arekm: agd5f: sometimes my radeon is unable to see tv (X.org log contains no modes found via EDID)
arekm: agd5f: http://pld.pastebin.com/f23b7b96b
arekm: docs documenting RADEON_BIOS_4_SCRATCH and such weren't released, right?
rx__: there's a mention of the bios scratch regs in the atombios files i think
rx__: other than that i don't think so
arekm: agd5f: and also another issue, bright backgrounds (like sky for example) have artefacts under tv-out
PSYCHO___: is it known bug that you heibernate/suspend text in buffer apears in app on which is focus (if it is editor or something where text can be placed)?
edgecase: crxdvr, ur using gentoo build system?
edgecase: maybe install full kernel sources, check for pci_register_driver in another header with grep, and add it
CRXLPY: yes full gentoo system
CRXLPY: edgecase I am unfamiliar with how to grep through the kernel tree if that is what you mean
edgecase: you might want to find a km irc channel or mailing list then
edgecase: or google some more
edgecase: i don't have any ati grabbers here, so i'm unmotivated just now ;<
CRXLPY: ok, thanks
sossosoo: I'm totally lost with Radeon flgrx, xorg drivers, ati drivers, drm AIGLX, MesaGLX, Opel GL
agd5f: CRXLPY: I suspect the gatos code is pretty bitrotten
agd5f: arekm: tv has no edid
edgecase: maybe arkem wants to turn on load detection instead?
CRXLPY: I got hopefule agd5f, the regular tar was looking for config.h which the kernel hasnt supplied since 2.6.20 at least. so I got the cvs, which gets past that, but fails on pci.h
arekm: agd5f: s/EDID/probed modes/
agd5f: arekm: tv load detection is unreliable at the moment so you'll have to add a mode using xrandr --newmode/--addmode if it's not able to detect it as connected
arekm: agd5f: the case is that it detects modes well until I suspend/resume (that's my guess; I'll be checking that). Always after fresh boot modes are detected correctly
agd5f: arekm: ok
arekm: my camera broke so I can't show how artefacts on bright areas look like
agd5f: arekm: I think there is an existing bug for that
crxdvr: edgecase: to be sure I am not trying to fix something that aint broke. this is Xorg.0.log . it loads alot of drivers for the various parts of the AIW http://rafb.net/p/mGsrpc63.html
crxdvr: where might I find docs on all the options in the Radeon's Device section of my xorg.conf? http://rafb.net/p/Dy2mdF40.html
LinasZvirblis: Hello all.
edgecase: cxdvr: looks like Xorg is detecting everything with Xv extension. you should try $ xvinfo
crxdvr: thank you. I will work on that now. what about that huge list of options? I am familiar with the ones that relate to the radeon it self. but the rest I am clueless
LinasZvirblis: I was wondering if I should spam agd5f blog with testing results of Xpress card, or is there a better place?
edgecase: cxdvr, well it might detect things without setting options. i'd suggest $ xvinfo, it should show a port for the Theatre
edgecase: then try the recommended Xv app and see what you get!
crxdvr: here is the output. http://rafb.net/p/sSAdx861.html just an fyi edgecase , I am digging too. it shows alot of stuff but I havent matched it up with respective options yet. (I did find in my searching that /dev/video wasnt needed for reading from the tuner. I guess that dev is for composite-in)
crxdvr: thanks alot. I know you are limited in what you can do since you cant look at the hdwr or similar locally
edgecase: sure np
edgecase: when using Xvideo extension, you don't necessarily need /dev/video, that's for v4l supporty by km driver
CRXLPY: ahh , but is I want to capture, I bet I will need km, what would you think?
CRXLPY: */is/if
LinasZvirblis: DRI seems to be working with Xpress 200M now, but I am getting "Error: couldn't get an RGB, Double-buffered visual" the SECOND time I start an OpenGL application. Should I report this somewhere?
LinasZvirblis: And it is not somewhat working, but actually working well enough to play games. This just seems SO close.
agd5f: crxdvr: km is only needed for v4l compatibility. all the video in (composite, tuner, etc.) is available via Xv as well
agd5f: crxdvr: you need AVview or some other gatos tv viewer to use it though
agd5f: crxdvr: http://gatos.sourceforge.net/avview.php
CRXLPY: ok sweet, still trying ti learn which opts in xorg.conf I need to put a setting into and uncomment
CRXLPY: yeah I have been all over the gatos site. never hurts to have another pair though
agd5f: LinasZvirblis: I was mostly requesting testing for the display fifo stuff which ended up not working so I reverted it.
agd5f: crxdvr: TBH, I'm not sure the Xv code in the current radeon driver works anymore either. I've never had a AIW card. YMMV
CRXLPY: agd5f did you see my pastebins?
LinasZvirblis: agd5f, hi. I am just feeling too exited, because I got DRI working today. And it actually works better than fglrx.
agd5f: LinasZvirblis: cool :)
agd5f: crxdvr: yes. log looks fine. looks like it detects the tuner, connectors, etc. fine
agd5f: I don't think you need any options
CRXLPY: just leave the xopt #'d?
agd5f: yeah
agd5f: you only need the Xv options if the driver isn't able to detect the stuff on it's own
edgecase: crxlpy, just get avview and try it!
