Nightwulf|work: hi all
loki_666_: experiencing some problems with some menu in mythtv (Xorg locks and eat 100% cpu)
loki_666_: how to help debug this?
El_Angelo: join the right channel?
El_Angelo: like.... #mythtv ?
loki_666_: well, i think I'am, since that didn't happend when i was running the fglrx driver
El_Angelo: so what are you using now?
El_Angelo: and on what hardware?
loki_666_: radeonhd r6xx-r7xx-support branch
El_Angelo: then i'm not surprised :)
loki_666_: rs780, video on Xv is fine
El_Angelo: there is like a reason it's not released?
loki_666_: yeah, and I'd like to help to debug
TASADAR-F: I have a small problem. Someone can help me??
adamk: TASADAR-F, Only if you let us know what the problem is :-)
TASADAR-F: adamk, thanks, I install de new experimental branch for r600 and r700 with DRI and XV output suport, I create a symbolic link to the driver (/usr/lib/xorg/modules/drivers//radeonhd_drv.so) but this error appear (EE) AIGLX error: r600 exports no extensions (/usr/lib/xorg/modules/drivers//radeonhd_drv.so: undefined symbol: __driDriverExtensions)
TASADAR-F: (EE) AIGLX: reverting to software rendering
adamk: TASADAR-F, the r600 branch has support for the DRM. For AIGLX, and hardware acceleration in 3D applications, you need the Mesa r600 driver...
adamk: Which is not available yet.
adamk: Basically, the r600/r700 branch gives you 2D aceceleratioin via EXA and somewhat working Xv support.
TASADAR-F: I not want AIGLX. Only want EXA and XV support. but desktop are very slow and Videos have tears
adamk: TASADAR-F, Have you enabled EXA in your xorg.conf file?
adamk: Then I don't know why it's so slow for you. Either something isn't right with your setup, or the driver is just too new to do much good for you.
TASADAR-F: I think this error (EE) AIGLX: reverting to software rendering . break direct rendering exa
TASADAR-F: My card is 4670 stable radeonHD work very good (without EXA and XV), If ERROR AIGLX is not the problem I think the problem is the experimental driver
TASADAR-F: thanks adamk
adamk: TASADAR-F, That AIGLX error is not the problem.
TASADAR-F: ok very thanks.
adamk: TASADAR-F, Go to http://pastebin.com/ and paste in your /var/log/Xorg.0.log file.
TASADAR-F: Now I have fglrx. Now I will install all other time wait 5 min.
rindolf: Hi all.
bridgman: TASADAR-F; make sure that the drm module was installed & runs OK (check your xorg log for a start), but the current EXA code is not optimized at all so will be slower than shadowfb, and the Xv code does not include tear-free logic yet
TASADAR-F: adamk, Xorg log http://pastebin.ca/1306587
bridgman: the priority for the release was getting enough info and sample code out into the development community to let 6xx/7xx driver support happen in public
TASADAR-F: bridgman, in the log all is ok. See this http://pastebin.ca/1306587 (modules work drm and radeon and not important errors)
TASADAR-F: I think is fault of driver, shadowfb (with the stable driver) work lot faster than EXA
TASADAR-F: thanks for the answers
adamk: TASADAR-F, As he said, EXA is not optimized.
bridgman: shadowfb is as fast as or faster than most hardware acceleration these days; the reason it isn't used more is that it can't coexist easily with video or 3d acceleration
bridgman: shadowfb wants the "master" frame buffer in system memory, while hw acceleration needs the master fb to be in video memory
MostAwesomeDude: Yay for fast CPU.
bridgman: yep, for some things CPUs are a great way to process
bridgman: it's just when you need kilo-mega-gigaflops that things get sluggish
libv: guys... the shedule as will be printed in the fosdem booklet this year is already available from http://wiki.x.org/wiki/fosdem2009
revx: bridgman, I use shadowfb when using xf86-video-ati and it's very fast :)
revx: but right now I'm mesing with xf86-video-radeonhd
revx: I have to go up to Kalamazoo and I'll be back later, then I'm diving into this cursor issue
revx: the cursor set hasn't changed again yet
revx: and since the cursor is loaded in the same spot each time I'm not sure how I'll track it down yet
revx: but I'll look into that later
dmb: libv, https://bugs.freedesktop.org/show_bug.cgi?id=19114 :(
dmb: libv, is there anyone you think i can assign that too?
dmb: or add to the CC list?
Zbang: hey guys...
Zbang: i have a problem playing movies on vlc/mplayer
Zbang: i keep getting X11 error: BadAlloc (insufficient resources for operation)
nanonyme: Heh, apparently testing out r6xx-r7xx-support on Debian means compiling most of your X.org again (for the moment). :) Oh, well.
Zbang: any1? help?
yangman: Zbang: GPU? driver version? x11 or xv output?
Zbang: ati x1950 rv570, radeonhd 1.3.2(i think... how to verify) and it's when trying with xv output
yangman: Zbang: driver version will be shown in Xorg.0.log. you probably mean 1.2.3
Zbang: 1.2.3 yeah
Zbang: just checked
yangman: Zbang: check the X log for anything unusual. the BadAlloc error isn't very helpful by itself
Zbang: yangman... when trying to play a video with compiz up (opensuse 11.1) then this happens...
Zbang: and no video shows... although mplayer thinks it's playing.
Zbang: when trying to play when WM isn't compiz, the video plays as if from the 3d card
yangman: Zbang: so, compiz issue?
Zbang: i've asked on compiz, and they've told me it's a drive config err...
Zbang: let me try playing with compiz off
Zbang: well it works... but i don't understant where's the problem
yangman: compositing still has issues
yangman: part of the reason why DRI isn't enabled by default yet
da1l6: maybe try updating to 1.2.4?
