to?
soreau: Ronis_BR: See the wiki link in the topic. You will want drm-next, libdrm, mesa and xf86-video-ati -9999 builds
dutchfish: cxo, newer version of there software
cxo: wtf?
cxo: is totally lost right now
Ronis_BR: soreau: somebody said that I just need mesa -9999
soreau: Ronis_BR: They are wrong
cxo: mesa -9999?
soreau: cxo: It's a gentoo thing, don't worry about it ;)
cxo: oh ok
soreau: It's basically gentoo speak for the development version of the package
cxo: lame
soreau: Ok, next time I wont bother telling you
cxo: I just meant its a lame designation, not that you were lame
Ronis_BR: soreau: does X11 overlay have all this things?
soreau: If you used gentoo, it would make sense to you
Ronis_BR: cxo: it is a fantastic one acctually
soreau: Ronis_BR: Yes, it has everything you need except the kernel
Ronis_BR: soreau: ok, the kernel I will compile now
Ronis_BR: soreau: I need to get away from fglrx
cxo: I want to get away!
cxo: hums lenny kravitz
soreau: now that - is lame
cxo: Noooo
Ronis_BR: soreau: If I install it with mesa 7.5 and kernel 2.6.31 I will get no 3D accel?
Ronis_BR: I don't like -9999 versions
Ronis_BR: and RC kernels :D
twnqx: get 2.6.31.y + DRM-NEXT
twnqx: and don't mind the -9999, they work well.
soreau: Ronis_BR: You're card is too new
soreau: And twnqx is right
soreau: they work surprisingly very well
Ronis_BR: soreau: ok
Ronis_BR: so :)
Ronis_BR: soreau: Will I need to purge fgrlx?
soreau: Yes.
MostAwesomeDude: "Purge" is such a weak word.
Ronis_BR: purge for me means remove fglrx and its configurations
Ronis_BR: maybe it comes from debian time :D
soreau: After burning fglrx with fire and building libdrm, mesa and ddx from git, I would recommend also rebuilding xorg-server as well
Ronis_BR: soreau: I will, because I'll change VIDEO_CARDS
soreau: Yes, that too
Ronis_BR: soreau: is it possible to patch gentoo-sources with drm-next?
soreau: Ronis_BR: I would not even recommend trying that. Instead, I would suggest you pull drm-next on top on linus-tree
soreau: You should use git to clone the kernel and don't bother with portage in this case
Ronis_BR: soreau: so, I'll try 2.6.32_rc6 :)
soreau: vanilla-sources?
Ronis_BR: yes
soreau: You can if you want
Ronis_BR: and add tuxonice patch
Ronis_BR: I just wanting to have tux-on-ice
soreau: It should work I guess, but no guarantees. I'd just do it right the first time so there is no guessing
Ronis_BR: :) ok
cxo: MostAwesomeDude, disinfect
cxo: soreau, why build xorg-server again?
Ronis_BR: soreau: drm-next is already in 2.6.32 right, I just need to compile ATI DRM as module right?
soreau: cxo: Several reasons, but mainly to ensure you have the right libGL.so.1.2 which fglrx replaces
cxo: doesnt libGl come from mesa rather than xorg-server?
jcristau: cxo: it does
soreau: Ronis_BR: drm-next is on going. The latest code lands there; it's where development happens. So you can't say 'drm-next is in foo'
spreeuw: jep
cxo: so why rebuild xorg-server?
soreau: Ronis_BR: Periodically, linus pulls drm-next into his tree when it is ready
Ronis_BR: soreau: so, the drm I need is already on kernel 2.6.32-rc right?
cxo: a good portion of it is
cxo: but if you want the latest, you need to pull drm-2.6
soreau: Ronis_BR: Like I said, it should be but I cannot say for certain
Ronis_BR: ok
Ronis_BR: I'll backup my machine and try it
Ronis_BR: thanks!
cxo: I just checked Ronis_BR, 2.6.32rc7 is not as new as the current drm-next
Ronis_BR: cxo: but I need something that works :D
cxo: there is about 1000 lines of diff in the drm code
cxo: yeah 2.6.32rc7 is sufficient
Ronis_BR: ok
soreau: cxo: You need to rebuild xorg-server so the X drivers get built correctly
cxo: Ronis_BR, i'm using 2.6.32rc7 right now, with a hd4870
Ronis_BR: cxo: is 3d working 100%?
soreau: Ronis_BR: 100% is somewhat a subjective proposition. Short answer is no.
Ronis_BR: soreau: I mean, you can run all desktop effects?
Ronis_BR: (I just use 3D for desktop effects :D)
soreau: Ronis_BR: Compiz works
cxo: Ronis_BR, yes compiz works, but anything more challenging than quake3 doesnt
Ronis_BR: ok
Ronis_BR: thanks!
cxo: there is a merge conflict in drm-next and current 2.6.31.y
cxo: luckily its in the intel code :)
soreau: yea, that's why I say use linus-tree instead of 2.6.31.y
soreau: The wiki should be updated soon.. enough
cxo: so use linus' tree and then pull drm-next changes into it?
twnqx: rrr
twnqx: how again do i solve conflicts?
twnqx: so i fix the conflict, then... what?
Ghworg: git add the file, then git commit
twnqx: mmhmh
twnqx: so apparently i alreday did a git pull origin earlier
soreau: cxo: right
cxo: whats git url for linus' tree?
cxo: ah got it git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
Honoome: hi, does anybody kno why would I get "No packet3 for relocation for packet at 45" when starting compiz?
soreau: Honoome: What is your system situation? card, driver, kernel, mesa version, kms or non etc
Honoome: soreau: kms, r600-based, kernel 32r7, mesa from git (slightly outdated 'cause mesa fails to build with parallel make), drivers and libdrm from git master
Honoome: rebuilding (serially) mesa now
Honoome: worked fine with 32rc6
Honoome: [32r7 should have read 32rc7, losing characters over ssh is not fun]
soreau: Honoome: Do you have dri working? how are you starting compiz?