CRXLPY: hmmmm not in portage. more stale compiles lol
cromo: hi. I have some huge memory leak here. X takes ~500MB after an hour of running. 1.4.0.90, ati x300, 6.8.0
agd5f: crxdvr: it may be part of a gatos package
cromo: not sure wheter this is radeon or xorg related
LinasZvirblis: Do I really need openssl to build xserver?
agd5f: LinasZvirblis: IIRC, yes
CRXLPY: I have 2 2nd hand AIW's the other is rage pro (mach64) based ;)
CRXLPY: I wont be in here about the old one :D
CRXLPY: I am hunting through portage for an avview ebuild
ossososs: Hi I'm having a problem with direct rendering - glxinfo | grep direct says NO
CRXLPY: agd5f nope. there is no gatos pkg. since gatos drivers were merged with xorg as of 7.0 i am not surprised
agd5f: CRXLPY: well, not for the driver, but for the support apps
CRXLPY: whoa! I just entered X a bit ago to test the tv apps. nothing I havent run b4. but now all my VC's are seriously foob'd.
CRXLPY: the cursor stays at about 8 lines from the top. and the text on the screen will not "clear". the login prompt stays at the top. and there are about 15 lines of old text from one of the vc's that stayes on the screen regardless of the alt-F(x)
CRXLPY: it is like the screen buffer shifted it's start and end points. or worse
CRXLPY: the text I enter is invisible as well
CRXLPY: ssh terms look fine
LinasZvirblis: Ok, things stopped working. But it was nice while it lasted.
LinasZvirblis: Nope, still the same. The first time it works, and the second it fails with "Error: couldn't get an RGB, Double-buffered visual" (glxgears), "Error: couldn't find RGB GLX visual" (glxinfo), or similar. Has anyone seen this before?
LinasZvirblis: This is with drm, mesa, and radeon driver from git.
LinasZvirblis: Maybe I should be asking somewhere else?
prahal: I ported the bios code from radeon to radeonfb ... it works well ... except for the only "desktop" I use gnome ... tried with or without compiz, killing the screen saver ... any idea what could get wrong only for gnome ?
prahal: note that the current bios code is disabled in radeon ... so my next step is to try to port the current code which looks quite tied to xserver internals to radeonfb ...
LinasZvirblis: prahal, if I recall correctly GNOME has compositing manager built into Metacity.
prahal: LinasZvirblis, the thing is if I use kde or terminal session with compiz or metacity it works well :-/ wierd
LinasZvirblis: Well, that was only a wild guess.
prahal: well I ll add some of the mac code that seems unrelated to the mac to my bios one ... this one worked only it leaded to artifacts ... this is getting confusing .
arekm: agd5f: found one bug, #12170 about tv-out not working after resume
rahmen_: Hi! I can get help for my Radeon 9600 card here, right? I'm using this how-to: https://help.ubuntu.com/community/RadeonDriver
rahmen_: Anyway.. compiz doesn't work and I've followed the steps.. glxinfo says direct rendering: No, and the guide says that my card probably isn't supported by the open source driver.. But the examples are with Radeon 9600... so I just really want to know if it is supported or not.. I tried the ati drivers and everything worked well except for direct rendering... now nothing works :P not even compiz.. so I'm planning on rein
rahmen_: stalling ubuntu and try with a fresh install..
mattmatteh: rahmen_, might want to look at the trouble shooting info on xorg site. you might be missing dri permissions in xorg.conf
mattmatteh: rahmen_, line like mode 0666
mattmatteh: i was the other day
rahmen_: mattmatteh: nope.. got Section "DRI" with mode 0666
rahmen_: it's strange.. but all gui-like apps are lagging.. like when I scroll in firefox
mattmatteh: rahmen_, hmm, thats all i could think of off the top of my head
mattmatteh: rahmen_, i am not an xorg or radeon expert. i am trying to get xv working
rahmen_: mattmatteh: oki, but thnx for trying :]
mattmatteh: :)
rahmen_: I'm no expert either :P
rahmen_: but I think I'm learning pretty much from it :]
rahmen_: I "forced" a friend to get ubuntu.. and he has a nvidia graphics card.. everything's working fine there.. annoying :P though his wireless on his laptop died recently with the warranty out.. :P
mattmatteh: oh
mattmatteh: i had nvidia, just got this radeon so i can get rid of it
mattmatteh: it being the nvidia
rahmen_: :]
mattmatteh: it works, but dont like using the proprietary drivers. and my laptop has radeon and works perfectly
rahmen_: you think reinstalling ubuntu will give me a better chance fixing it?
mattmatteh: rahmen_, thats guessing. it might, but then you have no idea what the problem is, and if it happens again you will not know how to fix it. and this is not microsoft, we do not need to reinstall once a day.
mattmatteh: rahmen_, i dont think you should, see if you can find the problem.
