Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

RadeonHD IRC Logs For 2009-4-08

Search This Log:


nicark|: Hi. Is it possible to rotate a screen to the right with the radeonhd drivers and with a radeon hd 4870??
Zajec: nicark|: you need more-or-less experimental EXA for that
Zajec: nicark|: then just xrandr --output VGA-1 --rotate 'normal', 'left', 'right' or 'inverted'
Zajec: about not-yet-stabilized EXA http://wiki.x.org/wiki/radeonhd:r6xx_r7xx_branch
Zajec: and it's quite stable actually ;)
nicark|: weird. I do enable exa in xorg.conf,restart x, and then i type "xrandr --output DVI-I_2 --rotate right" but nothing happens :(
agd5f: nicark|: exa and xv require drm support
nicark|: agd5f, I accomplish drm support by enabling dri in xorg.conf?
agd5f: nicark|: you need to build and install updated drm modules for your kernel
agd5f: nicark|: see teh link Zajec posted
nicark|: ooh, will do that then. Thanks
nicark|: agd5f, Got the lasest drm modules (on archlinux through yaourt xf86-video-radeonhd-git-r6xx-r7xx and more importantly drm-radeon-module-git-r6xx-r7xx). Restarted x, but still won't let me rotate my second screen, why could this be?
Zajec: nicark|: show your /var/log/Xorg.0.log (pastebin it)
nicark|: Zajec, http://pastebin.com/d41a22b6a
Zajec: nicark|: first of all read top of page I posted
Zajec: you SHOULD NOT use r6xx-r7xx-support branch of xf86-video-radeonhd
Zajec: and then part of your Xorg.0.log: (II) RADEONHD(0): Direct rendering turned off by default. Use Option "DRI" to enable.
Zajec: you really didn't check that page (there is info about Option "DRI")
nicark|: Zajec: Sorry, felt scared by doing it by something else than pacman and tried using something more familiar. I'm a little new to linux. Anyway, I am in the ./autogen --prefix=/usr step now and I get a "Note: DRI support is disabled", is this normal?
Zajec: hm, probably not
Zajec: first you can check if there is xf86-video-radeonhd package than containt current git
Zajec: or even a few days old... but not 1.2.4
Zajec: for compiling you may need to install libdrm-devel
mmm: nice to see debian 5.0 is using radeonhd by default after installation
mmm: having trouble figuring out how to set up dual monitors though -- been a while since the last time I configured this
rehabdoll: there are examples in the wiki
mmm: trouble searching online for answers is "radeonhd" comes up with everyone who talks about their Radeon HD video cards :)
rehabdoll: (topic)
mmm: rehabdoll: did try to do as the wiki states, by adding a Virtual line in xorg.conf to the screen section, but that caused X error on start
rehabdoll: its not required, as long as all displays are on when you start X
rehabdoll: what was the error?
rehabdoll: you can just use xrandr to set it up too
mmm: actually, now that I think about it - I might have used "Virtual 3370x1050" (using an "x" and not a space!)
mmm: will try it again and get back to you
mmm: okay - thinking I see the problem - the Virtual keyword must be in the "Display" subsection of te "Screen" section
mmm: had not made a subsection in "Screen"
kaouete: mmm: Why use Virtual ? If your monitors are always plugged in, I think it is not needed
mmm: oh? just try the xrandr ?
kaouete: yes, or just use the randr keyword in your xorg.conf
kaouete: I think there is examples on the wiki
kaouete: on the main page, section 11
mmm: I run the command, as root user, "xrandr --output DVI-I_2 --right-of DVI-I_1" and it produces no errors, but there are no changes either
kaouete: try xrandr --auto ? And you should just run it with the user who started X
mmm: run the command "xrandr --auto" and again it produces no errors and makes no changes
kaouete: and if you run "xrandr", is there all the monitor you want with their resolutions ?
mmm: kaouete: yes http://pastebin.ca/1386043
kaouete: first of all, don't forget the /digital at the end of the monitor names
kaouete: the name is DVI-I_2/digital, not DVI-I_2
kaouete: and when I look at the output of xrandr, I read that both monitor are enabled, but maybe the problem is that the max resolution is 1680x1680 ...
kaouete: you started X with the monitor plugged ?
mmm: yes - both monitors are plugged in at all times
kaouete: ok, and it is in clone mode currently ?
