Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-12-09

Search This Log:

cxo: Desktop Death
spstarr: :(
airlied: ah well I'll push that tree to drm-radeon-testing and see what hpapens
airlied: anyone wanting to bleed feel free to test drm-radeon-testing when it updates in about 10 mins
airlied: its head should be
airlied: d63ca9dcdfd39d8a0cbdf3a5861e81a558852e5f
DanaG: oh yeah, random thing: topic in #ati is way out of date. Somebody should update that. =þ
DanaG: That is all.
MostAwesomeDude: DanaG: What's wrong with it?
DanaG: It says Catalyst 9.9 is the latest.
MostAwesomeDude: What should it say?
DanaG: 9.11.
MostAwesomeDude: Fix'd.
DanaG: Cool, thanks.
cxo: haha as if anything actually changed
Nightwulf|work: hi all
spstarr: i would hope our goal is to kill off fgrlx one day :)
spstarr: fglrx
spstarr: so then all those devs in fglrx can work on the floss driver
spstarr: thus giving agd5f, airlied, glisse more help also
flyback: notes "fiction" by orgy is still a kick ass song and video after all these yrs
glisse: airlied: yeah i spoted the eviction bug yesterday
glisse: for the erestart i think i was testing wrong kernel got too many tree here
airlied: oops, test the restart one more ;-)
spstarr: glisse: morning
airlied: I'm not sure my patch got it all
airlied: I think the radeon_object.h still needs more cleanups
dileX: drm-radeon-kms will go out of staging-area with 2.6.33, is that correct what I read?
airlied: most likely after the merge window closes
dileX: great
glisse: airlied: the ttm validation failing is likely not due to ttm change but to radeon change to adapt to ttm
glisse: what happen is that if it's vram i only give tt as eviction candidate and if tt is full the it fails
airlied: glisse: wierd I doubt TT is full here
glisse: btw on which computer do you see eviction failing ?
airlied: this is my 256MB VRAM r600
glisse: PCIE ?
airlied: yup on 6GB RAM machine
airlied: and I haven't had that much running
glisse: so tt is 512M
airlied: yup
glisse: shouldn't happen
glisse: i will dive into that today
airlied: can you ack the fence patch as well
cornair: drmGetDeviceNameFromFd
cornair: just FYI
airlied: cornair: need a newer libdrm
glisse: airlied: i will look into it but i think the fence patch is ok
spstarr: glisse: congrads on the job with Red Hat :)
glisse: spstarr: thx
spstarr: you're welcome :)
lordheavy: got this when running glxgears : File r700_render.c function r700TryDrawPrims line 955 Rendering was 3 commands larger than predicted size. We might overflow command buffer.
taiu: lordheavy: yeah, i know
taiu: I'm thinking http://pastebin.ca/1708084 but maybe should just revert dc0777d3e
Glooom: Привет всем! Я тут новенький, у меня вопрос, ноут инспирон 1501 + убунта 9.04, с видеодрайверами как обстановка подскажите плз
glisse: airlied: of course i can't repeat the bug
AstralStorm: so did anyone have luck reproducing this mesa fail?
AstralStorm: maybe it's just me?
dileX: which mesa fail?
AstralStorm: the one vs mplayer -vo gl:yuv=2
AstralStorm: I have no idea why would it break
AstralStorm: the fragment program is all fine
honk: noone else is using gl for playing video ;p
AstralStorm: honk: yeah yeah
AstralStorm: everyone is using xv which uses bilinear crummy scaler
honk: ^_^
AstralStorm: try playing some low res video on 1920x1200 screen
AstralStorm: bicubic with mplayer's spline would be fine
AstralStorm: or even Casull-Rom spline
honk: :}
AstralStorm: -vo gl allows picking a shader scaler
AstralStorm: and unsharp mask as well :>
honk: I'm not into playing lowres videos anyway ;p
AstralStorm: ha, ha
AstralStorm: you don't have any dvds?
AstralStorm: replaced them all with bluray already?
AstralStorm: or is all of your collection pirated? ;p
honk: well.. lemme rephrase that
AstralStorm: as an additional insult, changing resolution == bilinear scaling
AstralStorm: someone should implement a better scaler in the driver
honk: the lowres videos I play are usually of such a bad quality that it doesnt matter ;P
AstralStorm: for flat panels
AstralStorm: oh, you want some screenshot?
honk: nah, thanks =)
AstralStorm: for the *huge* difference it makes
AstralStorm: the scaler is one thing
AstralStorm: another I have is a bunch of other effects
AstralStorm: work very well
AstralStorm: (of course, a high res movie needs them far less)
AstralStorm: (or rather, a high quality codec)
honk: another I have is a bunch of other effects <-- huh?
AstralStorm: simple deblocking, dringing, temporal denoising
AstralStorm: *deringing
AstralStorm: plus scaler with sharpness filter (software!)
honk: that stuff doesnt work with xv? O.o
AstralStorm: it does
AstralStorm: but software scaler is crap
AstralStorm: too slow
honk: lol
AstralStorm: and I get to hardcode resolution
honk: we're talking low-res movies here.. cpu usage doesnt really matter ^^
AstralStorm: (for Xv to work. x11 is just sad)
AstralStorm: it does matter
AstralStorm: it matters because sometimes you want to play 720p
AstralStorm: now, the major problem is not this
AstralStorm: it's that I get to use x11 or hardcode the scaling resolution
AstralStorm: x11 is plain slow at high res (even Xv eats some important % of CPU)
honk: like I said.. I cant really make out any difference between xv and gl wrt scaling anyway ;)
AstralStorm: huge one, again
AstralStorm: read mplayer manpage
AstralStorm: the default is the same, but gl has optional shader scalers. yuv option.
AstralStorm: I mean, yuv=[2-4] and {l,c}scale=
honk: you do realize of course that the mplayer manpage is like 8k lines long :]
honk: like I said.. I cant really make out any difference between xv and gl wrt scaling anyway ;)
AstralStorm: because they're the same *BY DEFAULT*
AstralStorm: use /
AstralStorm: search for lscale, there you have it.
AstralStorm: they're missing better bicubic though
AstralStorm: someone should write a lanczos and batter bicubic
honk: I did try those settings some time ago
honk: I was never able to tell the difference :]
AstralStorm: impossible
AstralStorm: although that bicubic *is* blurry
AstralStorm: it's b-spline one
AstralStorm: try the unsharp scaler
AstralStorm: even at 0.0 strength
AstralStorm: that one is better bicubic
honk: seriously.. I cant tell the difference at all ;P
honk: though it's prolly the source material ^^
AstralStorm: impossible
honk: lol
AstralStorm: the difference is readily apparrent even on real bad material
AstralStorm: unless those are cartoons
honk: ^_^
AstralStorm: now, someone good at writing GLSL could make a better filter
AstralStorm: and port the preprocessing one too
AstralStorm: :)
AstralStorm: (pp filter)
AstralStorm: as well as deinterlacer
AstralStorm: kerndeint is far too slow to run on CPU here
AstralStorm: and I heard VDPAU can admirably run it on modern GPUs
AstralStorm: alternatively, get us VDPAU and/or OpenCL already ;p
AstralStorm: gallium could come with such a frontend...
honk: yeah, go ahead and write a gallium x264 decoder =D
honk: I dont think anyone would object :]
AstralStorm: except 1) I'm not getting paid for that 2) don't have enough time otherwise
AstralStorm: can't dip into my weekend time for that or I'd burn out
honk: pff, you're just being lazy now ^^
AstralStorm: actually, working 8-23:30 every day (with some minor exceptions like shopping) is anything but lazy
AstralStorm: and I'm still behind my schedule
honk: ^_^
Rikki-Tikki-Tavi: I use an onboard AMD 780G-Chipset with. I updated my Kernel to 2.6.32-rc8 (3D worked fine with rc6). Compriz crashed after the update. Google pointed to a Mesa Problem (though mesa was unchanged). I updated mesa from 2.7-devel to 2.8-devel. Now the system defaults to software rasterizer. Ideas? (I know support is not a right and I am very grateful for any help)
tormod: AndrewR_, I can confirm xserver git is broken (massive rendering corruption), did you report it?
Rikki-Tikki-Tavi: http://pastebin.com/d7f15955f
Rikki-Tikki-Tavi: debug from startx
AstralStorm: Rikki-Tikki-Tavi: /var/log/Xorg.0.log is more precise
AstralStorm: pastebin that
Andrew_R: tormod, no ... (time to file bug, i guess)
Rikki-Tikki-Tavi: ok
adamk: Rikki-Tikki-Tavi, mesa 2.7 to 2.8? Do you mean mesa 7.7 to 7.8?
AstralStorm: Rikki-Tikki-Tavi: and use a non-sucky pastebin
edt: Rikki. Update if you know how pull Airlie's drm-next tree on top of 2.6.32, in any case rebuild liddrm, xf86-video-ati & mesa in that order.
AstralStorm: Rikki-Tikki-Tavi: one that doesn't drop 403
Rikki-Tikki-Tavi: yes
Rikki-Tikki-Tavi: http://pastebin.com/d2a3d0c1c
Rikki-Tikki-Tavi: sorry about the 2.7
Rikki-Tikki-Tavi: 7.7
AstralStorm: yeah, there have been some changes
tormod: Andrew_R, I did not have time to revert 955b9f23 but will try tonight
edt: I am using xorg 1.7.3 here
AstralStorm: but it should still start
AstralStorm: except 3D
edt: with a 790gx chipset
adamk: Rikki-Tikki-Tavi, I don't see anything wrong with that log file. Can you pastebin the output of 'LIBGL_DEBUG=verbose glxinfo' ?
Rikki-Tikki-Tavi: on it
edt: (II) RADEON(0): [dri] Found DRI library version 1.3.0 and kernel module version 1.31.0
tormod: Andrew_R, please ask mrcooper about it if he shows up
edt: update libdrm
AstralStorm: that's one
Rikki-Tikki-Tavi: I'll update libdrm
AstralStorm: but there is another proble
AstralStorm: m
AstralStorm: it works in Coherent mode
AstralStorm: it looks like it looped infinitely
Rikki-Tikki-Tavi: here's the output of glxinfo for anyone still interested
Rikki-Tikki-Tavi: http://pastebin.com/d41286fe9
adamk: Rikki-Tikki-Tavi, I see nothing in the log file that suggests you need to update libdrm, frnakly.
adamk: frankly, even.
Rikki-Tikki-Tavi: i x
adamk: Haha... I take it back :-)
Rikki-Tikki-Tavi: i c
Rikki-Tikki-Tavi: ok
adamk: edt was right. I think libdrm needs to be updated.
Rikki-Tikki-Tavi: to git I presume?
Rikki-Tikki-Tavi: I use gentoo
adamk: Yes, most likely it needs to be from git.
edt: yes then -9999 is a good idea
edt: gentoo here too
edt: espically if you add drm-next to 2.6.32 then it libdrm wants to be git for it all to work
edt: if you add drm-next and a firmware patch IRQs work
rah: Michael Larabel's spelling and grammar is atrocious
rah: sometimes it's shockingly bad
rah: you'd expect more from someone who's bread-and-butter is writing
adamk: His reporting is often shockingly bad, too :-)
edt: edt has the same grammer & spelling issues
rah: adamk: aye
rah: that whole "who uses the VESA driver?" thing was quite condescending
adamk: Yeah, seriously, there are plenty of GPUs that don't have accelerated 2D drivers.
Rikki-Tikki-Tavi: autogen of libdrm shows this:
Rikki-Tikki-Tavi: http://pastebin.com/m10df3d04
Rikki-Tikki-Tavi: Is that ok?
adamk: Not if you're using KMS.
adamk: There should be an option to ./configure to enable the experimental radeon api.
