Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-6-23

Search This Log:

Tatsh: you guys have any ideas?
Tatsh: http://forums.gentoo.org/viewtopic-t-774647.html
Tatsh: i know i posted on gentoo forums but that's only because it's the distro i use
airlied: try with no xorg.conf
Tatsh: yeah i'm not sure what the story is but i think i'm going to try reverting to an older kernel
arkascha: I get massive artefacts when scrolling in text editors using Xorg and the radeon driver.
arkascha: I appears that the artefacts appear only in dynamic texts, like in editors doing live syntax highlighting. The slower the application, the more massive the artifacts. Ergo most in Java apps like neatbeans and eclipse.
arkascha: Is there some fix known, this is _really_ annoying...
nanoki: Which card was this again?
arkascha: nanoki: sorry ! ! ATI RV505CE (Radeon X1550 64-bit6)
airlied: AGP or PCIE?
arkascha: airlied: good question, PCI-E (I think, one minute)
arkascha: yes, PCI-E.
Zajec: glisse: I fixed not getting modes for my PANEL, but this didn't change much
Zajec: glisse: after loading radeon I get black screet with bottom-flickering and loading fbcon doesn't bring console
Zajec: glisse: can tips how can I track this? what may cause that problem?
airlied: Zajec: if you can ssh dmesg would help
Zajec: airlied: for now I was rebooting to working kernel and checking /var/log/messages
Zajec: that should be the same, right?
airlied: not always, it doesn't write debug msgs usually
Zajec: ah, ok, thanks
Zajec: right now I recompile Linus's git, will get dmesg after that
Zajec: I was using DRM_ERROR for debugging until now, that's probably why /var/log/messages was enought for me :)
airlied: so can you see it picking no mode?
Zajec: airlied: KMS couldn't get anything from EDID (that's normal for my PANEL) but with some patch I got KMS reading native mode from atombios:
Zajec: Jun 23 10:55:56 linux-aodr kernel: [drm_mode:drm_mode_debug_printmodeline], Modeline 11:"1600x900" 59954 88540 1600 1614 1626 1630 900 902 904 906 0x48 0x0
airlied: Zajec: we should have a panel mode reading code already
airlied: like we do in the DDX, but maybe it isn't ported
Zajec: airlied: it just isn't choosed for my PANEL... moment
Zajec: airlied: check out radeon_add_atom_encoder(...) in radeon_encoders.c
airlied: ah cool I was going to suggest that :)
suokko: airlied: About commit a9f7fb134b9 to kms-support: Wouldn't it be better to always use RADEONSwapCopy and then just implement it differently depending on host endianess?
Zajec: still after fixing that (KMS gets my native mode fine) I get black screen ;) i'll get dmesg output now to debug that
airlied: Zajec: not really much advantage, sicne the non endian one wouldn't swap
Zajec: suokko: not really much advantage, sicne the non endian one wouldn't swap
Zajec: :P
airlied: Zajec: so do you get radeon_atombios_get_lvds_info
airlied: called?
suokko: :)
airlied: oops :)
suokko: But I think it would be cleaner code if not having endian check everywhere were is copy from/to vram
suokko: But I don't know if it is used in many places
Zajec: airlied: yes, with that fix adeon_atombios_get_lvds_info is called
suokko: It could even named RADEONCopyToVRAM :) Then everyone would remember to do endianess check automaticaly
airlied: so it gets a mode and doesn't set it? or it fails to set it
airlied: suokko: it won't aleays be correct
Zajec: airlied: gets mode but don't know what happens then
airlied: VRAM has these swapper things so its best to be explicit
Zajec: airlied: glisse told me that not working screen after loading radeon in expected...
Zajec: airlied: is that sth you agree on?
airlied: Zajec: if ou dont' have fbcon loaded
Zajec: airlied: ok, so loading fbcon should tell DRM to set my native mode, right?\
airlied: Zajec: you should build fbcon into the kernel
airlied: to avoid any issues :)
Zajec: damn, yesterday i was told i should use it as module :P
airlied: well as long as you load it
airlied: preferably before radeon
Zajec: airlied: .....
