adam_yarter: zcrayfish: They're good when you're single. lol
adam_yarter: here we go, lets go break the system
dmb__: whats the name of that x server written from scratch by someone from redhat/fedora?
amarks: wayland?
airlied: dmb__: not an X server, a display server, wayland
amarks: does that project have momentum?
dmb__: airlied, does that work with radeon kms yet?
airlied: dmb__: I tihnk it needs some userspace accel code that no-one knows the proper solution to yet
dmb__: oh
dmb__: i would love to see a non x server :)
amarks: how would this help you?
airlied: you get bored of running one app pretty quick
airlied: or maybe he has two going
dmb__: amarks, x is very complicated
amarks: complicated for a reason
dmb__: seems like there are some design issues with the x protocol
dmb__: but i don't know much about that
zcrayfish: Ever tried KDrive? (TinyX)
dmb__: eh
dmb__: i'v seen it
EruditeHermit: airlied, did you have a chance to look at kms on rv350 at all?
airlied: EruditeHermit: I've been away, haven't done anything since a while
EruditeHermit: ah ok
amarks: when using cgit to say browse drm-next, is there an easy way to see where in the shortlog the sync point of the last git pull for the kernel?
airlied: usually wherever it merges linus tree but it hasn't synced in ages
amarks: airlied: are you planning a sync soon?
airlied: amarks: not really, generally it works the otyher way
airlied: people pull my tree into theiers
amarks: airlied: will i see a green label for the merge in cgit?
amarks: don't you ask linus to pull it in when your ready?
airlied: amarks: yes as I said people pull my tree, I just keep adding patches on top of it
airlied: sometimes it conflicts with Linus then I fix it up for him
airlied: I should probably pull something newer into drm-next but I'm lazy and I have it on about 10 test boxes
amarks: hmm ok i assumed you would push when ready (for the kernel)
amarks: i'm not across the process
amarks: just trying to understand how your fixes end up in the kernel
airlied: I just push the patches that are ready in git, and Linus gets those patches
airlied: it doesn't matter if I never sync my tree with his
airlied: he just gets the patches I've added since he last pulled
amarks: right yes so there is a manual step
amarks: does he pull from a number of places regularly?
amarks: during the rc cycle
airlied: no I send a request to him via email when I'm ready
amarks: gotcha
amarks: i need to get across git properly, our development mainly lives in clearcase
dmb__: any of you know how important the amp requirement is on the cards?
slavka: hello,all!
ajami: Ever since the 7.6 mesa merge, I have been getting hardlocks with my video card. rv610, mesa git, drm git. I have been using KMS, but I just disabled it right now to see if that will fix anything. Also, with that merge, I have been getting bad corruption with flash videos/games/etc.
ajami: kms disabled, no video corruption.
ajami: still using mesa git, drm git, though I am using the latest (2.6.31.5) rawhide kernel.
ajami: I didn't have corruption before though.
ajami: commit 46a784b7fd2dcab97c26d5ca02fb8f74945ba157
ajami: 2.6.31.5-96.fc12.i686.PAE is the 'uname -r' of my kernel
spreeuw: me too ajami
spreeuw: [drm:radeon_cs_ioctl] *ERROR* cs->dwords too big: 25412
spreeuw: getting these
spreeuw: in dmesg
ajami: lol was going to check my dmesg, but kms is disabled, won't do any good. lol
glisse: airlied: it seems we have a lot of r6xx/r7xx issues
ajami: glisse:with the merge, but kms disabled, everything seems to be working smoothly. Although, having kms would be nice.
airlied: glisse: which issues -)
glisse: ajami: aside your issues, i have few bugs report of freeze just after gdm
glisse: airlied: :)
airlied: glisse: I think I have a fix for that one
airlied: glisse: if its FEdora kernel
airlied: if its upstream not sure
glisse: what is the fix ?
airlied: workaround is to boot with aspm turned off
airlied: kdekorte tracked it down for me
airlied: normal Linus kernel doesn't enable aspm by default
glisse: it doesn't seems to work with everyone
airlied: but Fedora kernel does
glisse: at least i learned the existence of this aspm
airlied: the kernel I built today should be a good test
ajami: aspm?
glisse: ajami: automatic somethings power management
airlied: 2.6.31.5-97.fc12
ajami: ah
airlied: glisse: get them to try with that, in theory r600/r700 should disable aspm now
ajami: I am using 2.6.31.5-96.fc12.i686.PAE
airlied: ajami: corruption should also be fixed in -97
airlied: http://kojipkgs.fedoraproject.org/packages/kernel/2.6.31.5/97.fc12/
ajami: where is the kernel-firmware package?
ajami: or the kernel-PAE-headers package?
ajami: airlied: at least the kernel-firmware package is needed.
airlied: ajami: all in there
airlied: noarch subdri
ajami: ah, thanks.
ajami: what changed from -96 to -97?
airlied: ajami: more radeon driver fixes
ajami: ah
glisse: airlied: we also have a couple of crtc monitor detection issue with kms
spreeuw: are the fixes also in mainline kernel?
spreeuw: or is this kms specific, I'm using dri1
airlied: spreeuw: all kms
airlied: glisse: any bug numbers?
airlied: glisse: I left secondary dac detect enabled by default
airlied: to hunt down the broken list better
glisse: damm can't find back the bz
glisse: it was a monitor setup in 1600x1200
glisse: highest crtc resolution
glisse: but best resolution is 1400x1024 or somethings
glisse: so use was complaining that it didn't pickup that mode automaticly
glisse: and there is another one about broken edid monitor
airlied: glisse: ah yes crtc monitor res picking is nearly always wrong
airlied: I think ajax has some ideas we might ask him
glisse: anyway right now i don't think this bugs hurt too much :)
airlied: if something appears on screen \o/
airlied: if nothing appears thats serious ;-)
DanaG: sorry about that... stupid Pidgin let me try to start it like 10 times at once, with no launch feedback.
spreeuw: launch feedback is for noobs
spreeuw: oops
DanaG: I was clicking "start pidgin" in avant-window-navigator, thinking it would pull up the buddy list. And when it didn't... I clicked it again. And it gave no feedback that anything was happening.