Rikki-Tikki-Tavi: ok
adamk: Hold on a sec...
edt: firmware patch if you pull libdrm: http://pastebin.com/m19ad1db2
adamk: Here it is... --enable-radeon-experimental-api
Rikki-Tikki-Tavi: thank you
[Enrico]: Rikki-Tikki-Tavi: http://xorg.freedesktop.org/wiki/radeonBuildHowTo can be usefull :D
Andrew_R: tormod, bug 25536
tormod: Andrew_R, great, thanks
edt: strange using -9999 I get the it built with --enable-radeon-experimental-api without any package.use setting for libdrm
edt: let me check make.conf
[Enrico]: edt: that is normal
[Enrico]: edt: 9999 forces --enable-radeon-experimental-api
edt: ok
edt: Rikki I also have KMS enabled here
Rikki-Tikki-Tavi: edt: what file do I need to patch that into?
Rikki-Tikki-Tavi: the firmware patch
edt: Rikki You _only_ need that patch if you use git to update the kernel source with Airlie's latest drm-next
Rikki-Tikki-Tavi: is that necessary?
Rikki-Tikki-Tavi: for drm?
edt: it take the firmware stuff post by afd5f, and puts in into a form that the kernel will compile and build.
edt: it add IRQ support for R600/R700
Rikki-Tikki-Tavi: I see
edt: 3D will work fine with just 2.6.32 with the pulling drm-next
edt: without IRQ support
edt: notes his grammer is once again bad...
edt: with just 2.6.32 without pulling drm-next
Rikki-Tikki-Tavi: I think I'll try to first get drm working and then get to IRQ
edt: just to be less confusing
edt: Rikki good plan
Rikki-Tikki-Tavi: ok
edt: If you pull drm-next (unless they added a firmware patch like mine) the kernel will stall when KMS tries to initialize as it will not be able to find the third firmware file.
honk: stall? when I tried last it just didnt initialize kms O.o
edt: has not updated since Sunday since everything here is working without any corruption - I have had one race when starting a 3D app but it has not reoccurred so I have not yet opened a bug (I do not have all the info needed to make a good report)
Rikki-Tikki-Tavi: edt: should I install libdrm to /usr/lib/ or /usr/local/lib/?
edt: I use usr/lib
Rikki-Tikki-Tavi: ok
Rikki-Tikki-Tavi: I'll look into the differences some other time :)
Rikki-Tikki-Tavi: restarting X now
edt: Rikki gentoo has an overlay that will pull git versions of some packages if you unmask then
edt: see the xorg overlay
edt: hi ho hi ho its off to work I go
edt: later
gnu_d: Hi, what odes this means> fb: conflicting fb hw usage radeondrmfb vs VESA VGA - removing generic driver ?
gnu_d: does*
adamk: gnu_d, You shouldn't have the vesa framebuffer and radeon DRM (with KMS enabled) at the same time.
gnu_d: adamk: how to I find the module which is loaded ?
gnu_d: do I*
gnu_d: stupid keyboard
adamk: Not sure I understand. Where do you see that message?
gnu_d: adamk: with dmesg | grep -i radeon
gnu_d: adamk: afterword it says> fb0: radeondrmfb frame buffer device
adamk: Oh, to me that sounds like fbcon is simply saying that it is disabling the generic vesa framebuffer and using the radeondrmfb instead.
adamk: Other than that message, is everything working fine?
gnu_d: So the problem is there're artifacts, the video memory gets overloaded, the opengl programs are very slow, the fbcon is missing I can't find it to use it with early start i.e to load it with the kernel as a module.
adamk: Alright, let's check to see if everything is installed properly first. What's the output of 'glxinfo | grep -i render'?
gnu_d: Blender crashes and the menu text is blurred all over the screen, when I render something the image in the rendering window is blurred, after that when I close it the program crashes.
gnu_d: adamk: OpenGL renderer string: Mesa DRI R200 (RV280 5964) 20090101 x86/MMX/SSE2 TCL DRI2
gnu_d: adamk: direct rendering: Yes
adamk: OK, so that's normal. What kernel and what version of mesa?
gnu_d: adamk: I removed almost every warning and unused module/section/option from xorg.conf
gnu_d: adamk: kernel version> 2.6.31-ARCH
gnu_d: adamk: mesa version> mesa 7.6-2
gnu_d: adamk: ati-dri 7.6-2, libgl 7.6-2
adamk: Hmmm... Well, honestly, I have no experience with KMS/DRI2 on r200 hardware. I don't believe that the problems you are seeing are related to that message from the kernel. I think that's just an informational message showing that the kernel is not going to use the vesa framebuffer.
adamk: Do the problems persist if you disable KMS?
gnu_d: adamk: currently is loaded as radeon.agpmode=-1, I think, how to check which mode is on ?
gnu_d: adamk: the screen is locking if I disable it, like blank screen, no kernel SysRq are functioning too.
adamk: Previously I believe glxinfo would show the AGP speed in the renderer string. If it's not showing up, I'd guess that AGP is disabled.
adamk: So if you boot up with KMS completely disabled, the kernel just gives you a blank screen?
gnu_d: adamk: if I start a opengl program> like a 3D game, Blender or OpenOffice it locks with a blank screen
gnu_d: an*
adamk: With KMS disabled? It really sounds like you've got quite a few serious problems going on here, with or without KMS.
adamk: I'd suggest upgrading your kernel and seeing if the problems persist. Maybe try Mesa from git. Check the freedesktop bugzilla and, if you can't find matching bugs, create new ones.
gnu_d: adamk: when KMS is disabled if this is it> nomodeset, screen locking when 3D program ....
gnu_d: adamk: the problem persists since the first KMS arrived.
Rikki-Tikki-Tavi: Ok, after some fiddling about I managed to get DRM working again
Rikki-Tikki-Tavi: Thank you all
Rikki-Tikki-Tavi: GLXgears displays this: Rendering was 3 commands larger than predicted size. We might overflow command buffer.
Rikki-Tikki-Tavi: bad?
adamk: Known issue.
Rikki-Tikki-Tavi: ok
adamk: Not sure if it breaks anything.
gnu_d: adamk: and when I tried to use> Option "AGPMode" "4" in xorg.conf the X server Added a (WW) that the Option is ignored.
adamk: gnu_d, Correct, that option doesn't work with KMS.
gnu_d: adamk: if I recompile the kernel, mesa and friends with no KMS can I accomplish something ?
Rikki-Tikki-Tavi: I swiched on Compriz: Everything A-OK. My windows wobble once more ;). A big thank you to everyone who helped
adamk: gnu_d, Compiling the kernel without KMS should be the same as using 'nomodeset'
gnu_d: adamk: so I would screw up everything than.
adamk: It's possible. My first suggestion would really be to try a newer kernel.
eosie: cool, some visual errors in compiz seems to be fixed
soreau: eosie: oh hai :)
soreau: eosie: Did you end up finishing the patches you were working on?
eosie: yes, I did, they're in MAD's email
eosie: just checking what started to work in compiz
soreau: eosie: btw, what on your r300g todo list will it take to get alpha blur working?
eosie: it depends on why it doesn't work (provided it STILL doesn't work, let's see)
eosie: hm... it doesn't
taiu: hmm, are the texture resources different on r600 for VS and PS ?
soreau: eosie: Which card are you testing with there?
eosie: RV530
soreau: ok.
gnu_d: adamk: hey, I rebooted the system with nomodeset now the video memory gets corrupted.
gnu_d: adamk: (WW) RADEON(0): MC_FB_LOCATION was: 0xf7fff000 is: 0xf7fff000 (WW) RADEON(0): MC_AGP_LOCATION was: 0xffffffc0 is: 0xfc7ffc00
gnu_d: adamk: glxinfo writes in the kernel> glxinfo:1766 freeing invalid memtype fc262000-fc272000, ten or more lines.
gnu_d: adamk: now I got AGP 8x> OpenGL renderer string: Mesa DRI R200 (RV280 5964) 20090101 AGP 8x x86/MMX/SSE2 TCL
Zajec: glisse: you're really patient :)
glisse: i am not in front of bug ;)
Zajec: glisse: ok, what you mailed is simple and clear
Zajec: glisse: i also mailed with detecting disconnecting one display
Zajec: that's more problematic I'm afraid
Zajec: (described in mail)
glisse: yeah we might need to hook up in dpms_off but i need to double check what we want to do on disconnected monitor
Zajec: well, just detect it and compute clocks again :)
Zajec: btw. I'd prefer to use "clocks" in implementation :)
glisse: i think in such case dpms_off on the crtc is call
Zajec: yeah, but earlied you said you don't like my hooking dpms_off :)
glisse: anyway it would call same path: func_compute_clock which will call schedule which wil in the end lead to reclock
glisse: i would like to minimize the number of different path but sometimes it's not possible
Zajec: yeah, that should be same path
glisse: too much path leads to more possible way of things going wrong :)
Zajec: yup
Zajec: also if we want to disable PM for 2 heads I think I'll have to stick to my PM's state enum
glisse: why would we want to desactive PM ?
glisse: there is no reason for this
glisse: except for debugging purpose
Zajec: glisse: for two heads only
Zajec: glisse: we can not sync reclocking with 2 VBLANKs
Zajec: that was said many times...
Zajec: so i'm pretty sure of that...
glisse: oh, for that case we need some cleverness
Zajec: ?
Zajec: what I can imagine is waiting to 2 VBLANKs occurring at (almost?) the same time...
Zajec: do you mean that? or sth else?
Zajec: it may be tricky as VBLANKs are quite short, aren't they?
glisse: need to play with that but my idea was to blank on of the output
TBBle: And if both monitors are running the same refresh rate, then the vblanks are going to stay in the same phase with each other.
Zajec: sorry?
glisse: but this would mean flicker for the enduser on one screen
glisse: so need to discuss that
Zajec: ah, syncing to just 1 screen's VBLANK?
glisse: TBBle: not necesarily
Zajec: TBBle: that would mean we could never meet 2 VLBANKs from them both as the same time?
glisse: TBBle: vblank of one can happen in the middle of the displaying of the other
TBBle: Yeah, that's what I mean. And if one's vblank is out of sync with the other, it'll stay out of sync if they're the same refresh rate.
glisse: anyway, yeah as first approach easiest is to disable reclocking if 2 screen are connected
TBBle: I presume you can't push one output up to 61Hz or something.
adamk: So it's just not possible to compile R700_rlc.bin into the kernel? Which means that if radeon is compiled into the kernel, you *have* to use an initrd now, even if it only contains that one firmware file?
soreau: adamk: idk if its different for the r3xx fw, but I compile everything into the kernel and dont use initrd
adamk_: soreau: There's new firmware for r6xx/r7xx that won't get accepted into the kernel now.
gnu_d: adamk: I manage to make OpenOffice to start without locking the screen, but all the opengl applications lock it.
adamk_: BioTube: Thanks, I'll give that a try. What exactly does that do?
soreau: adamk: ah. that sucks
BioTube: adank_: compile outside firmware into the kernel
adamk_: Ahhh... That makes sense.
adamk_: gnu_d: Sorry, but there's really nothing more I can do for you. I stick by my previous suggestions, though.
gnu_d: adamk_: downgrading 5 versions of kernel ???
TBBle: Why won't the _rlc firmware be accepted into the kernel? Isn't it the same license as the other r6xx and r7xx blobs?
adamk_: gnu_d: My suggestion was to try the latest kernel, and even Mesa from git.
adamk_: gnu_d: I didn't say anything about downgrading the kernel
gnu_d: adamk_: my brain did, sorry
adamk: BioTube, Do you know how to specify multiple firmware blobs? Are they supposed to be separated by commas, semicolons, etc?