Zajec: i was doing modprobe radeon && sleep 3s && modprobe fbcon
airlied: try the other way
Zajec: ok, will do so
suokko: ouch. That was real hard lock. Even power button died :(
suokko: It is 10x easier to write user space memory management code :'(
suokko: (In case of error one doesn't have to remove battery and unplug power)
MrCooper: suokko: are you testing KMS on a big endian machine?
suokko: No just plain old 32_x86 machine
MrCooper: so what was the endianness discussion about?
suokko: I just looked the commit to video-ati and tought about it would be easier not to write #ifdef everywhere were one has to do swap because of different endianess in GPU and CPU
MrCooper: there's many possibilities, e.g. I was considering a loop of cpu_to_le32()s
MrCooper: it's gonna be tricky to handle endianness consistently, too bad at least the legacy display hardware can't byte-swap :(
nanonyme: MrCooper: Completely distinct code paths for different endianness out of which the right one is chosen on module loading isn't doable?
MrCooper: suokko: that was just a quick'n'dirty fix for the cursor because it tends to be mostly invisible when the byte order is wrong
suokko: ok :)
suokko: I just had idea that maybe it could be done differently
MrCooper: nanonyme: that's not really the problem, the big problem is that the display engine only supports little endian but e.g. the X server only supports native byte order
MrCooper: so we have to convert somewhere, somehow
nanonyme: Egh.
nanonyme: And only on some CPU's?
nanonyme: Sounds like pain.
MrCooper: sure, only on big endian CPUs
suokko: Sounds like display engine needs some love
nanonyme: MrCooper: Are there any big endian GPU's, btw? :)
suokko: But I guess writing endian netural display driver would be hard
MrCooper: nanonyme: no idea
suokko: At least not from ati :)
nanonyme: Right. Then probably a non-issue with this particular kernel driver.
suokko: (bridgman comes soon to tell that they have some bigendian firegl cards)
nanonyme: Just wondering if you'd have to do the conversion the other way around if you had a small endian CPU and a big endian GPU.
MrCooper: suokko: pretty sure there isn't one from any current big vendor, but e.g. SGI GPUs might have been big endian
Zajec: where can I find fbcon in make menuconfig?
airlied: framebuffer console
Zajec: somehwere in Device Drivers?
airlied: under graphics
Zajec: got it
Zajec: thx
Zajec: if I don't use vesafb/radeonfb should I cae about this option?
airlied: shouldnt' matter
MrCooper: pixman regression bisect time, bbl
glisse: airlied: for me with linus head modesetting doesn't work it picks wrong encoder dunno why, so maybe Zajec is having similar issue
Zajec: glisse: I get ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA and this is quite similar to radeonhd:
glisse: Zajec: this is not where the issue is
glisse: the issue is in drm modesetting code
Zajec: er, ups
glisse: on digital output it always pick analog encoder
Zajec: ah, at modesetting stage... didn't start hacking that yet
glisse: which of course doesn't work
airlied: glisse: ah I thknk the EDID changse
airlied: glisse: in Zajec's case that won't matter
airlied: he has LVDS conector
airlied: it'll always pick LVDS encoder
glisse: airlied: dunno why i didn't think to edid before
glisse: i have been starring at code all morning
airlied: apply MrCooper latest patch
airlied: and see i it helps
glisse: yeah edid fixes definitly help
glisse: will go through them see if i can stop bitfield error
airlied: glisse: you got the laetst patch in reply to yours?