mmm: yes correct
kaouete: and you want them to have an extended desktop instead ?
mmm: yes again
kaouete: ok, so maybe the problem is the "maximum 1680 x 1680"
kaouete: either your card is not powerful enough, either there is a problem somewhere :)
kaouete: but I can't help you for this
mmm: yes, that's why in past I have needed to do a virtual in xorg.conf
kaouete: maybe check the xorg.log to see if it says something about it
adamk: Did you configure the Virtual?
adamk: It is necessary.
adamk: In many situations, still, anyway.
kaouete: adamk: are you sure ? Here it is not, maybe because I am smaller monitors ? I can't check now what are the right values ...
kaouete: s/am/have/
adamk: Yes.
kaouete: ok :)
adamk: And, even if I'm not, what's the harm?
mmm: the wiki says I can configure "multimonitor setup statically" by using Option "RightOf" inthe monitor sections
kaouete: adamk: none :)
kaouete: mmm: you can try statically yes, follow the example section 11, this is what I did and it worked ...
mmm: adamk: the wiki says virtual should be used for dynamic configs only
kaouete: mmm: but when using xrandr, you are doing it dynamically !
mmm: although virtual is how I have always done it in the past
kaouete: this is why
adamk: Right, if you set it up in your xorg.conf so that starts up with the monitors positioned, the Virtual isn't needed.
adamk: If you use the xrandr command, it is.
mmm: okay - giving the static confguration a try - will report back again soon
mmm: http://pastebin.ca/1386051 this xorg configuration looks right ?
mmm: will try dynamic config now
kaouete: mmm: I don't think, I don't even see the driver specified, nor the output binded to monitors in the device section
kaouete: start from there : http://wiki.x.org/wiki/radeonhd#head-0b4fdb3bbfdbdcdc68c8d712f75be6741e149685
mmm: kaouete: yes - I am a bit unclear where all this configuration has been moved in new debian distributions
kaouete: Option "monitor-VGA_1" "External" <= this is important for the static conf, for you it will be "monitor-DVI-I_1/digital" for example
kaouete: and "Configured Monitor" instead of "External"
mmm: I see....
mmm: thanks - saw it but did not think it was pertinent
rehabdoll: yum.. bootstraping an entire distribution is fun
rehabdoll: now i know why people like gentoo :)
kaouete: rehabdoll: yes but you do it only one time, after it is no more fun :)
Zajec: agd5f: is EXA enabled by default in radeon? for r6xx/r7xx
Zajec: I ask about 6.12.2
agd5f: Zajec: yeah, 6.12.1 and 6.12.2
Zajec: thx
Zajec: btw, can you tell what is status of EXA for r6xx/r7xx?
Zajec: is there something that still needs optimization?
Zajec: will this be faster?
agd5f: Zajec: pretty much everything that can be accelerated by exa is implemented
Zajec: sounds nice :)
mmm: yes - all set now - thanks for your help
iws: hello all. I've got an FireMV 2260 card which I'm trying to make work. (I'm running Ubuntu Hardy). I built radeonhd-1.2.4 myself. In my xorg.0.log, I get "card information has invalid connector information"
mmm: was able to set up a static configuration by defining the monitors and using Option "RightOf" "LeftVDU"
mmm: turns out that I did not need to use Option "RROutputOrder" at all
kaouete: yes, it is not needed normally
kaouete: but sometimes, it is, I don't remember exactly the explanation, if you want you can take a look at man radeonhd
mmm: glad X has such a readable log file - was stuck once with a typo but the log made it clear what was wrong
Zajec: agd5f: what is needed for enabling EXA on more than one X session?
Zajec: some DRM magic? DDX? DRI?
agd5f: Zajec: not possible right now
agd5f: yeah, new drm bits
agd5f: kernel memory manager, multi-master drm stuff
Zajec: oh, doesn't sound like one little patch ;)
agd5f: Zajec: mostly already done for r1xx-r5xx. just need to fill in the bits for r6xx/r7xx
Zajec: "memory manager" sounded (is that past of "sound"?) like something really big :)
agd5f: yeah, it is
iws: I'll be right back, if anyone can help me with my problem. I'm going to restart X with radeonhd-git instead of 1.2.4
Zajec: agd5f: do you think you will hit 2.6.31?