BioTube: i don't know
adamk: Let's try commas.
adamk: Alright, so no comma.
adamk: Well, for now I guess I can just compile in R700_rlc.bin as that's the card I'm using.
AstralStorm: btq
AstralStorm: btw
AstralStorm: why can't we have non-irq and irq code side by side?
AstralStorm: so that the speedup firmware is optional for now
AstralStorm: until AMD decides to release it under a better licence
AstralStorm: adamk: no, you can compile R700_rlc.bin into the kernel
AstralStorm: but it's not GPL-compliant then
TBBle: so don't distribute it to anyone.
AstralStorm: exactly
TBBle: Or let the FSF catch you with it on your person.
adamk: AstralStorm, Right, BioTube showed me how I can compile in R700_rlc.bin. Which is great. But I wanted to also compile in R600_rlc.bin, but can't see to figure out how to compile in both.
AstralStorm: they tend not to hunt people
AstralStorm: adamk: the option is space separated afaicr
adamk: There's probably some way of passing two firmware files to CONFIG_EXTRA_FIRMWARE but I don't know what it is.
AstralStorm: spaces.
adamk: Ahhh... That's simpler than I had expected :-)
AstralStorm: hey, it's a makefile - it is simple
adamk: I tried colons, spaces and, I believe, semi-colons :-)
adamk: s/spaces/commas/
AstralStorm: but it makes me wonder
AstralStorm: why they removed the rlc-less option
TBBle: What'd the patch look like?
AstralStorm: ?
AstralStorm: a cross between revert and a new version
adamk: Hmmm.... Alright, even with CONFIG_EXTRA_FIRMWARE pointing at the file, and recompiling the kernel, I still get a message that it can't load that firmware (and the kernel completely hangs for 30 odd seconds during the boot).
adamk: Which sounds like a bug report I saw on dri-devel recently.
AstralStorm: you have to set that FIRMWARE_DIR
AstralStorm: and then point radeon/.bin
AstralStorm: otherwise it'll fail
adamk: Let me try this again.
AstralStorm: the firmware parser could be a bit more flexible :)
AstralStorm: or the linking in part at least
honk: [12:37:00] the one vs mplayer -vo gl:yuv=2 <-- btw, what's supposed to happen when I do that?
AstralStorm: honk: movie should just play
honk: it does
AstralStorm: here, the fragment program fails to compile
honk: 1 and 5 give a b&w image, everything else just works
honk: uhh, and 6 is broken ;p
AstralStorm: I know
AstralStorm: 1, 5 and 6 aren't supposed to work
AstralStorm: but why am I getting a failure with the fragment program
AstralStorm: my card is RV670
AstralStorm: mesa is 7.8 latest git
AstralStorm: who built your mplayer btw?
honk: mhh.. r700 + a few days old mesa, libdrm, xf86-video-ati and drm-radeon-testing :}
AstralStorm: your own or?
honk: yeah, git
AstralStorm: hm
Wizzup: Works here too, r700
AstralStorm: I can't say anything about r7xx
AstralStorm: my r6xx fails though
tobydeh: hello, i have an RV710 card running on Ubuntu 9.10. How do i enable dvi out?
tobydeh: xrandr: http://pastie.org/735496/ dvi and vga are both connected to the same screen, a Sony 32" bravia
tobydeh: when i switch the screen to the dvi input i get a black screen
adamk: tobydeh, So the screen currently only displays something if the VGA port is plugged in?
tobydeh: i havent tried using one without the other
tobydeh: but when i boot they both work until gdm starts, then only the vga works
BioTube: this includes when the monitor is already set to DVI?
kdekorte_: tobydeh, are you using KMS?
tobydeh: kdekorte_: whats KMS?
tobydeh: BioTube: yeah the monitor goes blank if i boot on DVI
kdekorte_: Kernel Mode Setting
tobydeh: kdekorte_: i dont know i just installed 9.10 and im trying to avoid the prop. drivers
tobydeh: :S
kdekorte_: what kernel is 9.10?
kdekorte_: uname -r if you don't know
adamk: He's not using KMS. It's disabled by default in Ubuntu 9.10
tobydeh: its running: 2.6.31-16-generic
kdekorte_: Old kernel for r7xx support too
adamk: 2.6.31 is just fine for 2D, which is apparently what he's asking about at the moment.
kdekorte_: tobydeh, might try the Fedora 12 live cd and see how that works, the drivers are a little newer
adamk: tobydeh, I'd suggest simply trying to boot up with the VGA disconnected.
tobydeh: adamk: trying now
adamk: tobydeh, According to xrandr, the DVI port detects the screen.
kdekorte_: probably EDID infor issue from the TV over DVI
tobydeh: kdekorte_: thanks ill bear that in mind
adamk: tobydeh, So, hopefully, it will just work.
tobydeh: adamk: it boots and shows the bios info etc then goes blank before the login screen
soreau: tobydeh: Did you try looking at the settings in sys>prefs>display?
tobydeh: soreau: yeah it says both monitor are there but could get it to do anything
kdekorte_: tobydeh, can you post your /var/log/Xorg.0.log to pastebin.ca or somewhere like that
tobydeh: adamk: sorry, im talking rubbish, it doesn't finish booting with vga disconnected, when i plugged it back in it finished booting
tobydeh: kdekorte_: yep comming up
kdekorte_: use the one that you boot when both displays are connected
tobydeh: http://pastie.org/735496
tobydeh: line 17 onwards
kdekorte_: tobydeh, yeah you are going to need newer drivers or a custom mode line when using DVI
kdekorte_: personally I would try a newer xf86-video-ati driver...
tobydeh: kdekorte_: ok whats the easiet way to go about installing a newer driver?
kdekorte_: Well for Ubuntu there is the xorg-edgers packages
kdekorte_: Since this is a new install you might want to try them... worst case is a reinstall of the OS
tobydeh: so just apt xorg-edgers?
tobydeh: lets see what happens...
kdekorte_: https://launchpad.net/~xorg-edgers/+archive/ppa
tobydeh: thanks i think i get it...
tobydeh: not 100% sure
tobydeh: ok, im installing them now
tobydeh: if everything falls apart did you mention that fedora has more recent drivers?
kdekorte_: Yes, Fedora 12 is where alot of the ATI drivers went first
tobydeh: any debian based distros?
adamk: tobydeh, No, the primary developers all use F12, and commit directly to that project :-) There is a live CD for it, as kdekorte_ said above. So you could try it without having to wipe out anything. You could also try compiling an newer driver from source.
soreau: kdekorte_: So xorg-edgers does not supply a kernel..?
soreau: nor xserver-xorg-video-ati package?
tobydeh: crashed and burned!
tobydeh: hangs on the ubuntu loading screen
TBBle: You want xserver-xorg-video-radeon don't you?
tobydeh: kdekorte_: any thoughts why it might have hanged?
tobydeh: im going to download fedora 12 whilst im messing around
soreau: tobydeh: You could try booting with radeon.modeset=1 but no guarantees
tobydeh: soreau: how would i do that?
soreau: at this point, that kernel is somewhat outdated
soreau: tobydeh: Use E at the grub prompt and add that to your kernel parameters
tobydeh: ok
hifi: hm, massive font corruption on RV570
adamk: So on FreeBSD, Xorg.0.log shows: Mapped VideoRAM: 262144 kByte (128 bit DDR SDRAM) -
adamk: On Linux, dmesg shows: [drm] RAM width 64bits DDR
adamk: So is it 64 bit or 128? :-)
gnu_d: adamk_: I just upgraded the kernel to> 2.6.32-ARCH still no change.
agd5f: tobydeh: the DVI issue is fixed in xf86-video-ati git master
agd5f: taiu: predict patch seems fine for now
adamk: gnu_d, Then you probably need to start reporting bugs.
tobydeh: agd5f: great, how do i go about getting it?
agd5f: taiu: each shader as a set of texture resource regs
hifi: can I force the DDX to use shadowfb
agd5f: tobydeh: I think there are ppas you can use
tobydeh: agd5f: thanks ill do a search
BioTube: if not: git clone git://anongit.freedesktop.org/xorg/driver/xf86-video-ati
gnu_d: adamk: that won't be necessary if you tell me how to debug it ?
adamk: gnu_d, I've given you all the suggestions I can.
agd5f: hifi: Option "DRI" "false" for non-kms r6xx+
soreau: gnu_d: You should use latest kernel from drm-radeon-testing, libdrm compiled with kms support and mesa, dxx compiled against that libdrm
gnu_d: soreau: could you point me to the testing repo, please ?
hifi: agd5f: what about R5xx?
hifi: on KMS
adamk: Considering he gets lockup whenever he doesn't use KMS, I'd almost recommend getting that resolved first. It sounds like there's a lot going wrong on his machine.
hifi: I have a huge corruption which came from some update
soreau: gnu_d: repo for which component?
gnu_d: soreau: kernel + ati drm drivers
gnu_d: soreau: the thing you said
agd5f: hifi: Option "NoAccel" "TRUE" maybe
gnu_d: adamk: so every 2D programs runs with the speed of light, but with 3D active + KMS I got no 2D
gnu_d: speed.
soreau: gnu_d: For kernel, you can use: git clone git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git; cd drm-2.6; git checkout --track -b drm-radeon-testing origin/drm-radeon-testing
hifi: agd5f: got a segfault with that
soreau: gnu_d: For libdrm, mesa and ddx, see the link in the topic here
hifi: my xorg is pretty broken, evdev driver segfaults every second time I try to start x
adamk: gnu_d, Right, and you said that all your 3D applications were causing your machine to lockup when you weren't using KMS, right?
gnu_d: adamk: when no KMS 2D is great, but 3D is locking, with KMS 3D is poor, 2D has artifacts and fonts are blurred.
agd5f: glisse: is there an option to force shadowfb with kms?
lordheavy: here i've got lot of stalling .... overwise now video playing is really good
agd5f: AstralStorm: no new ucode or firmware is allowed in the kernel, the old ones got grandfathered in
soreau: agd5f: So the license doesnt matter?
agd5f: soreau: nope
glisse: agd5f: no but we could add one
agd5f: glisse: I think is we checked the NOACCEL option in RADEONIsAccelWorking() that would do the trick
agd5f: s/is/if/
glisse: yup should do the trick
kdekorte_: agd5f, can you merge 33a120e4761a661736ea64a3efc2e3831ac5600a into master, should fix bug 25514
kdekorte_: it is currently in the 7.6 branch
tobydeh: kdekorte_, adamk: ok so im running the fedora 12 live cd, and its great - picked up the correct resolution on dvi. but its overs canning slightly, i can only see half of the top panel
agd5f: kdekorte_: 7.6 gets merged into master on a regular basis
agd5f: kdekorte_: in the meaintime, you can cherry-pick that commit
kdekorte_: tobydeh, interesting, however, there are newer drivers in Fedora 12, so it is possible the fix in a newer package, but you will have to install F12 to get it
kdekorte_: agd5f, so how would I do that with git? git merge 33a120e4761a661736ea64a3efc2e3831ac5600a ?
agd5f: kdekorte_: git cherry-pick 33a120e4761a661736ea64a3efc2e3831ac5600a
kdekorte: agd5f, thanks for the git tip... worked great
rah: has anyone got the firmware files for a KMS radeon kernel module to load with debian?