rahmen_: oki
mattmatteh: rahmen_, i have done the reinstall method before but only when the system is really toast and i have tried to fix
rahmen_: I'll try the how-to again.. since my card _should_ be able to run it..
rahmen_: "Accelerated 3D support (r300 series) ..... 9600 / rv350 or rv360 based cards."
rahmen_: and lspci says "9600 / rv350 or rv360 based cards."
rahmen_: oops.. I meant VGA compatible controller: ATI Technologies Inc RV350 AR [Radeon 9600]
rahmen_: :]
mattmatteh: i have r350 9550 here
rahmen_: and mattmatteh... you know the difference between "Accelerated 3D support" and "Full 3D support" ? I should be able to get direct rendering on, right?
mattmatteh: for the most part it works, except i get strange dot noise at first. and xv has a blur with dark greens and dark blues
rahmen_: k
mattmatteh: rahmen_, glxgears was working ok. thats not a real test.
mattmatteh: but still messing with it, just got it the other day
rahmen_: mattmatteh: whoa.. I tried running glxgears and it logged me out :P
mattmatteh: rahmen_, x crash ?
rahmen_: yeah I guess
mattmatteh: rahmen_, might want to take notes with detail. i think you are going to have to post on a forum, mailing list, or wait for an expert here
mattmatteh: crashing is the worst
rahmen_: ok
rahmen_: I'd rather wait for an expert :P
rahmen_: but I've just installed ubuntu so don't really care if I need to reinstall :P so I'll just goof around :P
mattmatteh: rahmen_, havent used ubuntu, thinking about installing it to try it
rahmen_: I think it's nice.. as long as you're satisfied with the standard stuff... but I really want to see if I can play games like warcraft and so.. I don't play it, just curious :P
mattmatteh: rahmen_, warcraft for linux ?
rahmen_: yeah
rahmen_: mattmatteh: I play EVE Online a bit.. but if that works I guess warcraft should work
mattmatteh: i didnt think they made warcraft for linux, perhaps i am thinking of WoW.
mattmatteh: i dont play games much
rahmen_: nah they don't
rahmen_: but with wine I think you can play a few games
rahmen_: starcraft for example
mattmatteh: oh dont use wine, thats cheating
rahmen_: I play it on my laptop
rahmen_: haha
mattmatteh: and wine wont really work here anyway
rahmen_: without wine or cedega I don't think you can play any win-games
rahmen_: k
rahmen_: I had problems with fullscreen on my laptop in wine
mattmatteh: i dont play windows stuff
rahmen_: and now it seems I've lost my starcraft CD so I can't install it on this computer :P
prahal: rahmen_, do you have a virtual size in your config ? btw I wonder why 9600 would not work ... rv280 (radeon 9200) works great for years same for M24 rv370
rahmen_: I hardly do either.. but if it works I know windows sucks even more.. and that feels good.. :P
prahal: because the only way I found to crash the current driver is to have a virtual size bigger than max texture and then it only crashes in corner cases
prahal: 6.8.0 ?
rahmen_: prahal: nope.. don't have virtual.. but maybe I'll try adding it :]
mattmatteh: prahal, how do you find out if there is a virtual ?
mattmatteh: or what the virtual size is?
prahal: this is a line in the screen section telling "Virtual YYYYxYYYY" where YYYYxYYYY is a "virtual" screen size (usually bigger than the real one).
rahmen_: restarting X...
mattmatteh: prahal, i have never set that
mattmatteh: prahal, didnt know if there was a default or a way to find out the currect
mattmatteh: current
prahal: to find out the current you have to look at /etc/X11/xorg.conf and look after a line with Virtual (in the Screen section)
prahal: ubuntu may add one automatically
prahal: but it is highly unlekely that the "default" one would be bigger than 2048 (the maximum texture size) ... though I don't know ubuntu that much .
mattmatteh: prahal, that wont tell the current. that says what you want. i have never set a virtual srceen size. and when i first set up my radeon it was all messed up
ndim: shouldn't xdpyinfo report the current one?
ndim: (I'm without X11 currently, so I can't check myself)
mattmatteh: ndim, do you know what the line would look like so i can grep it ?
prahal: grep Virtual /etc/X11/xorg.conf
ndim: mattmatteh: I don't remember. And I have no working X11 on this machine...
prahal: oups sorry
ndim: I thought the issue was to find out the current VIrtual at run-time?
prahal: I am a bit slow thinking this evening ... too much hacking this week end
mattmatteh: ndim, its was
scobby: hi all
rahmen_: Anyone here who's skilled when it comes to radeon and glx? :]
rahmen_: and can anyone explain what the difference is between "glxinfo -display :0.0" and "glxinfo -display :1.0" ??
rahmen_: 0.0 has direct rendering and 1.0 hasn't.. when I type "glxinfo" (without -display) the result of 1.0 is printed.. so I guess that's the primary/standard display..?
airlied: rahmen_: what is the second carD?
rahmen_: airlied: http://pastebin.com/m1524a230
rahmen_: so I guess.. it's the same?