Honoome: soreau: dri is reported working, compiz is started through fusion-icon like usual
soreau: Honoome: Which distro?
Honoome: gentoo
soreau: Can you pastebin the output of 'compiz-manager'?
Honoome: difficult right now, given my startx dies :)
cxo: get a pastebin script
soreau: You can't run compiz without X running..
Honoome: soreau: as soon as compiz is started up (fusion-icon is part of the session), X dies, telling me to check dmesg, and that's the error in dmesg
Honoome: now I'm trying to see if it might be the outdated mesa at fault but rebuilding it serially is *definitely* slow
cxo: why can't X always be running? And the conventional X is just a client that connects to it
soreau: cxo: Read up on wayland, a new idea for an xserver
Honoome: didn't say "impossible", I said "difficult" ;) trying the mesa rebuild first since that's the only piece of outdated code I have around, then I'll definitely get a log with a manual X session
Honoome: that actually seemed to be it, mesa rebuilt, and it starts
cxo: when you build drm make sure you have that --experimental param
cxo: has forgotten that on more than one occasion
Honoome: cxo: luckily, the ebuild takes fare of that for me :)
Honoome: [and as I said, it worked fine before, seems like after rc7 it needs a newer mesa than I had... fun]
soreau: points Ronis_BR to Honoome's issue
soreau: Ronis_BR: You want -9999 packages for libdrm, mesa and xf86-video-ati, you need them
cxo: Everything works for me...thats why i'm trying to fix it some more now by building drm-next :)
soreau: If it aint broke, fix it anyway
cxo: exactly, nothing can be too fixed anyways
soreau: heh
cxo: there doesnt seem much difference to linus' tree at the moment, http://pastebin.com/m42b6ef27
fredl: hi guys, http://pastebin.ca/1672566 trying to assign a Monitor section to a specific output port
fredl: is that not the correct syntax to do that??
fredl: that 'Option "HDMI" "HDMI Monitor"
adamk: Just name the Identifier "HDMI-0"
Ronis_BR: soreau: thanks :D
fredl: adamk - was just instructed to use http://pastebin.ca/1672584
fredl: (on #xorg)
adamk: And I'm instructing you not to :-)
fredl: still says Output VGA-0 using monitor section Tatung V42
fredl: oh the identifier...
adamk: If the Identifier for the monitor is the port output name from xrandr, you don't need to include it in the Device section, either.
fredl: hang on, didn't catch your intention
adamk: So change "Tatung V42" to simple "HDMI-0"
fredl: ok
fredl: and then obviously use Monitor "HDMI-0" in the Screen section right?
airlied: agd5f: btw do we need to save CLOCK_CNTL ref div select over s/r?
adamk: fredl, Or skip it entirely.
airlied: it looks like we fail to restore it and we only toucj the bottom 8 bits
fredl: I can leave out the Option lines in the Monitor/Device sections altogether as well?
adamk: airlied, Is it possible to find out the mode line that the console framebuffer is using in KMS so that I can force X to use the same modeline?
airlied: adamk: boot with drm.debug=2 I think
airlied: it should print out modelines in dmesg
adamk: fredl, Yes, I believe so. I'd just comment them out for now. They certainly aren't necessary with DVI and VGA outputs, so I doubt they are necessary with HDMI.
adamk: airlied, Thanks, I'll give that a try.
fredl: adamk - it seems they were actually working against me. Taking them out gives Output HDMI-0 using monitor section HDMI-0 now
adamk: airlied, It took me a while to realize that setting video=1280x1024@60 wouldn't necessarily force both monitors to use the same modeline :-)
cxo: multicore processors are really useful when you change kernels more often than change underwear
adamk: Just the same resolution and refresh rate.
fredl: I now just have http://pastebin.ca/1672594
fredl: going to check downstairs if it picks up that modeline now.
fredl: brb
cxo: airlied, hey i wanted to ask if there is anything worth while getting from drm-next that isnt in 2.6.32rc7 thats gonna make a non-running app suddenly start working?
airlied: adamk: no it still reads the modes from the monitor
airlied: cxo: its mostly hiding in drm-radeon-testing at the moment
airlied: most of what was in drm-next is gone to Linus now
adamk: airlied, What happens then if 1280x1024@60 isn't an option in xrandr for a monitor? Is there simply a way to tell the DDX to use whatever modeline KMS is using?
airlied: not yet, its on the list of todo things but I keep forgettnig
airlied: just need to add X server support to get the initial mode from the kernel insetad of picking its own
cxo: So can i, git remote add -t drm-radeon-testing airlied_drm_remote git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git && git pull airlied_drm_remote drm-radeon-testing
cxo: if i'm using linus' tree?
cxo: Anyways, so is there anything worth getting from drm-radeon-testing that has any end user application benefits over 2.6.32rc7?
Ronis_BR: soreau: man, it seems I can't test it now :( tuxonice isn't working on 2.6.32...
fredl: adamk - got that pretty close to working now. http://pastebin.ca/1672609 but how do I make the 1920x1080 mode the default mode?
Ronis_BR: It seems I'll have to wait 2 months :D
adamk: airlied, drm.debug=2 doesn't appear to have displayed any extra info: http://pastebin.com/m11761bfe
cxo: Ronis_BR, whats wrong with the default power management and suspend to disk code?
adamk: fredl, http://pastebin.ca/1672610
Ronis_BR: cxo: it just doesn't work here :D
Ronis_BR: cxo: tuxonice works almost out of the box
fredl: adamk - awesome, thanks so much for your help!
adamk: No problem. I probably should have mentioned that earlier, but completely forgot about it :-)
fredl: going to run downstairs to see if it worked :)
fredl: log says it did, but as I'm overriding what the monitor reports I hope it'll stick
Ronis_BR: cxo: tuxonice is better by far!
adamk: airlied, drm.debug=0xf (which I pulled from a log file of this channel) seems to have produced something: [drm:drm_mode_debug_printmodeline], Modeline 44:"1280x1024" 0 109000 1280 1368 1496 1712 1024 1027 1034 1063 0x0 0x6
fredl: damn, it does 'work' but the monitor doesn't like it.
cxo: Ronis_BR, i've only tried suspend to disk and suspend to ram once, and both worked, but i was using a desktop pc
adamk: airlied, But that doesn't look like a modeline I can just feed to xorg.conf and, in fact, when I tried, it just failed to parse this line:
adamk: Modeline "1280x1024_1" 0 109000 1280 1368 1496 1712 1024 1027 1034 1063 0x0 0x6
adamk: airlied, Any thoughts?
cxo: isnt there a util that comes with X that makes modelines for you
Ronis_BR: cxo: hum, laptop is a little bit harder
fredl: adamk - I got that Modeline by hooking up the same gfx card to the DVI input of the monitor, hoping it would work on the HDMI input as well. I know the monitor can do 1080p through HDMI because I've done it with the PS3 as well.
