Radeon IRC Logs For 2009-4-06
l3iggs: this is what happens:
spstarr: l3iggs: interesting
l3iggs: it pretty much sucks
spstarr: its not even repainting the window decoration on move
l3iggs: no it's not
airlied: l3iggs: don't suppose you have bandwidth to pull down the F11 beta liveCd and try it :)
l3iggs: ha, i have bandwidth coming out my ears
airlied: l3iggs: btw that is EXA or XAA?
l3iggs: that's XAA, exa locks up hard
airlied: wierd since XAA is probably only using the 2D engine
l3iggs: can you give me a link to that beta CD please?
spstarr: airlied: is radeon/r200/r300: fix up the whole buffer space checking ready for koji? i'll be testing this in a custom RPM
airlied: spstarr: I'm about to build a whole kernel/drm/mesa combo
airlied: spstarr: which breaks the API
spstarr: i will hold off then
spstarr: looks at the changes in cvs
spstarr: oh, not committed in cvs yet
spstarr: airlied: your personal code changes aren't on cgit?
spstarr: like this change coming
airlied: spstarr: I'm just building it all locally first
airlied: spstarr: I thought I pushed the git bitsz
l3iggs: haha woo hoo! pulling the fedora 11beta at 10MB/sec :)
spstarr: i see a DDX change but cgit isn't showing the change (37 minutes ago)
airlied: radeon-gem-cs3 is updated and libdrm
airlied: or drm, modesetting-gem
spstarr: I see those DDX changes yes
spstarr: lots of code ripped out
l3iggs: burning F11 Beta
spstarr: it looks like you're moving more stuff into the drm side out of the DDX since
spstarr: memory stuff
spstarr: airlied: when you go here http://cgit.freedesktop.org/~airlied/ it doesnt show you making any changes
spstarr: so i get confused :)
l3iggs: ok, booting F11 beta live cd
airlied: crosses fingers/toes :)
l3iggs: splash screen works
airlied: l3iggs: one down :)
spstarr: is that booting with KMS?
airlied: should be
spstarr: does buntu even use KMS yet?
l3iggs: this is cool, i've never tried fedora
l3iggs: i don't know what that is
airlied: spstarr: its not upstream yet
airlied: I wouldn't expecty any distro to use it until it its
airlied: its upstream
spstarr: l3iggs: Fedora is where all the future is :)
airlied: l3iggs: fail when gdm starts? :)
l3iggs: pretty much
l3iggs: there was a little spinny thing around the cursor for a bit
l3iggs: then that stopped, now it's locked, black screen cursor in the middle
l3iggs: there he is
l3iggs: oops, wrong window
spstarr: does a double take
spstarr: airlied: the drm changes look almost identical to the DDX changes or very close to
airlied: spstarr: its all the same, moving stuff to the new APi
spstarr: ahh ok so this is being combined in the drm now only?
spstarr: or thats all just gone now
airlied: we keep a copy of the drm header in each place at the moment.
airlied: glisse: okay my drm-rawhide tree has the fully cleaned up API
airlied: I merged my further changes into the patch with your name by mistake :)
airlied: I make rawhide mmap the same also
spstarr: and so 6.12.1-5 comes
spstarr: massive libdrm changes
spstarr: looks at kernel cvs commits
airlied: I hope it all works once the kernel builds
airlied: I'm not sure I'll have to rebuild mesa or not
spstarr: doesn't mesa depend on libdrm anywhere?
airlied: it does but I didn't break any APIs to mesa hopefully
terracon: kernel building in Koji? See you in 3 hours from now when it's done
spstarr: powers on quad box
spstarr: terracon: build it locally :)
terracon: build it locally than upload it to Koji. Riiiight. I got it!
spstarr: it's only 2am here anyway, and with 25cm of snow coming, hell if im going into work today :)
airlied: of course I built it locally (well in Westford first :)
airlied: but it failed on powerpc on koji then.
