kcodyjr: /usr/include/xorg/miscstruct.h:53:20: error: pixman.h: No such file or directory
airlied: kcodyjr: wierd.
kcodyjr: yeah. that's why i try to avoid using unstable gentoo packages...
kcodyjr: and i believe it's because pixman-0.12 decided to put its includes in /usr/include/pixman-1/???.h
airlied: sounds like pkgconfig should take care of that
kcodyjr: yeah, if it was being told to in configure.ac...
kcodyjr: dammit, still missing dri2.h
airlied: oh maybe you need 1.6, I fail on when things ger merged
kcodyjr: hmm. no ebuild for that it seems.
MrCooper: agd5f: you forgot the post-release version bump ;)
airlied: anyone remember what the issues with depth readback are on radeon?
airlied: I'm guessing I need to detile my the rv200
MrCooper: oddly enough, I think marcheu is still the expert on that
agd5f: MrCooper: still an rc release. I'll post bump after I roll 6.11
MrCooper: agd5f: but now it's no longer trivially possible to recognize Git snapshots, defeating the purpose of the scheme I proposed
agd5f: ok, I can bump it
fpoibaf: Hi, it would be nice to have https://bugs.freedesktop.org/show_bug.cgi?id=19923 fixed before 6.11
MrCooper: fpoibaf: I think we'll need more eyes to look at the change, as I completely fail to see how it can make any difference for apps like Oo.o that don't use RepeatPad or RepeatReflect
MrCooper: in the worst case the change should just be reverted before release
ernstp: does tear free Xv work with rs780?
agd5f: ernstp: it should
ernstp: on the r6xx-r7xx branch that is
ernstp: agd5f: cool, gotta try!
ernstp: then I would finally be happy with my ATI card :-)
ernstp: no options, it's on by default
agd5f: you can toggle it with the XV_VSYNC attribute
ernstp: agd5f: you're working a lot on both ati/radeon and radeonhd, why is that.. ?
ernstp: agd5f: how do I toggle an attribute? :-)
agd5f: ernstp: xvattr
ossman: agd5f, how difficult is it to improve the texture uploading routines?
agd5f: ossman: what do you mean?
ossman: we discussed some improvements to the current memcpy version the other day. I wondered if it would be something I could take a stab at, or if I'd just get frustrated with insufficient documentation :)
agd5f: I implemented textured uploads and downloads over the last couple days
agd5f: only one not currently supported is packed YUV formats
ossman: that should teach me to git-pull before asking stupid questions :)
agd5f: since we have to munge the layout since we don't have a native shader for packed yet
ossman: have you written shaders for planar formats?
agd5f: ossman: yes
agd5f: packed still converts to nv12
ossman: don't you sleep or something? :)
ossman: yuv420 should cover 90% of the material out there
agd5f: all we need now is a packed shader so we can do textured uploads
ossman: are you doing DMA uploads right now?
agd5f: memcpy to gart buffer then texture from gart
ossman: ok. and the first version was memcpy to mapped vram?
ossman: why can't you do textured uploads right now for everything except packed data?
agd5f: we can and do
agd5f: downloads as well
ossman: ehm.. now I'm completely confused
agd5f: I can't do packed because it requires munging to nv12 since we don't have a native packed shader
ossman: there are three methods of getting the texture to the card, right? memcpy to vram, the gart and the texture upload thing?
agd5f: two: memcpy to mapped vram, or memcpy to gart and textured blit to vram
ossman: the way you talked about textured uploads, it sounded like textures had their own fancy dma engine with some special abilities
agd5f: nope. you can texture to/from anywhere in the card's address space so that means gart or vram
ossman: but currently there are two copies then. first to the gart space, then to the card vram?
ossman: will the kernel memory manager allow us to get rid of the first copy?