Ronis_BR: to set
adamk: cxo, Yeah, but that's not what I need.
adamk: fredl, Sorry, that's the extent of my knowledge... I've never touched HDMI.
fredl: adamk - no problem. Guess this is going to be hard, as the HDMI input for the TV underreports the capabilities
fredl: the screen goes black with this modeline :(
chithead: fredl: maybe set displaypriority to high
fredl: what does that do chithead?
chithead: fredl: this reduces display buffer underflow, see radeon manpage
chithead: fredl: or try with lower resolutions like 1600x900 first
fredl: ah okay, looking at the radeon manpage.
DanaG: hates how some HDTVs fail at edid.
DanaG: Fail as in, lying.
gimzo: radeon x700, mesa git master (yesterday), openarena input lags on 1920x1200, works ok on lower resolutions, but my monitor doesn't like 75hz then, any solutions ?
fredl: yeah me too...
DanaG: Oh hey, I'm 1366x768... but the only input I accept is 1280x720! /banghead
fredl: it just flat out underreports.
adamk: airlied, Also, do you know how I can specify one mode for one monitor and another mode for another monitor?
fredl: on the DVI port is does report 1920x1080, so perhaps there is a good reason for it.
DanaG: google for debian xstrikeforce
fredl: after all, HDMI is different circuitry than DVI
DanaG: It shows how to make xrandr permanent.
adamk: We already configured his HDMI with the modeline he wants in /etc/X11/xorg.conf
adamk: It just didn't work :-)
fredl: well I'll go downstairs in a few minutes to try out what chithead suggested but I've little hope that that will work.
adamk: (ie. the monitor didn't like the modeline).
fredl: isn't buffer underrun dependent on the performance of the system?
fredl: I mean, this is a Core i7 system...
fredl: I really would like it to go through HDMI so I can also utilize HDMI audio at some point
fredl: through the optical cable it doesn't do Dolby 7.1 but through the HDMI port it does.
twnqx: with what driver?
fredl: what audio driver?
fredl: Intel HDA on the gfx card
twnqx: mh
twnqx: i didn't get the audio on my displayport to produce *anything* yet
twnqx: looks in the developers general direction
fredl: does it have it's own audio chips or looped through?
twnqx: it's a laptop
fredl: coz if it has it's own chips it's a matter of Alsa config
twnqx: it has it's own codec though
twnqx: but i tried all of them
twnqx: nothing.
fredl: HDMI on your laptop?
twnqx: displayport.
fredl: cool.
twnqx: converted to hdmi, though
fredl: I dunno much about displayport, somebody pointed me to some webpage about it though
fredl: damnit, this pisses me off that that modeline doesn't work.
fredl: it's also not the most common monitor so finding a working modeline will be next to impossible :(
twnqx: oO
twnqx: just grab a 1920x1080@60
fredl: I did, which is a modeline that I grabbed from a working DVI setup.
fredl: (the TV also has a DVI in)
fredl: but it don't work on the HDMI :(
fredl: I wonder how the PS3 makes it work though.
cxo: X just hung on me, while playing openarena
fredl: on automatic probing the PS3 also doesn't see the 1080p mode
fredl: but forcing the PS3 into 1080p works fine.
lordheavy: cxo: yesterday got same problem
cxo: sound kept repeating, screen froze, and X took up 100% of cpu, couldn't recover, had to reboot
cxo: not bad, i've got X to crash twice now. So open source doesn't automagically translate to bug free drivers
lordheavy: cxo : got this in kernel messages http://pastebin.fr/5998
Ghworg: It translates to drivers that anyone can fix, without having to rely on the whims of a third-party
fredl: Hmmm. Sony 'supports' yellow dog linux on the PS3. YDL uses xorg...
gimzo: ...and has no gpu access
cxo: you know the xbox360 has an r500, i wonder if radeon will run on it
cxo: gimzo, isnt there a fully fledged gallium driver for the cell?
cxo: wasn't that the first driver they ever wrote even?
fredl: gimzo - just thinking of a way to get an HDMI modeline that will work for my monitor.
gimzo: yes, but for the cell, not the gpu
MostAwesomeDude: Not the first, but IIRC it was the first accelerated driver.
cxo: I thought the Cell is the GPU and CPU all in one in the PS3
gimzo: nope
MostAwesomeDude: No.
MostAwesomeDude: PS3s have an nVidia chipset inside as well.
cxo: So why is there a gallium driver for the Cell if its not a GPU?
fredl: gimzo - it'll still be near impossible IMO as I would still need to make a monitor specific modeline
MostAwesomeDude: cxo: Because somebody paid TG to write it, and it was open-sourced later.
MostAwesomeDude: Protip: There are non-open Gallium drivers out there.
cxo: Ok i think i meant to ask "How" not "Why"
gimzo: I think you can run less demanding games on cell without gpu
Dr_Jakob: MostAwesomeDude: not quite the cell driver was made as project for the DOE
MostAwesomeDude: Well, it's like llvmpipe. It's a software rasterizer that's been heavily optimized for a certain CPU.