Zbang: da1l6, this means re-compiling kernel...?
da1l6: no, just the radeonhd driver
da1l6: it is not part of the kernel
Toksyuryel: there's a few hoops to jump through on the user-end to actually make DRI work
Toksyuryel: if they have elected to wait on that, the driver would error if DRI was enabled
Toksyuryel: and they would have to disable it manually first
yangman: Zbang: actually, pastebin your Xorg.0.log
Toksyuryel: since they are already jumping through those hoops to get it working at all, it is not a huge requirement to also ask they add the option in the driver section; compared to everything else they need to be doing, this is relatively easy
da1l6: is DRI needed for Xv on r500?
yangman: da1l6: yes
Toksyuryel: meanwhile those who can't use or would rather wait on DRI and just plug it in and go
nanonyme: Toksyuryel: I dunno, I think it's fairly simple to set it working if you've a relatively up-to-date system. :)
Zbang: yangman, http://pastebin.com/d23e76274
nanonyme: (Which I do not, Debian testing with xorg server 1.4.x >.<)
Toksyuryel: nanonyme: a lot of people don't have up-to-date systems :P
nanonyme: Simple then too though, just requires compiling loads and loads of stuff.
Toksyuryel: hoops :)
Toksyuryel: wasn't complaining, just pointing out if you're doing all of that anyway, how hard could it be to add one simple line to your xorg.conf
nanonyme: It's not.
nanonyme: Especially since your xorg log *says* you have to manually set DRI on.
Toksyuryel: yup; you're editing the conf anyway because the module isn't even loaded by default
Toksyuryel: (which amounts to just deleting a single # symbol)
Toksyuryel: and you have to edit the conf to load this driver
nanonyme: Not exactly.
nanonyme: At least here radeonhd was automatically chosen by xorg.
Toksyuryel: if you use xorgconfig you do
Toksyuryel: xorgconfig doesn't have radeonhd in it :(
nanonyme: I'd assume it's chosen because it's currently the best available default driver for r6xx and r7xx.
nanonyme: Toksyuryel: You don't need a xorg.conf on modern xorg servers. :)
Toksyuryel: you don't?
nanonyme: It will auto-detect the driver.
Toksyuryel: how do you configure it without the conf file?
nanonyme: You don't need to configure it.
nanonyme: Xorg server will automatically choose it.
nanonyme: At least for r6xx and r7xx.
nanonyme: It might choose radeon for r5xx, dunno.
yangman: Zbang: you don't have DRI enabled and running XAA. enable DRI so Xv actually works and use EXA instead
Toksyuryel: so it just automatically knows I want DRI, GLX, 1280x1024 32-bit color, knows all my monitor hsync and vsync, etc?
nanonyme: DRI and GLX modules are automatically loaded anyway, they're just not enabled for the driver.
Toksyuryel: has an old old OLD monitor and a r5xx card
nanonyme: The modeline probably is necessary for that monitor.
nanonyme: And since it's an r5xx, it's hard to say whether X server would choose radeon or radeonhd so it's best to define it.
nanonyme: (Which is why merging would be a Good Thing (tm) :P)
Zbang: yangman ? option dri is in the xorg.conf
Zbang: wait maybe it's old... i'll paste xorg.log again
yangman: it's probably old then
nanonyme: Toksyuryel: I think the current ideology calls for deprecation of xorg.conf altogether. Then if you have proper working opensource drivers, stuff just works out of the box no matter which card you junk in.
nanonyme: And mouses and keyboards are hotplugged and so on. :)
yangman: Zbang: (**) RADEONHD(0): Selected XAA 2D acceleration.
yangman: Zbang: use EXA instead. XAA isn't very well supported
mjt: heh. EXA does not work here...
Zbang: what's the option exactly on xorg.conf?
mjt: Option "AccelMethod" "exa"
yangman: Zbang: post your xorg.conf as well. I have a feeling it needs to be pruned
yangman: hm. strange how it's selecting XAA by default
yangman: Zbang: anyways, what mjt said. put that line in the Device section
Toksyuryel: what is EXA and does r5xx support it?
yangman: Zbang: I also recommend you remove the Files and Module sections entirely. it's unrelated, but they're not required anymore
yangman: Toksyuryel: it's one of the three 2D acceleration frameworks xorg supports
nanonyme: Including UXA?
Zbang: yangman thx.. i'll try
yangman: nanonyme: UXA is intel specific. it's going to be folded back into EXA anyway
nanonyme: Ah, right.
nanonyme: yangman: Just wondering what you counted the third one. I know only XAA, EXA and UXA.
yangman: nanonyme: shadowfb
Zbang: yangman, selected EXA 2D acceleration :(
mjt: is it bad?
Zbang: what's wrong with the configuration??
Zbang: well everything's slow now :(
Zbang: slow. very.
Zbang: wrong? i don't know
yangman: Zbang: post your xorg.0.log again
mjt: woops. i tried exa again here. and got 100% garbled fonts again.. lol, nothing changed ;)
nanonyme: Hmm, wasn't there some exacomposite or something like that you had to disable to get working fonts?
mjt: it has no effect
mjt: Option "NoEXAComposite"
mjt: bah. and xaa is dog slow here (shadowfb is much faster)
nanonyme: I'm mostly waiting for the patch which makes me able to start X at all with DRI on. :)
mjt: i can watch how an xterm window redraws when it becomes visible
mjt: slowly, from top to bottom
mjt: an effect like filling a bottle with water, just in reverse. nice :)
nanonyme: Sounds more like the vesa driver to me. ;)
mjt: no, vesa works very fast
mjt: significantly faster
nanonyme: Well, true, it has gotten faster. It used to be a real slug.
mjt: (i used vesa before fglrx)
agd5f: Zbang, et. al: XAA + Xv + composite generally doesn't work since XAA provides no mechanism to migrate the Xv pixmap into video ram so it can be accelerated
agd5f: use EXA
mjt: RADEONHD: version 1.2.3, built from dist of git branch master, commit ed532a70
mjt: is it right?
agd5f: the other option is to allocate scratch space and manually copy the pixmap into vram for XAA which could be done, but no one's implemented it yet
mjt: agd5f: do you know by a chance what to do with garbled fonts with exa?
mjt: (if at all ;)
agd5f: mjt: with r6xx/r7xx?
agd5f: if so I'm working on it now
agd5f: turn off exacomposite for now
mjt: funny thing i've not seen any mention of this prob
mjt: it's turned off
agd5f: Option "EXANoComposite"
mjt: i used NOEXAComposite
Zbang: agd5f... why should i need EXA/XAA?|
agd5f: check your log to make sure composite is disabled
mjt: Zbang: because you need acceleration?