Zajec: what ML do you use for KMS?
airlied: Zajec: dri-devel and lkml
glisse: airlied: yup this one
suokko: Some statistics about pool size when only using 2D with exa and short run of glxgears: http://pastebin.com/m4ffbf286
suokko: Total runtime before hard locking machine was about 30 seconds (from X start to hard lock)
suokko: That pool size was output everytime a ttm_tt object was freed
suokko: (after collecting the pages to pool)
glisse: suokko: i am not i understand the design of your pool
glisse: usually a pool have a size
nanonyme: Static?
glisse: and when you get request you start by using page from the pool if pool get empty then you realloc a new pool
suokko: static 1024 is max size
suokko: I jsut collect pages when they are freed to pool and then reuse them
glisse: ie pool size = 1024, you allocate 1024 page and set them wc
glisse: you get a request of 2000 page
nanonyme: Oh, neat.
glisse: you give the 1024 pages from the pool, reallocate 1024 and gives 976 from this new pool
glisse: when bo get free you add page to the pool
suokko: yes. I just didn't implement that allocate 1024 pages part :)
suokko: I just reuse freed pages
glisse: if the pool size is bigger than some level, for instance 2048, you free enough page and reset them with normal caching
glisse: idea is to never stole too much unused page
suokko: I know that half of features are missing from my implementation but I tought that should give some performance gains too
glisse: but still avoid having to reallocate page if they are gonna be needed soon again
suokko: But I have there some bug that causes instability
nanonyme: glisse: Hmm, as in compromises between memory consumption and performance?
airlied: suokko: you do know my code does this alreadY?
suokko: And the worst part is that I never get any error messages to error logs after hard lock :(
airlied: suokko: you could port it in 5 mins :)
suokko: airlied: That wouldn't be fun ;)
glisse: nanonyme: yup
suokko: I want to learn how ttm works and what I do wrong with page storing
glisse: debugging kernel is pretty much never fun
suokko: It would be better if I could run kernel in qemu and keep the underlying stable kernel runnig all the time
nanonyme: Would at best help with initial phases of development, I think. Always have to test on real hardware too.
nanonyme: (that is, if that idea involves a virtual card with GPU, registries and mem ;)
nanonyme: Still remember the C64 emulators that had to implement hardware bugs for things to work properly. :p
suokko: I was more like thinking direct acccess to GPU if possible (That would require to to cards in one machine)
nanonyme: Right.
suokko: But that kind of setup could help a lot if debugging kernel bugs
suokko: t least I learned something new today :) How to write generic algorithms in C. That list implementation is very clever :)
suokko2: A bit improvement :) Now I'm only causing GPU hand :)
Zambezi: I have Debian Squeeze (testing), Radeon HD 2400 PRO (PCI, not PCIe), kernel 2.6.26-2-686, 32 bit bit. I want big desktop to work again as it did with Debian Lenny (stable) using fglrx and amdcccle. But fglrx won't work due to a faulty kernelmodule. Will big desktop work with two 2048x1152 (4096x1152 in total) using radeon or radeonhd?
suokko2: Zambezi: If you don't need 3D then yes
suokko2: - You have to just remove fglrx.
suokko2: - Clean Xorg.conf (dpkg-reconfigure -phigh xserver-xorg)
suokko2: - Set Virtual 2048 1152 to display subsection of screen
suokko2: - Set correct output ordering/rotating with xrandr command
Zambezi: suokko2: I don't play video game, but will 2D effect resolution if I watch video? Cause now the video is like 1/16 part of my monitor.
suokko2: *Virtual 4096 1152
Zambezi: suokko2: Do I have to remove gambas2-packages?
suokko2: There is textured video that handles accelerated video scaling
suokko2: No idea about gambas
suokko2: I think no
Zambezi: suokko2: It's OpenGL-packages.
suokko2: You will have software implementation of 3D so you wil lhave oepngl but it will be slow
Zambezi: suokko2: Might be the reason why mplayer showing a smaller window?
suokko2: I don't know
suokko2: sounds like not very likely but
Zambezi: suokko2: But when I played 720p, it downscaled it or something. I have a P4.
suokko2: For me mplayer doesn't downscale even tough I get horrible tearing because my system isn't capable to play HD video
suokko2: http://wiki.debian.org/XStrikeForce/HowToRandR12 <- Here is good explenation how to use xrandr
Zambezi: suokko2: That link is opened, but first I try manual configuration.