Wizzup: Should I be using gallium on r600/700? (I haven't tried it.)
hifi: no, if you want anything working
Wizzup: ok
rah: :/
AstralStorm: agd5f: hmmm
AstralStorm: agd5f: new policy? how fun
tobydeh: Hello, I've just installed Fedora 12 and I'm using an ATI RV710 card. Fedora has found my TV on the DVI output but i need to reduce the overscan. How should i go about it?
rah: Dec 9 16:59:22 myrtle kernel: [drm] 2: conflicting encoders switching off 9-pin DIN-1
rah: Dec 9 16:59:22 myrtle kernel: [drm] in favor of DVI-I-1
rah: >:|
agd5f: tobydeh: turn off overscan on your tv
agd5f: rah: that dac drives both the analog portion of the DVI port and the TV connector
tobydeh: agd5f: My TV doesnt have that option?!
agd5f: tobydeh: it might have a strange name like fit or exact mode or something
agd5f: tobydeh: if, not, you'll have to adjust the modeline to underscan to compensate
tobydeh: agd5f: modelines make me cry. Could you help me with it please?
rah: agd5f: I don't know which dac you're referring to
rah: agd5f: but anyway, https://bugs.freedesktop.org/show_bug.cgi?id=25520
rah: agd5f: what you're saying would explain that
lordheavy: http://www.x.org/wiki/FAQVideoModes ?
rah: agd5f: now if only it were fixed :P
lordheavy: tobydeh: here too http://en.gentoo-wiki.com/wiki/X.Org/Modelines
tobydeh: lordheavy: thanks
adamk: oga, Does radeondrm in OpenBSD support r6xx/r7xx?
spstarr: drm: disable all the possible outputs/crtcs before entering KMS mode
spstarr: does this mean loading KMS prior to X starting does not show multi-colours on screen?
spstarr: intel's KMS switch will work if you dont have KMS on startup until X loads.
honk: oh btw.. when viewing images with "display", my x server jumps to 60ish % cpu usage and becomes unresponsive. the image flickers a bit while doing this. is there anything I can do to prevent that? ;)
spstarr: drm changes
spstarr: builds kernel
spstarr: looks to be serious IRQ changes in linus's linux-2.6
spstarr: builds master + drm-radeon-testing
spstarr: hmm
spstarr: drm-next conflicts drm-radeon-testing
spstarr: CONFLICT (content): Merge conflict in drivers/gpu/drm/drm_fb_helper.c
spstarr: thats ok
spstarr: i dont need drm-next if drm-radeon-testing has it
leifer: agd5f: i'm trying to test out git for bug 19960
leifer: agd5f: i'm on fedora 12, and i'm getting the following in xorg log
leifer: (II) Loading sub module "fb"
leifer: (II) LoadModule: "fb"
leifer: (WW) Warning, couldn't open module fb
leifer: (II) UnloadModule: "fb"
leifer: (EE) RADEON: Failed to load module "fb" (module does not exist, 0)
leifer: what i did was get the latest from git, build it, install it to /opt/xorg and add the ModulePath to my xorg.conf
leifer: "latest from git" == xf86-video-ati
agd5f: leifer: I guess it's looking for the xorg fb module in /opt/xorg
leifer: agd5f: bah, /usr/lib64
leifer: ahhh there we go, gdm lights up laptop, but not DVI yet
leifer: so sadly, no git didn't fix things for me
agd5f: leifer: can you update the bug and attach your xorg log from git?
spstarr: wow
spstarr: Linux is removing BKL!
spstarr: (Big Kernel Lock)
maligor: Do the implement the ODP? (Open Door Policy)
MostAwesomeDude: As long as it's not BFWT.
MostAwesomeDude: (Big Fuckin' Waste o' Time.)
Zajec: agd5f: could you help with macros? http://bugs.freedesktop.org/show_bug.cgi?id=24505
agd5f: Zajec: what's the problem
agd5f: all you need to do is make install
rah: spstarr: where did you hear that?
tobydeh: How can i find out what version of xf86-video-ati ubuntu 9.10 is using?
spreeuw: apt-cache
tobydeh: thanks spreeuw
maligor: as in: apt-cache show xserver-xorg-video-ati
maligor: it has some git snapshot
spstarr: *********************************WARN_ONCE*********************************
spstarr: File r700_render.c function r700TryDrawPrims line 955
spstarr: Rendering was 3 commands larger than predicted size. We might overflow command buffer.
spstarr: ***************************************************************************
spstarr: glxgears running
Zajec: agd5f: i thought some kind of make (compiling) is needed
agd5f: Zajec: nope. it's just macros
Zajec: ah, ok
spstarr: glisse / agd5f, airlied: I am seeing less IRQ stalls, but still happening
spstarr: I note: X is using 30% CPU with glxgears
spstarr: and glxgears is using 61%?!
spstarr: uh oh
spstarr: oh
spstarr: its using accel
spstarr: agd5f: I thought using accel != using CPU?
spstarr: or is it falling back to sw paths alot
rah: spstarr: where did you hear that the BKL is being removed?
spstarr: rah: see git tree from linus
spstarr: BKL is being removed in drivers all over now
spstarr: i just removed it in nvram.c
spstarr: since it didnt get removed (broke my build of kernel)
rah: spstarr: I see BKL usage being removed from specific areas, but not the BKL itself being removed
spstarr: drivers
spstarr: brb
rah: pfft
lordheavy: http://kerneltrap.org/BKL for example :-)
rah: big deal
rah: people have been removing BKL reliance for months now
lordheavy: it's not really new
rah: years, even
leifer: how do i verify the version of the radeon driver that's loaded? i thought it used to print out a git checksum thing in xorg log
tobydeh: Im trying to update the xf86-video-ati drivers on ubuntu to the git master, but ubuntu seems to have two packages... xserver-xorg-video-ati and xserver-xorg-video-radeon
tobydeh: are they the same thing?
tobydeh: should i just install the xf86-video-ati source and everything will sort itself out?
maligor: xf86-video-ati builds both
tobydeh: maligor: perfect thanks
spreeuw: libdrm, xf86-video-ati, mesa
spreeuw: in that order
spreeuw: if oyu were aiming for kms
tobydeh: spreeuw: what that aimed at me? What is kms and do i need to install all of those packages?
agd5f: spstarr: the CPU is still used to copy vertices, fill command buffers, etc.
spstarr: I see
spstarr: agd5f: im back to no-KMS for now
tobydeh: how do i build xf86-video-ati? standard ./configure?
spstarr: agd5f: the stalls are still making things sluggish
agd5f: tobydeh: ./autogen.sh --prefix=/usr or wherever you want to install it
tobydeh: agd5f: ah thanks
tobydeh: agd5f: im getting: xorg-macros version 1.3 or higher is required
agd5f: tobydeh: need to install newer macros
tobydeh: ok how do i go about satisfying all dependencies?
agd5f: tobydeh: see my instructions here: http://bugs.freedesktop.org/show_bug.cgi?id=24505
tobydeh: thanks
agd5f: tobydeh: also see the instructions in the topic of this channel
WhiteRabbit56: leifer how did you fix the modules loading problem you had earlier
tobydeh: agd5f: ok, I've install macros but the fx86- install doesnt see them? where should i install them to? the instructions dont mention it
agd5f: tobydeh: whereever you set the prefix
agd5f: when you ran autogen
tobydeh: recommended prefix?
Zajec: i'm going to buy notebook, looking for Samsung R522 series as it includes all I need for nice price
Zajec: there are models with ATI (of course) 4330 and 4650
Zajec: i know 4650 is more powerful and I can afford that but not sure about heat and power consumption
Zajec: assuming we will have power management soon, will 4330 eat less power and do less heat than 4650?
spreeuw: http://yfrog.com/iyscreen003yp I have broken smoke in spring rts
spreeuw: on present git
spreeuw: and the menu tiles have a weird blob of corruption too
spreeuw: on the positive side the water relection shaders seem to be working, looks awesome
spreeuw: http://yfrog.com/iyscreen004xp
tobydeh: agd5f: i have installed the macros with --prefix=/opt/xorg
tobydeh: but when i install fx86- with --prefix=/opt/xorg it still doest find them
tobydeh: do i need to pass --macros or something?
WhiteRabbit56: scrolling in firefox when it doesnt have focus and there is a window over the top causes the webpage to distort and i couldnt find any bugzilla reports on it so i figured id ask here if its been fixed
airlied: export ACLOCAL="aclocal -I /opt/xorg/share/aclocal"
agd5f: tobydeh: you need to adjust your ACLOCAL and PKG_CONFIG_PATH vars
tobydeh: thanks airlied worked a charm
DanaG: http://www.notebookcheck.net/Samsung-R522.16468.0.html
DanaG: doesn't look like ATI.
DanaG: oh, "series".
ossman: is it possible to feed Mesa a GLSL program and get it to spew out the if?
Zajec: DanaG: yup, there are models with ATIs
Zajec: DanaG: and as well with Intels as you noticed :)
Zajec: DanaG: do you have idea about power/heat?
DanaG: hmm, note the "this rating is bad" bit.
DanaG: I know my EliteBook can go down to 19 watts or so when on battery and in low-power mode.
Zajec: DanaG: yeah, i know... :/ will rish
Zajec: *risk
DanaG: Mine's 3650.
Zajec: how do you measure watts?
DanaG: My battery tells me. =þ
DanaG: Granted, it may be wrong, but eh, I don't have a kill-a-watt thingy.
Zajec: sysfs?
honk: 19W is quite a lot actually ;P
DanaG: It also depends on what I'm doing.
DanaG: 16 or so is the lowest I've seen, browsing the web under compiz with fglrx.
Zajec: DanaG: where do you read that?
airlied: glisse: why don't you use debugfs?
Zajec: i don't see battery class in /sys/class
airlied: printks are generally ugly
DanaG: Powertop is a nifty way.
airlied: ah for state dump might be useful
glisse: airlied: because by the time i get to debugfs the situation have change
glisse: btw just send a patch which will fix most of the eviction issues
dileX: airlied: with drm-testing-radeon I got Xorg frozen - here the trace
glisse: i still eviction fails with small gtt when evicting big bo
airlied: glisse: see dileX
glisse: but this situation was already failing before the ttm rework
airlied: did you fix that oops
tobydeh: could someone please help me update ubuntu 9.10 xf86-video-ati drives to the ones used in fedora 12?
dileX: airlied glisse: this is w/ X1300 pcie
airlied: glisse: oneliners are fine, makes it easier to fix
glisse: airlied: don't think i encouter this bugs
glisse: dileX: were you doing anythings special ? many firefox tab, 4 hd movie playing ;) ?
DanaG: ah, review of one with the ATI: http://www.notebookcheck.net/Review-Samsung-R522-Aura-Notebook.22234.0.html
dileX: glisse: 4-5 tabs in ffx and freeze came while loading a png
glisse: dileX: GPU ? AGP ? agp aperture size ? VRAM size ?
glisse: oh pcie
glisse: hhhhmmm
dileX: glisse:
airlied: glisse: you aren't trying hard enough ;-)
kdekorte: tobydeh, might try following the guide in the subject
airlied: I spent all day yesterday trying to reproduce that bug, and it happened then when I gave up trying
tobydeh: kdekorte: i did and nothing has changed after installing?
airlied: dileX: how much RAM you got?
glisse: airlied: i guess i need to leave all my geeky website with no picture and start browsing all this site with 4096x4096 picture :)
airlied: wonders ig highmem matters
kdekorte: did you do the full think or just the xf86-video-ati part?
dileX: airlied: 2GByte RAM
tobydeh: kdekorte: just the xf86-video-ati part
airlied: dileX: PAE kernel?
bjaglin: tobydeh: did you add a ModulePath in your xorg.cong?
kdekorte: did you install it in the right place? Xorg.0.log will tell you where it is looking for it... make sure you new one is there
airlied: glisse: google reader is good for helping it crap out
tobydeh: bjaglin: when i do that i get unsupported output on my screen, so i had to go to tty1 and remove the xorg.conf
tobydeh: kdekorte: i installed with --prefix=/opt/xorg is that correct?
dileX: airlied: how do I check for PAE? cat /proc/cpuinfo?
bjaglin: tobydeh: unsupported output?
glisse: airlied: ok so my 1 liner patch will likely fix dileX issue
tobydeh: bjaglin: yeah as in my tv didnt like the resolution my card was sending
glisse: what happen is that placement->num_placement is not 0 and placement->placement point to wrong ptr
DanaG: grr, lately my laptop has started making coil-whine noise.
kdekorte: tobydeh, not sure, but probably not... where is the current radeon_drv.so
glisse: due to me not copying stuff
glisse: i guess to avoid such driver issue a ttm patch is needed
glisse: will do one
kdekorte: on fedora it is in /usr/lib64/xorg/modules/drivers so I use a --prefix=/usr
kdekorte: and --libdir=/usr/lib64
Zajec: DanaG: yeah, negatives do not sound so badly
Zajec: DanaG: as I said, I'm money limited
DanaG: Anyway, check notebookreview.com, too.