MostAwesomeDude: Dr_Jakob: Oh, really? I didn't know that.
cxo: hates how this MAD dude's nick messes with his chat window indentation
gimzo: why would anyone pay to have software renderer made when they could pay sony for gpu access ?
Dr_Jakob: it wasn't specificaly made for the PS3, but it was the cheapest cell dev platform out there.
fredl: gimzo - perhaps you can help me out somehow. With my PS3 hooked up to my HDTV, the automatic detection finds a resolution lower than what the screen can actually do. I've tried to manually set it to 1080p and that works, I've played 1080p games on the screen as well.
gimzo: I don't have any hdtv, I played with gentoo on friend's ps3 (on pal crt TV) and so I have no idea :D
cxo: The Xbox360 has 3x3.2Ghz 64bit PPCs and a suped up R500. That would be so cool as a gaming machine if xf86-video-ati ran on it
fredl: gimzo - by that I know that the HDTV *can* do 1080p through HDMI. Using that knowledge I tried to figure out and configure a modeline based on the DVI modeline that I used in the past (the HDTV has a DVI in as well)
fredl: (this is for a PC that I also want to hook up to HDMI instead of DVI)
gimzo: fredl: and what happens if you try that modeline
gimzo: ?
fredl: gimzo, the screen turns black
gimzo: no messages like "invalid signal" from tv ?
fredl: nope.
fredl: not even any flickering, nothing.
Dr_Jakob: I know that the PS3 requires the monitor to support HDCP in order for it to switch to 1080p
gimzo: cxo: I think there's a project to make linux work on x360, it that's working just get one to driver dev and maybe it'll work :)
fredl: the Modeline I found seems like a pretty standard one: ModeLine "1920x1080" 138.50 1920 1968 2000 2080 1080 1082 1087 1111
Dr_Jakob: fredl: ^^ does your monitor support HDCP
gimzo: cxo: I guess used x360 can be found pretty cheap after those bans lately
fredl: Dr_Jakob, I believe it does but I'm not certain.
gimzo: Dr_Jakob: I don't think linux requires HDCP :)
gimzo: If we are still talking about PC, and not PS3
fredl: yeah we're still talking PC
Dr_Jakob: ah never mind then
fredl: the PS3 I only used as evidence that 1080p through the HDMI port works on the HDTV
Dr_Jakob: gimzo: its the hypervisor that sets the real mode on the PS3 and it enforces that limitation
Dr_Jakob: or it could be a hw limit.
cxo: gimzo, its a sub $200 product these days, and thats getting you 3x3.2Ghz 64bit PPCs, and a R500, cant beat value for money
gimzo: cxo: actually, getting 5 or more used / banned ones would be nice for running a cluster :)
gimzo: I guess banned ones can be found under $100
cxo: there arent any nice generic cluster kernels anymore
cxo: used to run openmosix back in the day
gimzo: there is MPI :)
gimzo: there is also blender rendering farm :)
cxo: I think there is also something that allows gcc to spread around nodes
gimzo: distcc
gimzo: cxo: http://www.free60.org
gimzo: this seems nice
fredl: hmmmmm
fredl: oh my gosh
fredl: could this be true... :)
fredl: http://martin.linuxnerds.net/?page_id=98
fredl: hmm why does that modeline keep saying the mode clock is too high?
gimzo: fredl: have you tried "use edid" "false" with your previous modeline ?
fredl: gimzo - nope I didn't
fredl: it's 'Option "IgnoreEDID" "false"' isn't it?
fredl: ehr "true"
gimzo: ignoreEDID should be true
fredl: that's a good idea...
fredl: hang on
fredl: (BTW, the monitor in that URL is exactly my monitor, however the modeline the guy posted is for the VGA input)
fredl: runs downstairs
fredl: *sigh* no such luck gimzo
gimzo: still dark ?
fredl: yup
gimzo: what does xorg.0.log say ?
fredl: and I'm just looking through the log but I don't see any indication that it actually picked up the IgnoreEDID line
fredl: looks like this now: http://pastebin.ca/1672712
fredl: log says it's using the 1920x1080 mode
gimzo: um, that's xorg.conf ?
fredl: yeah
fredl: wanna see the whole log?
gimzo: yup
fredl: http://pastebin.ca/1672714
agd5f: airlied: possibly. I thought we explicitly set it at some point, but maybe not (pll divider selection)
gimzo: fredl: I don't see anything odd here. Does the tv work in any other resolution ?
fredl: it works somewhat in the 1280 mode but it looks like the timings are off somewhat there.
fredl: those may be adjustable though, it's just that part of the pixels is outside the screen.
gimzo: shouldn't dvi/hdmi screens not have timing problems ?
fredl: I would think so. the 1920 mode works perfect through DVI
fredl: but haven't tried any other modes in DVI, never had a need to.
gimzo: tv has dvi in or do you use the same hdmi in ?
fredl: tv has dvi in, vga in and hdmi in
fredl: the HDMI is hooked up through my Sony AV BTW
gimzo: can you try connecting it directly ?
fredl: I think I should also try hooking up the HDMI in directly but I'll keep that for tomorrow.
fredl: it's getting rather late here in .nl :)
gimzo: then you can see the modeline :)
gimzo: it's late here in .hr too :)
fredl: LOL, ok, lemme run downstairs again then :)
fredl: brb
fredl: heh, good thing I did that!
fredl: it works!
airlied: agd5f: we seem to clear it then program DIV_3
fredl: it has some similar pixels-outside the screen problems
fredl: but essentially it works fine. So it's the Sony AV that's fucking things up
fredl: it's always Sony fucking up.
gimzo: is sony rated as 1080 compatible?
fredl: I have no clue, is there even such a thing?
fredl: Sony makes 1080p equipment yeah :P
fredl: I bet it does it's own checking and won't allow the signal through or something when it's outside the specs
airlied: agd5f: are those ref div changes for doing LVDS reclocking or something like that?
airlied: always wondered why you get 4 ref divs on crtc 0
fredl: which is strange though, the PS3 also does 1080p through the Sony AV
fredl: maybe I should hook up the PC to the same HDMI in port the PS3 is connected to.