airlied: hmm I appear to have broken compiz or something will do some research later
spstarr: uh oh
spstarr: maybe kwin works now
airlied: haven't made any code changes
airlied: at least no intentional ones, api changes and breaks only
l3iggs: well thanks for your help guys
l3iggs: i'm going to bed
l3iggs: i'm gonna tell myself I've got a hardware problem
l3iggs: that makes me feel better
terracon: accuweather says the changeover to snow will happen around 3pm
terracon: so, you're going to work in the morning
spstarr: terracon: im not in Toronto :)
terracon: Where are you?
spstarr: north of city
terracon: You're screwed!
terracon: grrr. Racoon outside working over the garbage
terracon: yay. Mess to clean up tomorrow
spstarr: it is mind you warmer than it should be
spstarr: oh wrong cvs dir
spstarr: not that I wouldn't like trying .30 yet :)
twoerner: airlied: ping
spstarr_wap: airlied: epic fail cannot start X with or without KMS
spstarr_wap: airlied: ill rollback later today its almost 4 am :-)
airlied: spstarr_wap: did you get all the packages.
airlied: twoerner: pong
twoerner: airlied: the driver is still working for me
twoerner: airlied: so we can close the bug
airlied: twoerner: excellent
spstarr_wap: airlied: should be libdrm ddx kernel
airlied: spstarr_wap: yes, anything in the logs?
spstarr_wap: airlied: nothing I could find nothing in messages nor xorg logs even after don't sysctr magic
spstarr_wap: Er after I did use the magic keys
nanonyme: I'm impressed by http://www.x.org/wiki/RadeonFeature ^^
nanonyme: Huge areas covered by done. If you also count in the wip and mostly, a clear majority of the matrix gets coloured. ^^
arekm: suspend support on r600 marked as unknown, works fine here (to ram) on mobile HD 3400
arekm: unless "suspend" is about someting else
MrCooper: arekm: with DRI enabled?
airlied: MrCooper: on radeon it should be fine, rhd has issues.
arekm: MrCooper: yes
osiris__: glisse: so now to test newttm I just need drm-next-radeon branch from your kernel tree. the ddx can be from airlied's tree and mesa from main tree, right?
glisse: osiris__: let me check if airlied pulled my patch
glisse: osiris__: you need ddx from my cs3-newttm branch
airlied: glisse: I just pushed info instead :)
airlied: so I need to make my DDX use info
glisse: airlied: what is the bin name of oprofile gui under fedora ?
airlied: glisse: not sure I always build my own sysprof :)
glisse: sysprof is better ?
airlied: much easier to use, you have to build a kenrel module
airlied: oprofileui is the package
airlied: or oprofile-gui
b0le: with dri2, when running glxgears and I trigger scale (compiz), the background of glxgears goes transparent. Apparently this was fixed for intel?
airlied: b0le: using radeon-rewrite?
b0le: Another problem I have had when trying out dri2, is that having a desktop larger than max 3d texture size no longer works, the part that goes over the max texture size is always corrupted. The problem is that it used to work, though maybe it shouldn't have?
b0le: airlied: yes
airlied: I think it used to corrupt pre-dri2 just differentyl
airlied: I used to see blank space or sometyhin
osiris__: glisse: does your new ddx work with old ttm?
b0le: I previously used the wallpaper plugin with an image smaller than max texture size that was tiled, so it worked fine - no graphical glitches. If I didn't, it would still show the windows above the 'extra' space (but the background would draw incorrectly), now it won't draw the windows above the corrupted area. Is this the problem that shatter is supposed to fix?
glisse: osiris__: yes
glisse: airlied: sysprof can see where time is spend in the kernel ?