MrCooper: glisse: any reason for r300_asic to use r100_copy_blit rather than r300_copy_dma? The latter seems faster in both directions here according to radeon.benchmark=1
nanonyme: Zambezi: 1080p might still be slow with that setup with open drivers due to the card acceleration not being completely used yet. Just thought to say since your CPU isn't really bleeding edge. :)
Zambezi: nanonyme: I haven't even tried 1080. I'm still "happy" with mine here, but I want big desktop to work. I'm preparing a pastebin cause this is too hard for me.
glisse: MrCooper: well it's not faster on all config i tested it on
glisse: MrCooper: ppc ?
MrCooper: yeah, with PCI GART
suokko: Zambezi: It should be mostly automatic if you just set the virtual size
glisse: ah, don't think i tested it with pci gart
MrCooper: VRAM -> GART is 4-5x faster with the latter, the other direction about 2x
glisse: pcie/agp where definitly slower with cpdma
Zambezi: suokko: Running xrandr as root or my regular user?
suokko: regular
nanonyme: Zambezi: I didn't notice it on all 1080p content. It's mostly on some very tightly packed h.264, I think. :)
glisse: ouch 4-5 time is a lot
MrCooper: glisse: maybe it should always benchmark and use the faster one?
glisse: MrCooper: maybe i should fire a thread at load which will do benchmark and picks the fastest
MrCooper: heh
nanonyme: Zambezi: But yeah, with luck Gallium + VDPAU will get done before you run into speed issues. ;)
nanonyme: (Or rather, get your hands on the really high-quality stuff)
glisse: MrCooper: btw out of curiosity what the number ?
MrCooper: radeon: dma 1024 bo moves of 1024kb from 2 to 4 in 1093ms (959kb/ms 959000kb/s 936M/s)
MrCooper: radeon: blit 1024 bo moves of 1024kb from 2 to 4 in 4513ms (232kb/ms 232000kb/s 226M/s)
MrCooper: radeon: dma 1024 bo moves of 1024kb from 4 to 2 in 1060ms (989kb/ms 989000kb/s 965M/s)
MrCooper: radeon: blit 1024 bo moves of 1024kb from 4 to 2 in 1903ms (551kb/ms 551000kb/s 538M/s)
Zambezi: If anyone would like to check, I'll appreciate it cause I'm new with Xorg-configuration so I have no idea where to look. http://pastebin.com/d570b470a
suokko: What is the problem?
suokko: They both are showing same image?
suokko: Then you might try: xrandr --output VGA_1 --right-of DVI-I_1
suokko: Zambezi: Only difference for me is that I don't even set the graphics driver there. Correct driver is auto detected in startup.
fpoibaf: MostAwesomeDude: the problem I had yesterday (radeon lock) vanished today with the same code...
fpoibaf: suokko: were you having the assertion failed with torcs ( http://bugs.freedesktop.org/show_bug.cgi?id=21582 )?
fpoibaf: is it fixed for you now?
suokko: fpoibaf: yes
suokko: Let me test. How long time ago fix was commited?
fpoibaf: I am still having it with sauerbraten, but with DRI1
suokko: ok. I did report another bug because it was different bug causing that torcs crash
suokko: I couldn't look backtrace when I posted it because my fonts were messed so I couldn't know if it was same or not
fpoibaf: which bug?
suokko: http://bugs.freedesktop.org/show_bug.cgi?id=22371
fpoibaf: ok, I think it's the same bug I just filed at http://bugs.freedesktop.org/show_bug.cgi?id=22438
fpoibaf: is it fixed now?
suokko: no
suokko: CAn yo uget bactrace from that crasg?
suokko: crash
fpoibaf: ok, it not only me then.
fpoibaf: I'll try now
Zambezi: suokko: Then I remove that and restart X.