agd5f: airlied: yeah, could do lvds reclocking. we actually use if for vga in the bios IIRC. div 0 and 1 were for common vga modes and 3 was for the driver
fredl: well I'll keep that for tomorrow now that I know the modeline essentially works.
airlied: agd5f: so we should be programming ref div 3 for the driver?
agd5f: well, bios programs 0 and or 1, but once the driver loads, I guess it doesn't really matter
tsamolotoff: Hi guys
agd5f: but the driver has always used 3
airlied: agd5f: I've seen 3 programmed on I think EH's logs
airlied: agd5f: well we program 3 but we seem to select 0\
tsamolotoff: Is it worth to update to rc6 from rc7 concerning KMS ?
airlied: as we clear all the bits of CLOCK_CNTL_INDEX
tsamolotoff: err, upside down
airlied: unless I'm missing something
airlied: or what I am
airlied: I fal at reading WREG32_P
airlied: might make that actually say PLL_DIV_SEL_3 for ease of reading
tsamolotoff: Hello?
agd5f: tsamolotoff: best bet is drm-next or radeon-drm-testing
cxo: TOP GEAR STARTS TODAY!!!!!!
tsamolotoff: agd5f: Thanks, I'll try the linus' tree first... There's many changes (and as I hope, compressed textures fix) in the LKML log for rc7
tsamolotoff: Thus, is it fixed?
airlied: agd5f: another q, if you are near an r600, put some app over firefox and scroll fierfox (may need to turn off auto-raise)
airlied: (also no compositing)
gimzo: Would it be useful to anybody if I continue doing this every day ? http://global.phoronix-test-suite.com/?k=profile&u=gimzo-11227-9088-27957
cxo: I was just thinking... now that we have all these cool KMS and DRM and stuff happening, can we play Quake without X?
cxo: ie just using the framebuffer
gimzo: cxo: I think plymouth guys were implementing gl on console or something like that
cxo: that would be so sweeeeeeeeeet
cxo: in the future everything is CUI and openGL
gimzo: but I don't think running games was their idea
cxo: GUIs are so 1999
gimzo: :D
cxo: I just noticed that mplayer doesnt work on the console anymore
cxo: sdl, fbdev or fbdev2, neither works
cxo: directfb fails too :(
airlied: cxo: fbdev has no accel
airlied: so its mostly pointless since you don't use the gpu
airlied: but I think I have some fixes I need to queue up
gimzo: airlied: can doing benchmark on git master every day help anyone somehow ?
airlied: gimzo: probably doing one with mesa could help spot regressions alright
cxo: airlied: how come none of the fb mplayer video outputs work anymore with the radeon driver?
factor: What is the best GL driver for ATI the default one that comes with ubuntu 9.10 or another one in the repo
gimzo: ok, because I'm planning to do it waiting for framerates to become usable, so it will be around
airlied: cxo: not sure when it broke, there are patch on the mailing list to fix it
airlied: I just need to test them
cxo: gimzo: hey, sdl has a fbcon video output driver!
cxo: but it doesnt work right now with radeon :)
gimzo: cxo: well, I never had mplayer working on console, but now that I have 120x1200 console it's worth a try
cxo: SDL_SetVideoMode fails
cxo: I always had mplayer working on the console with the vesafb
gimzo: I had mplayer fighting with console :)
gimzo: and leaving it broken after it finishes
cxo: this is so cool, SDL supports a whole lot of different console output drivers
factor: is the fglrx the best over the standard ati driver
Ghworg: Depends on your definition of best, but probably yes.
factor: I tried that last itme but messed up my computer but may have been the GL progam
factor: had to reinstall ubuntu ext4 fs did not hold up to well
adamk: So how do I translate 'Modeline 44:"1280x1024" 0 109000 1280 1368 1496 1712 1024 1027 1034 1063 0x0 0x6' into a modeline for xorg.conf?
chithead: looks suspiciously similar to "cvt 1280 1024" output
adamk: Ah ha... 10900 = 109.00
adamk: Damn.
adamk: Thanks for that...
adamk: Now to see if that works.
adamk_: Still get the delay when switching from Xorg to console on the CRT.
factor: I get GLX error: Can not get required symbols
factor: and I put fglrx in the xorg Driver slot
factor: is that all I should have to do
factor: I get no devices detected via X -config /etc/X11/xorg.conf
factor: but I have an ati 9200
adamk_: factor: fglrx does not support your GPU.
adamk_: It hasn't supported it for probably 4-5 years.
factor: ok
factor: which one does
adamk_: And, besides, this channel is for the open source drivers.
adamk_: The open source drivers :-)
adamk_: radeon
factor: ok
factor: if I use radeon all I get is a black screen
factor: does another driver support the card besides radeon
factor: and will give me GL
yangman: mesa gives you GL
yangman: and really only Mesa
factor: can I put mesa in the Driver section of xorg
yangman: factor: http://yangman.ca/blog/2009/10/15/linux-graphics-driver-stack-explained/
yangman: if radeon is problematic for you, try radeonhd. in either case, Mesa is what provides 3D
adamk_: No. Mesa is a set of libraries that provide 3D acceleration. Only 'radeon' provides 2D acceleration.
adamk_: yangman: You're recommending radeonhd for a radeon 9200?
adamk_: factor: Do *not* try radeonhd.
factor: heh
yangman: ah, 9200. yeah, don't try radeonhd
factor: I want a hd card but will wait for a few days
factor: I have three systems this one has only agp
factor: which should I try
factor: radeon just is not working
factor: can I use some frame buffer driver?
MostAwesomeDude: factor: Do you still have fglrx installed? It'll interfere with open drivers as long as it's there.
factor: I took it out of the xorg file
biotube: you need to completely remove it
factor: it was custom compiled though upon install of the package.
factor: ok
factor: both xorg ati and xorg radeon black screen
adamk_: Did you completely remove fglrx yet?
adamk_:
factor: yes
factor: all of the fglrx stuff
factor: looking for a frame buffer way to do it
[Enrico]: factor: let me bet..... ubuntu?