ossman: great :)
ossman: at that point it should be as good as it gets
agd5f: we'll be able to page stuff into the gart and texture from it directly
ossman: I'm seeing some performance issues with the r6xx and 720p video. the cpu isn't doing much so my next best guess is that the data shuffling is taking too long
ossman: but I'll need to retest with your updates now then
ernstp: I'm not getting anything in /dev/dri/ when I load the radeon module
ernstp: how can I debug that?
ernstp: remember it worked a few weeks ago
ernstp: the module is loaded
ernstp: dmesg doesn't say anything about radeon
ernstp: is there some option I can set to the kernel module for debug?
ernstp: not getting any /dev/dri device with the latest radeon module from r6xx-r7xx
ernstp: how can I debug that?
bridgman: have you built and installed the 6xx-7xx branch of mesa/drm ?
spstarr_work: hey bridgman
ernstp: the module loads correctly
agd5f: ernstp: you need to install the drm kernel modules from the drm, not libdrm
bridgman: do you have Option "DRI" in your xorg conf ?
bridgman: hi spstarr
ernstp: yeah, but that's a later problem
ernstp: it complains because /dev/dri is empty
ernstp: is there some option to modprobe I can use so I get debug output from the radeon module for example?
agd5f: ernstp: run dmesg after loading the modules and starting x
ernstp: agd5f: it just says [drm] Initialized drm 1.1.0 20060810
agd5f: ernstp: you son't have the new drm modules installed properly
ernstp: it loads properly... ?
bridgman: no, *some* drm loads properly (an old one ;))
chithead: ernstp: or maybe your pci id is not recognized, what does lspci -nn say about your card?
ernstp: ok, find radeon.ko, remove, install, try again
bridgman: are you getting messages about microcode loading etc ?
ernstp: oh, it installs in extra/, right, I had two versions installed
ernstp: thanks, this shoudl solve it!
ernstp: yey, card0!
ernstp: "XV_VSYNC" (range 0 to 1): (current value is 1) , yey, let's try!
ernstp: agd5f: yes, vsync on video, finally!
ernstp: agd5f: it's you and ossman, right? :-)
ernstp: let's see if I can crash it...
agd5f: ernstp: :)
ernstp: agd5f: no, 1080p working fine at least
ernstp: Xorg at 32% cpu
agd5f: ernstp: excellent
ernstp: damn fglrx, no vsync. finally the opensource-part of my ATI card came to benefit!
ernstp: agd5f: any options I can enable, tiling, dma, flipping?
agd5f: ernstp: nope
loki_666_: how to enable audio on hdmi output
loki_666_: i'm on rs780
ernstp: loki_666_: the rs780 seems popular, and problematic :-)
ernstp: loki_666_: "ati" driver from r6xx-r7xx working fine here, but I'm not using hdmi
loki_666_: ernstp, well here, video are playing way to fast
loki_666_: like if it doesnt care about the fps of the video, but play it at max speed
ernstp: loki_666_: wierd, never seen that, with any driver...
loki_666_: does ati driver requires drm kernel modules from branch r6xx-r7xx-support?
ernstp: loki_666_: yes
loki_666_: well i did install them, could it be an issue with XV_SYNC?
agd5f: loki_666_: nope
agd5f: probably something with your player
loki_666_: well mythtv internal player and mplayer does that
ernstp: loki_666_: same thing as I'm using, no problems here
ernstp: loki_666_: dmes, xlog on pastebin?
kcodyjr: airlied, package updates are done, but i'm -still- missing dri2.h
kcodyjr: hrm. changing it from "dri2.h" to "X11/extensions/dri2proto.h" may have satisfied it
endschranz: hi, after updating to mesa 7.3 , when i start ut2044 i get http://pastebin.com/m16181a7e. Does anyone something know about this?
loki_666_: ernstp, xorg.log http://dpaste.com/119074/
loki_666_: dmes says nothing
agd5f: loki_666_: log looks fine, however, you've disabled render acceleration
agd5f: Option "EXANoComposite" "true"
loki_666_: should i comment this?