suokko: Zambezi: yes. (Unless you want to force radeonhd on and radeon driver is default for your card) But radeon might have better video (tear free)
Zambezi: suokko: Both monitors shows same picture. I try your command.
suokko: yes. Default is to have monitors in "mirror" mode but xrandr can place monitors to different location in virtual desktop
suokko: (and rotate and flip the output)
Zambezi: suokko: Flip like using pivot? Cause one of the monitors is pivot.
suokko: yes
Zambezi: suokko: This virtual-line, how should I write it cause Xorg won't start with the one I wrote and not with 4096x1152 either. Do I have to write 4096 and 1152 on different lines with vertical, horizontal something?
suokko: Zambezi: http://pastebin.com/m1d2ca119
suokko: I copied my screen section and changed the resolution to match yours
Zambezi: suokko: I'll reset my last changed so I can enter Xorg. A minute.
suokko: links2 is good fast browser ;)
nanonyme: And links2 with framebuffer graphics is even more awesome. :)
Zambezi: suokko: I didn't think about that.
Zambezi: I like w3m best.
suokko: Zambezi: direct download link for wget/curl http://pastebin.com/pastebin.php?dl=m1d2ca119 :)
Zambezi: suokko: That was the problem, to write of the link. Nevermind now. I just comment three lines.
nanonyme: suokko: Mostly if it only wasn't missing JavaScript (and thus couldnt' do AJAX), could use it for an everyday browser. ^^
fpoibaf: suokko: backtrace attached to http://bugs.freedesktop.org/show_bug.cgi?id=22438
suokko: Where you need ajax? :) Most of good site anyway have a pure html version too ;)
Zambezi: nanonyme: Oh, if that could use Ajax, it would be awesome.
nanonyme: suokko: Facebook?
Zambezi: suokko: www.flashscores.com :-P
suokko: doesn't do facebook
nanonyme: It's heavily built upon AJAX.
Zambezi: suokko: Should I just add you lines?
suokko: Zambezi: Seems like your problem could be same as the compiz bug
suokko: fpoibaf: that is
suokko: Zambezi: Yo ucould replace your own screen section with it
Zambezi: suokko: My Xorg0.log is quite confusing. Randr enabled, ignore the disable-messages. And then it says disabled. A guy in #debian said I have to re-complile a module.
suokko: Zambezi: randr works like the first message sayes
suokko: fpoibaf: What mesa are you using?
fpoibaf: current mesa master
suokko: I think r20 0driver has different bugthere beucase call to bo validation comes form r200 specific code
fpoibaf: suokko: my bug seems indeed related to http://bugs.freedesktop.org/show_bug.cgi?id=21582 , but that last one is confirmed fixed while I am still having the problem
Zambezi: suokko: xdm won't start automaticlly, but it works with startx. An errormessages occur, same as with fglrx: FreeFontPath: PFE "/usr/share/fonts/X11/misc" refcount is 2, should be 1, fixing.
Zambezi: suokko: Interesting. Now the monitors named VGA-0 and DVI-0 so I just changed your command and now it says: Screen 0: minimum 320 x 200, current 2048 x 1152, maximum 4096 x 1152
suokko: I don't know what could cause that kind of problems. What did xdm log say about error?
Zambezi: suokko: No screens found it said.
suokko: I can't guess what is wrong. I did have multi-head working out of box 2 weeks ago when I have to use projector with my laptop
suokko: That virtual line and xrandr command was all that I have to do
Zambezi: New randr-output: http://pastebin.com/d35389a8a
suokko: But is output still in mirror mode?
Zambezi: suokko: Yes.