scobby: hi all
airlied: rahmen_: why have you got at :1.0?
airlied: rahmen_: are you doing multi-desktop?
rahmen_: airlied: What do you mean? I don't know why there's two :]
scobby: have installed newest free radeon driver but flash video are using near to 100 % cpu usage. divx dont have such a bad performance. anyone have an idee or is this normal?
airlied: rahmen_: pastebin your xorg log.
rahmen_: nope, just installed ubuntu and I'm trying to get direct rendering working
airlied: or xorg conf
rahmen_: airlied: http://pastebin.com/m4302ede7
rahmen_: can't remember the logs dir
airlied: rahmen_: are you starting two X servers?
airlied: or doing fast user switching or something?
airlied: like :1.0 shouldn't exist for you
airlied: or you should even care about it
edgecase: airlied: aside, would that work, multi-seat X w/ radeon for example
airlied: edgecase: depends..
airlied: edgecase: multiple cards vs two heads on one card etc.
rahmen_: airlied: I really don't know how many x servers I start.. it's nothing I've made.. it's the installations fault ;P I have a s-video plugged in.. could that have caused it?
edgecase: airlied, multiple cards
airlied: rahmen_: why did you use :1.0 at all? or where did you read that?
airlied: rahmen_: I'm not sure what you are trying to fix :)
airlied: edgecase: it should work ...
rahmen_: airlied: this is what glxinfo outputs: http://pastebin.com/m8aa4661 as you can see on the first row there's "name of display: :1.0" so I just tried "glxinfo -display :0.0" to see the outcome :]
rahmen_: airlied: I don't know what it means or so.. I just tried it.. I'm not that good with linux yet :]
airlied: rahmen_: are you running Xgl or something? I've no idea what you've managed or ubuntu has managed to do
airlied: it looks like two X servers are running..
prahal: airlied, thanks about the hint about porting the radeon bios code to radeonfb . Works well now and reliably. The code is disabled in radeon so I guess it is not eligible for the kernel modesetting drivers'work , isn't it ? I took a look at the active code in radeon though it sees to be much tid to xserver infrastructure
prahal: if anyone has started work on kernel modesetting driver I d be glad to have a look at the code
airlied: prahal: we'll use that code in the kernel
airlied: prahal: it not really useful in userspace yet, as we mostly use int10 to post the cards
rahmen_: airlied: I fixed so I had the ati drivers.. but since I couldn't get direct rendering to work I decided to try the open source drivers.. so I don't know, maybe that's not so good or something? =) but I think I've always had 2 diplays (or x servers or whatever :P)
airlied: rahmen_: hmm no idea why you are having two server.. I'm lost..
rahmen_: airlied: ok.. no suggestions? :P reinstall maybe? I could reinstall ubuntu and see if I still have 2 X servers..? I'm not 100% sure what a X server is.. so I'm prolly more lost than you are :P so just any advice is appriciated :] it's newly installed so I don't have anything to loose :P
airlied: rahmen_: it might be worth reinstalling and not installing the fglrx drivers initially
rahmen_: airlied: k, and then strictly go for https://help.ubuntu.com/community/RadeonDriver ?
rahmen_: you think you'll be here when I'm back and running again? :]
airlied: rahmen_: probably.. be careful with those help pages, I don't trust random ubuntu pages.
airlied: you shouldn't need to do most of that stuff it should happen by default
airlied: Option "TripleBuffer" "true" #This *might* help if you use something like Beryl and have slow video playback.
airlied: hahaqha..
airlied: radeon doesn't even have that option.
airlied: Ubongo..
rahmen_: airlied: I've been sent there by some ppl from irc.. but I prefer chatting on irc.. since many guys know what they doing here :]
rahmen_: anyway.. I'll be back later then :]
rahmen_: thnx for the help!
prahal: ok might be of interest so . http://prahal.homelinux.net/~prahal/patches/ati/radeonfb_bios_hacks.diff , this is a rough attempt at porting the code (would have been better to split the bios code in a radeon_bios file , I didn't knew how to convert nanoseconds to milliseconds in kernel for the 150us I rounded it to 1 ms and worst for the RADEON_TABLE_FLAG_DELAY: case I use nanseconds as milliseconds :( , I kept for the functions the na
prahal: me of the radeon x driver... and well I added in radeon_reinitialize_M24_bios a bunch of code that came from the mac reinitialize function and still need to narrow down which part of it is really critical ... without that bunch of lines gnome does resume with a black screen ...
prahal: and when I tell gnome I am totally lost because kde or anything else than gnome with or without compiz running works :-/
edgecase: prahal, there is work on a new kernel modesetting driver for radeon, FYI
prahal: yup I knew but didn't knew where and I needed/wanted something that worked asap ... so I did this attempt . May be of use or not for the project at stake
edgecase: cool, sounds like you've discovered a fix for something
prahal: how s that ? too late ?