DanaG: But nope... it waited until I'd clicked about 10 times... and then OPENED about 10 times.
spreeuw: dont you love DEs
DanaG: goes right back to gnome-panel.
DanaG: Heartbeat LED is very handy for telling when things are dead.
DanaG: I was moving stuff from ntfs-3g to a truecrypt volume on the same ntfs-3g partition.... and the system essentially had a "heart attack".
DanaG: That is, it got so far bogged down that the heartbeat LED stopped... for like 10 or so seconds.
spreeuw: hehe
ajami: airlied, glisse, -97 fails as KMS should be starting. I cannot boot with -97
DanaG: Anyway, off to bed I go.
ajami: rhgb or not, it fails
DanaG: Mon Oct 26 02:14:09 PDT 2009
ajami: screens go blank and the box just sits there.
spreeuw: sleep well
airlied: ajami: damn that isn't the desired result
airlied: ajami: what gpu? agp/pcie?
ajami: rv610 agp
ajami: I know, agp, bad, but I haven't gotten a new box in almost 5 years....
airlied: does it boot with radeon.agpmode=-1 on the kernel command line?
ajami: didn't try. will try now, brb
ajami: airlied: shall I try with kms?
airlied: ajami: that shoud be kms
ajami: k, brb
mikkoc: airlied: could u help me with these errors im seeing that cause corruption? [drm:radeon_cp_indirect] *ERROR* sending pending buffer 17
mikkoc: here's the bug i opened http://bugs.freedesktop.org/show_bug.cgi?id=24300
ajami: airlied: no go
ajami: wait, crap i didn't see the "-" in =-1, brb.
airlied: mikkoc: thats non-kms I don't deal with that.
mikkoc: airlied: ok, then please help with with making kms work :D
mikkoc: http://bugs.freedesktop.org/show_bug.cgi?id=24218
mikkoc: since X doesn't work when kms is enabled
airlied: wonders if MSI broke on AGP
airlied: shouldn't have though
mikkoc: btw, my chipsed is rs880, but it loads rs780.. which one should i report when opening bugs?
airlied: rs880, it loads rs780 microcode
airlied: trying drm-linus or drm-next tree might be worth it again
airlied: I've pushed a bunch of changes toda
mikkoc: airlied: i just did, no go
mikkoc: well, i patched 2.6.32-rc5 with all the patches up until today's
airlied: mikkoc: up to the msi patch?
mikkoc: until drm: Add async event synchronization for drmWaitVblank
airlied: ah cool thats latest then
mikkoc: yes
airlied: I should test on my rs780 tomorrow
mikkoc: could it be that my rs880 needs some special treatment?
airlied: shouldn't
mikkoc: it's a fairly new asus mobo
mikkoc: hm ok
airlied: wonders if ASPM could be involved there as well
airlied: try booting with pcie_aspm=off
airlied: thou thats just a stab in the dark
airlied: dinnesr &
ajami: airlied: radeon.agpmode=-1 is a no go...
ajami: airlied: shall I now try with pcie_aspm=off?'
glisse: ajami: doesn't hurt to try
ajami: glisse, airlied, still no go. Just a blank screen.
glisse: ajami: did you opened a bug for it ?
ajami: not yet, I can
glisse: ajami: it's easier for us to trakc bug than to track irc
glisse: ajami: does the computer freeze or can you log in through ssh ?
ajami: hmm, didn't even think to try ssh, I will try that before I report the bug.
glisse: ajami: if you can login through ssh please open a bug and attach dmesg of failed kms
glisse: with output of lspci -vnnxx
ajami: kk
ajami: If I can't...?
DanaG: Serial console, maybe?
DanaG: If possible.
ajami: nope, not for me
glisse: ajami: well reboot in non kms and kms dmesg might be in /var/log/messages
glisse: hopefully
ajami: ok rebooting, brb
mikkoc: airlied: in case you were referring to me, my kernel already has # CONFIG_PCIEASPM is not set
ajami: glisse, airlied
ajami: glisse, airlied couldn't ssh in.
airlied: wonders what we regressed
airlied: ajami: so -96 boots, 97 doesn't?
ajami: -97 does with nomodeset
ajami: without nomodeset, -97 fails to boot
ajami: airlied: I'm opening a bug on it in fedora's bugzilla atm.
airlied: glisse: how well do your r6xx agp cards work ;-)
ajami: Now that I think about it, I probably should have named the bug better, oh well; here it is: https://bugzilla.redhat.com/show_bug.cgi?id=530961
ajami: airlied, glisse nothing in /var/log/ to help out either.