DanaG: 4650 for 600 bucks, though -- pretty good.
DanaG: Though, lighted touchpad? weeeeird.
tobydeh: kdekorte: /usr/ib/xorg/models/drivers
Zajec: DanaG: lighted? what does it mean?
kdekorte: so use --prefix=/usr and that should be enough
tobydeh: ok
DanaG: Check the second link.
kdekorte: use that with ./configure
Zajec: ah, that lighted
Zajec: DanaG: yeah, weird... i guess this can be disabled in BIOS if u don't like
Zajec: DanaG: anyway it's notebook for girlfriend so she will love it... you know them ;)
dileX: airlied: /proc/cpuinfo lists in "flags" PAE
Zajec: ah, lovely, shiny ;)
airlied: dileX: yeah the kernel might not have it enabled
Zajec: DanaG: shortly tested that notebook in shop already... i quite like it
Zajec: DanaG: just not sure about GPU
dileX: airlied: you mean via CONFIG_xxx?
Zajec: DanaG: maybe I should choose less powerful 4330 for better battery life and lower temperature...
DanaG: Find benchmarks of comparable desktop cards, perhaps?
glisse: airlied: ok sent patch for ttm
airlied: dileX: yup I think its CONFIG_HIGHMEM_64G
DanaG: wonders how the ranking of 3650, 4330 / 4350, and 4650 would go.
airlied: glisse: so I need that + the oneliner?
glisse: yeah, most of my other today patches fixup others issue
dileX: hmm, 'zgrep CONFIG_HIGHMEM_64G /proc/config.gz' gives no output back
glisse: the ttm patch is mostly to allow us to catch driver stupidity
airlied: try CONFIG_HIGHMEM
chithead: DanaG: http://www.tomshardware.com/charts/gaming-graphics-cards-charts-q3-2009-mainstream-quality/Sum-of-FPS-Benchmarks-Totals,1503.html
glisse: if driver doesn't screw then the ttm patch is useless
DanaG: tomshardware is not extremely reputable, in my mind.
dileX: # zgrep CONFIG_HIGHMEM /proc/config.gz
dileX: # CONFIG_HIGHMEM64G is not set
chithead: DanaG: they may be not, but you will get a general idea
glisse: maybe we can add a check after the driver callback and add system ram if driver didn't set any eviction placement
glisse: airlied: so for eviction failure in front of small gtt i think we need to have this scratch space
glisse: will try to do a patch tomorrow
glisse: will likely use the ibpool as scratch space
DanaG: 4350 < 3650.
DanaG: hmm.
DanaG: Also: consider screen resolution, use of compiz (if any), and video RAM.
kdekorte: The 3650 is a pretty decent chip
airlied: glisse: ah yes ibpool is good plan
tobydeh: kdekorte: how can i check it has picked up the new install?
airlied: do 16k at a time transfer
DanaG: Coming from a GeForce Go 7600, the 3650 is badass.
airlied: glisse: for migrate to VRAM use memcpy,wc should be fas tenoough
kdekorte: tobydeh, the version in Xorg.0.log of radeon might help clue you in
tobydeh: kdekorte: thanks ill have a look
kdekorte: tobydeh, the version mine shows is 6.12.99
kdekorte: you can also check the data time stamp on the module xorg says it has loaded
tobydeh: ubuntu 9.10 default? or master?
kdekorte: tobydeh, I'm using git master
chithead: 6.12.99 is a git snapshot which distributors package for lack of kms capable release
kdekorte: chithead, but when you build from git, you always get that version
chithead: indeed
glisse: airlied: yeah but i am thinking to the unmappable vram too :)
kdekorte: chithead, sometimes a build date code/git version would be nice somewhere in there
airlied: ah yes
glisse: anyway i will look into that tomorrow then i need to shrink down my agp todo list
DanaG: http://www.driverheaven.net/reviews.php?reviewid=729
DanaG: wishes he had a kill-a-watt on hand.
tobydeh: kdekorte: im so confused, both ubuntu 9.10 and fedora 12 use version 6.12.99 but ubuntu doesnt recognise my tv on dvi and fedora does??
tobydeh: kdekorte: ive installed the master as it says in the instuctions but nothing seems to have changed
tobydeh: :(
DanaG: I'd say, go with 4650... it's more future-proof, in terms of performance.
kdekorte: tobydeh, well Fedora uses KMS and Ubuntu doesn't that may be the difference
kdekorte: KMS uses the kernel to set the video modes and it might be better at it
DanaG: ah, best review: http://www.sysopt.com/features/graphics/article.php/3778251
tobydeh: kdekorte: does any debian distro use KMS?
tobydeh: kdekorte: if not is it difficult to install?
Zajec: DanaG: sure, 4650 is better than 4330... but it's hard to know how much power it eats in notebook I am interested in
tobydeh: kdekorte: i installed fedora12 which was fine but i couldnt get mythtv installed
glisse: Zajec: maybe if you can find review of desktop equivalent you can get power consumption figure btw this 2 gpus
glisse: it's often easier to find review of normal desktop gpu board
spreeuw: DanaG: dunno but get a 4670 ;p very speedy card
DanaG: http://www.tomshardware.com/news/ATI-Radeon-4350-4550,6432.html
glisse: than btw laptop GPU
spreeuw: 1GB memory
DanaG: Weirdest desktop video card I've seen: mobility 9600... on a desktop board. I mean... why bother?
DanaG: well, weirdest almost-normal, as opposed to the 5x-voodoo2 sort of "weird".
spreeuw: nice card
spreeuw: as igp?
spreeuw: sounds very compatible and mature
spreeuw: and cool
airlied: DanaG: lots of mobility chips get used on desktop cards
DanaG: http://www.newegg.com/Product/Product.aspx?Item=N82E16814153028&cm_re=mobility_9600-_-14-153-028-_-Product
airlied: the mobility bit doesn't mean what you think
DanaG: It means you need to use mobility modder, for one. =þ
DanaG: Do desktop 4xxx cards have PowerPlay?
airlied: mobility is mostly just a marketing term
airlied: it used to be used to distinguish whether LVDS was available
DanaG: I remember Mobility 9800 used to be based on an X800... wasn't it?
DanaG: Or something like that.
DanaG: http://forum.beyond3d.com/showthread.php?p=1212337
tobydeh: kdekorte: is their a faster alternative to fedora that still uses kms?
Zajec: according to http://www.tomshardware.com/news/ATI-Radeon-4350-4550,6432.html 4650 eats much more power
bjaglin: is there any patch available i could try for the suspend/resume broken on RV250 - https://bugzilla.redhat.com/show_bug.cgi?id=531825 ?
DanaG: hmm, another thing to consider: what will be run on it?
kdekorte: tobydeh, the xorg-edgers or follow the full guide in the subject
Zajec: DanaG: no often 3D
Zajec: DanaG: kde4 effects for every day usage
Zajec: DanaG: and less often 3D games
DanaG: I've found kwin, specifically, way slower than compiz on my 3650 (though using fglrx).
DanaG: And I mean like 5 FPS versus 45 FPS.
Zajec: DanaG: yeah, but compiz does not integrate really great with KDE4
Zajec: DanaG: i just won't enable all effects ;)
DanaG: For me, even just drawing a desktop passively, is slow on kwin.
DanaG: And higher GPU usage.
Zajec: maybe i'll compare to compiz
Zajec: for now I've to choose GPU :
Zajec: :)
Zajec: i'd like to get more powerful 4650... it's described as quite better then 4330
DanaG: What was the screen resolution, again?
Zajec: 13...x768
Zajec: and plugging 1920x1080 for movies :)
spreeuw: Zajec: yeah better get a fast chippy
spreeuw: I like to get as fast as passively cooled
spreeuw: so got a 4670
Zajec: it's for notebook
Zajec: so don't want to get this too hot
DanaG: has 1920x1200 on his 15" laptop directly. =þ
Zajec: but hopefully it won't with PM implemented
DanaG: But no blue-ray drive.
Zajec: DanaG: i wouldn't like to use that res on 15'' :0
DanaG: And I _refuse_ to call it "blu" like a dumbass marketer.
DanaG: Actually, in Linux, it's really nice... very few things break, compared to Windows.
DanaG: REally tiny pixels == awesomely smooth fonts.
Zajec: DanaG: forcing DPI I guess? :)
Zajec: normally you'd get really small fonts :)
DanaG: Nope, no forcing, as long as EDID is correct... it knows what DPI it is.
soreau: spies eosie's mesa patches
spreeuw: lol wish adobe flash wasnt taking my browser down every frikking hour
Zajec: i got some problems with font size... maybe broken edid
yangman: some DEs have an override DPI
DanaG: Or you can add displaysize in xorg.conf.
WhiteRabbit56: I just followed the radeonBuildHowTo and I get X to start
WhiteRabbit56: but no direct rendering and my Xorg.0.log is giving me this error
WhiteRabbit56: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch.
WhiteRabbit56: [dri] radeon kernel module version is 2.0.0 but version 1.17.0 or newer is needed.
WhiteRabbit56: [dri] Disabling DRI.
DanaG: "or newer" is still badly worded.
WhiteRabbit56: meaning?
DanaG: Meaning, one is 2.0.x and one is 1.17.x -- one was built with KMS, the other was not.
DanaG: Or something like that.
ossman: what does it mean when an extension is prefixed by GL_MESAX_ instead of GL_ARB_?
AstralStorm: is that KMS going to white instead of blanking bug known?
AstralStorm: LVDS light is not disabled on blank
AstralStorm: only LCD
dileX: WhiteRabbit56: recompile libdrm with --enable-radeon-experimental-api
rah: Dec 9 20:11:06 myrtle kernel: radeon 0000:01:00.0: object_init failed for (15974400, 0x00060004)
rah: Dec 9 20:11:06 myrtle kernel: [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (15974400, 4, 4096)
rah: :/
WhiteRabbit56: thanx dileX
rah: and also
rah: Dec 9 20:54:06 myrtle kernel: [drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation !
dileX: airlied: just digging deeper in PAE: looking into arch/x86/Kconfig. to have PAE-support (NX-bit) I have to enable HIGHMEM64G - this is independent of installed RAM. found this older posting .
AstralStorm: dileX: PAE has nothing to do with NX
AstralStorm: NX can be (and is) used regardless of PAE
mjg59: AstralStorm: 32-bit x86 requires PAE in order to have NX
AstralStorm: wrong.