airlied: glisse: yes
glisse: i am must be doing somethings wrong then
airlied: oh you might need framepointer kernel
airlied: I don't run 64-bit too often :)
glisse: i do have framepointer
glisse: 61% is in the kernel but u can't see what :(
glisse: maybe i shoud disable oprofile
glisse: might be getting in the way
airlied: well unload it, or stop it runnig
MrCooper: glisse: I think the sysprof kernel module just doesn't support kernel profiling on amd64, apparently this is the reason anholt is still running 32 bit kernels
MrCooper: unfortunately the sysprof ftrace Git branch doesn't seem to work either
arekm: about wiki/RadeonFeature - what "Console restore" is about? restoring txt console after resume?
airlied: vt switch
arekm: that also works well on r600 (with and without dri)
glisse: MrCooper: that suxx oprofile isn't giving information neither
arekm: updated wiki then
MrCooper: glisse: with oprofile you need to point it to the uncompressed vmlinux file
glisse: i do
MrCooper: seems to work here
nanonyme: MrCooper: Are there some tutorials on that? I can think of at least one usage scenario where I could produce useful results. ^^
MrCooper: tutorials on what?
MrCooper: dunno, there probably are
nanonyme: I wouldn't even know if they were daemons running in the background or kernel modules or what. ^^
MrCooper: is your friend
nanonyme: Oh, right.
nanonyme: I was trying to use oprofile even at one point.
nanonyme: Now I remember. It was just that it was broken in Debian. ^^
MrCooper: nanonyme: that's fixed now
nanonyme: MrCooper: Doesn't matter, I changed distro anyway.
nanonyme: (Not because of that though)
Netzpython: hi :)
Netzpython: i experience reproducable xserver-crashes (1.5) with this nice radeon/r300 + exa + dri while compiz is running if i use some app-switcher plugins (alt+tab)-things on an ati m22. similiar bug already exists in bugzilla; but is the a way to find out whats going wrong?
mattst88: glisse, does your commit to radeon-rewrite bring it in sync with your kernel tree branch drm-radeon-next?
glisse: mattst88: yes
mattst88: OK. Dave said he thought my writeback failure was probably in the ttm code
mattst88: so I'll try yours later and hope it magically works :)
glisse: use pci
glisse: no agp code in it
agd5f: glisse: for rs400_gart_enable() treat rs740 the same as rs690
glisse: agd5f: for agp base right ?
agd5f: glisse: yeah
glisse: damm it seems oprofile doesn't fully work neither on amd64
glisse: MrCooper: me no getting info is likely due to lack of support for amd64 :(
MrCooper: glisse: it works here on amd64
Netzpython: hmm that sounds not too promising ^^
glisse: MrCooper: dammm i don't see what's wrong then
glisse: i use opreport -p /pathtokernelmodule/
glisse: but i can't have anyinformation on time spend in module
MrCooper: glisse: you have to specify the path to vmlinux to the oprofile daemon, there's an entry for it in the GUI configuration
glisse: i did that too
glisse: i can see vmlinux symbol
MrCooper: hmm, not sure what needs to be done for modules
glisse: now it works... i am bit puzzled
MrCooper: glisse: better puzzled but working than the other way around :)
osiris__: glisse: unfortunately it doesn't work. screen doesn't come up
Netzpython: has r300 development stopped?
osiris__: here's the log http://pastebin.com/m794afc02
osiris__: but nothing unusual there
osiris__: Netzpython: nope
glisse: osiris__: according to log everythings is fine
glisse: modprobe fbcon doesn't bring anythings ?
Netzpython: so why nobody is interested in what i said? ^^
osiris__: glisse: I loaded fbcon before radeon
agd5f: Netzpython: known issue for some users, don't have a fix yet
glisse: osiris__: load it after
osiris__: glisse: I can't, I don't have second machine to ssh in
Netzpython: ah ok thx :)
osiris__: glisse: I also tried running you gem-cs3-newttm ddx branch with airlied's drm-rawhide kernel, but it oopses badly: http://pastebin.com/m3272ca3e
glisse: osiris__: you can do a small script to load radeon and the load fbcon :)
nanonyme: I guess r300 eventually reaches a point where it is fully featured and everything after that will be bug fixes. ^^ I guess someone could define that as in "development stopping". :)
osiris__: glisse: I can try, but with airlied's drm-rawhide kernel I load modules in the same order (drm, fbcon, radeon) and it works
glisse: osiris__: your modesetting-gem libdrm is outdated
glisse: and your rawhide module too
glisse: you need bleeding edge of everythings
nanonyme: agd5f: Anything new on the Mesa r600? :) (Sorry, I know I'm a bit impatient but it'd be cool to be able to test out what's ready soon)
osiris__: glisse: I have all the latest stuff
agd5f: nanonyme: not yet
osiris__: megari: did you manage to test my patches?