glisse: airlied: they worked quite well when i did the support for them
glisse: well at least suspend/resume
glisse: glxgears
glisse: + compiz iirc
ajami: glisse, airlied, I don't know how un/common this issue has come up, but a lot of visual elements seem slower/more sluggish with KMS enabled. Is that known about, or is it just me?
glisse: it's know
ajami: WOW, gnome-shell does not work without KMS...
airlied: ajami: so -96 modeset works?
ajami: airlied, yes, but there is flash corruption and the occasional hardlock, but yes, I can boot with modeset in -96
airlied: can you attach dmesg from that kernel to that bug?
ajami: -97 dmesg or -96?
airlied: the -96 one, and -97 nomodeset as well
ajami: So -96 modeset dmesg and -97 nomodeset dmesg...?
airlied: yup
ajami: ok be back in a minute.
ajami: airlied: both are attached to the bug
ajami: airlied, glisse, why does KMS being enabled slow things down?
airlied: in theo -96 and aove should speed back some stuff up
glisse: no optimization
airlied: since the kernels have a lot less debugging enabled
airlied: kms also has a bit more overhead
ajami: ah
airlied: though for r600 we really need to do better ddx optimisation
airlied: we emit wa too much now
ajami: So long as it's usefull, then leave it.
ajami: I forgot to mention, with KMS enabled, flickering occurs when resolution is being changed on my CRT, but not on my LCD.
virtuald: does x set the last activated monitor as the primary monitor?
airlied: nothing picks primary by default
virtuald: ok
airlied: though the first crtc is normally used
virtuald: i'm just wondering what puts my gnome-panels on my second monitor (dvi-1) when i remove the configuration or switch between ums and kms
airlied: its just ordering of when connectors are added to the list
ajami: airlied, glisse, any reason for the CRT flickering on res change, but not the LCD with KMS enabled?
airlied: ajami: CRTs are crap :-)
virtuald: dvi-1 is always below dvi-0 in xrandr
airlied: virtuald: xrandr --verbose to see which is crtc 0 assigned
virtuald: i mean the opposite :>
ajami: airlied, but without KMS, it doesn't flicker...
airlied: glisse: if you get a chance can ou try and boot an agp r600 with -97 and/or drm-next
airlied: I booted my agp rv280 to fbcon at least
glisse: airlied: i will
virtuald: restarted x with minimal conf (just virtual size set)
virtuald: xrandr says it's using CRTC 0 on DVI-1 and CRTC 1 on DVI-0
virtuald: i'm on ubuntu 9.10 btw
stumped42: should i be able to use 3d acceeration on linux if i have a 4850?
virtuald: airlied: does it matter?
ajami: airlied: find anything useful in the dmesges?
ajami: mmmhmm delicious 4am, only 4 hours to go and I will be sitting in class bored...
airlied: ajami: nothing to explain where it all went wrong
virtuald: i guess i'll just move the cables
lordheavy: http://www.phoronix.com/scan.php?page=article&item=fedora_r600_3d&num=1 looklike a pretty good success :-)
agd5f: eosie: you should use num_z_pipes for all chips. see the classic mesa driver
glisse: agd5f: i am having trouble with blit, i am trying to blit 256x256 to 1280 but i have to set blit dst pitch to 1280 / 64 instead of (1280 *4) / 64
glisse: even doing that i got 4 256x256 square
agd5f: glisse: do you have the depth set correctly?
glisse: changing dstx,dsty seems to work properly and set things to the right place
glisse: i think i do
agd5f: glisse: also, IIRC, radeons usually use aligned width for struide rather than bytes, but it's been a while since I looked at 2D
glisse: what's strange is that Xorg configure the crtc slightly differently
agd5f: glisse: crtc?
glisse: i am doing a standalone blit test program with kms
glisse: to hunt down agp corruption issue
glisse: btw do you have any info on AGP_FLUSH* reg ? :)
agd5f: glisse: I asked a while ago, and apparently we don't use them
agd5f: glisse: the pitch_offset regs take pitch in 64 byte units
glisse: yup i saw that i think it's the crtc programming i got wrong
glisse: damm i spend 1 day looking at the wrong place for the bug :(
agd5f: airlied: you mean the kernel disables msi by default on rs480?
Kano: agd5f: is there a fedora iso with .32 kernel?
agd5f: Kano: don't know
glisse: Kano: lastest fedora kernel got pretty much all the lastest drm/radeon kms bit of 2.6.32
Kano: i need intel too
glisse: http://kojipkgs.fedoraproject.org/packages/kernel/2.6.31.5/97.fc12/
glisse: i think same apply for intel
Kano: good
kdekorte: Is anyone else seeing screen corruption in a qemu guest with the latest fedora r600 drivers. It appears to be similar to the corruption seen in flash apps
tobydeh: Hi, could someone help me use the radeon driver for my ATI 4550 hd please?
tobydeh: I cant get the DVI output to work, its not reporting the correct resolutions on EDID
ajax: what is it reporting? can you pastebin your X log?
tobydeh: hmm ill try, im on my laptop you see...
Rikki-Tikki-Tavi: hi - small question ...
tobydeh: ajax: just getting it for you now
tobydeh: ajax: Also, when using the os drivers, i dont have an xorg.conf file
Rikki-Tikki-Tavi: when I run glxinfo as a user I get "software rasterizer"
tobydeh: when i try to make one gdm wont start
ajax: tobydeh: that's fine, i just want the log
tobydeh: ok 2 mins
Rikki-Tikki-Tavi: when I run it as root, I get_ Mesa DRI R600...