AstralStorm: it doesn't at all
AstralStorm: only has to be set in the compatibility mode
AstralStorm: windows for instance doesn't use PAE at all
AstralStorm: yet uses NX bit
dileX: here an extract from arch/x86/Kconfig:
soreau: eosie: Fisheye mode now works, and I must say it is very responsive (much more so than when it used to work with classic mesa) :D
mjg59: AstralStorm: It's bit 63 in the page table entries. How do you get it without PAE?
AstralStorm: ...
AstralStorm: PAT is independent of PAE
AstralStorm: PAE allows for extra level of page tables
AstralStorm: just that.
soreau: eosie: but mipmaps for Expo are still broken
airlied: AstralStorm: NX requires PAE
chithead: on 32 bit
airlied: yse, 64-bit sort of doesn't need PAE
airlied: since its got 64-bits
ossman: airlied, what's with the GL_MESAX_ prefix for the texture_float extension?
airlied: so having 36-bits would be useless
AstralStorm: airlied: you can cheat
AstralStorm: and put 64-bit cpu in 32-bit compat mode
ossman: Just different enough to not be compatible?
AstralStorm: to have NX w/o PAE :)
MostAwesomeDude: ossman: That's the idea.
AstralStorm: NX uses then one of reserved bits in page table
airlied: AstralStorm: 32-bit compat mode means you need PAE for NX from what I rememer
AstralStorm: not.
soreau: eosie: and alpha blur looks the same (still not working). this is on my rv350
airlied: we ship execshield for thos reasnos
AstralStorm: it uses the same old PAT
AstralStorm: there were 2 reserved bits
AstralStorm: one of them became NX
ossman: MostAwesomeDude, so what's the difference? the MESAX variant isn't in the ARB registry
WhiteRabbit56: i recompiled libdrm but im still getting that error message in my xorg log
MostAwesomeDude: ossman: That's also the idea. :3
dileX: my question was to clarify how to set PAE - I was irritated by the HIGHMEM*64G*. as far as I understood there speaks nothing against to activate PAE on a 2GByte-RAM machine?
airlied: AstralStorm: thats not 32-bit compliant though
DanaG: I think you need to rebuild the display driver against the new libdrm.
AstralStorm: airlied: it is
AstralStorm: the bit was reaserved
AstralStorm: *reserved
airlied: MESAX generally is experinemental mesa extension I think
WhiteRabbit56: alright ill try that
ossman: MostAwesomeDude, you trying to be dark and mysterious today? :)
AstralStorm: meaning "don't write"
AstralStorm: now it became unreserved
airlied: AstralStorm: reserved means don't use it
MostAwesomeDude: ossman: Think of it as a testing-the-waters approach; in order to support that extension, we obviously need a floating-point setup throughout the entire lib.
AstralStorm: airlied: which means still compliant
AstralStorm: you detect if the CPU has NX
airlied: it doesn't mean pretend its NX
AstralStorm: then that reserved bit means NX
AstralStorm: :)
AstralStorm: just that.
MostAwesomeDude: Which means that if, for some reason, somebody *happened* to add ARB_texture_float, then it wouldn't be as big of a deal.
ossman: MostAwesomeDude, it's fairly new and/or ignored part of mesa?
MostAwesomeDude: ossman: What's big, green, and likes to toss around stupid graphics patents? :3
ossman: I see
MostAwesomeDude: Or, in this particular case, it might be SGI; can't remember.
airlied: AstralStorm: NX is bit 63, I fail to see how you can fit that into a 32-bit pte entry
AstralStorm: airlied: it is. in 64-bit mode
AstralStorm: in 32-bit mode, it's one of the reserved bits
AstralStorm: I fail to remember which of the two
dileX: airlied: so I can enable PAE via CONFIG_HIGHMEM_64G on a 2GByte-RAM machine?
AstralStorm: with PAE, it's done cleanly
airlied: wierd I don't see that documentd anywhere
AstralStorm: airlied: it's only in CPU manuals
AstralStorm: it's not "the correct" way to use NX
ossman: MostAwesomeDude, xbmc uses a floating point texture to do bicubic interpolation, which then did not work on top of mesa
mjg59: dileX: Yes
kdekorte: Any idea when GLSL is going to be enabled by default... I've been running with it enabled for a couple of days and it seems ok
dileX: mjg59: thx
AstralStorm: kdekorte: it isn't?
AstralStorm: I need some option?
MostAwesomeDude: ossman: Ugh. Is it not possible for them to do 16-bit?
AstralStorm: that might be why fragment_programs fail for me
ossman: MostAwesomeDude, no idea. I've only briefly looked at it so far.
ossman: It also uses GLSL, even though it has no control flow
kdekorte: I have to edit r600_context.c and remove the comment for the GLSL_TEST define
ossman: so I'm thinking I could redo it with ARB programs
MostAwesomeDude: ossman: Well, the HW can do it in many cases (including ours), there's just not the corresponding support in core MESA.
AstralStorm: ossman: you can steal mplayer fragment programs
AstralStorm: they have a cheap B-spline bicubic
dileX: airlied mjg59 AstralStorm: which advantages gives activated PAE for radeon OSS driver :-)? (to come to a conclusion)
AstralStorm: dileX: none.
mjg59: dileX: None
ossman: MostAwesomeDude, isn't there fallbacks that need to be able to grok the texture format though?
mjg59: dileX: It'll be slower
airlied: dileX: I just wondered about it being enabled to help glisse find crash
ossman: AstralStorm, I'll have to have a look at that. thanks
AstralStorm: PAE is pure performance loss (compared to 64-bit)
AstralStorm: I recommend using a 64-bit kernel instead
AstralStorm: (it can run 32-bit userland)
dileX: OK, thank you guys for clarifications
MostAwesomeDude: ossman: Right, so it's not going to be feasible without Mesa support.
ossman: MostAwesomeDude, IOW, "fixing" xbmc is probably easier...
soreau: eosie: neverball: radeon_cs_gem.c:121: cs_gem_write_reloc: Assertion `bo->space_accounted' failed. <-- neverball fails at runtime
kdekorte: AstralStorm, GLSL is enabled if glxinfo says you are using OpenGL 2.0. If it says 1.5 it is not
mjg59: AstralStorm: Uh. What makes you think dilex has a 64-bit CPU?
Wizzup: Is 2.0 available yet?
AstralStorm: kdekorte: so?
AstralStorm: kdekorte: ARG_fragment_program is not 2.0
AstralStorm: *ARB
MostAwesomeDude: ossman: IOW, yes, although be prepared as XBMC does *not* follow good GL practices.
kdekorte: I'm just saying that is the difference I see when GLSL_TEST is uncommented... quick way to tell
MostAwesomeDude: e.g. card-specific quirks.
AstralStorm: kdekorte: still, my mplayer fails for some reason
ossman: MostAwesomeDude, they come from a Windows background. you've got to give them some slack ;)
AstralStorm: the fragment program fails to validate in mesa
AstralStorm: despite looking ok
dileX: mjg59 AstralStorm: I ran a Debian/sid 32bit host and have an Intel Core2Duo T7200 CPU.
AstralStorm: and working before
AstralStorm: dileX: doesn't matter
kdekorte: AstralStorm, what vo are you using?
AstralStorm: you can run 64-bit kernel with 32-bit compat.
AstralStorm: kdekorte: RV670, mplayer some SVN. I've filed a bug on fd.o already
AstralStorm: latest mesa (or close to it)
kdekorte: I have mplayer SVN and both gl, gl2 and xv all work for me
AstralStorm: -vo gl:yuv=2
kdekorte: I'm using mesa as of a few mins ago
dileX: AstralStorm: me thinks "real" 64bit-host is the right way :-)
soreau: eosie: OA is a bit laggy but playable with seemingly all visual features in tact
AstralStorm: that will enable fragment program (for yuv conversion, lscale and cscale affect fragment program scaling)
AstralStorm: dileX: yes, it might be slightly better
AstralStorm: dileX: but there is a cost to those 64-bit pointers
MostAwesomeDude: ossman: If you wanted to make a positive contribution, you could rip out their fglrx-specific code and make it respect GL extensions.
kdekorte: AstralStorm, using a vo of gl:yuv=2 works for me
AstralStorm: if you don't need >2 GB space for an app, there's no reason to run 64-bit right now
AstralStorm: except the kernel
AstralStorm: kdekorte: RV670?
kdekorte: rv635
AstralStorm: hmm
AstralStorm: why then doesn't it work for me
AstralStorm: shader fails to validate
ossman: MostAwesomeDude, all the code I've checked so far is very good at checking extensions. No driver specific hacks yet.
AstralStorm: try with -msglevel vo=7
kdekorte: using drm-radeon-testing branch of airlied's drm-next kernel, libdrm, mesa and ddx all from git
AstralStorm: that will dump the shader
AstralStorm: pastebin it to me
AstralStorm: kdekorte: same, same and same.
MostAwesomeDude: ossman: There's at least ARB_npot/fglrx hax in there IIRC.
MostAwesomeDude: ossman: Could be that somebody already got to it. :3
DanaG: I went from 32-bit to 64-bit, and really didn't see much change, aside from things such as Flash suckage, and such.
AstralStorm: DanaG: flash doesn't suck anymore
AstralStorm: there's a 64-bit version
AstralStorm: same of Java.
ossman: MostAwesomeDude, possibly. There is a LinuxRenderedATI.cpp, but I haven't had any reason to poke around in that one yet
spreeuw: flash still crashes alot here
kdekorte: AstralStorm, http://pastebin.com/m16318915
DanaG: yeah, but going from nspluginwrapped flash to non-wrapped flash means that you can no longer kill just Flash.
AstralStorm: (now, it doesn't suck more than 32-bit one. that doesn't mean it doesn't suck at all)
DanaG: Instead of taking down npviewer.bin, flash crashes Firefox as a whole.
DanaG: In fact, when I was on 32-bit, I even used nspluginwrapper there, because it provided that layer of isolation.
AstralStorm: kdekorte: use a non-suck pastebin
AstralStorm: kdekorte: got a 403
AstralStorm: I recommend dpaste.com
Wizzup: Tried gnash?
Wizzup: AstralStorm: Works here
AstralStorm: Wizzup: I know
ossman: MostAwesomeDude, time for some sleep now. I'll see if I can get rid of the ARB_texture_float dependency in the coming days...
AstralStorm: Wizzup: it bans all tor nodes
AstralStorm: Wizzup: I'm running one
Wizzup: aha,didn't know
kdekorte: AstralStorm, http://dpaste.com/131423/
soreau: eosie: enemy territory doesn't look much different - still too messed up and slow to be playable
AstralStorm: kdekorte: could you try with a video thats >512x384
AstralStorm: e.g. 640xsomething
soreau: but at least fisheye works and is very fast!
Wizzup: big buck bunny? :)
AstralStorm: so that -vo gl is to allocate 1024x512 texture
kdekorte: AstralStorm, http://dpaste.com/131426/ 848x480
AstralStorm: weird, why then does it fail here...
AstralStorm: kdekorte: mesa from master, right?
kdekorte: Elephant Dream at 1920 barfs
kdekorte: eys
Wizzup: wild guess: tried reinstalling/compiling mplayer?
AstralStorm: kdekorte: you can check my pastebin in the bug report
AstralStorm: Wizzup: of course I did
kdekorte: No video... cause it is bigger than my screen
AstralStorm: even added extra debug
AstralStorm: kdekorte: huh
AstralStorm: downscaling fails? interesting
AstralStorm: but that's not my problem (1920x1200 screen)
kdekorte: what bug?