MostAwesomeDude: Could you pastebin your dmesg?
factor: huh
factor: just a black screen
MostAwesomeDude: Take the output of the dmesg command and put it in a pastebin.
factor: I will put it in here just one line
factor: that is worthy
factor: but running the radon driver thru to get fresh errors
factor: says something about freeing invalid memtype d0102000-d0112000
MostAwesomeDude: Well, if you're not going to give the whole thing, have you at least tried doing "modprobe radeon" from your VT?
hnsr: what's the difference between drm-radeon-testing and drm-next? I see the former has more recent commits
factor: will try
factor: install it , then run X -config xorg.conf just black screen
chithead: factor: you don't need any xorg.conf, just rename or remove it
airlied: hnsr: one contains commits that need testing
airlied: I'm trying ot move radeon dev out of drm-next where it only ended up by accident
hnsr: ah, ok
hnsr: will switch to drm-radeon-testing
yangman: factor: do you need to configure a xorg.conf? does xorg.conf-less setup not work for you?
factor: humm let me see vesa work by the way
factor: but vesa does not have GL support
cxo: This is interesting.
cxo: I just switched from vanilla 2.6.32rc7 to 2.6.32rc7+drm-radeon-testing and my Quake 3 FPS dropped by 10% but glxgears went up 20%
factor: and no a confgiless X does not work for me
[Enrico]: and this is strange
cxo: factor, conflicting drivers?
[Enrico]: factor: Xorg.0.log and dmesg output can help a lot if can provide them to us
MostAwesomeDude: factor: Seriously. Pastebin Xorg.0.log and dmesg.
cxo: factor, seriously
factor_: install debian maybe something wrong with ubuntu
[Enrico]: factor: i don't think there is something wrong with ubuntu, but if you really want to reinstall i suggest you fedora
biotube: why are kernel merge conflicts _always_ in intel's code?
factor: dont like rpm too much
factor: I am a debian fan anyway, more comfortable with it and apt
factor: I sat on a plane for linuxworld next to Donnie Barnes telling hm how much I hated rpm
factor: heh
factor: had no idea it was him though
[Enrico]: factor: understood :D
airlied: biotube: because we don't merge their trees into radeon trees
factor: I like fedora setup happy for them , keep on keeping on
factor: just like debian
biotube: at least it's fairly easy to tell what to do to fix it
[Enrico]: factor: i see everybody has its personal feeling with every distro so just use the one you enjoy much :D
factor: will the new ati HD class card worj with a 300 watt powersupply
factor: work^
factor: read somewhere you needed 500watt
[Enrico]: factor: well the most powerfull r800 uses 325 watt TDP
[Enrico]: do your cumputing :D
[Enrico]: the less powerfull 86
[Enrico]: factor: so a 300W psu is not raccomended imho
factor: wow
factor: will it work though
MostAwesomeDude: I have a 500W that does the job.
factor: if I got one and upgraded later.
factor: well I know a 500 would
factor: but can a 300 do it for a time being
[Enrico]: MostAwesomeDude: what TDP has you card?
MostAwesomeDude: Sure, but you could burn it out pretty fast.
MostAwesomeDude: [Enrico]: Oh, no idea. I just got a PSU that I expected wouldn't burn out quickly.
[Enrico]: MostAwesomeDude: then can i ask you the model? i can find the tdp :D
MostAwesomeDude: It's all about power *ratings*, not actual consumpsion.
MostAwesomeDude: *consumption, even.
MostAwesomeDude: [Enrico]: Sure. It's a Gigabyte 780G board, small factor, with an RV410 plugged in.
[Enrico]: oh rv410 so small TDP
MostAwesomeDude: The onboard GPU is disabled, although I bet that there's not a number for that.
MostAwesomeDude: I've got an HD 4850x2 sitting beside me, but I haven't put it in yet. :3
[Enrico]: MostAwesomeDude: i have a mobility radeon hd 3470 works great with git stuff eheh (but i have to force lowpoermode, so no kms for now :'( )
[Enrico]: it become a bit hot without powersave and you know this is a laptop so the fan is shared -> also the cpu becomes hotter
[Enrico]: but wow the 3d is just awesome
[Enrico]: it is not so far from fglrx (at least in nexuiz) and wow it is very very quickier then fglrx in almost everything (expecially composite) so congrats guys
factor: well I have a 300 watt running a nvidia 9800gtx which seems to work fine.
factor: but also recomends a 550 watt
factor: I want both streams cards to test my code on for openCL
[Enrico]: lol demo5 in nexuiz just frozed my system rofl
cxo: didnt even know there was a demo5!
cxo: immediately drops everything hes doing and tries demo5
[Enrico]: cxo: multiplayer -> demo
[Enrico]: rofl
cxo: system waaaaaaaaaaaaaaaaay more robust!
cxo: it just crashed on mine, but the system didnt freeze
cxo: drmRadeonCmdBuffer: -22. Kernel failed to parse or rejected command stream. See dmesg for more info.
[Enrico]: cxo: are you running xorg-server 1.7 ?
cxo: [10713.495403] [drm:radeon_ib_get] *ERROR* IB 3 scheduled without a fence.
cxo: [10713.495405] [drm:radeon_cs_ioctl] *ERROR* Failed to get ib !
cxo: [10730.093664] [drm:radeon_cs_parser_init] *ERROR* cs IB too big: 16424
[Enrico]: cxo: couse i'm still on 1.6.4
cxo: i'm using whatever Ubuntu 9.10 has
cxo: 1.6.4 apparently too
[Enrico]: cxo: well a friend of mine with xorg-server 1.7.whatever say the same as you
cxo: Well we should ask airlied then
cxo: I'm running his drm-radeon-testing branch on 2.6.32rc7
[Enrico]: and i'm using .31 zen7 kernel + drm-next (from few days ago) and libdrm video-ati and mesa from git
[Enrico]: oh well i don't play nexuiz normally so who cares :D
cxo: I want to play Modern Warfare 2! if i had an Nvidia card, i could be playing it already! suckage
factor: do the HD cards work fully with the open source radeonHD driver.
factor: seems to have the streams SDK for it but only 64 bit match lib
factor: so 32 bit stuff cant use streams
biotube: 2d support should be the same as radeon, minus KMS
factor: not to any usefull level
factor: what about 3d support
biotube: that's separate
biotube: mesa
factor: separate?