agd5f: render accel should work fine now
loki_666_: ok i'll give it a shout w8
loki_666_: nope videos are still to fast
loki_666_: and X eats 100% cpu
agd5f: loki_666_: can you try a different video player?
agd5f: all Xv does is render and scale a frame handed to it. it's up to the player to keep time
loki_666_: well, i tried to different players, mythttv internal player an mplayer, w8 i'll try xine
loki_666_: agd5f, mmh xine seems to works ok, yet i dont experience this while using radeonhd
loki_666_: any idea?
agd5f: loki_666_: make sure you ahve the latest git bits
loki_666_: yep I'm sure
loki_666_: agd5f, just to be sure how to check (git pull report Already up-to-date.)
agd5f: assuming you have the right branch checked out
loki_666_: git status reports # On branch r6xx-r7xx-support
loki_666_: agd5f, by the way, when using radeon driver xrandr --addmode doesnt report BadMatch (invalid blablabla)
kmacy: is anyone running with an rv770 based card?
kmacy: my rv630 based FireGL works fine and dandy
loki_666_: how difficult would it be to port xv vsync code to radeonhd?
kmacy: but my rv770 based HD 4850 has been an insta-lockup
agd5f: kmacy: should work. are you using the latest bits?
kmacy: which driver?
kmacy: the radeonhd didn't recognize it
kmacy: the radeon claimed to recognize it but well
agd5f: kmacy: either should work
kmacy: I was running ati 6.10.0
kmacy: rnoland recommends trying ati 6.10.99
agd5f: kmacy: try that
kmacy: do you know what changed between the two (to bolster my confidence a bit)
agd5f: kmacy: I rewrote most of the output code
kmacy: uhm k
kmacy: why not bump it to 6.11?
agd5f: kmacy: I will when it's ready
agd5f: it's a release candidate
kmacy: okie dokie
kmacy: grabbing my new card to swap back in
m03sizlak: where can i find instructions on howto build a working radeon-modesetting setup
m03sizlak: and why is i915 getting all teh attention
nha: m03sizlak: to the latter point, perhaps a question of money and dedicated developers
nha: for the first point, I don't know if it's written down anywhere
chithead: m03sizlak: see http://airlied.livejournal.com/62269.html for instructions
nha: ah :)
agd5f: m03sizlak: easiet mehtod is to use fedora 10 or 11
glisse: nha: hey how are you doing ? :)
nha: glisse: I'm fine, thanks, and you?
nha: "real" (aka PhD) life is keeping me quite busy, but it's fun :)
glisse: nha: fine too, lot better since i passed my phd :)
nha: congratulations :)
nha: so what are you doing now?
glisse: was writting an article and now looking for a job
glisse: but i am coming at quite a bad time for job seek (keep getting ERESTART ;))
nha: yeah, I suppose the times could be better...
nha: what's your phd in?
glisse: nha: biology
glisse: which also kind of confuse people when i talk about computer :)
nha: hehe neat
nha: just pure biology or something computational on the side? I heard some stuff about people from biology working with computer scientists and math people on some problems
glisse: well i was doing computer simulation but on pure theoritical subject
glisse: anyway this was fun :)
glisse: from my pov this is the most important aspect :)
kmacy: agd5f: ping
rnoland: kmacy: ?
nha: glisse: yup, if you don't enjoy your work, you'll never do a really good job of it
agd5f: kmacy: pong?
kmacy: agd5f: radeon still causes a lockup
agd5f: kmacy: can you pastebin the log?
kmacy: radeonhd can be made to work if I change the options in Xorg.conf
kmacy: I'm running radeonhd on the new card now
kmacy: I was running radeon on the rv630 last time I was on
kmacy: I'll paste the log from Xorg -configure
kmacy: radeonhd will also cause the system to lock up if I try Ctrl-Alt-Backspace
kmacy: but otherwise works so long as you don't try to kill X while in X
agd5f: can you boot into single user mode and copy the log?