DaemonDragon: Hi. Everytime I startup my computer I get this problem "fail to load nvcpl.dll". This has happened since I've uninstalled Nvidia and installed ATI.
edgecase: this channel is about linux ATI driver, not nvidia or windows
DaemonDragon: Would you be able to suggest a place to go?
rx__: DaemonDragon; you can try #ati.. but no guarantees there they might only deal with linux ati driver
scobby: anyone have an idee why flash is running so bad ?
DaemonDragon: Thanks.
scobby: i think its a radeon problem
edgecase: i have flash running slow too sometimes
edgecase: i haven't tried linux + nvidia tho
rahmen_: airlied: hi! :]
rahmen_: newly installed now
rahmen_: only thing I've done is adding xchat
rahmen_: so you recommend me to follow that guide or should I take orders from you? :P
PSYCHO___: it really uses cpu for playback, i got sometimes even 80% cpu usage when playing vid from youtube (cpu 2gh p-m)
rahmen_: airlied: good thing is that glxinfo shows display name :0.0 now.. so only one X server running now :P
airlied: rahmen_: so what do you want to do :)
rahmen_: airlied: fix direct rendering
rahmen_: glxinfo still says "direct rendering: No"
rahmen_: so I guess I want to install the open source drivers :]
airlied: rahmen_: they should be all installed by default
airlied: have a look in /var/log/Xorg.0.log
airlied: pastebin that.
airlied: I won't be here long have to get bus to office in about 2 mins..
rahmen_: airlied: http://pastebin.com/m7cfd39a6
rahmen_: airlied: "(EE) RADEON(0): [agp] AGP failed to initialize. Disabling the DRI." that's no good..?
m3thos: hi there, I have a radeon x1600 (macbook pro, 1st generation) and I was trying to use xf86-video-ati driver on a ubuntu 7.10
m3thos: did a git clone of the repo, but the configure fails on "./configure: line 20303: syntax error near unexpected token `XINERAMA,'
m3thos: log: /configure: line 20303: `XORG_DRIVER_CHECK_EXT(XINERAMA, xineramaproto)'
glisse: m3thos: you need xorg devel package of your distribution
glisse: m3thos: on debian apt-get build-dep xf86-video-ati
m3thos: unable to find source package for xf86-video-ati .. sorry.. this is my debian lameness
glisse: m3thos: on debian apt-get build-dep xf86-video-radeon
m3thos: same thing
m3thos: its xserver-xorg-video-ati
rahmen_: airlied: you're getting anywhere with the log? seems to me the problem is AGP.. "(EE) RADEON(0): [agp] AGP failed to initialize. Disabling the DRI."
rahmen_: Anyone who's good at Radeon and/or glx.. I just installed ubuntu and the x log file says: "http://pastebin.com/m7cfd39a6"
agd5f: rahmen_: do you have agp modules for your kernel?
rahmen_: agd5f: I don't know.. What does that mean? :]
rahmen_: agd5f: lsmod | grep agp shows: "intel_agp 25620 0
rahmen_: agpgart 35016 2 drm,intel_agp"
agd5f: rahmen_: you need to kernel modules for AGP support. if you built a kernel without them they won't work
rahmen_: agd5f: ok.. but the ubuntu installation cd prolly set up the kernel.. can I edit the kernel somehow?
agd5f: if you are using a stock distro kernel it should support AGP
agd5f: rahmen_: anything in dmesg?
m3thos: glisse: I've done that apt-get build-dep that you mentioned, but still same problem
m3thos: ./configure: line 20303: syntax error near unexpected token `XINERAMA,'
m3thos: ./configure: line 20303: `XORG_DRIVER_CHECK_EXT(XINERAMA, xineramaproto)'
m3thos: I probably have to instal some xinerama library
glisse: m3thos: apt-get install xserver-xorg-dev
agd5f: m3thos: you need the xserver macros. I don't know what package includes those
rahmen_: agd5f: dmesg: http://pastebin.com/m1b0f6a1b
m3thos: argh.. that pkg is allready installed, thanks anyway glisse,
rahmen_: agd5f: only line with agp in: "[ 44.882031] Linux agpgart interface v0.102 (c) Dave Jones"
agd5f: rahmen_: drm:radeon_cp_init] *ERROR* radeon_cp_init called without lock held
agd5f: [drm:drm_unlock] *ERROR* Process 5254 using kernel context 0
m3thos: glisse: ahh.. thanks.. I just needed to do a ./autogen.sh
m3thos: god damn I have autotools
m3thos: s/have/hate
rahmen_: agd5f: ok.. so.. I don't know what that means :P
m3thos: humm.. hell.. now what do I do.. I don't find any .so, .ko or any binary that looks like a driver
m3thos: just did make on the dir..
rahmen_: agd5f: you're working on it or.. well.. what should I do? =)
m3thos: ./autogen.sh && make -j3 .. now what ?
m3thos: Is copying the radeon_drv.so to /usr/lib/xorg/modules/drivers, setting up xorg.conf to use radeon and restart X all that is needed ?