MostAwesomeDude: factor: If you want OpenCL, you might as well do fglrx on an HD card.
factor: oh ook that will be nice then
cxo: i think StreamSDK will only work with fglrx
factor: it will be for openCL
cxo: MostAwesomeDude will code an openCL and CUDA driver for gallium one day and then we'll have that instead of StreamSDK
factor: well cuda and fireStreams should work with only a openCL driver
factor: write once work on all
MostAwesomeDude: cxo: mesa/clover
factor: I can write CUDA streams workingon openCL since it is pretty new
factor: which is why I want a ati stream card
factor: to test how open cl works on no CL hw and cuda and fireGL
factor: working on the no CL box right now ati 9200
factor: see how debian works with the default radeon driver and xorg
factor: on no openchrom GL works for looks really bad and out of sync
factor: debian radeon driver also worked openchrome was form my on board ati
factor: found GL mesa driver as well now
factor: and glx gears working fast now
factor: now I am happy everything I wanted in a no stream environment
factor: just would not work under ubuntu suno
factor: dunno
basilgohar: Is there any support for video capture through AIW cards through the radeon driver?
basilgohar: I have an All-in-Wonder 8500 DV, by the way.
basilgohar: References on the GATOS website mention that the code for video capture was merged into Xorg, but I didn't know if that meant the ati driver or not..
MostAwesomeDude: Most video-capture-only cards should work. The ones with video capture chipsets on the same board as the graphics probably don't though.
MostAwesomeDude: GATOS really should be re-examined and merged once and for all, I'm pretty sure that AMD doesn't mind anymore. :3
airlied: some of them used to with userspace modesetting drivers
airlied: really need to port them all to kms
airlied: and just use in-kernel v4l interface
basilgohar: Just an interesting observation...my 8500DV has far better 2D and 3D performance than my 9600 Pro...
basilgohar: I assume that is because of the support for the older, R200 chipset compared with the 300 series...
MostAwesomeDude: airlied: I totally agree.
basilgohar: I forgot how good performance under GNU/Linux could be...:D
MostAwesomeDude: I has a dedicated ATI decoding card which I assume already works.
MostAwesomeDude: I don't think any of my cards are AIW though.
basilgohar: So, is there anything I can do to help getting that to work?
basilgohar: I think GATOS was discontinued after the 8500DV was released, so it probably doesn't even support it...
basilgohar: I'm mostly interested in video capture, not the TV tuner...
cxo: I had an 8500LE and boy did it kick
basilgohar: Nevermind...
basilgohar: All-in-Wonder Radeon 8500DV (Radeon200)
basilgohar: Should work fine with 4.2.0 or later drivers. XvImage (YUV->RGB overlay and scaling) should work fine. TV-in works in NTSC. It is expected that PAL and SECAM will be easy to made working once there are people willing to test and debug TV-in with these formats. Video capture should work. 3d acceleration is not implemented as of this writing. Firewire port (at least the one on the card itself) has been observed to work in
basilgohar: Linux. Wireless (RF) remote control should work.
basilgohar: Appears the 8500DV is supported...
basilgohar: Oops, multiple line paste, sorry about that...
cxo: How do you set the refresh rate for the tty console?
cxo: I'm guessing its some sort of param to pass to drm or radeon?
MostAwesomeDude: basilgohar: 3D works fine.
basilgohar: MostAwesomeDude: yeah, I know...I'm using that card now. I was pasting what was on the site, which I shouldn't have done.
basilgohar: I am using the 8500DV right now and I'm well pleased with the 2D & 3D performance.
basilgohar: it's just the video cap that I was hoping to enable, but it seems like what it says on the website hasn't happened yet, at least completely...i.e., merging of GATOS into Xorg.
airlied: there should be info in the xorg log
airlied: pastebin it
basilgohar: airlied: Are you talking to me?
basilgohar: I'll take a look...
basilgohar: airlied: http://pastebin.com/m2bcdc04c
airlied: it can't find the table in the bios by the looks of it
basilgohar: airlied: Anything I can do about that?
airlied: ah you are using F111, try booting with nomodeset on the grub command line
airlied: we don't have any AIW support in the kms code
airlied: which F11/F12 use
dmb_: they still make all in wonders?
basilgohar: airlied: I am using F11. When I tried nomodeset earlier for another issue, X never came up...
basilgohar: My monitor is not being detected properly, or rather, its capabilities are not.
airlied: basilgohar: oh wierd, not sure why non-kms doesn't work with it
basilgohar: The issue with resolutions is odd...
basilgohar: I'm running right now at 1600x1200x85Hz on a 19" KDS CRT, looks beautiful.
basilgohar: But, there is a section in the log prefixed with "Supported established timings:"
basilgohar: And then it lists a very limited number of resolutions and refresh rates...
basilgohar: But in other parts of Xorg.0.log (the same one I pastebinned), it lists the full capabilities of the monitor as it should be (up to 1920x1440x60Hz, for example).
basilgohar: I had to xrandr to add the high res mode myself.
basilgohar: *use xrandr utility
airlied: I think we fixed that in F12, not sure though
basilgohar: I suppose I can try nomodeset again, I guess...
basilgohar: I see.
airlied: anyways AIW won't work with KMS enabled
basilgohar: Hmm...I don't have a LiveCD handy...
cxo: airlied, how do you change the refresh rate for the console?
basilgohar: It's already midnight here, too, I need to get some rest for work tomorrow.
airlied: if you geta log from nomodeset we can see what the driver thinkgs
basilgohar: airlied: Okay, I'll keep that in mind. Thanks for the insight!
basilgohar: airlied: Okay, I'll try that...maybe I'll give it one more shot for tonight...