app4des: guys is there ant way to force shadowfb acceleration with the ati driver (not rhd)?
agd5f: app4des: only on r6xx/r7xx at the moment
glisse: osiris__: i am pretty sure that is due to a mismatch in ioctl size
glisse: i saw that too
nanonyme: agd5f: Could you please put it in the channel topics immediately when it's out? :)
glisse: had to make clean everythings
glisse: before rebuilding
agd5f: nanonyme: you'll know :)
glisse: first libdrm, than ddx, than mesa
glisse: and also rebuilding radeon.ko
glisse: rm *o files in the directory
osiris__: glisse: ok, will rebuild everything
glisse: osiris__: does cs3 updated branch of airlied works ?
osiris__: glisse: yeah, I'm using it now
glisse: with lastest rawhide >
osiris__: glisse: looks like I have 3 days old airlied's kernel
glisse: ok so that's explain it
osiris__: will rebuild everything and try again
osiris__: glisse: hmm, these warnings looks important: radeon_object.c:52: warning: integer constant is too large for ‘unsigned long’ type
osiris__: radeon_kms.c:86: warning: cast to pointer from integer of different size
osiris__: glisse: no luck, when I loaded radeon module, it changed resolution to native one. but loading fbcon turned off the screen
bobbens: feenableexcept(FE_ALL_EXCEPT); in my opengl app triggers SIGFPE signals from r300_dri.so, is that to be expected?
agd5f: glisse: http://www.botchco.com/alex/xorg/0001-Fix-vram-setup-on-IGP-cards.patch
agd5f: against your drm tree
agd5f: glisse: whoops, typo, i sec
agd5f: 1 sec
agd5f: glisse: http://www.botchco.com/alex/xorg/0001-radeon-Fix-vram-setup-on-IGP-cards.patch
kdekorte: I just ran a torture test of mplayer using xv on my 3650 using the ATI driver in F11 Beta
kdekorte: I had 6 videos running, 1 was 720p, 2 were DVD quality and a couple of smaller ones
kdekorte: not a single frame drop, although CPU usage was using 2.5 of my 4 cpus (Q6600)
kdekorte: so that is pretty cool
kdekorte: screen shot of the test http://www.flickr.com/photos/25272759@N07/3418707652/
spstarr: airlied: new DDX you pushed works
spstarr: in KMS right now
spstarr: tries kwin
spstarr: airlied: !!!!!!!! kwin works!
megari: osiris__, not yet. Ran out of space trying to clone airlied's drm git.
megari: ... and now that it's no longer weekend, I'm in a constant hurry again.
spstarr_home: airlied: VT switching with composite ON locks up hard
spstarr_home: airlied: not enabling KMS causes X to not start, so im in KMS now w/o composite on
t1ger: having issues with the Radeon driver and XBMC
t1ger: X Error of failed request: BadMatch (invalid parameter attributes)
t1ger: Major opcode of failed request: 1 (X_CreateWindow)
t1ger: Serial number of failed request: 28
t1ger: Current serial number in output stream: 29
t1ger: 2.6.28-11-generic - Ubuntu 9.04
t1ger: ATI Technologies Inc M56GL [Mobility FireGL V5250]
t1ger: xorg-server 2:1.6.0-0ubuntu9
t1ger: (II) Loading /usr/lib/xorg/modules/drivers//radeon_drv.so
t1ger: (II) Module radeon: vendor="X.Org Foundation"
t1ger: compiled for 1.6.0, module version = 6.12.1
t1ger: Module class: X.Org Video Driver
t1ger: I guess there some OpenGl or XV issues ?
agd5f: t1ger: what are you trying to do?
t1ger: Just start XBMC
agd5f: you mean XvMC?