Rikki-Tikki-Tavi: normal?
adamk: Rikki-Tikki-Tavi: Nope... What group owns /dev/dri/card0?
adamk: And do they have read-write acceess to it?
Rikki-Tikki-Tavi: root
Rikki-Tikki-Tavi: brb
Rikki-Tikki-Tavi: only root an group have read/write
adamk: Right, but what is the group? You said 'root'. Is that the name of the group?
tobydeh: ajax: log is here: http://pastie.org/670149
Rikki-Tikki-Tavi: oh, I c now
Rikki-Tikki-Tavi: I'm on gentoo, the group is video, and appearently my user is not in that group
Rikki-Tikki-Tavi: thanks a bunch
ajax: tobydeh: mmm. looks like your monitor wants 1080i, and we're too dumb to give it that.
tobydeh: it also does 720p
tobydeh: its on hd ready not full hd
tobydeh: only*
tobydeh: the dell 17" on vga works fine
tobydeh: but the dvi 32" sony isnt playing ball
spreeuw: only sony
tobydeh: i dont know where 1920x540 has come from?? thats not a supported res so i dont know whay its in the edid
ajax: that's what 1080i is.
ajax: 540 lines, interlaced up by two.
tobydeh: oh ok
tobydeh: i thought it was 1920x1080
tobydeh: i see
ajax: yeah, edid parser bug.
tobydeh: what are my options?
ajax: which means it's my fault. i should work on fixing that today.
tobydeh: :(
tobydeh: the binary drivers SUCK
tobydeh: video is waaay to slow
spreeuw: sure its a video problem?
tobydeh: honestly - no
spreeuw: you have enough decoding power?
spreeuw: you need an i7 or so
Kano: or vdpau ;)
tobydeh: mythtv plays back video ok (not amazing but ok)
spreeuw: with high clock
tobydeh: but live tv judders
tobydeh: and it only improves when i lower the quality
tobydeh: so i assumed the drivers for the card wernt up to much
tobydeh: vdpau in myth didnt work at all for me, just crashs
Kano: works fine with mplayer. i guess your signal is bad
Kano: use latest beta driver of course
tobydeh: no, signal is ok
tobydeh: where can i get that?
Kano: from nvidia?
spreeuw: what stream are you playing?
tobydeh: no, ati
tobydeh: uk freeview stream
Kano: vdpau is nv only
tobydeh: oh ok
spreeuw: and thats a real 1080 stream?
tobydeh: no im trying to use 720p
Kano: best solution maybe nv g210 card
spreeuw: 720 should be ok on slower cpus
tobydeh: its an old pentium 4
spreeuw: thats not enough
spreeuw: it needs to be 3+ghz
happycube: isnt' freeview just mpeg-2?
spreeuw: then you may stand a chance
Kano: and pci-e ;)
spreeuw: displaying power you have enough
spreeuw: just not decoding of the stream
spreeuw: if its x264
spreeuw: if its mpeg2 it should work
tobydeh: ok, could that be why then?
tobydeh: what proc would you reccomend
tobydeh: Kano: g210 looks nice but i need a fanless card
spreeuw: you say it looks ok on your CRT?
happycube: a core 2 something
tobydeh: ok
tobydeh: thanks
spreeuw: core2 8400 is nice
tobydeh: when i use the binary drivers live tv is choppy - videos play ok
spreeuw: its not too expensive and has nice virtualisation extension
happycube: in that case check the stream buffering
happycube: (if videos *of* the live tv are ok)
spreeuw: but tobydeh if your other monitor works ok then it is something in the displaying
tobydeh: £116!!
tobydeh: jeees
happycube: is that an e8400 or q8400?
tobydeh: erm
Kano: e8400 is 3ghz
tobydeh: 38400
tobydeh: e8400*
tobydeh: £116
happycube: look at the celeron e3xxx series if you're on a budgert
tobydeh: the E7600 is 3.06ghz
happycube: those usually get to 3ghz np
tobydeh: thats £95
happycube: ...
happycube: oh. :) that should be fine
happycube: how much are i5-570's over there?
tobydeh: i thought it may be stream buffering but i wouldnt know where to look?
happycube: what player are you using?
tobydeh: i played with the playback profile in myth
tobydeh: mythtv
spreeuw: amd x2 5600+ is nice too or the 6000
tobydeh: i570?
spreeuw: if you swap the cooler
happycube: i dunno myth, but if you're tuned "cat /dev/video0 | bfr | mplayer -" might be an interesting test
spreeuw: sufficient for 720
Kano: tobydeh: MSI N210-MD512H
tobydeh: yeah i tested the card outsude myth - it seemed ok
happycube: the i5's $200us, which isn't much more than an e8400
Kano: tobydeh: fanles
happycube: prolly myth settings then - or you're just at the cpu's limit
tobydeh: thanks Kano
tobydeh: yeah cpu was my worry
tobydeh: my mplayer test was much a smaller window
happycube: that and if you're worried about noise p4 is *not* the way to go
tobydeh: if that makes a difference?
happycube: usually not, but this is fglrx we're talking about
spreeuw: yeah p4 is noisy
spreeuw: needs alot of air
tobydeh: ive got a silverstone fanless heatsink
tobydeh: only 1 fan - in the powersupply
Kano: tobydeh: but you know that you usually need case fans then
spreeuw: put a big fan on low rpm on it
tobydeh: yeah, ive got a 14dba fan
spreeuw: you have to give it a little air
tobydeh: i was going to leave it until it blew up, then put fans in :P
happycube: well, not noisy as much as hot - esp if you have a 1mb l2 p4
Kano: gfx cards will get very hot without active cooling
tobydeh: yeah, could that be the problem? better plug my fan back in :0
happycube: oooh yeah - p4's throttle when they get too hot
tobydeh: how can i determine if its a myth setup / cpu overload problem?