glisse: m3thos: make install should do it for you but basicly this what needed
leio: agd5f: can I take you up on the claim that r100/r200 have full EXA render support and get you to fix some missing accelerations there or help me with it? :)
rahmen_: agd5f: https://launchpad.net/ubuntu/+source/linux-source-2.6.20/+bug/78684 <-- reading this.. same problem.. and seems like old kernels work but not newer.. haven't read all though
rahmen_: agd5f: and btw.. if you've given up - tell me =) otherwise I'll keep spamming you ;P
m3thos: glisse: should I rmmmod the fglrx driver, or modprobe any other one?
glisse: m3thos: depend on your card if r3xx/r4xx then removing fglrx module and add radeon module will give you 3d otherwise doesn't hurt
glisse: ubuntu should have a clean way to uninstall fglrx
glisse: you can reinstall after if you want
m3thos: glisse: my card is radeon x1600
rahmen_: agd5f: "I've removed the following modules and agp works again with the kernel 2.6.20-14:" <-- is this smart?
glisse: m3thos: then no 3d so no need to bother too much about fglrx kernel module
m3thos: glisse: I think I have a potential bug report
m3thos: :D
glisse: m3thos: what is it ?
m3thos: I have X, at 1280x800 (native resolution is 1440x900), but every pixel has a something like a ghost image 10 pixels to the right
m3thos: like.. for instance the mouse pointer.. has two "tails"
glisse: likely apple issue, which won't be fixed, blame apple
glisse: fglrx works ?
m3thos: yest
glisse: then radeon is doing somethings wrong
m3thos: at native resolution
m3thos: even that vesa doesn't do this
glisse: do you got xorg log, pastebin
m3thos: will do
agd5f: rahmen_: sorry, busy with something else. that bug looks like the same issue.
rahmen_: agd5f: yeah, I'm reading every post.. so hopefully they come up with something :] thnx for the help anyway!
agd5f: rahmen_: sure
m3thos: http://pastebin.com/m37fc7a35
m3thos: glisse: I've remove the fglrx driver, and changed the monitor definition on xorg.conf .. and everything looks really good!!
m3thos: opensource 1440x900
agd5f: leio: sure
m3thos: YeY!
leio: agd5f: that's on which one? :) Basically the problem is that with a modern GNOME it hits two fallbacks, killing any benefits EXA could give over XAA
leio: (at least that's why I think it's sluggish, especially with "always" heuristics)
agd5f: leio: what chip are you using?
leio: 01:00.0 VGA compatible controller [0300]: ATI Technologies Inc RV280 [Radeon 9200 PRO] [1002:5960] (rev 01) (prog-if 00 [VGA controller])
agd5f: ok
leio: R200CheckComposite: Component alpha not supported with source alpha and source value blending.
leio: R200CheckCompositeTexture: NPOT repeat unsupported (8x22)
leio: I'm being told the NPOT one _might_ be fixed or relatively better behaving with a git xorg-server (think it was something about it redoing the op to something else that can be accelerated, but that isn't so good as the straight thing)
leio: the hex in the NPOT one is varying. Can the component alpha one be fixed on that chip?
agd5f: the npot repeat should be better in xserver git
m3thos: humm.. I guess I have a new bug report
leio: too bad xserver RC2 is over a week overtime. Would getting NPOT fixed make it much better for now? Out of the fallbacks in total component alpha happens 80% of the time, and NPOT 20% of the time, roughly
agd5f: possibly. we'd need to take a look at the pixel shader in r200. I suspect it's not capable if we didn't already implement it
rahmen_: I need help (probable expert help) with my direct rendering. I get the same error as they do in this thread: https://launchpad.net/ubuntu/+source/linux-source-2.6.20/+bug/78684 I'm not sure what to do - test a few of their tips or keep on looking for other answers (the threads are pretty old (from 2007) so I'm a bit worried it won't work..)
agd5f: are those really the problems though?
leio: agd5f: you mean me? Well, with "greedy" it's better, with "always" totally awful (though somewhat better with EXAOptimization). How would I find out what the problem is?
agd5f: leio: use oprofile and see where most of the time is spent
leio: I used sysprof and found with "always" heuristics half of the time is spent in doExaMigration
agd5f: leio: you can also remove those checks and see how that affects thing (rendering of certain things will be wrong)
leio: checks to fall back?
agd5f: leio: in radeon_exa_render.c
agd5f: R200CheckComposite()
leio: my biggest problem is the unmap/map of stuff due to workspace switches with metacity. I use that often and it takes 3 seconds or more and I can see stuff drawing from top to bottom
agd5f: leio: what size desktop?
leio: 1600x1200, with framebuffer at 1600x1600
leio: (I have a pivot, that I'd like to eventually have auto-rotate in GNOME from the DVI interrupt that's supposed to happen)
agd5f: ok
leio: agd5f: should I do all those tests with "always" migration heuristics or?
agd5f: leio: remove the option and stick with the default
leio: ok, that's what gives 50% in doExaMigration, I'll see with EXAOptimization true
leio: EXAOptimizeMigration*
leio: agd5f: 15.73% in exaComposite, and takes 2-3 seconds to switch workspace fully (rectangles draw visibly one by one as said before), but CPU isn't 100% used up interestingly
agd5f: leio: you you have the full profile?