AstralStorm: https://bugs.freedesktop.org/show_bug.cgi?id=25505
AstralStorm: whee
AstralStorm: commas are invalid
AstralStorm: I bet mplayer uses locale
AstralStorm: commas are our local decimal separator
AstralStorm: while GLSL expects points/dots
AstralStorm: needs a fix
AstralStorm: har, har
kdekorte: AstralStorm, yup... probably it....
kdekorte: might try with LANG=C
WhiteRabbit56: still getting :(EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch.
WhiteRabbit56: [dri] radeon kernel module version is 2.0.0 but version 1.17.0 or newer is needed.
WhiteRabbit56: [dri] Disabling DRI.
AstralStorm: indeed it works with LC_ALL=C
AstralStorm: blames mplayer
MostAwesomeDude: WhiteRabbit56: You need to update your DDX (xf86-video-ati).
AstralStorm: I'll report the bug there
MostAwesomeDude: AstralStorm: You think they localized their GLSL? :3
WhiteRabbit56: as in git pull
WhiteRabbit56: ?
kdekorte: WhiteRabbit56, libdrm with --enable-radeon-experimental-api and then recompile the DDX
WhiteRabbit56: kdekorte i did that
kdekorte: make clean or make distclean on both?
AstralStorm: MostAwesomeDude: they did
AstralStorm: they used printf
WhiteRabbit56: ill do that
AstralStorm: I'm reporting the bug there then
AstralStorm: what a funny bug
AstralStorm: so not yours at all :D
AstralStorm: I've expected so, since fragments worked in mesa test case
AstralStorm: new validator in mesa caught it
AstralStorm: it has worked before, heh
AstralStorm: (bet mesa used some kind of scanf
AstralStorm: or similar function to parse the string)
AstralStorm: that also explains why it worked before
AstralStorm: I had LC_CTYPE=en_US.UTF-8
AstralStorm: (by mistake)
AstralStorm: uhm, LC_ALL=.. that
AstralStorm: <3s the scaler
lordheavy: got this when runnng code with shaders : Program InfoLog: linking with uncompiled shader
lordheavy: shaders doesn't looklike to work :-/
Zajec: agd5f: i heard of programming notebooks LVDS with some lower PLLs
Zajec: agd5f: for reducing power consumption
Zajec: agd5f: do we have that implemented?
agd5f: Zajec: no yet
eosie: soreau: 1) fisheye still doesn't work here, it segfaults compiz instead of drawing random pixels 2) mipmaps in expo seem to work here 3) neverball works here, do you have latest libdrm?
soreau: eosie: Ah, I guess I should update everything else too.. Maybe it will break fisheye :)
soreau: eosie: fwiw, before this mesa update, fisheye would just draw a square with what you could tell was pixels from whatever was under it, but it two triangles going from corner to corner of the square
soreau: I'm grabbing latest radeon-drm-testing now and will update libdrm and ddx too
eosie: well, I have karmic's stock kernel and ddx, so it's probably not needed to update these
soreau: Well I'm going to update my entire system while I'm at it, so it'll take some time anyway
soreau: eosie: Maybe radeon-drm-next kernel will help make it work there? idk :P
soreau: testing even
WhiteRabbit56: I'm still getting that dumb error message... and i did:cd drm; make clean; ./autoconfig.sh; ./configure --enable-radeon-experimental-api --prefix=/usr/; make; make install;
adamk: WhiteRabbit56: Did you recompile xf86-video-ati after building libdrm with that option?
eosie: --prefix=/usr
DanaG: put the -- on autoconfig, also.
WhiteRabbit56: yah i did
DanaG: at least, I do the params on both, just for the heck of it.
adamk: WhiteRabbit56: When you ran ./configure for xf86-video-ati, did it say that kernel modesetting was enabled?
WhiteRabbit56: not a clue
adamk: Well you might want to find out by checking config.log or trying again and paying attention this time :-)
WhiteRabbit56: im looking thru config.log right now
adamk: If it doesn't say that kernel modesetting is enabled, it's not picking up the new libdrm.
adamk: bbl
WhiteRabbit56: i didnt find anything
AstralStorm: seems vsync doesn't work properly
adamk: I don't know how it shows up in config.log, just how it displays at the end of ./configure :-)
AstralStorm: I'm getting a nice tearing artifact with -vo gl
AstralStorm: (at high res)
WhiteRabbit56: http://pastebin.com/d5067d6a0
WhiteRabbit56: ran make clean in xorg-video-ati
WhiteRabbit56: then ./autogen.sh --prefix=/opt/xorg
WhiteRabbit56: then ./configure
WhiteRabbit56: and im still not seeing anything about kernel modesetting
[Enrico]: mhm i know there is a new firmware included in the kernel for interrupts support. is it included in drm-next ? couse a friend of mine said it should be added manually, and since i'm going to update my git build of radeon stuff well i ask :D
AstralStorm: WhiteRabbit56: fix your PKG_CONFIG_DIR
airlied: [Enrico]: its not included in the kernel
airlied: it needs to be picked up separately
WhiteRabbit56: AstralStorm?
[Enrico]: airlied: so what should i do? the radeon build how to wiki say nothing about it
[Enrico]: airlied: i guess license problems isn't it ?
roysjosh_home: agd5f, that patch fixed my monitor (fdo 19960) - thanks
AstralStorm: WhiteRabbit56: to point to where you have libdrm
[Enrico]: airlied: i know i have to use the EXTRA_FIRMWARE kernel config var (or something like that) but where i can pick up the firmware ?
AstralStorm: say, you installed libdrm in /opt/xorg
agd5f: roysjosh_home: cool! finally!
AstralStorm: then it probably installed its pkg-config files there as well
agd5f: roysjosh_home: I sent a kms patch to dri-devel
roysjosh_home: agd5f, indeed :) thanks for looking into it! woot. now to compile a kernel.
airlied: [Enrico]: kernel doesn't accept new firmwares anymore generally
WhiteRabbit56: AstralStorm: meaning?
DanaG: hmm, that mplayer lc_all bug.... where did you post the mplayer bug? I wanna' see it, for my own amusement.
roysjosh_home: agd5f, what in the world is the algorithm you are using in that patch, out of curiosity?
AstralStorm: WhiteRabbit56: likely /opt/xorg/lib/pkgconfig
airlied: [Enrico]: dri-devel archives
airlied: distros should pick it up eventually
[Enrico]: airlied: eheh i use gentoo here :D
AstralStorm: e.g. export PKG_CONFIG_DIR="/opt/xorg/lib/pkgconfig:/usr/lib/pkgconfig"
agd5f: roysjosh_home: something I dug up internally
WhiteRabbit56: thanks
airlied: [Enrico]: then someone should package it for Gentoo
[Enrico]: arekm: but i think i can include it with CONFIG_EXTRA_FIRMWARE :D
airlied: I suppose we could put the files into a tarball somewhere for sources
[Enrico]: airlied: i can package it don't worry thanks, i will google on the dr-devel to find the sources
[Enrico]: airlied: yes a tarball is just a good idea :D
agd5f: I'll put them on my fdo space
AstralStorm: :-)
[Enrico]: airlied: btw it is needed @boot if i use kms or can i load after udev start ?
airlied: [Enrico]: genearlly we load it in initrd
airlied: udev start is fine you just get a mode change then which can be ugly
agd5f: is there some magic script to set up public_html on annarchy?
[Enrico]: airlied: but with gentoo i don't use an initrd since i compile my kernel manually. ok i have to use CONFIG_EXTRA_FIRMWARE (well i already use it for other radeon firmwaer but since they are included it is easier
[Enrico]: airlied: thanks for all infos, i'm searching for it on google
airlied: agd5f: not it should just mkdir public_html
[Enrico]: but if you do a tarball pls paste a link in radeon build how to wiki :D
agd5f: cool
DanaG: AstralStorm: mind linking to that mplayer bug? I'm curious to see it.
AstralStorm: DanaG: not yet
AstralStorm: I'm trying to find their bug reporting machinery
agd5f: http://people.freedesktop.org/~agd5f/radeon_ucode/
AstralStorm: DanaG: there's none
AstralStorm: they take bug reports via mailing list
AstralStorm: I'll send a mail there
AstralStorm: no, wait
AstralStorm: there IS a bugzilla
AstralStorm: nicely linked in a third level page
rah: agd5f: why does this firmware have a different license to the other firmwares?
agd5f: rah: lawyers
AstralStorm: likely not yet cleared
AstralStorm: I bet it will be fixed some time soon
agd5f: all new ucode will have the license
AstralStorm: still, that's no reason to remove support for people lacking this file
rah: agd5f: this is bad
agd5f: rah: why?
[Enrico]: airlied: wow that firmware is very propetary as i can understand in the license..... :'(
[Enrico]: no reocmpile this time i guess
[Enrico]: recompile*
rah: agd5f: doesn't the license preclude it from being included with the other files?
agd5f: it's a similar license to every other firmware used by the kernel
rah: hmm
rah: umm
rah: agd5f: why isn't it included with the other files? :)
flyback: prepares for the worst due to high wind warning by charging lots of batteries and ponders getting a natural gas heater at walmart or home depot
WhiteRabbit56: still no luck...
spstarr: heh
airlied: rah: the kernel isn't accepting new firmware
[Enrico]: agd5f: i guess, well it is a pity but well better then fglrx anyway
spstarr: that storm hit here
WhiteRabbit56: although on reboot it seems that I have kms working because my terminal is native res
spstarr: flyback: some rebound winds though tonight/tomorrow
Ghworg: As long as it can be packaged in each distros equivalent of firmware-linux which AIUI it can it's fine
agd5f: you can still redistrubute it
lordheavy: intel firmware are available separatly
lordheavy: *firmwares*
airlied: all wireless fw is separate and it seems fine
rah: airlied: the other files are included with the kernel?
flyback: notes this would have been awesome day to test his windmill generator
rah: rah@myrtle:~$ dpkg -S /lib/firmware/radeon/R600_cp.bin
rah: firmware-linux: /lib/firmware/radeon/R600_cp.bin
airlied: rah: they were included in src code and moved to firmware bins
rah: hmm
AstralStorm: yup
AstralStorm: this is ok
AstralStorm: someone should update /topic with the link to the new file I think
rah: airlied: are the _rlc.bin files going to be included with the kernel?
agd5f: rah: they can't
agd5f: rah: http://git.kernel.org/?p=linux/kernel/git/dwmw2/linux-firmware.git;a=tree
rah: airlied: ie, is the kernel not accepting new firmware at all, or just because of the merge window?
airlied: rah: at all
airlied: there is another tree
DanaG: that's stupid. Seems entirely arbitrary. NO FIRMWARE FOR YOUUUUuuu!
DanaG: oh, I see.. it's just a different package.
rah: I see
agd5f: rah: the onld ucode only happens to be in because it was there before the decision was made
agd5f: s/onld/old/
DanaG: Then why doesn't the old firmware get moved out?
DanaG: It seems unfair to me.
DanaG: Especially if licenses are comparable.
agd5f: DanaG: it will eventually
rah: how will the kernel load the ucode if a driver is compiled in and the firmware is in a separate package?
MostAwesomeDude: request_firmware
rah: MostAwesomeDude: and what will that do?
chithead: rah: need to put in initrd
DanaG: and then, of course, you'll have the zealots who'll insist on not having any firmware at all.