Kano: tobydeh: also: p4 runs with throtteling when you dont have enough cooling, you know that?
tobydeh: heard about it, how can i check the temp?
tobydeh: the heatsink is barely warm, if thats of any interest?
tobydeh: bios says cpu temp is 38 deg c
tobydeh: is that ok?
Kano: sure without load
Kano: interesting it the temp under load
tobydeh: how can i check the temp whilst running myth
dmb__: tobydeh, that temp is fine
dmb__: usually 30's are great idle temps
dmb__: tobydeh, if you want to see how hot your cpu will get, try mprime
tobydeh: ok
dmb__: mine goes up to about 80C
dmb__: which is pretty damn hot
dmb__: http://mersenneforum.org/gimps/mprime259.tar.gz
dmb__: thats 32bit version
tobydeh: should i really be using an nvidia card? s vdpau better?
dmb__: i like ati a lot better, but thats just me
tobydeh: hmmpphhh i wish i could get it to work, im going to give the binary dirvers another go i think
tobydeh: should i use the hardware drivers tool, or download a more up to date version?
Kano: tobydeh: which distro?
tobydeh: ubuntu 9.10
Kano: card is?
tobydeh: 4550 hd
tobydeh: saphire radeon 4550 hd
Kano: ok
tobydeh: dmb__: what dirvers do you use
dmb__: tobydeh, well, i'm using fglrx right now, but when the open source driver works on my new card, i'll switch
dmb__: use the fglrx from ati site, have it generate debs, and install it
Kano: look at query
tobydeh: query?
spreeuw: I have a 3450 passive
tobydeh: last time i tried from the net the install failed
tobydeh: what are the best xorg options for smooth video
spreeuw: none at all
spreeuw: driver uses sane defaults
tobydeh: ok
spreeuw: use xv for your mediaplayer
tobydeh: can i configure myth to use xv?
Kano: tobydeh: opengl output with vsync enabled in amdcccle, xv with fglrx is crap
spreeuw: it probably does already
spreeuw: fglrx is another channel
tobydeh: ah yeah sorry, ill head over now
tobydeh: are the ati drivers being updated for 9.10?
happycube: yeah
happycube: i don't think the 3d driver's included, but 2d should be quite snappy
Kano: 9.10 has 8.660 fglrx not 8.661
dmb__: unlinks Kano
happycube: my wild phoronix-like guess is that r600 will be in pretty good shape for 10.04 ;)
happycube: nm, if i was really phoronix i'd say it'd be playing doom3 in two weeks
_KAMI_: hello
_KAMI_: I have radeon 3470 in my laptop how is it supported?
Kano: happycube: you can use software rendering too ;)
_KAMI_: I am using Ubuntu Karamic
dmb__: can doom3 play on any mesa driver?
dmb__: like intel driver
happycube: you should have pretty good 2d, but r600_dri.so is WIP and isn't in karmic iirc
taiu: doom3 plays nice over here
dmb__: taiu, intel?
spreeuw: how about quakewars?
taiu: r600, had it working on r500 also
_KAMI_: I would like to use Impress 3D transitions
dmb__: someone should see if maya works with the radeon mesa driver :)
taiu: somehow r600 needs low setting othrwise some texture corruption
_KAMI_: Maybe can i get and build a never version of radeon driver, mesa or drm?
_KAMI_: dri
_KAMI_: sorry
happycube: probably need to build it from source
happycube: or xorg-edgers?
_KAMI_: xorg-edgers?
_KAMI_: what is it?
happycube: yeah they do unofficial ubuntu packages of the latest code
taiu: ahh, forgot theres one longstanding radeonTexSubImage bug, for which i commented out some code
happycube: https://launchpad.net/~xorg-edgers
_KAMI_: happycube: I am following this guide:
_KAMI_: http://www.x.org/wiki/radeonhd%3Aexperimental_3D
happycube: that should work
_KAMI_: I build drm without problem
_KAMI_: now I am on mesa
_KAMI_: is this enough for basic 3D functionality?
_KAMI_: or should I do more with radeon/radeonhd driver itself?
spreeuw: http://zerowing.idsoftware.com/linux/etqw/
spreeuw: taiu: you just fixed something?
taiu: spreeuw: nope
Rikki-Tikki-Tavi: Just played a round of quake 3 on my HD3200. Ran perfect. Thanks again for your help. Last time I offered to drop 5$ in a Paypal account, to help the project. Offer still stands.
happycube: KAMI - that oughta be fine
_KAMI_: okay
_KAMI_: hmm mesa is so huge :)
_KAMI_: happycube: Source: xserver-xorg-video-ati, Version: 1:6.12.99+git20090929.7968e1fb-0ubuntu1
_KAMI_: happycube: what is gallium?
happycube: gallium's the next generation 3d driver system - it's very unfinished for r300 and nonexistant for r600+
happycube: in about a year it'll be quite awesome ;)
happycube: (or less. who knows?)