agd5f: hopefully
Zajec: :)
mmm: on another topic, while I am here... I have an R635 chip card (3650HD I believe)... any estimate the improved 3D support - I think it is the DRI modules
Zajec: mmm: "improved"? currently there is NO 3D support :)
Zajec: mmm: bridgman told few days about about IP reviewing... and *maybe* waiting a little to match changes in radeon-rewrite
mmm: the radeonHD has been a real workhorse for my 2D, dual screen desktop, but I look forward to when I will have access to the 3D applications
Zajec: but "hepefully soon" ;)
agd5f: mmm: 3D work has already started. we are still going through the IP review process before we can release it publically.
mmm: so it is still in the "IP wait and see" stage ... sigh
mmm: well - your work here is sure appreciated - think it is awesome what you are doing for the linux desktop
mmm: thanks again for your help !
iws: ok, I tried radeonhd-git, but I still have the problem: "(EE) RADEONHD(0): Card information has invalid connector information" and "(EE) RADEONHD(0): Screen(s) found, but none have a usable configuration"
agd5f: iws: what card?
iws: ATI FireMV2260 (RV620)
agd5f: iws: can you send me your video bios?
agd5f: iws: does radeon work any better?
iws: agd5f: I haven't tried radeon yet. vesa is working fine
iws: agd5f: here is the output of rhd_conntest -d on my card, with a single DVI display connected (through a DisplayPort adapter): http://www.mmarray.org/~iws/95CF.1002.2143.vga.rom
iws: I do get: "(II) RADEONHD(0): Unknown card detected: 0x95CF:0x1002:0x2143." in my Xorg.0.log
agd5f: iws: no display port support yet
iws: agd5f: :( do you know if radeon might work?
agd5f: iws: ajax got DP to DVI ports working as long as they were attached at boot
iws: agd5f: yeah, that's how mine is set up
agd5f: but we need to get link training sorted out for it to work on the fly
iws: agd5f: do you know if that was radeon or radeonhd that ajax used?
agd5f: iws: radeon
iws: agd5f: is that xf86-video-r128 in xorg git? I can't find one called radeon
agd5f: iws: xf86-video-ati
iws: agd5f: ah, I didn't think of that :)
iws: agd5f: I'll try building the git version, and see if that works
Zajec: agd5f: do you plan to implement scaled modes in radeoon?
agd5f: Zajec: already supported
iws: agd5f: here goes nothing
Zajec: agd5f: oh, great
iws: agd5f: so, radeon kinda works. The Xorg.0.log makes it look like everything is working, but the monitor shuts off when X starts
iws: agd5f: I can still jump back to VT1 with Ctrl+Alt+F1 without problems
iws: agd5f: X is running fine, I can open a terminal (blind) and type "find /" and see the disk drive start up
iws: I just can't see anything :)
agd5f: iws: might try the other DP port
iws: agd5f: yeah, I tried that. I moved my DP adapter to the other port and rebooted. No luck.
iws: agd5f: I've got an extra monitor, I was thinking of connecting both at the same time
agd5f: iws: looks like it needs more work
ech0s7: how can i revert the latest commit of radeon?
hatseflats: evening everyone :)
hatseflats: I have a setup with dual 3850's, and three connected monitors, one of which is connected by itself on a single card, over a DVI to HDMI cable (my plasma tv)
hatseflats: I have been running radeonhd stable in xorg zero-conf mode
hatseflats: but I can't ever seem to get the third screen to show up anywhere
hatseflats: xrandr reports of only two dvi interfaces, those of the primary card I would assume
hatseflats: anyone got pointers on where to go from here?
agd5f: hatseflats: xrandr only supports single cards at the moment
hatseflats: okay, is there another way to, for example, start a seperate X instance on that second card? Do I need to run them in crossfire-x mode, or do I not?
agd5f: hatseflats: you should be able to run zaphod style multi-head, but you'll need the radeon driver
hatseflats: that's a shame, radeonhd is the only driver I found to be stable enough-ish
hatseflats: and what exactly do you mean with zaphod style, I vaguely remember it being a type of X11 server, correct?
agd5f: hatseflats: separate instance of the driver for each head
hatseflats: I see
agd5f: dualhead pre-randr
Oskar_Linux: I wonder, how is the roadmap for the opensource RadeonHD driver this year ?
yangman: none-existent, as usual ;)
Oskar_Linux: I have some problems with the driver from ATI, therefore I'm waiting for the opensource driver. The screen went black with the ATI driver.
rehabdoll: unless you need 3d its great
Oskar_Linux: Is it good with Wine to play 3D games ?
yangman: 3d support's only available for r5xx atm
Oskar_Linux: full 3D support with r600, I have R600 (Radeon 2600 XT)
rehabdoll: nope
Oskar_Linux: I think it's hard to code the driver, I have looking at the source. And there is so many registers and 600 pages to read through.