Zbang: that's the engine of the 3d?
agd5f: Zbang: the underlaying accel method dicates the memory manager
mjt: not only
mjt: agd5f: http://www.corpit.ru/mjt/exa01.jpg for some laugh ;)
mjt: (that's a screen shot)
nanonyme: Interesting rendering.
mjt: not very much if you're trying to read something :)
nanonyme: Maybe I should consider myself lucky there's a bug that prevents me from starting X with DRI at all. :D It doesn't look like it'd be that much use. :)
Zbang: agd5f... now xv output works on vlc (cause of EXA i believ) but everything is slow as hell...
Zbang: any way to fix that?
agd5f: Zbang: what xserver are you using?
Zbang: what version
yangman: agd5f: < Zbang> http://pastebin.com/d5268b8a3
Zbang: check that pastebin..
agd5f: Zbang: yeah. that should be fine
agd5f: what desktop environment?
Zbang: desktop? gnome..
Zbang: windows manager? the problem is with compiz
mjt: Option "NoEXAComposite"
mjt: Option "EXANoComposite"
mjt: Option "EXAComposite" "off"
mjt: (WW) RADEONHD(0): Option "NoEXAComposite" is not used
mjt: (WW) RADEONHD(0): Option "EXANoComposite" is not used
mjt: (WW) RADEONHD(0): Option "EXAComposite" is not used
Zbang: i mean i'm trying now...
Zbang: computer stuck!
agd5f: mjt: those are EXA options. look to see if this line exists: (II) Composite (RENDER acceleration)
agd5f: (II) EXA(0): Driver registered support for the following operations:
mjt: solid, copy, uploadtoscreen, downloadfromscreen
mjt: no composite
mjt: (II) RADEONHD(0): EXA: Disabling Composite operation (RENDER acceleration)
mjt: aha. got it
Ge0rG: whats the difference between NoEXAComposite, EXANoComposite and EXAComposite off?
mjt: but effect is the same ;)
mjt: i just tried everything ;)
mjt: the real thing is EXANoComposite
nanonyme: Ge0rG: It's likely at least NoeEXAComposite and EXAComposite off are equal.
nanonyme: Without the extra e.
mjt: yeah - both are wrong :)
Ge0rG: is it possible to start a second X server with EXA on, when I have already one running without?
nanonyme: mjt: Yeah, iirc both are parsed to be exactly the same thing (which in this case would apparently be nothing).
mjt: seems not
Zbang: for me it's NoEXAComposite...
Zbang: but video is still fucked... mjt, agd5f... any ideas?
nanonyme: Zbang: That's the right one, yes.
mjt: Zbang: "it" (NoEXAComposite) = what?
mjt: Zbang: you mean, for you that option does something but EXANoComposite not?
Zbang: just option "NoEXAComposite"
mjt: and what's the effect/
Zbang: that option is loaded and recognized by xorg... no effect though
nanonyme: I'm keeping misreading, I'll leave you to deal it. Will educate myself with the logs afterwards.
mjt: Zbang: see the logfile
mjt: Zbang: xserver does not abort on unknown options
agd5f: Zbang: that option isn't relevant for you
agd5f: it'll just make things slower in your case
Zbang: agd5f... how do i make my video play fast?
mjt: run windows?
Zbang: no way!
agd5f: Zbang: are you running HD video?
mjt: "you does not exist. go away"
Zbang: by HD you mean?
mjt: high definition
Zbang: agd5f, when compiz isn't working the video is working perfectly...
Zbang: same video... same xvideo output
agd5f: Zbang: sounds like an issue with compiz then
Zbang: compiz-fusion told me that for everyone it's working ok...
Zbang: could you please help by guiding me where to look for the wrong config?
mjt: interesting. not all fonts are that garbled with exa. firefox/thunderbird works just fine.
mjt: client rendering vs server rendering, it seems
mjt: with exa, switching between firefox and thunderbird windows makes the system 100% busy for ~5..10 sec
mjt: (nothing similar with xaa)
mjt: wug. had to reset the PC as it freeze while starting X the next time. Not really freeze - it worked but not the keyboard/mouse...
squids: can anybody here help me set up my xorg.conf to run radeonhd to its full potential?
mjt: takes a wild guess: full potential is still "implementation pending"... ;)
squids: im just not sure which options to put in exa xaa??? dri??
mjt: s/takes/makes/ but anyway
mjt: try and see which works for you
squids: what all kinds of options are there so that i can try everything out?
mjt: see the manpage
kcodyjr: summary: you're not likely to get more than one feature working at a time.