leio: from sysprof-1.0.9 that doesn't see kernel level stuff
agd5f: yeah, that's fine
m3thos: I get hard lockups (no ssh, no screen/keyboard response, no pings) when trying to play videos on a radeon x1600
leio: agd5f: want a copy?
agd5f: leio: sure
agd5f: can you pastebin?
leio: well, it's 100k lines
leio: of XML, so :)
agd5f: m3thos: are you using the drm?
leio: so you probably want it by file to load up in sysprof
agd5f: leio: I really just wanted to see the top 10 or 15 on the list, but go ahead and send it
leio: the XML itself isn't ordered it seems. I don't have your e-mail address offhand
agd5f: alexdeucher at gmail dot com
agd5f: leio: if you remove the check in R200CheckComposite() how does the performance change?
leio: I'll send the XML off and give it a try
leio: on its way
leio: agd5f: I'll define RADEON_TRACE_FALL again too I suppose
airlied: MostAwesomeDude: hey.. how's things going?
leio: agd5f: very interesting fonts with that ;p
leio: (exclusively white)
agd5f: leio: is it any faster?
leio: oh, now it's full of R200TextureSetupCP: Width 8 and pitch 64 not compatible for repeat
leio: doesn't really feel like it, still 1-3 seconds or so
agd5f: those fallbacks are probably not the issue then
m3thos: agd5f: yes, I modprobed radeon ...
leio: can it be found out what is? :)
agd5f: m3thos: does teh log show direct rendering enabled?
agd5f: leio: oprofile
leio: seems like a lot of time goes to glyph rendering, I thought that's supposed to be a lot better now with carl's work
leio: agd5f: sysprof from SVN fine too? That sees kernel, though I don't see system time being used
airlied: leio: it is better, still more better to be done.. also I doubt radeon is tuned so well.
agd5f: leio: whatever. just something to see where time is spent
leio: I just wish the time spent on X so far, so I can work on optimizing GNOME better
leio: err, X so much*
leio: weren't spent in X so far. /me gets more coffee
leio: so much*, gah
agd5f: leio: exa and glyphs are currently a loss
agd5f: leio: fredrik and MrCooper were looking at improving the migration checking with a rolling counter
leio: shouldn't I be seeing exaDoMigration at the top then?
agd5f: the EXA core needs better optimization in many places
m3thos: agd5f: should I try without dri ? it rebooted now I'm not with dri enabled now (no radeon kernel driver loaded).. should I try now ?
agd5f: sure
woffle: Hey
woffle: Using `dmesg | grep drm I get this: [drm] Initialized drm 1.1.0 20060810
woffle: Does that mean it's working or not?
m3thos: agd5f: hard lockup on login.. dri is not enabled
agd5f: woffle: look for a message about radeon
woffle: agd5f, there isn't one
m3thos: agd5f: I can profile a xorg.conf and Xorg.0.log
leio: brb with readable fonts
agd5f: woffle: then the radeon drm isn't loaded
woffle: hmm
MostAwesomeDude: airlied: Yo. Real life is a drag sometimes. Noticed you guys have true vert progs now.
airlied: MostAwesomeDude: that's just cleanups from the docs mostly..
agd5f: m3thos: are you sure the dri was enabled?
m3thos: agd5f: I'm shure that dri was _disabled_ this time
airlied: MostAwesomeDude: r300 and r500 vertex engines are very similiar
agd5f: m3thos: right, but it probably has a better chance of working if you ave the dri enabled
m3thos: I have to use drm from git, no ?
agd5f: yes
m3thos: I'm guessing that the kernel drivers on the stock ubuntu 7.10 kernel won't have what it takes
agd5f: m3thos: not for r5xx
m3thos: okay, will do that
m3thos: what's the svn info equiv of git ? (or the equiv of hg path)
leio: to see what? git config --list maybe?
m3thos: the uri used in git pull etc..
woffle: Just how big is git://anongit.freedesktop.org/git/mesa/drm ? It's moving at stupidly slow speeds here, so I'm trying to guess how long it'll take to complete
agd5f: git clone git://anongit.freedesktop.org/git/mesa/drm
agd5f: looks like someone needs to kick fdo
agd5f: woffle: it's not very big
woffle: ok
woffle: Now, I also need to enable this in my kernel right?
agd5f: woffle: if you are using a distro kernel, you can just replace the kernel ones with the ones from git
woffle: I need to build it myself, source based distro
agd5f: woffle: you can build the drm against whatever kernel you are currently running as long as the drm is enabled as a module in that kernel
woffle: Ok, must be a module? Not builtin?
m3thos: module
woffle: Ok
agd5f: woffle: much easier that way
agd5f: if it's built in you have to rebuilt the entire kernel to update the drm
GhotiPhud: how would I do that on Ubuntu 8.04? or would that be included already?
m3thos: GhotiPhud: would be included, has a module
woffle: I might wait to rebuild, because currently it's builtin
GhotiPhud: would I need to built from git to test out the latest radeon driver?