Zambezi: suokko: But maximum is now changed from 2048x2048 to 4096x1152.
suokko: Now it should be able to move one of your display next to another with xrandr
Zambezi: suokko: I tried with grandr, but it's just jumping back.
suokko: https://bugzilla.redhat.com/show_bug.cgi?id=480752 <- There is at least a bug report in Red hat bugzilla
suokko: man xrandr has a example how to set display next to each other and rotate one of them
suokko: But I hve to reboot
Zambezi: suokko1: Should I use radeon or radeonhd as driver? It's using radeon now.
adamk: Zambezi: Either one supports your GPU. If you have problems with one, you can always try the other.
suokko1: IT should be large difference. You should use one which works better. radeonhd has extra sound feature if you need it
Zambezi: adamk, suokko1: At least xorg0.log says (II) RADEON(0): Max desktop size set to 4096x1152 so I switch to radeonhd now.
suokko1: That is because of virtual line in xorg.conf
Zambezi: Still grandr says 2048x1152 is the biggest. Hopefully radeonhd will do it. Lets see soon.
suokko1: grandr says per monitor. You could also try gnome-display-settings
Zambezi: This from /var/log/Xorg0.log might be a clue? (II) RADEONHD(0): Setting screen physical size to 510 x 287
adamk: No, that has nothing to do with xrandr.
adamk: I missed your earlier discussion. What, exactly, is the problem?
suokko1: adamk: dual head doesn't work (It works only in mirror mode)
adamk: xrandr shows the correct resolutions for both monitors?
suokko1: yes
adamk: And you've tried to use the xrandr command with the --left-of and --right-of options?
adamk: Oh, and it's not even you that's having this problem, right? It's Zambezi? Just want to make sure.
Zambezi: adamk: Yes suokko1 gave me that, but no difference.
adamk: Did you check 'xrandr --verbose' to see if each output is using a different crtc?
Zambezi: Woho!
adamk: And, presumably, there were no error messages when you tried the --left-of and --right-of options?
Zambezi: Almost there. This command "disabled" VGA-monitor: xrandr --output VGA_1 --right-of DVI-I_1/digital I can move the mouse around on VGA-monitor, but not drag and drop windows or open anything. Most likely a setting.
adamk: Zambezi: Can you use a service like http://pastebin.com/ to show the full output of 'xrandr --verbose' ?
Zambezi: adamk: Will do. A minute.
Zambezi: http://pastebin.com/d245c480d
adamk: Well they appear to be running on separate CRTCs.
adamk: Which means that I am effectively out of ideas.
Zambezi: adamk: As me, but I had no idea alot earlier. And I now it Ion3 works on both. I used it with fglrx, but that driver won't work with a kernel module (or opposite way).
adamk: If you are unable to get it working properly, you may want to ask on one of the mailing lists, or even open a bug report.
Zambezi: adamk: Bug report is the last option, but I give it a day. I'm fed up with Xorg for today. :-)
adamk: So KMS has really come a long way. One issue that I still have is that if I switch to text console (which happens super fast) and then switch back to Xorg (also super fast) one of my two monitors goes into standby mode.
adamk: I used to be able to force both monitors off with 'xset dpms force off' and then turn them back on by hitting a key, but I can't even force them off now.
adamk: Any ideas?
adamk: Oooh.. Using xrandr with --off and --auto on the output gets it working again.
adamk: Still quite annoying.
ufoz: hi all
davi: What library must I install to get be below as "yes"?
davi: checking for glBegin in -lGL... no
davi: checking for glBegin in -lMesaGL... no
davi: I have installed mesa-common-dev, and so I have got
davi: checking GL/gl.h usability... yes
davi: but something is missing yet to turn the above "no" as "yes"
suokko: apt-get build-deb ?
davi: I try to build the old xmms (etch package) on Debian lenny. Because I need some of the old features.
davi: I am trying this: apt-get build-dep xmms2
davi: thanks suokko
davi: still "no"
davi: checking for glBegin in -lGL... no
davi: checking for glBegin in -lMesaGL... no
moo_: davi: apt-file search /libGL., apt-file search /libMesaGL.
davi: is installing apt-file an initializing its data base ...
davi: apt-file search /libGL
davi: lot of files
davi: trying to install libglw1-mesa-dev
suokko: libgl1-mesa-dev and libgl1-mesa-dri?