kmacy: you're assuming that the syncer writes it to disk
kmacy: I'll paste what I have
kmacy: but I'll try that to humor you
agd5f: kmacy: locks up before you get a log?
kmacy: it gives me a white screen which is unresponsive to input and the system no longer responds to ping
kmacy: so I would bet money that no log makes it to disk
kmacy: but to be honest
agd5f: you never know
kmacy: I haven't tried
kmacy: so I will
kmacy: agd5f: this is the -configure log:
kmacy: will bbiab
kmacy: agd5f: I appear to have gotten it
kmacy: I can also paste the radeonhd log as well if that helps
agd5f: what option did you need?
kmacy: for radeonhd?
kmacy: I needed to take out Load "radeon" and I *think* I needed either the Sync information or to remove: Virtual 2560 2560
kmacy: which I needed to make the RV630 work with my 30"
agd5f: does radeon work with the virtual stuff removed?
kmacy: not at 2560x1600
kmacy: or at least didn't with the version of X I was running previously
kmacy: I don't have a working 30" at the moment
kmacy: so I can't say
agd5f: but with your current monitor
kmacy: my current monitor is only 24" it doesn't need it
agd5f: yes, but does it make a difference if you remove it
kmacy: it only made a difference with the 30", it is currently commented out while I'm running radeonhd
kmacy: I could uncomment it and see if radeonhd works
agd5f: do modes other than 1920x1200 work?
kmacy: this is the radeonhd log:
kmacy: don't know
kmacy: haven't tried any other modes
agd5f: setting your virtual to 1024 768 for example
kmacy: can try
kmacy: will that info help you in some way?
kmacy: not inclined to do anything for the sake of intellectual curiousity alone :D
agd5f: that will limit the modes to 1024x788 or smaller
kmacy: restarting with
kmacy: Virtual 1024 768
kmacy: is radeonhd with that useful?
kmacy: or do you need me to try with radeon?
agd5f: nope, just radeon
kmacy: probably be back in 5
kmacy: agd5f: lockup
kmacy: but when I do something I think will break I have while (1) sync running as root in another virtual console :)
agd5f: kmacy: weird
agd5f: kmacy: does the other port work better?
agd5f: well, either
kmacy: radeonhd works
kmacy: radeon doesn't
kmacy: for the rv770
kmacy: for the rv630
kmacy: they both work
agd5f: or both
kmacy: I can use either driver for my rv630 based FireGL
agd5f: kmacy: you card has two connectors right?
kmacy: you mean the other DVI out
kmacy: I can try that
felipec: airlied: do you need any help testing stuff for 18397?
airlied: felipec: thats agd5f's bug :)
agd5f: felipec: we really need proper fencing
felipec: agd5f: probably, but why does it only happen with this card and smaill pixmaps?
agd5f: felipec: timing and bad luck
felipec: agd5f: how long would it take to implement fencing?
agd5f: felipec: major rework of the entire driver stack
felipec: crap, I was fearing that
agd5f: that's part of what airlied and glisse have been working on in the kms/mm trees
felipec: agd5f: cool, I thought so... is it related to iommu?
kmacy: agd5f: log for switched port: http://pastebin.com/m2b9d7bcc
agd5f: kmacy: still hangs?
kmacy: instead of making the screen go white it goes in to power save
kmacy: agd5f: yes lockup
kmacy: very much DOA
kmacy: can try this port with 1024x768
kmacy: if it is worthwhile
kmacy: don't know enough about graphics to know whether or not that is voodoo debugging
felipec: I'm on F10, so I could try with the gem branch
kmacy: agd5f: I hope the logs at least provide you with some insight
agd5f: kmacy: they look fine. very strange
kmacy: just my luck
kmacy: I had issues with the FireGL at first too a year ago
mattst88: before I really have to learn to use git properly, maybe you could take a guess where this problem lies, agd5f
mattst88: with 6.9.0, everything is fine on my DEC Alpha, DRI works too
mattst88: with 6.10.0, upon X startup, X, ssh, hald, et al all run out of memory and die
mattst88: definitely reproducible.