_KAMI_: Requested 'libdrm >= 2.4.15' but version of libdrm is 2.4.3
_KAMI_: I got this error durin mesa configure
happycube: ok you need to set environment variables to point at the libdrm you just built
Kano: _KAMI_: -dev package
happycube: karmic doesn't have 2.4.15 iirc
_KAMI_: hmm Source: libdrm Version: 2.4.14-1ubuntu1
happycube: i don't remember what the env. variables are for using your freshly=-built libdrm, i think it's in the gallium build instructions
happycube: nope, not in there...argh
happycube: uninstall the libdrm-dev package and do "make install" on the fresh libdrm
dmb__: hello spstarr
spstarr: hulo
spstarr: hullo
dmb__: i think i'm going to be the only one in this channel that has the 5770 :)
_KAMI_: happycube: no way, it has dependancy on all xserver files
_KAMI_: :(
_KAMI_: glcontextmodes.c:42:23: error: GL/glxint.h: No such files
_KAMI_: happycube: do you have idea?
_KAMI_: x11proto-gl-dev
_KAMI_: ./../../include/GL/internal/dri_interface.h:45:17: error: drm.h
_KAMI_: :(
_KAMI_: moved rm.h drm_mode.h to /mesa/include/GL/internal
happycube: humm - yeah you need the requirered build packages
happycube: i wonder if someone's compiled a r600_dri.so for karmic
happycube: fglrx would probably work ok if you don't mind a binary blob
_KAMI_: I have issu with impress presentations
_KAMI_: e
_KAMI_: happycube: https://bugzilla.novell.com/show_bug.cgi?id=536016
_KAMI_: I just vanted to try how OS drivers works
_KAMI_: in this case
happycube: bummer
happycube: anyone in #ati have any ideas?
_KAMI_: happycube: no I haven't
_KAMI_: Also I wanted to try OS driver
happycube: i wish ubuntu made a r600_dri side package like fedora's done :P
kdekorte: happycude, why can't you use xorg-edgers?
kdekorte: happycube ^^
happycube: :) in my case, i want to rebuild the drivers to hack on em ;)
happycube: in his, i dunno.
_KAMI_: Okay
_KAMI_: I get it
Kano: airlied: isnt there xdm/gdm or something in live mode?
agd5f: woot! r6xx/r7xx irqs working
hnsr: agd5f, cool :D
MostAwesomeDude: Woot!
Kano: agd5f: xv faster that way?
agd5f: Kano: no, less cpu usage with dealing with fences
agd5f: and vblank support
MostAwesomeDude: agd5f: Should I add fences to r300g at some point?
Kano: what do you mean by fences?
agd5f: Kano: it's a flag you put in the command stream that buffers pend against
MostAwesomeDude: A fence is, simply, a point in the command stream. When you flush, you can specify a fence to flush to.
Dr_Jakob: MostAwesomeDude: not quite...
spreeuw: ERROR: The current video card / driver combination does not support the necessary features: GL_ARB_occlusion_query
spreeuw: for et quake wars
Kano: ah
spreeuw: a doom 3 engine game
agd5f: spreeuw: there's a patch of r6xx/r7xx OQ on the mailing list
MostAwesomeDude: Dr_Jakob: How's it work in Gallium then?
spreeuw: http://wiki.x.org/wiki/RadeonProgram
spreeuw: hey nice list
spreeuw: I made a similar page once for r300
spreeuw: platinum hehe
Dr_Jakob: MostAwesomeDude: "when you flush you can specify if you want to get a fence you can syncronize on or not"
MostAwesomeDude: Dr_Jakob: Oh.
MostAwesomeDude: Dr_Jakob: So it's an opaque type. Where else can you use it, besides flush?
Dr_Jakob: In theory you should be able to inject fences into the command stream at any point, but we never saw the need for it so we just have it at the end of the command stream (at flush)
Dr_Jakob: MostAwesomeDude: in the fence functions of the screen.
Dr_Jakob: singaled/finish/destroy
Dr_Jakob: I think it only used for glFinish
MostAwesomeDude: Ah, I see.
Dr_Jakob: might be other use cases for it
MostAwesomeDude: So at flush, we're blocked while our command stream gets added to the RB... The fence is a way to wait until the commands are actually finished, a few ms later, right?
Dr_Jakob: no a fence is signaled when the hardware passes that point in the command stream
MostAwesomeDude: Ah.
bragon: Hi *!*@*
bragon: Which ATI card, can i buy (a new card) in order to be sure to have the maximum performance with an ATI free drivers.
Lutz_Ifer: http://www.free3d.org/
Lutz_Ifer: @ bragon
bragon: For the moment i have a x850, but i need to change (My card is old and don't work very well (the hard))
Lutz_Ifer: may be that's a little old information
MostAwesomeDude: Lutz_Ifer: Don't recommend that site, it's horribly old and useless.
bragon: I need two DVI for have a dualscreen
bragon: MostAwesomeDude: +1
MostAwesomeDude: bragon: Anything that's not an HD5xxx will work.
bragon: ok
MostAwesomeDude: 3D is unstable for HD2xxx, HD3xxx, and HD4xxx, but it's getting better daily.
bragon: but i have around ~6000 FPS with my x850
bragon: it is better or not ?
bragon: ok
bragon: with radeonhd drivers ?
adamk: Your x850 probably outperforms the r600 driver on most HD cards.
adamk: THough that is changing.
bragon: many time ago the x850 was > *
bragon: (when i buy it)
adamk: I still happily use my x850.
bragon: me too but, my computer reboot (hardware issues) when i use the 3D for long time .... with my x850.