mjt: and if your card is >=r6xx, you have to use r6xx_r7xx_support branch of both dri and radeonhd code from git
squids: yeah didnt know i have rs780
kcodyjr: which i've got running, stable in unaccelerated 2d mode, but lotsa stuff breaks the minute you turn some speed on - not the least of which is text rendering
mjt: that's r6xx
kcodyjr: when these guys say 'aimed at developers' they're not kidding
kcodyjr: yes, r6xx
kcodyjr: i've got an RV610
mjt: have that same rs780 here - does not work with radeonhd so far
squids: so how do know how to "git" this done
mjt: (radeon (without hd) and vesa works)
kcodyjr: if i remember what i've read right, the very latest 7xx series changed some register locations
squids: pun intended
kcodyjr: squids, patience, unless you're interested in hacking on the driver
mjt: don't try to if you dont know how to use git ;)
squids: kcodyjr: how hard can that be
squids: let me tack a hack at it
kcodyjr: hacking on the driver? depends what you want to accomplish and what your knowledge level / learning curve speed is
mjt: how hard to hack the driver you mean?
kcodyjr: for instance if you want to help with 3d, you've got a LOT of reading to do ;)
mjt: knowlege level of the hardware is very important too
mjt: (how well you know the chipset internals etc)
kcodyjr: the drivers are currently being imlemented by writing the hardware code inline with the X code, so knowing X's internals is also a big help
squids: im just kidding, partly, i would love to do it, but im just a beginner if its not obvious
kcodyjr: squids, so am i with video hardware, but that ain't stopping me ;)
squids: kcodyjr: could yall give me a tip on how to start?
kcodyjr: find something and play with it. break it and see how it breaks. put it back together again.
kcodyjr: r600_demo is probably a good place to start just to find out how deep the wter is
kcodyjr: otoh if you're listening to mp3's on linux, it's probably playing via libao into an alsa driver, which means you're running my code several hundred times per second ;)
squids: i wish i could say somthing like that
mjt: i wonder why so many ppl thinks music == mp3...
lupine_85: huggs his oggs
kcodyjr: mjt, point well taken
lupine_85: vorbis ftw
kcodyjr: in that case, i saw something broken, managed to understand what it was supposed to be doing, and, in that case, rewrote that module from pretty much scratch
squids: so say i want to start fiddling with r600_demo, where do i start
kcodyjr: and actually i've evaluated mp3, ogg/vorbis, and flac. unless you're an OSS evangelist, mp3 and ogg are coke and pepsi. if you think you need flac though, something's wrong with your audio stack. mp3 is more than capable of bringing the distortion below human perceptible levels. and i -am- an audiophile.
kcodyjr: squids, pull it down via get, and build it
kcodyjr: you'll need the latest r6xx branch's drm and x driver, and you'll want to explicitly disable accel in the x config
squids: ill try that out for my first "task"
kcodyjr: being able to get your hands on the code and run it does seem like a good place to start ;)
squids: i guess everybodys gotta start somewhere
kcodyjr: i spent a month tearing apart libao when i did that
kcodyjr: unfortunately, the -really- good result wasn't possible without changing libao's api...
_Steve_: i'm trying to setup the radeonhd driver from git r6xx-r7xx-suport branch so i can have at least 2d accel with my 4850 card.
_Steve_: i'm new to git but managed to get the branch checked out. when I run autogen.sh I get http://pastebin.com/d5113f9c
kcodyjr: _Steve_, unless you plan to hack, you aren't going to be satisfied with what percentage of it works...
_Steve_: i'm using FreeBSD, BTW. can anyone tell me what I might need to do to get it to run?
squids: so i guess kcodyjr is the designated recruiter
kcodyjr: ;) self appointed. for the moment.
_Steve_: kcodyjr: i may be willing to do that, if i have time.
kcodyjr: believe me i'd encourage anyone to hack, but bear in mind, the only thing i've done for radeonhd thus far is a bunch of compile warning fixes.
squids: kcodyjr: i guess every little bit helps
_Steve_: i can write C, so i may be able to help.
kcodyjr: all i can say is take a look at the code and see what comes to you ;)
_Steve_: heh, well, i'll have to get it to build first, or at least figure out these basics, then look at the docs a bit and learn what needs to be done and how to do it.
_Steve_: the hardware docs
kcodyjr: ati's released some, but not all, from what i understand
_Steve_: there are hardware docs or these now, right?
kcodyjr: i think the instruction set docs are out, but they're missing register docs
kcodyjr: others in here would have a more sure answer
_Steve_: *nod* ok
Ge0rG: how do I enable HDMI audio on r7xx?
kcodyjr: that should be handled entirely by alsa
kcodyjr: it just shows up as a pci audio device (at least on my RV610)
Ge0rG: yup, here too. do I need an alsa patch for that?
kcodyjr: you'll know you've found the right device when alsamixer reports exactly one control - an IEC958 on/off switch. no mixers.
kcodyjr: it's an snd_hda_intel
kcodyjr: a? an? my english teacher should shoot me for that one. ;)
mjt: everythin's snd_hda_intel nowadays :)
kcodyjr: quite a lot is, yes... and it ain't a bad chipset, but not a great one either
_Steve_: woo, installing xorg-macros helped
mjt: just like ahci in sata front ;)
kcodyjr: well, ahci is an actual standard
Ge0rG: hm. probably I've got one too many snd_hda_intel here. onboard realtek chip, onboard radeon gpu/chipset, pcie radeon card
kcodyjr: yes there are probably sound chips enabled but not connected to anything
mjt: ONE too many?
mjt: i see no single "too many" :)
kcodyjr: for example, i've got the radeon hd 2400 xt. it has hdmi on the chip, but not on the back of the card. the audio device exists but there's no way to use it.
osiris_: agd5f: are you around?
squids: ah ok i found man gittutorial, ill start here
mjt: solder some wires :)
kcodyjr: why? there isn't a tv or stereo under $30K with better audio than a xonar d2x ;)
mjt: squids: you're at the start of a long and interesting way... ;)
kcodyjr: squids, just remember the 'long' part and don't get frustrated ;)
agd5f: osiris_: yeah
squids: im not the type who can stop, even if i wanted to
osiris_: agd5f: i'm trying to run rv620 alongside with rs690
osiris_: the rs690 is the primary vga
osiris_: but there's no PCI resource with BIOS data for it
osiris_: so radeonhd cannot init it
agd5f: osiris_: I think multiple cards are currently busted on newer xservers
Ge0rG: wow. the only mixer control I have on this "ATI R6xx HDMI" is "IEC958 Mute"
_Steve_: so i've got radeonhd r6xx branch building and then it spit out something about needing pciutils and zlib stuff to build rhd_conntest
kcodyjr: Ge0rG, you've found the right one then
Ge0rG: kcodyjr: what about pcm playback?