DanaG: Hope they use coreboot... or else they're hypocrites.
agd5f: rah: include it in your initrd
MostAwesomeDude: rah: request_firmware() is a kernel interface for modules to load firmware at runtime.
rah: I don't have an initrd
chithead: there is a kernel option to include things in a built-in initramfs
agd5f: then it will load it from the filesystem
airlied: will do it
rah: agd5f: that's what I tried earlier; the kernel with radeon compiled in and the rlc_.bin files in /lib/firmware/radeon
rah: agd5f: it didn't work
AstralStorm: DanaG: the zealots are left with intel chips ;p
AstralStorm: and only gfx ones
[Enrico]: rah: if the raedon driver is builtin you have to include the firmware in the kernel using CONFIG_EXTRA_FIRMWARE
AstralStorm: or making their own (hard work, really, as OpenFirmware guys found out)
DanaG: Anticompetitive thing Intel does: make their wifi cards not work on non-Intel chipsets.
DanaG: At least, I've heard they do that.
DanaG: My system has ATI graphics, Intel wifi and such, and HP UEFI firmware (it's HP, not outsourced to Insyde).
rah: so I have to copy these .bin files around myself?
DanaG: goes off to try xfce newness.
lordheavy: yes, in /lib/firmware/radeon
devh: rah: maybe try to move the new firmware to firmware/radeon in your kernel tree
devh: before compiling
rah: that sucks severely
rah: does CONFIG_EXTRA_FIRMWARE_DIR include subdirectories?
AstralStorm: I'd like a proof of that "Intel wifi not working on non-intel chipsets"
DanaG: Or rather, in Windows, specifically.
AstralStorm: since I have a few and they work with AMD and even Via chips
mmp: DanaG: I've personally seen intel wireless 5000 series working on non-intel chipset
AstralStorm: hmm.
mmp: (one PowerPC SoC from freescale)
AstralStorm: even in windows they work
AstralStorm: but the driver is buggy as hell
DanaG: Hmm.
DanaG: Google for Intel Wifi Code 10.
AstralStorm: (unlike the Linux one which is now fine)
DanaG: Also, I can't use the wireless functions of AMT, or it breaks wireless in Linux.
eosie: soreau: where did you get enemy territory?
AstralStorm: DanaG: AMT being?
DanaG: Intel Active Management Technology.
AstralStorm: ah power saving
DanaG: Most useful part for me is serial-over-LAN.
AstralStorm: no, it works fine
AstralStorm: ahh, that junk
DanaG: It's a sideband management thingy.
AstralStorm: don't ever use it
AstralStorm: it's broken
AstralStorm: it doesn't work in perfect conditions often either
DanaG: I find it's nice in small-business mode for serial-over-lan, but I don't do any more than that.
DanaG: Good for getting logs of kernel panics, and such.
AstralStorm: serial over lan is done with PPPoE
AstralStorm: :)
AstralStorm: alternatively, PPP VLAN
DanaG: This one is firmware-based, though -- even if OS hard-locks, it's still available.
AstralStorm: ha, ha
DanaG: Does AMD have some serial-over-LAN sort of thing>
AstralStorm: you wish
AstralStorm: if it hardlocks, it can never empty its buffer
AstralStorm: it can't even reboot the machine
AstralStorm: making it less useful than wired ethernet power-on signals
DanaG: Well, I know I've gotten at least kernel logs from panics.
AstralStorm: kernel logs will work with normal netconsole
DanaG: I use it with a laptop with no serial port; otherwise, I could just use just plain serial.
AstralStorm: even with (mostly) dead kernel
AstralStorm: as long as tcp/ip stack works and mac80211 does
AstralStorm: and irqs are sent
AstralStorm: which means like, 80% of the oopses
mentor: No
AstralStorm: but only 20% of the real bad ones
mentor: It bypasses the TCP/IP stack
AstralStorm: mentor: it does... partially
AstralStorm: :)
mentor: meh
DanaG: It's also good when ssh dies, for some reason.
AstralStorm: yes, but that's same with usual netconsole
AstralStorm: no need for "serial over LAN"
rah: rah@myrtle:/usr/src$ du -sh attic/
rah: 2.0G attic/
rah: rah@myrtle:/usr/src$ ls -1 attic/*.deb | wc -l
rah: 130
rah: lol
AstralStorm: esp. with faulty solutions
AstralStorm: rah: dusty attic
rah: indeed
DanaG: hmm, would usb-serial adapter be better, or worse, than netconsole?
airlied: worse since usb is balls
DanaG: Oh, and IDE Redirection is just plain amusing.
airlied: ehchi debug port is probably the best thing
airlied: ehci even
DanaG: Or firewire.
AstralStorm: firewire is more prone to failure though
AstralStorm: ehci debug port or normal ethernet is your best bet
airlied: firewire is good since you it doesn't require CPU
chomwitt: hi. i have a x1650 ato card and radeon installed in my ubuntu 9.10 . how can i enable accelarated graphics?
AstralStorm: (firewire controllers are known to be moody)
AstralStorm: chomwitt: it is enabled by default. x1650 being r2xx right?
chomwitt: x1650 is rv535 ?
AstralStorm: mhm, r5xx then
rah: urg
AstralStorm: doesn't change much
AstralStorm: the drivers in ubuntu do 3D out of the box
[Enrico]: airlied: i done the package for gentoo, i only need an host for the tarball now (and a dev who push it on the overlay :D )
rah: is getting artifacts
DanaG: Anyway, SOL works for me. Oh, and random thing: if I set up a VPN server, I could use the enterprise mode as a hackish anti-theft tracking thing.
[Enrico]: a gentoo dev indeed
rah: when is it going to work? :(
AstralStorm: DanaG: indeed
AstralStorm: but what for again?
chomwitt: AstralStorm: how can i check if i have accelerated graphics?
DanaG: Tracking a laptop.
AstralStorm: I'm not getting your use case
AstralStorm: chomwitt: run glxinfo
AstralStorm: read the output, look for OpenGL vendor string
AstralStorm: whether it mentions your card
DanaG: But then, there's the caveat that they'd have to connect to wired ethernet for it to be trackable.
soreau: eosie: I will have to look up a link for ET later. Will get it to you sooon
AstralStorm: ET is also available on Gentoo mirrors
AstralStorm: (the licence allows that)
airlied: agd5f: one bug in your atom pll patch, you divide a 64-bit freq by 10
airlied: that needs to be done with do_div
AstralStorm: chomwitt: any luck?
chomwitt: AstralStorm: OpenGL version string: 1.5 Mesa 7.6
AstralStorm: http://dpaste.com/131453/ mine.
AstralStorm: your Mesa is likely a tiny bit too old
chomwitt: AstralStorm: OpenGL vendor string: DRI R300 Projec
airlied: agd5f: I'll fix it up locally and push to drm-radeon-testing
AstralStorm: chomwitt: r3xx
agd5f: airlied: cool. thanks
AstralStorm: not r5xx. but it should work.
AstralStorm: :)
AstralStorm: with mesa 7.6 as well
agd5f: airlied: btw, on that ucode fallback patch where I send the wrong version, do you want me to make a new patch?
AstralStorm: OpenGL Renderer string I meant.
airlied: agd5f: lemme check where I pushed it so far
chomwitt: AstralStorm: http://dpaste.com/131455/
airlied: has a baby climbing on him
AstralStorm: chomwitt: looks fine
AstralStorm: airlied: heh, good that it's not climbing on the machine via cables
AstralStorm: chomwitt: might want to try glxgears to be extra sure
AstralStorm: chomwitt: what doesn't work for you?
chomwitt: AstralStorm: tha cube in compiz
airlied: agd5f: new patch is fine
AstralStorm: chomwitt: hmm.
AstralStorm: chomwitt: I suspect old mesa
airlied: AstralStorm: yeah I have no machines at home naymore other than a laptop ;-)
AstralStorm: I bet there is a package for Ubuntu and version 7.7 somewhere
roysjosh_home: agd5f, should I be worried about the freq/10 instead of do_div? will it boot/function?
chomwitt: AstralStorm: thanks! i'm not sure i can find 7.7 though for ubuntu.
agd5f: roysjosh_home: should be ok, but airlied said he'd fix it up when he merges it
roysjosh_home: agd5f, ok, thanks
agd5f: roysjosh_home: works fine here
airlied: just pushed drm-radeon-testing now with it in it
AstralStorm: chomwitt: there should at least be 7.7 release candidates available
AstralStorm: in testing or somewhere around
speps: i've tried the last 7.7-rc2 mesa3d implementation with latest radeon with r300 gallium implemntation, and it goes very slow. Latest working mostly well is the 7.6.1-rc3. What's the gallium status for status radeon drivers?Thanks
AstralStorm: gallium is... slow, yes
AstralStorm: for gallium to work *somewhat*, you need 7.8
AstralStorm: that is, latest master
AstralStorm: it's still very devel code
AstralStorm: speps: if you need it to work, you're still better off w/o gallium
chomwitt: AstralStorm: i cant find it as a ubuntu package in its repos. if i download mesa 7.7 (not ubuntunized :-) ) should i built it?
AstralStorm: well, why not
AstralStorm: it shouldn't hurt anything
AstralStorm: 7.7 is about to go stable I think
roysjosh_home: agd5f, the kms patch works as well. thanks!!! my monitor looks great now, and I can show it off :)
agd5f: roysjosh_home: cool
Sekkai: Hello i am having some wierd crashing issues with a game called Quantz avaible here http://dl.dropbox.com/u/2159868/QuantZ-beta_1.1.1_i386.tar.gz
Sekkai: when entering action mode i am being resulted this
Sekkai: QuantZ: radeon_mipmap_tree.c:117: compute_tex_image_offset: Assertion `lvl->size > 0' failed.
Sekkai: Aborted
Sekkai: Ive googled around this and it seems that all opensource radeon users suffers this
Sekkai: tried on an nvidia card with a more or less identical installation and it works fine
shadowmaster: does the radeon driver attempt to store the graphics controller status on startup? (e.g. before switching to graphics mode from a Linux text console)
MostAwesomeDude: shadowmaster: Yeah.
airlied: not really
airlied: we genearlly redetct and reprogram it all
MostAwesomeDude: Doesn't it, for UMS?
airlied: oh fo rUMS yes
airlied: not really
airlied: it stores a bunch of state and restores it yes
airlied: is half asleep her
shadowmaster: well, I'm not using KMS yet. I've noticed that without KMS radeon (on a RS780-based controller) occasionally breaks the text consoles and they turn into pitch black screens :)
shadowmaster: I can restore them with vbetool but then I need to kill and restart X or radeon does it again after the next switch
agd5f: shadowmaster: save/restore is often fail
shadowmaster: right, and nothing I can do about it? :(
agd5f: shadowmaster: you can fiddle with it, but some systems never quite work
agd5f: shadowmaster: that's why kms is better since it programs the hardware explicitly
shadowmaster: eh, well, it got worse with KMS for me because the "occasionally" turned into "always after the first X-to-textconsole" switch.
shadowmaster: although maybe it's because I had the kernel drm/radeon modules inserted on X's startup rather than before.
airlied: agd5f: hey div by 0 in your new pll code somewhere on my machine
airlied: will see if I can debug
soreau: eosie: The file you would want is et-linux-2.60.x86.run for the installer http://www.google.com/search?q=et-linux-2.60.x86.run+intitle%3A%22index+of+%2F%22&ie=utf-8&oe=utf-8&aq=t&rls=org.mozilla:en-US:official&client=firefox-a
soreau: Sorry it took so long to get back
tavl: anyone here using opensuse + KMS + radeon module?
airlied: agd5f: ah I think the div/0 was my do_div patch
eosie: soreau: thanks
airlied: agd5f: cool I fail at using do_div correctly, fixed it now and repushed drm-radeon-testing
tavl: anyone here using opensuse + KMS + radeon module?
AndrewR: hey all. drm-radeon-testing is really _fast_ moving target ... but i got oops/segfault on modprobe with this tree up to "drm/ttm: Initialize eviction placement in case ..." and rs780