bragon: and impossible to buy a new x850 :'(
glisse: bragon: it's a driver bug
bragon: realy ?
bragon: (sorry for my english)
glisse: bragon: yes, lockup a driver bug 99% of the times
glisse: s/a/are
bragon: i thought that it's a hardware issue because i allready test to upgrade xorg xf86-video-ati my kernel and mesa etc ....
glisse: to test it's hw issue try with windows
bragon: sorry :)
glisse: if there is no lockup after intensive use
bragon: it's not possible here :D
glisse: than hw is fine
bragon: i'm not able to test with Windows
spreeuw: bragon: for the free drivers x850 is still a good choice
bragon: ok
bragon: so the best choice is the x850 :x
bragon: i ever try twice to buy it on ebay.
bragon: but i receive two buggy card
adamk: NOt necessarily the best choice. The x1950 is certainly better.
bragon: (one bip bip with my motherboard)
bragon: (and the other have a T° issues ...)
bragon: it's very difficult to buy a ever use graphic card
happycube: there's a us seller on ebay that has 3 new x1950's
happycube: newegg has no r300-r500's anymore... they have a few 7000's and 9250's :P
happycube: one would think x300/1050's would be available forever like fx5200's are ;)
spstarr: looks at drm-next and builds a new kernel
spstarr: (assuming no conflicts with git master kernel.org)
Ghworg: is eagerly awaiting the git commit for r600 IRQs now
ajami: MostAwesomeDude, r6/7xx just took a huge hit. The merge from 7.6 caused a lot of issues.
ajami: airlied, has anything been figured out as far as -97 failing on boot?
agd5f: ajami: try make realclean and rebuild
agd5f: mesa
ajami: agd5f: why would I need to do that if the kernel is failing to boot?
agd5f: ajami: you just said the merge from 7.6 caused a lot of issues
ajami: agd5f: yes, but the -97 was supposed to fix the kernel side KMS issues from the 7.6 merge which actually caused my kernel to not boot.
ajami: agd5f: https://bugzilla.redhat.com/show_bug.cgi?id=530961
agd5f: ajami: there shouldn't be any kernel changes needed for mesa since 7.6 got merged into master
ajami: airlied and I were talking about it a few hours ago. I didn't understand why the kernel needed to be altered, but that's what I was going off of.
ajami: agd5f: ^
ajami: agd5f: Will you enlighten me? I seem to have forgotten the ./configure options for specific dri drivers.
agd5f: ajami: something like: ./autogen.sh --prefix=/path/to/install --with-dri-drivers=swrast,radeon,r200,r300,r600 --with-dri-driverdir=/path/to/dri/driver --disable-gallium
ajami: k, building right now
ajami: agd5f: still is a kernel issue so I can't boot 2.6.31.5-97
ossman: I got a crash in mesa with r600
ossman: bad pointer reference in radeonFlush
ossman: bt: http://pastebin.com/m3fefbc97
ossman: unfortunately I don't know the stuff well enough to know if there simply should be a pointer check there, or if radeonFlush should never have been called in the first place without that pointer
ossman: (it's ctx->DrawBuffer that's NULL)
ossman: pokes at agd5f
thinkpaduser: with desktop effects enabled i get some static on the right side of the screen- xf86-video-ati, X1300, KDE 4.3
thinkpaduser: but only when dragging windows or opening menus
thinkpaduser: here's my xorg.conf: http://pastebin.com/m2d08e163
airlied: agd5f: yes it seems to be mildly broken on the chipset, it might be overly agressive (i.e. it might just be the IDE controller)
agd5f: airlied: MSI don't seem to work rs780
agd5f: so I have a patch to disable it on igp chips
agd5f: other than that irqs are working on r6xx/r7xx with and without msi on discrete cards and without msi on igps
airlied: agd5f: it worked on my rv370/rv530/r600
airlied: the rs690/780 were my only worries
agd5f: airlied: I'm trying to get some answers internally about msi on igp chips
zicada: im getting out of range (lcd just shuts off) when booting with radeon.modeset=1 on fresh install of karmic with xorg edgers and rc5 kernel
ossman: airlied, agd5f, any thoughts on the radeonFlush() crash above?
zicada: booting with single, and then modprobe radeon modeset=1 works
zicada: using DVI -> CRT dongle thingy
zicada: is this a known bug or should i report it somewhere
zicada: 4850
agd5f: ossman: sounds like a mesa bug
airlied: agd5f: any aspm ideas ;-)
spstarr: agd5f: when do you expect PM stuff to be started in the drm?
ossman: agd5f, it's easily provoked here using "./glean -v -t vertProg1 -r foo"
roysjosh: zicada, http://bugs.freedesktop.org/show_bug.cgi?id=24456 , http://bugs.freedesktop.org/buglist.cgi?quicksearch=radeon+dvi+range
agd5f: ossman: does it work with dri1?
ossman: agd5f, how do I switch? :)
agd5f: ossman: turn of kms
agd5f: off
ossman: ah, give me a sec and I'll test
agd5f: boot with radeon.modeset=0
ossman: agd5f, nomodeset should do the same I guess?
agd5f: ossman: yeah
ossman: agd5f, got another bug now :)
ossman: glean: radeon_lock.c:65: radeonGetLock: Assertion `drawable != ((void *)0)' failed.