_Steve_: so i installed pciutils and i already had zlib stuff, then ran autogen and configure again, but it still can't find them... i wonder if i have to help it find zlib...
osiris_: agd5f: I cannot run any of them if both are enabled in BIOS.
kcodyjr: Ge0rG, iec958 is the signalling type, it should be able to handle pcm as well as dolby and dts, if your display's sound hardware supports those formats
agd5f: osiris_: what do you mean if both are enabled in the bios?
osiris_: and that pretty sucks, because to be able to use rs690 I need to pull the rv620 out of machine
_Steve_: anyone know about this?
kcodyjr: _Steve_, what distro?
_Steve_: kcodyjr: not Linux, FreeBSD
_Steve_: (I could have just said "FreeBSD", but FreeBSD is not a distro of Linux...)
kcodyjr: _Steve_, no idea what to tell you. the BSD's always gave me headaches about command line option standards. ;)
osiris_: agd5f: in bios I can select the primary vga. if I select PCIE only the rv620 will be visible, if I select onboard both are visible in lspci
kcodyjr: does ports have an option to install/not install headers?
_Steve_: depends on the port, but for most, no, you install the program, you get the headers.
_Steve_: (and you always have a compiler installed...)
kcodyjr: well, that was the only instinct i had, was missing or misplaced headers...
_Steve_: (can't think of any ports that let you not install headers actually, although many let you not install docs and some let you not install examples, etc.)
kcodyjr: so what's the argument for freebsd over linux anyway?
_Steve_: how much time have you got?
kcodyjr: not trying to start a flame fest, what's the bird's eye view? ;)
osiris_: agd5f: is there any other way to read bios besides from PCI resource?
agd5f: osiris_: probably have to read the rs690 bios via the legacy vga segment since the rom is probably part of the system bios
_Steve_: kcodyjr: i like it better for many reasons
_Steve_: hmm, funny enough, the pciutils port doesn't seem to install headers...
_Steve_: just the utils and man pages
kcodyjr: sounds like that's your first culprit
adamk: _Steve_, The DRM changes have not been ported to FreeBSD.
adamk: _Steve_, Apparently there are quite a few changes in the DRM between r500 and r600/r700 and, of course, they are linux specific at the moment.
kcodyjr: thinks if you want to be on the bleeding edge, you'll have generally better luck on linux...
_Steve_: adamk: does that mean the code i'm trying to build won't work?
_Steve_: i recall seeing someone talking about porting the DRM changes the other day actually....
adamk: _Steve_, The 2D driver might build, but theres still no 2D acceleration via EXA without the DRM.
adamk: _Steve_, rnoland has been looking at it, but I don't know what progress he has made, if any.
_Steve_: yeah, i recall seeing him discuss it the other day, i can ask
adamk: _Steve_, And, just a quick FYI, EXA is not yet optimized and still slower than the shadowfb acceleration via XAA.
kcodyjr: here's an abstract question... since the drm alone can establish a full communications path to the gpu hardware, why would any PCI code be necessary?
_Steve_: so even if i get it working it will be slower?
adamk: _Steve_, Correct.
_Steve_: weak. :)
adamk: Well, the release really is for folks who want to work on the driver.
_Steve_: right, makes sense
kcodyjr: what'd i say? "for developers" is a serious statement around here ;)
osiris_: agd5f: do you have any particular function in mind? or I'd have to write one by myself?
Enverex: hrm, I got DRI working, glxinfo says direct rendering yes and it shows R300 as the OpenGL renderer, but if I try and use 3D in Wine libgl gives an error saying it fell back to software rendering. Any odeas why?
_Steve_: so the question is where to find docs on optimizing EXA...
_Steve_: adamk: thanks for the info, btw.
agd5f: osiris_: xf86ReadDomainMemory() from 0x000c0000
osiris_: agd5f: ok, will try
adamk: _Steve_, No problem.
adamk: _Steve_, Use FreeBSD'ers have to stick together :-)
agd5f: osiris_: take a look at radeon_read_bios() in radeon_bios.c
_Steve_: haha, true :)
osiris_: agd5f: one more thing. are you familiar with rhd_i2c code?
agd5f: osiris_: nope
_Steve_: adamk: do you know if the info currently available is enough info to add the EXA acceleration or if mroe is required?
osiris_: libv: ping
Enverex: I noticed you have to force 3D on the x1xxx series at the moment because it says you can't have 2D and 3D accel at the same time. Is that likely to be fixed any time soon?
agd5f: _Steve_: there is enough info
adamk: _Steve_, I know that the developers working on the drivers have access to all pertinent information via NDA.
adamk: _Steve_, Ideally the docs will be available without an NDA shortly.
_Steve_: ok. so even if i was willing to hack, it's better to not get in the way and just wait for the folks who do this to do their thing, i guess.