airlied: cxo: video=DVI-0:1024x768@60
airlied: cxo: something like that
cxo: as a kernel param? or do I need to do radeo.video=.... ?
airlied: kernel param and no
cxo: holy! half life 2 works... well i get 0fps but it works
factor: heh
cxo: hl2 says the hardware only supports dx6
ossman: airlied, could you have a look at https://bugs.freedesktop.org/show_bug.cgi?id=24611 ?
ossman: I spent last evening doing some tracing, but I could use some ideas from someone familiar with the code
cxo: Holy! i got about 20fps on hl2 now, (turned down all the details)
cxo: check it out!!! http://img35.imageshack.us/img35/5584/screenshot3yl.png
factor: nifty
dmb_: it ugly
dmb_: now i want to see someone get borderlands working :P
dmb_: that is a GREAT game
airlied: ossman: that is very strange
ossman: indeed :)
ossman: considering the behaviour, I'm guessing the fb is not setup properly somewhere
ossman: perhaps some assumption that DDC will be available?
cxo: sheeeit, i couldnt care about hl2, i just wanted to see what worked with the open source drivers
airlied: ossman: it should oops then
airlied: ossman: or at least give something of use, I might suspect the vbl irqs maybe
ossman: do you have guard pages around the fb?
airlied: ossman: the fb is in VRAM
airlied: it doesn't matter if you trash all around it
ossman: is there some hack I can do to kill vbl irqs?
airlied: just comment out the vbl irq enablses in r100.c:r100_irq_set
ossman: that code is shared, despite the naming? :)
airlied: yup
cxo: Is radeon_legacy_* old code or just referring to old hardware?
ossman: non-atombios from the looks of the files
ossman: seems to be everything pre-r400
ossman: airlied, both masks removed. let's see what happens...
airlied: r400 is blurry it uses both legacy and atom code
ossman: the TV-out limitation of only supporting 800x600, is that only for the non-atom cards?
airlied: ossman: should be in theory
airlied: also for r400
ossman: you fill me with confidence ;)
ossman: I guess I'm one of the few holdouts still using tv-out...
cxo: http://www.semiaccurate.com/2009/09/15/nvidia-gt300-yeilds-under-2/
basilgohar: airlied: http://pastebin.com/m25159013
basilgohar: That's starting with nomodeset
ossman: airlied, still locked up with vbl irqs disabled
airlied: basilgohar: so it sees most of the devices on it
airlied: I've no idea what tools you need to use to play with them
basilgohar: airlied: Okay, thanks for looking for me.
airlied: ossman: ouch
basilgohar: airlied: So, nothing is registered standard V4L2 devices, though, right?
airlied: basilgohar: no for kms we can do that
basilgohar: airlied: Sounds great...when do you think it'll be ready? :p
airlied: but with gatos it was all a hack if I remember correctly
ossman: airlied, I take it there is no gpu recovery code present yet?
airlied: basilgohar: when someone writes it ;-)
airlied: ossman: thats a bit worse than a gpu crash
airlied: ossman: I think the whole bus is gone or cpu is stuck in a loop
basilgohar: airlied: Well, you'd better get started on it! Snap snap! :D But seriously, thanks for helping me at least understand what's going on better.
airlied: basilgohar: I actually started trying to expose the i2c bus so we could probe the chips but I got distracted by higher priorities
ossman: airlied, so waiting for a gpu completion, even for the fb console, should never stop userspace?
airlied: ossman: there is no gpu completeion its not doing anything
airlied: either we program the mode and it gets stuck somewhere dumb
airlied: or somehow it crashes the bus
cxo: catalyst's gpu recovery is rock solid on windows
airlied: it might be worth finding a few while loops or maybe enabling the kernel nmi watchdog
ossman: you don't think it crashes as a result of fbcon trying to draw stuff?
airlied: cxo: nothing to do with this
airlied: cxo: gpu recovery for the accel engines is very different
airlied: ossman: no
ossman: do you remember off hand where the watchdog is in the config menu?
ossman: goes to pull the plug on the machine
airlied: ossman: kernel hacking somewhere I think
ossman: ok
ossman: somewhat related, is it possible to control which CRTC is used for OpenGL's vblank syncing?
airlied: ossman: it should be automatic
airlied: whichever the window is mostly on
ossman: in clone mode it isn't as clear :)
airlied: ah in clone mode not sure
ossman: for Xv, I believe there is a property controlling it
ossman: but I haven't seen anything similar for GL
ossman: I was hoping the new --primary would do something :)
airlied: you'd have to hack RADEONBiggerCrtcArea
airlied: at least with DRI2
ossman: airlied, found the watchdog. apparently it's always in as long as you have APIC suppor
hagabaka: I'm using 2.6.32 kernel to get KMS for radeon 9800 Pro; there are flickering lines about 1/2 inch apart on the right side of the screen. I have to change the refresh rate from the default 60 to 75 to get rid of the flickering, but it comes back when I reboot. any suggestions?
ossman: you just need an extra kernel parameter
cxo: hagabaka, here, video=DVI-0:1024x768@75
airlied: ossman: see if it can trigger for you
hagabaka: do you happen to know where to put it on ubuntu?
ossman: airlied, patience grasshopper. this machine isn't that fast. it's compiling in some extra hang checks :)
cxo: hagabaka, /boot/grub/menu.lst
ossman: rebooting...
hagabaka: but then I need to do it again when I add a new kernel right? there's something called initramfs, would that work?
MostAwesomeDude: hagabaka: No, you don't need to do that.
ossman: airlied, nmi:s are ticking, so this seems promising...
ossman: airlied, machine does not reboot
ossman: so I'd say a hung bus rather than a cpu loop
cxo: a hung bus
cxo: sounds funny - a hung bus
hagabaka: ok
hagabaka: strangely my monitor VG150 is supposed to have only 60Hz refresh rate...
cxo: sounds like humbug - a hung bus
hagabaka: can I just put the video= line in the top section of menu.lst, not on a kernel line?