davi: Radeon hardware here, of course.
davi: suokko, After installing them, and trying ./configure again I got "yes"
davi: thanks a lot suokko
Tatsh: i figured out why i had no 3d finally
Tatsh: first of all, somehow libdri.so disappeared completely
Tatsh: so i reinstalled xorg-server
Tatsh: commented out Option "Xinerama" "true"
adamk: Please take a look at this video: It's flickering I see at 1600x1200 when KMS is enabled.
adamk: It only happens with KMS enabled, and happens both in X and in the console.
adamk: It is, for me, one of the last two big issues I have with KMS. Ideally, it wouldn't flicker and, secondly, there would be a way to specify modes for KMS to use.
airlied: adamk: is it flickering because it icks a different mode
airlied: than ums?
adamk: airlied, I tried replicating the problem in Xorg, with UMS, at both available modes (1600x1200@75 and 1600x1200@85). No flicker
adamk: Let me use xvidtune to grab the specific modes that Xorg is reporting with KMS and without.
adamk: Switching computers... brb.
mcgreg: hehe just tested a 1080p video (1920x816) h264 ... used up 40% of my system cpu time :) not that bad
mcgreg: while xorg consumes about 10% of one cpu... thats ok with such a big resolution :)
adamk: airlied, xvidtune shows the exact same mode: "1600x1200" 229.50 1600 1664 1856 2160 1200 1201 1204 1250 +hsync +vsync
airlied: adamk: must be the memory bandwith setup then
suokko: I think it would be good idea to run "indent -linux *.c *.h" in mesa/drivers/dri/r200/ Current indention is so bad mess that reading code is a bit hard in some places
MostAwesomeDude: You'll have to get everybody to sign off on that.
suokko: That might need still a few extra parameters to match mesa coding style
MostAwesomeDude: Well, TBH there is not a uniform Mesa coding style.
suokko: MostAwesomeDude: Who would oppose it? :)
MostAwesomeDude: suokko: Anybody who doesn't want to see several dozen thousand lines of change on the diffstat. :3
MostAwesomeDude: Gallium code and Mesa code look *very* different, and each individual backend driver looks different too.
suokko: They just have to use diff -b ;)
suokko: But r200 looks mess
suokko: I jsut proposaled to run it in one subdirectory to improve that mess ;)
MostAwesomeDude: But then it'll get run in other dirs, too. :T
MostAwesomeDude: Anyway, NAK unless at least airlied and MrCooper are okay with it.
suokko: I know. That why I wanted to comment here if there is huge opposition. Also there needs to be decision what parameters are given for indent
suokko: (I do now =G in vi nearly every time I open any r200 file)
otaylor: suokko: though r200/ isn't really a separate directory in radeon-rewrite
suokko: find -type l -delete && indent -linux ... *.c *.h :)
airlied: you'd have to do radeon/r200/r300 in one go
damjan: is there any experimental KMS on r600 (hd3200) I could try ?
suokko: http://pastebin.com/m582f2412 <- Indent example what I though could be good
airlied: F-11 has some basic kms in it, but we need to do a lot to get it properly working upstream.
damjan: maybe it's time to update my F-10 ... I'm dual-booting 32bit Ubuntu/ 64bit F-10 on my (well, girlfriends) desktop
damjan: I don't understand why Fedora suggests upgrading via a CD and not with yum
airlied: damjan: they also suggest preupgrade
airlied: mainly due to incompatiblities at a kernel level
damjan: what's preupgrade?
airlied: http://fedoraproject.org/wiki/PreUpgrade
airlied: greanted I mostly just use yum myself
airlied: but it can break and leave you in a bad place from time to time
airlied: glisse: btw also ned to move any function decls out into header files
airlied: esp the ones in r300.c
airlied: as if someone changes a function signature, the compiler won't catch it
airlied: dagnammit, tiled backbuffers makes me get some gear garbage