airlied: hmm have they r500 drm support in their kernel.. it not in a release kernel yet..
airlied: I think 2.6.25 is the first one it'll be in.
airlied: but maybe I put it in 2.6.24 I don't remember
GhotiPhud: I built the radeon driver from source a few days ago to test out Textured Video
GhotiPhud: would there be a benefit to building the mesa/drm?
leio: agd5f: got sysprof from SVN now. There's an "Everything" now that is 100% Total
m3thos: on r5xx (macbook pro, bios emulation), should I enable/disable something in xorg.conf ?
agd5f: m3thos: are you using bootcamp of EFI?
m3thos: I'm using bootcamp, efi loads the bios emulation on bootup
agd5f: ok, should be fine. like I said, try with the frm from git
agd5f: *drm
m3thos: okay
leio: ExaOffscreenMarkUsed at 10%, hmm
agd5f: leio: that's the current bottleneck. fredrik and MrCooper were discussing some improvements
leio: but it's just 10% of self time, 10% is small in my book for such a slow thing, though maybe many things need small optimizations
leio: would it have to migrate at all if those two things were accelerated?
airlied: leio: it gets larger the more pixmaps you have.
leio: it does feel to be slower over time
leio: with "greedy" it got 1.6% of time in exaDoMigration only
leio: some days ago, still same old slowness
woffle: I wonder if my ISP is throttling git or something... I'm getting 5KB/s :/
m3thos: woffle: I've pulled from git just right now, no problem at all.. I'm in Portugal FYI
woffle: Is there an alternative?
m3thos: amazing
m3thos: radeon + radeon_drm (using exa)
m3thos: and I'm playing x264 flawlessly
m3thos: better that what I was ketting with fglrx
m3thos: moving windows and switching desktop
m3thos: is quite slow though....
m3thos: humm.. colors look worse
leio: guess why I'm here ;)
m3thos: it looks like its 16bit only
m3thos: well.. "fat props" for you hackers!
m3thos: leio: good luck with that profiling and testing :)
leio: trying git xorg-server now to hopefully get rid of that NPOT fallback and see how it is then
leio: a few obvious optimizations possible in evo and xchat code too to do in the just begun week
m3thos: totem
m3thos: totem is still slower than mplayer..
m3thos: this h264 movies I have..its really borderline... it stalls a bit on some scenes in totem
m3thos: just enough to be anoying
m3thos: but on mplayer... with the right flags... all fine & dandy
woffle: Is there somewhere you can download source snapshots?
leio: gst-plugins-ffmpeg doesn't have a good released version
m3thos: probably any gentoo mirror, in the distfiles
leio: for x264
leio: albeit 0.10.3 is much better already than before. CVS should be even better.
m3thos: its really borderline.. 100% cpu on totem
m3thos: 90% on mplayer
m3thos: something like that
leio: ffmpeg snapshot in gst-plugins-ffmpeg-0.10.3 is a lot older than mplayers snapshot, there have been a lot of 1-2% optimizations for H264 in it inbetween. And all this is off-topic here I think :)
m3thos: okay, sorry
leio: well, I dunno if anyone minds about that :)
m3thos: anything known about image quality issues ?
m3thos: it might be placebo effect, but radeon looks it has worse image quality...
m3thos: I'm starting to see thrashing in the images
m3thos: lost pixels
m3thos: lost something.. I don't know.. the look like movie render gliphs or something.. I don't know much about video stuff
leio: agd5f: still as many NPOT fallbacks with git xserver, it seems
m3thos: yey! it suspends!!!
m3thos: my laptop suspends (and resumes) with the radeon driver.. no problem what so ever
dli: m3thos, it suspends to ram for me, no problem
MostAwesomeDude: Hmm. Maybe I should try suspend again.
airlied: agd5f: https://bugzilla.redhat.com/show_bug.cgi?id=428515
airlied: agd5f: any ideas? can we fix it so it doesn't see two VGAs with different ddcs?
agd5f: airlied: looks like a bad connector table
agd5f: we could add a quirk
airlied: agd5f: where are the quriks hiding now?
agd5f: airlied: we don't have any at the moment
airlied: ah explains why I couldn't find any..
airlied: ah the XPRESS one is inline now.
airlied: I might splt them out into a function
agd5f: yeah. xpress chips don't have a crt2_ddc line
agd5f: it maps to monid_ddc
agd5f: but I haven't sorted out how the pins work in all cases yet
airlied: agd5f: I think I'm still seeing some mac related bugs, but I'm asking to retest with latest code.
agd5f: the connector table option may be busted since we merged the atom code. I haven't tested it in a while
leio: agd5f: whoops, 122MB Xorg.0.log here by now. Sad part is that I get the alpha component stuff even with ugly gtk (builtin I believe) simple theme. No NPOT from that though it seems, but still with clearlooks with git xserver
agd5f: leio: I don't think those are the cause of the problem.
leio: right, mainly saying the NPOT fallbacks don't go away with git xserver as was expected