Oskar_Linux: In the future when this driver and Wine is more mature then Linux is more ready for desktops and games
Oskar_Linux: 2 years maybe ?
rehabdoll: buy a ps3 or wii instead
rehabdoll: :)
rehabdoll: wine will always have to catch up
Oskar_Linux: wy, I use Linux to serious stuff also, DTP, math etc.
MostAwesomeDude: I'll write r600 stuff as soon as I'm finished with r300 and if there's nothing more important to take care of unless I get paid for it.
MostAwesomeDude: Or, just wait a few weeks; AMD's been working on it. :3
Oskar_Linux: I is able to write some C programs, but I wish I was better than that.
Oskar_Linux: It takes as well several years to be professional at C.
Oskar_Linux: If the Larrabee processor was finished then it was easier ?
MostAwesomeDude: Nope, the only reason that r600 3D is missing is because nobody's written it yet.
Oskar_Linux: The Larrabee processor is only software opengl/direct3D
Oskar_Linux: John Carmack was good when he wrote the engine to Quake 1, only software.
davidjheinrich_: good evening
davidjheinrich_: can anyone help me with installing radeonhd head? I'm having an error when trying to get the drm kernel modules using git checkout -b r6xx-r7xx-support origin/r6xx-r7xx-support
yangman: davidjheinrich_: what's the error?
davidjheinrich_: yangman, it says: git checkout -b r6xx-r7xx-support origin/r6xx-r7xx-support
davidjheinrich_: fatal: Not a git repository
yangman: davidjheinrich_: guide's missing the step where you cd into the cloned directory
edt: davidjheinrich_ another alternative is to download and build 2.6.30-rc1. It has the new drm builtin
rmrfslash_: Hey everyone: I just read on radeonhd:r6xx_r7xx_branch that I don't need to use a branch of radeonhd.
rmrfslash_: This is new.
rmrfslash_: Does the radeonhd driver w/ DRI enabled still conflict w/ hibernation?
rmrfslash_: I remember this was an issue several weeks ago
yangman: rmrfslash_: resume from suspend/hibernate is still broken
rmrfslash_: i c
rmrfslash_: I will continue to use the regular radeon/ati driver
rmrfslash_: this one was ok
rmrfslash_: yangman: do you know is EXA or DRI would help XRender?
rmrfslash_: is=if
rmrfslash_: XRender compositing that is.
yangman: any hardware acceleration on r6xx already requires DRI
rmrfslash_: EXA is hardware accel?
yangman: yes
rmrfslash_: This might be a terrible question, but is EXA in anyway related to OpenGL or do they serve common purposes?
yangman: unrelated
yangman: EXA is for accelerating basic 2D operations
yangman: fill, copy, etc
airlied: well Xrender
rmrfslash_: XRender is for accelerating 2D operations?
rmrfslash_: I should just read a bit more
airlied: well EXA is for accelerating Xrender operations, Xrender is a 2D accel API I suppose
rmrfslash_: I see
rmrfslash_: so OpenGL and XRender seem to be more closely related.... at least in their intent.
rmrfslash_: like an API for drawing stuff
rmrfslash_: So would I be on the right track by saying: a driver that has 3D acceleration support basically means that it has the ability to pass the execution of OpenGL functions directly to the GPU.
rmrfslash_: whereas something like Mesa 3D would use the system CPU
airlied: rmrfslash_: kind off, the gpu doesn't accelerate GL directly in most cases
airlied: rmrfslash_: mesa has hw drivers that translate GL into the GPU commands
rmrfslash_: I think I understand it a bit better now.
rmrfslash_: thanks :)
rmrfslash_: gotta go grocery shopping now
yangman: ugh. this resume issue is a stumper :\
yangman: I think I figured it out
yangman: \o/
yangman: yup, works
yangman: please give this a try: https://bugs.freedesktop.org/show_bug.cgi?id=20695#c14 r5xx folks as well, to make sure it doesn't regress anything