_Steve_: agd5f: did you mean public or under NDA?
agd5f: _Steve_: public
_Steve_: oh, great.
adamk: Ahhh, cool.
agd5f: the info in r600_demo, and the r6xx-r7xx-support branches of the drm and radeonhd
adamk: agd5f, How similar is EXA on r300 vs r600?
adamk: If that question actually makes sense.
osiris_: libv: there're possible null ptr derefs in rhdInitI2C function. in first iteration of for loop the I2CPtr var is null, and we're trying to get I2CPtr->scrnIndex in error messages
_Steve_: i've got the r6xx-r7xx-support branch checked out and built actually
agd5f: adamk: the basic idea is the same, the hw is of course different
agd5f: exa composite is basically the same pipeline setup regardless of chip. just need to adjust it to match new chip programming interface
osiris_: libv: I don't know the code enough, but I think that moving I2CPtr = xf86CreateI2CBusRec() to the top should do the trick
_Steve_: hmm, the configure wants pci/pci.h for pciutils to work, but i don't see that anywhere, even in pciutils source. i see lib/pci.h tho...
kcodyjr: where's the X11 entry into the radeonhd driver? If i wanted to pick apart the way drivers interface with X.
_Steve_: http://bugs.freedesktop.org/show_bug.cgi?id=13497 woo
Ge0rG: "aplay: set_params:918: Sample format non available" do I need some black dmix magic to make hdmi sound work? (is this channel even slightly on topic for radeon hdmi sound?)
kcodyjr: Ge0rG, try to get an ac3 sample and play that
_Steve_: this bug about pci stuff is pretty old, i guess the r6xx branch was created a while ago
_Steve_: and this change never got merged to it
_Steve_: hmm, no, it's in there... so why isn't this picking up pciutils... (and libpci)
Ge0rG: kcodyjr: probably I'm too dumb right now to let mplayer play the .ac3 file, and I don't know what other player should be able to do that
Ge0rG: "cat ac3_the_other_side_48khz.ac3 > /dev/snd/pcmC1D3p" -> write error: File descriptor in bad state
_Steve_: woo, got everything building now at least
_Steve_: ok, gotta head home, then i can play with this later...
Ge0rG: gives up on hdmi audio and puts two additional speakers on his table
osiris_: agd5f: libpciaccess-ed Xserver doesn't seem to provide that function anymore :/
Ge0rG: wow. that was like 30s of setup time after half an hour of xorg.conf, asound.conf, mplayerrc etc. fidling :<
osiris_: agd5f: will I be able to access atombios without vbios?
kcodyjr: Ge0rG, i've -never- used cat to play a file like that. mplayer should be able to do it if it's configured for ac3 passthrough.
Ge0rG: kcodyjr: it seems to mis-detect the ac3 file as mp2. probably just a bad compile ;)
Ge0rG: kcodyjr: when outputting regular sound with mplayer -ao alsa=device=hw:1.3 I get no errors, but also no hearable sound. the files in /proc/asound/HDMI/pcm3p/sub0/ indicate that there is something going on
Ge0rG: what are the perspectives of kernel mode-setting for radeon cards?
airlied: perspectives? -ENOPARSE
airlied: we have a working prototype in Fedora
airlied: we are waiting on better memory manager code and lots of testing code to push it upstream
Ge0rG: is it the one from http://jglisse.livejournal.com/852.html or an independent development?
airlied: different, it was more of a port of the current userspace, rather than a rewrite
Ge0rG: does it work for r7xx?
airlied: it will soon, I have most of the code in place just need to add r600/r700 drm support
Ge0rG: cool! what are the userland requirements (for people not running fc)
airlied: it'll be -ati driver, I have test code in a branch of my repo, and we need new mesa which I'm currently trying to work on
Ge0rG: would really really love to see a pixel-correct framebuffer on the 1920x1080 tft
Ge0rG: so if you need code testers... :>
kcodyjr: Ge0rG, what do you mean pixel-correct framebuffer
nanonyme: sytse: Hey, what did you say again that was the problem that prevented RV670 from starting with DRI? I might want to manually patch and try it out. :)
nanonyme: Being utterly bored. ^^
Ge0rG: kcodyjr: the only framebuffer I can use on my r770 is vesafb at 1280x960, nothing near the 1920x1080 panel resolution
kcodyjr: ahh. the RV610 is doing enough modesetting to get 1920x1080 right, it's just accel that's broken.
kcodyjr: what is it that AtomBios is supposed to handle?
sytse: nanonyme: I'll send a patch to the ml in a couple of minutes :)
nanonyme: Ah. ^^
nanonyme: Sorry, didn't mean to rush.
kcodyjr: could someone please explain why an X driver, radeonhd included, has any code whatsoever about the HDMI audio output? isn't it totally independent of video?
nanonyme: sytse: Hmmm, actually, I don't subscribe to any mailing list. Which one are you using and does it have an archive?
airlied: kcodyjr: HDMI audio is sent in the video stream
kcodyjr: airlied, but isn't it presented to the OS as entirely separate? or does X actually have to be running for HDMI audio out to work?
Ge0rG: kcodyjr: regarding 1920x1080, I have no idea how to activate that mode for the linux framebuffer - radeonfb has no support for r7xx
kcodyjr: Ge0rG, i wouldn't know for the linux framebuffer either, i've only been working in X
airlied: kcodyjr: the graphics driver has to set the GPU up to handle the audio
kcodyjr: but it's as simple as turning it on? the graphics driver has no further involvement?
sytse: nanonyme: no, it's actually good that you asked, I wouldn't have done it today otherwise ^_^
Ge0rG: kcodyjr: framebuffer is the main reason I'm interested in KMS. X is already working at 1920x1080 thanks to radeonhd :)
kcodyjr: ahh. i'm still content with the vga console. i'm never in it. ;)
nanonyme: Isn't KMS supposed to sort of replace the old-fashioned framebuffers?
sytse: nanonyme: http://speedy.student.utwente.nl/~sytse/drm-r600-ioremap-wc.patch
sytse: seeing as you're bored :P
Ge0rG: nanonyme: KMS should replace the mode setting in old framebuffers and X
nanonyme: Hmm. I was told it as in KMS would contain everything old-fashioned framebuffers do + modesetting.
nanonyme: *shrug* Maybe those explanations are the same, I'll start the compile.