mattst88: on a Radeon 9800, btw.
spstarr: does his daily browse at git bits
kmacy: mattst88: I have /had the same problem
kmacy: FireGL 3650 IIRC
mattst88: interesting. Good to know it's not Alpha-specific.
kmacy: I would run X until swap started to get used
kmacy: it may be 64-bit specific
spstarr: he has been very very busy too
kmacy: so I'd have to re-start X every few days
mattst88: I'll do a git bisect later to find the exact commit soon.
spstarr: airlied: you're just obliterating code
spstarr: -524 lines + 45 lines :)
kmacy: I'm running 6.10.99 - hoping that it fixes it
kmacy: but not holding my breath
mattst88: spstarr, wrt his radeon-rewrite branch? :)
mattst88: kmacy, I experience it upon X start. Instantly. I can't even use X at all.
kmacy: I find that it is a gradual leak
kmacy: how much memory do you have?
spstarr: just think once this is ready, we're going to have very nice r1/2/3 mesa DRI1/2 drivers
mattst88: maybe we're experiencing different problems.
kmacy: I have 2GB
mattst88: kmacy, 4 GB.
kmacy: maybe different
mattst88: I'll debug it more later, bbl.
kmacy: I see Xorg gradually using more memory
kmacy: it takes a couple of days before I run out
agd5f: mattst88: I haven't noticed any problems
spstarr: notices airlied's new commit 2 mins ago
spstarr: is just anxious to test, but its not ready for that yet ;/
airlied: spstarr: its nearly ready to replace legacy
spstarr: oh really!
spstarr: airlied: what bits remain?
airlied: just fixing regressions
airlied: then i have to make it work on kms.
spstarr: good progress, explosive progress
spstarr: airlied: i guess all the design is now starting to show given the code savings
airlied: no design, just merging all the code that looks the same!!
spstarr: really? some of it looks cleaned up
spstarr: oh well, at least code base shrinks
King_InuYasha: does the new release of the radeon driver have Gallium3D support or at least FBOs?
airlied: King_InuYasha: the radeon driver has nothing to do with Gallium3D or FBOs
airlied: when we ship a new radeon driver we are shipping the X.org DDX
airlied: we don't ship 3D code with it that comes from mesa
MostAwesomeDude: King_InuYasha: It's been decided that FBO support will not be hacked into classic Mesa so you'll have to wait for Gallium support.
airlied: MostAwesomeDude: I'll probably hack it in at some point :)
airlied: but raelly I'd rather work on gallium
MostAwesomeDude: airlied: If you wanna take a look at Gallium feel free. I know exactly what's missing and it shouldn't be too hard to put in although it's a lot of code.
MostAwesomeDude: Frag shader, vert shader, vert emit (needs vert shader bits).
airlied: you know I how feel about wrtiting shaders :)
MostAwesomeDude: Well, these are different. They're not shader compilers, they're shader *assemblers*.
airlied: MostAwesomeDude: yeah I don't like doing those :), I like textures and buffers and stuff :-P
MostAwesomeDude: airlied: What a coincidence, I hate textures and miptrees.
RTFM_FTW: ah shaders are fun
mattst88: haha, cgit says airlied's most recent commit to radeon-rewrite was (-462 minutes) ago
MostAwesomeDude: Epoch fail.
MostAwesomeDude: Or at the very least somebody needs to reset their clock.
airlied: I've got too many random machines with random clocks.
MostAwesomeDude: airlied: ntpd FTW.
airlied: that involves acutally caring about the machines
airlied: I install random fedora and just hack the crap out of it, then nuke, reinstall
MostAwesomeDude: There's a great mental image that goes with that.
airlied: hmm cube textures, why do you hate me so