MostAwesomeDude: XBox Media Center...
MostAwesomeDude: OpenGL or Xv rendering?
t1ger: Dunno, I don't see any useful options
MostAwesomeDude: Also I can't remember which card the
MostAwesomeDude: M56 is... isn't that an r5xx?
t1ger: Are there any svn ppa for radeon driver ?
MostAwesomeDude: t1ger: Should just work. Could you pastebin the entire log that you copied from earlier?
t1ger: Most: Sure.
t1ger: (Worked in 8.10, but that was with the "/&"/&"#/&"## ati driver)
agd5f: t1ger: ati and radeon are the same driver
agd5f: ati is just a wrapper
t1ger: No. The "other" ubercrapsucksblabla driver
t1ger: You newer used it I see.
t1ger: Sorry, 2 years of frustrations.
agd5f: t1ger: withou seeing a log, my guess is that you don't have the dri enabled, and hence no Xv
t1ger: AIK I do. Mythtv actualy works with ATI cards now.
t1ger: let me put up some debug in pastebin and I investigate
nanonyme: ponders when "3D features: Textures" will be marked as "DONE" in http://xorg.freedesktop.org/wiki/RadeonFeature for r500.
nanonyme: That is, mostly wondering how incomplete they still are. ^^
agd5f: t1ger: can you pastebin your xorg log? aslo if you run xvinfo does it list any adapters?
agd5f: nanonyme: they are pretty much done, maybe just texure tiling, but that's not enbaled yet on r3xx/r4xx either IIRC
MostAwesomeDude: nanonyme: Tiles, cubemap weirdness, small mipmaps sometimes don't work.
spstarr: airlied: oh im getting corruption now with _ 's again
spstarr: after a while of use with KMS + no-composite on
t1ger: stay tuned. It huge.. (xorg log)
MostAwesomeDude: t1ger: Pastebin plz.
t1ger: I know.. copy pasting..
t1ger: agd5f: http://pastebin.com/m23a43727 <- Kernel/xorg.0.log + xvinfo
agd5f: t1ger: looks fine
t1ger: Can't say I find any verbose option in xbmc to highlight any issues when it start. I know it news OpenGL.
t1ger: glxgears runs fine.
nanonyme: What would glxinfo say?
t1ger: compiz runs fine. (I don't have it enabled now)
agd5f: t1ger: sounds like an issue with XBMC
nanonyme: Oh, should be fine. :)
MostAwesomeDude: t1ger: Yeah, not our bug.
t1ger: nononyme: http://pastebin.com/m528c081e <- glxinfo
t1ger: agd5f/MostAwesome: I guess. Thanks anyway.
t1ger: Any thoughs on what it could be ?
t1ger: X Error of failed request: BadMatch (invalid parameter attributes)
t1ger: Major opcode of failed request: 1 (X_CreateWindow
t1ger: dri/xorg bug etc ?
MostAwesomeDude: Looks like an app bug.
t1ger: Any free hints to where the "X error" list should be ?
t1ger: erros.h in xorg code ?
agd5f: XBMC trying to create a window with bad attributes. might try #xorg
t1ger: Thx :)
t1ger: While I have you here
t1ger: I use this a lot
agd5f: doesn't work well with xrandr
agd5f: it's not multi-head aware
t1ger: I miss the "Apply"
t1ger: Like powerstrip in windows
agd5f: t1ger: you can specify your own custom modelines with xrandr however
t1ger: So I can swap resolution in realtime
agd5f: xrandr --newmode
agd5f: xrandr --addmode