Ge0rG: maybe :)
nanonyme: I personally try not to get too excited. It's hard not to considering what kind of a start for a year this has been and probably still will be. Not only on display driver front but also on WiFi front (that is, the opensource project for making a Broadcom-compatible firmware).
airlied: kcodyjr: I think it has to configure some link info for the auido
airlied: muich easier when drivers are all in the kernel
nanonyme: airlied: Couldn't you just use dummy X for it? :)
sytse: nanonyme: this is a patch for the r6xx-r7xx-support branch of drm btw :)
nanonyme: Yeah, noticed.
nanonyme: I had the branch already, applied nice and fine.
airlied: nanonyme: need to init the gpu so can't really use dummy anything
nanonyme: Ah. :/
nanonyme: sytse: I changed to Debian yesterday and noticed the xorg server there was a "little" too old for testing out r6xx-r7xx-support branch. Not that bad really since I managed to get the X.org compile script working a little earlier. :)
sytse: yeah, I guess you'll need x.org 1.5
sytse: most of the patch is just adding checks and debugging messages, the fix is simply changing drm_core_ioremap to drm_core_ioremap_wc
nanonyme: I think some helpful person talked about 1.5.2 dependency.
sytse: heh :)
sytse: and I changed the drm_core_ioremap_wc code; ioremap_wc really isn't 2.6.26+-only, but mostly x86-only (and only useful when PAT support is enabled), but a #define exists which allows one to check whether it's supported by the current arch :)
sytse: what's 2.6.26+ is an ioremap_wc #defined to ioremap_nocache for arches that don't support it
nanonyme: Odd. I don't seem to have libXdmcp.la.
sytse: nanonyme: hmm, that's odd indeed.. debian doesn't seem to have it
sytse: my gentoo host has it
nanonyme: sytse: Which version of libXdcmp do you have?
nanonyme: Same version for me.
sytse: it's a bit odd that libxdmcp-dev from debian doesn't contain the .la file
sytse: you should file a bug
nanonyme: Trying to get it solved on #Debian first.
sytse: I just posted a mail to the radeonhd and dri-devel mls btw
nanonyme: Nice. :)
sytse: and I should go to bed, it's 03:00 and I've had four beers ;)
nanonyme: Heh, it's 03:58 here. And yeah, night.
sytse: you're english?
sytse: in england it's one hour -earlier-
sytse: I know someone who lives in finland (and on IRC) at the moment :)
sytse: nanonyme: man, my writing actually gets -better- after a couple of beers.. so odd
sytse: at least, it looks that way, looking at the email I just wrote
nanonyme: sytse: It's a 64bit-only bug.
sytse: could be
sytse: I don't know
nanonyme: It is.
sytse: it seems to be a bug with some permission check concerning the PAT, don't know if it only hits on amd64
nanonyme: sytse: Erm, I meant the thing that prevented me from compiling libXv. ;)
nanonyme: A Debian user just got libXv compiled without problems on i386 and the .la file existed on his system.
sytse: you should file a bug for libxdmcp on bugs.debian.org
sytse: it's definitely a serious bug that the package maintainer should have a look at (I could look at it for you, and send you a working deb, but hey, it's 03:15, and I'm lazy :P)
nanonyme: Just sent a bug report mail.
nanonyme: Meh. I've been wanting to get Debian released asap and here I am reporting bugs. :P
sytse: mm, normally you should open a ticket on a bug tracking system, these things don't exist for nothing you know
nanonyme: Well, the Debian bug tracking system talked of a) some weird reportbug utility b) automatic mailing system.
nanonyme: I couldn't find a useful tracker. (except for already existing bugs)
nanonyme: Apparently the e-mail will be automatically parsed and put into the bug tracker. Somewhat traditional, I guess.
sytse: hmm, it should work just fine
sytse: sure :)
sytse: I like traditional methods :)
sytse: beats having to use some 'user-friendly' piece of crap
nanonyme: (Also one reason why I dislike the idea of slowing Lenny's release is that the more it slows down, the worse Debian lags behind in X.org versions :P)
sytse: just uses gentoy, suits me just fine
_Steve_: how do i enable EXA?
nanonyme: sytse: The what?
_Steve_: i'm trying the radeonhd driver from the r6xx-r7xx-support branch for the first time and want to enable it.
_Steve_: nanonyme: i think he means gentoo in a joking sort of way.
nanonyme: Do you have the drm kernel modules from r6xx-r7xx-support too?
_Steve_: nothing wrong with not taking your choice of OS too seriously...
_Steve_: no, i do not. i'm running freebsd btw...
_Steve_: perhaps i should see rnoland and his work on DRM bits...
nanonyme: Hmm, no idea what's the status with the BSD kernel modules.
_Steve_: i can load *A* drm module, but i doubt it's recent, it's from 7.1-release
_Steve_: let's pretend it is tho, what would i need to do then?
nanonyme: _Steve_: You'd need to compile new ones from r6xx-r7xx-support branch. drm.ko and radeon.ko
_Steve_: hmm, i have the branch checked out, not sure how to build the kernel modules tho
nanonyme: Hmm, I think I saw someone doing make drm.ko radeon.ko earlier, I personally just make the whole DRM, then left it uninstalled.
_Steve_: don't know how to make drm.ko
nanonyme: Well, if you just type ./configure && make, the files should be in bsd-core directory. Assuming BSD modules are as up-to-date as Linux ones.
nanonyme: sytse: Any idea on the matter?
sytse: nanonyme: yeah, but I would recommend against trying r6xx-r7xx-support right now if you want to see something useful :)
_Steve_: i did autogen.sh, then env PCIUTILS_CFLAGS="-I/usr/local/include -L/usr/local/lib -lpci -lz" ./configur
_Steve_: then make
sytse: (replying to your earlier message)
sytse: r6xx-r7xx-support is really devver-only atm
sytse: it works, and has working xv support, but is really slow and glitchy
sytse: is off to bed
nanonyme: _Steve_: And yeah, sytse had a point, best not to expect anything actually working at this point. :)
sytse: _Steve_: btw, to compile it, just do cd linux-core, and run make
sytse: _Steve_: and from that directory, make install as root to install the kernel modules
_Steve_: i have no bsd-core nor linux-core directories
_Steve_: perhaps my checkout is not correct?
sytse: oh, in the r6xx-r7xx-support branch of git://anongit.freedesktop.org/git/mesa/drm I meant
_Steve_: ah, i lack that. i just have that branch for radeonhd...
sytse: that's the git repository that contains the kernel modules (and libdrm plus headers)
_Steve_: i'll take a look, thanks.
sytse: using libdrm 2.4.3 plus headers is fine though, you only need the kernel modules from the r6xx-r7xx-support branch
_Steve_: libdrm-2.3.1 here...
_Steve_: guessing that won't suffice.
sytse: combine it with xf86-video-radeonhd branch r6xx-r7xx-support, use Option "DRI" "yes" and Option "AccelMethod" "exa" and you should be good to go
sytse: mm, no idea. You could definitely try, if xf86-video-radeonhd compiles fine, I guess it might be okay
_Steve_: one sec... :)
_Steve_: seems OK, and actually slightly faster than what i had
_Steve_: which was 1.2.4
sytse: and Xorg.0.log shows both DRI and EXA being enabled?
_Steve_: (II) RADEONHD(0): Unknown card detected: 0x9442:0x174B:0xE810.
_Steve_: i can pastebin the output, one sec
sytse: unknown card detected is fine
sytse: that's just a message that's been put in there to aid with support for specific video cards that break radeonhd
sytse: (WW) RADEONHD(0): RHDCSInit: No CS for R600 and up yet.
sytse: [drm] failed to load kernel module "radeon"
sytse: (EE) RADEONHD(0): RHDDRIVersionCheck: drmOpen("radeon", "pci:0000:02:00.0") failed.
sytse: (WW) RADEONHD(0): RHDDRIPreInit: Version check failed. Disabling DRI.
sytse: there's your answer..
sytse: it needs drm.ko and radeon.ko to be loaded or modprobeable
_Steve_: they're loaded
_Steve_: perhaps too old?
sytse: and you're using udev?
_Steve_: what's udev?
sytse: oh, and you need the branch r6xx-r7xx-support for the drm modules (radeon.ko) too
_Steve_: this is freebsd btw...
sytse: no idea if that's even supported, may not be
_Steve_: let me see...
nanonyme: sytse: Yet anyway. :)
nanonyme: (That's actually what I was trying to ask you earlier ;)
sytse: ah, I must have read that and then it slipped my mind ;)
_Steve_: it is late where you are, right? :)
sytse: that's the price for drinking too much beer I guess
sytse: 03:45 here
_Steve_: i have my own keg. :)
sytse: nanonyme is at 04:45 already
nanonyme: It's already getting early.
_Steve_: builds drm bits
nanonyme: Anyway, some sleep, new apartment contract, then more coding.
sytse: anyway, what you're doing is crazy too, _Steve_ ;-P
sytse: nn nanonyme
_Steve_: hmm, drm autogen bombs, but should be solvable
sytse: (WW) OS did not count PCI devices, guessing wildly
sytse: that sounds fishy :)
sytse: yeah, autogen.sh just runs autoreconf
_Steve_: gnome is installed... weird.
sytse: you only need autogen.sh/configure if you need to install libdrm and its headers btw
sytse: if you don't, you can just cd bsd-core, and make / make install from there
_Steve_: Makefile:5: *** missing separator. Stop.
_Steve_: if i do that
_Steve_: (yes, gmake)
sytse: gmake indeed..
_Steve_: oh, hah, it wants bsd make
_Steve_: build build build....
sytse: of course
sytse: it's a bsd kernel module :)
_Steve_: too much beer here... :)
sytse: damn, it's been a long time since i've meddled with anything BSD ;-)
sytse: how late is it where you are?
_Steve_: only 9:50pm
_Steve_: hmm, drm loads, but radeon bombs
sytse: ah, MST?
_Steve_: link_elf_obj: symbol r600_do_cp_stop undefined
sytse: that's a bug in the bsd driver
sytse: guess it isn't finished yet
sytse: oh wait
sytse: that should be symlinked to bsd-core/
sytse: well, maybe ;-)
_Steve_: even if the make file has SHARED= ../shared-core
sytse: (note that I'm not a radeonhd or drm developer ;P)
sytse: have a look at bsd-core/, most of the files are actually symlinks to ../shared-core/
sytse: but of course, I don't know if r600_cp.c currently compiles on bsd at all
_Steve_: i'm about to find out. :)
sytse: I see more inconsistencies
sytse: I think, at least in the current branch, the bsd-core directory is quite outdated, and doesn't work with the current version of shared-core.. so bsd support is out
sytse: should be fixable, but won't be just one or two fixes
_Steve_: it needs changes from the main branch?
sytse: it needs completely new changes
_Steve_: ok, gotcha
sytse: in other words, some bsd guy needs to look at it
_Steve_: well, i'm some bsd guy... i'll have to work on figuring it out, i guess.
sytse: if you're good with c.. :)
_Steve_: i've hacked C before, but never for hardware level stuff.
_Steve_: best i've done is some perl XS stuff to talk to openssl
_Steve_: and a bit of svgalib junk
sytse: anyway, I'm going to bed now.. it's getting -really- late
_Steve_: i have no idea what's required to make a video driver tho
kcodyjr: *sigh* what's it take to force X to -not- do a pci scan...
revx: xrandr --output DVI-I_2/analog --right-of DVI-I_1/digital
revx: err sorry