ossman: might be the same issue, just grabbing it earlier
agd5f: ossman: that usually means a realloc failed somewhere
ossman: I'll get a backtrace
ossman: seems to be roughly the same place
ossman: http://pastebin.com/m225ac4a8
lordvan: hi
lordvan: has anyone tried the OSS radeon driver with a HD 5750 / 5770 or 5850/5870 yet? ;)
agd5f: lordvan: not supported yet
lordvan: :(
agd5f: lordvan: I got analog working on friday, but haven't gotten back to it yet
lordvan: ;)
lordvan: to be honest i don't care if it is radeon, radeonhd or ati catalyst as long as it works :d
lordvan: as i'm considering getting one instead of some old nvidia GeForce 7600 ;)
agd5f: lordvan: fglrx should work already
lordvan: that's what i was told in #ati ;) but noone knows of anyone who got one and tried it ;)
ossman: agd5f, any luck sorting out those irq issues?
agd5f: ossman: git them working this morning
ossman: glee
agd5f: ossman: just waiting on approval to release the patch
lordvan: but thanks to you guys and #radeonhd guys i can use 2d on my laptop again without vesa(no-widescreen resolution of course ;))
lordvan: :d
ossman: agd5f, some secret sauce involved :)
ossman: then airlied just needs to pull it into fedora
agd5f: ossman: not really, just the usual approval for new IP release
ossman: agd5f, so doing all the work of vetting the docs is not enough to let you develop unhindered? :)
agd5f: ossman: evrything gets vetted. bigger/new stuff is just more involved
lordvan: anyway thanks for answering my question .. cya ;)
spstarr: kernel.org git master + drm-next == conflict free still
amarks: agd5f: I tried EXANoDownloadFromScreen + EXANoUploadtoScreen + DFSAccel all set to False, but i still see the minor icon pixel corruption
amarks: seems to be only at the top of the icon
agd5f: amarks: EXANoDownloadFromScreen + EXANoUploadtoScreen don't need the false
agd5f: it's already implied in the "No" in the name
amarks: do you leave out the boolean in the config?
agd5f: amarks: yes
agd5f: just Option "EXANoDownloadFromScreen"
amarks: will it ignore my extra "False"
ossman: agd5f, do you know if there is a lot of stuff for the r6xx compiler waiting to be released from AMD?
ossman: I figured I'd start with the opcodes not yet implemented to get comfortable
agd5f: amarks: when you check your xorg log, you should no longer see UTS/DFS listed under (II) EXA(0): Driver registered support for the following operations:
agd5f: ossman: richard is starting on a glsl compiler, but feel free to work on missing opcodes for the existing code-base
amarks: agd5f: they are registered, it must have ignored my config, i'll try again now
ossman: great. will do :)
Zayec2: agd5f: can change you could put your hands on locking HD 34xx?
amarks: agd5f: of the 3 nerd knobs, setting *only* EXANoDownloadFromScreen fixes it
agd5f: amarks: ok. then your hitting bug 18397
amarks: agd5f: seems like there's been trouble tracking it down
agd5f: amarks: yup
amarks: agd5f: i can reproduce within a few seconds of starting up so if i can test anything...
Pallokala: I must say that compiz-fusion with stuff from git works quite nicely
ajami: airlied, glisse, agd5f, XVMC works damn well with r600 classic.
agd5f: ajami: there is no xvmc support for any radeons at the moment
ajami: interesting. If I have VLC use it, video playback is the best...
ajami: agd5f: any reason for that?
agd5f: ajami: no one's implemented it yet
agd5f: ajami: probably just falls back to Xv
ajami: agd5f I mean for VLC using XVMC video to be the best playback.
agd5f: ajami: vlc probably falls back to Xv if there is no XvMC lib
ajami: I see. Well it's damn good.
ajami: Plays back 720p better than proprietary drivers in windows. :)
ajami: agd5f KMS enabled allows for better Xv playback as well. Should that actually matter?
agd5f: ajami: shouldn't, but ttm might handle memory better
ajami: agd5f: well with kms/ttm the playback is drastically better. I am on an older box which really shouldn't be playing 720p, but as of right now with XVMX fallback to Xv with KMS/ttm I nearly can.
ajami: agd5f: I didn't try w/o compositing on, that might actually have allowed it to playback just fine.
ajami: I really, really would like to get involved in working on radeon drivers, however I don't really have an idea of what to do. I know there are occlusion queries and such that need to be implemented, but I don't know where to begin (aside from reading docs) or even if I do get a start, where to go. Would someone be willing to essentially take me under their wing and help me get up to a level where I can help out with driver development? I missed ou
ajami: airlied, glisse 2.6.31.5-97 nomodeset, X seems to crash randomly.
airlied: ajami: -96 nomodeset do the same?
laska: airlied: i have some news about multiseat configuration with 2 radeon videocards and second videocard initialisation
laska: 1) i found a program that run on 1-st Xserver and destroy picture on second X-server. This is etracer. When i start it (on first X-server) the screen on second Xserver becomes blank and monitor types "signal out of range"
laska: 2) It looks like this bug fix the second videocard initialisation. After it i can normally switch to 1280x1024@75 without image defects
laska: airlied: https://bugs.freedesktop.org/show_bug.cgi?id=24536
ajami: airlied, yeah
airlied: woot my rc410 suspends/resumes now
soreau: airlied \o/
soreau: Great success :)
EruditeHermit: airlied, what is rc410?
EruditeHermit: did you mean rv410?
EruditeHermit: or is rc another card
airlied: no rc410, its like an rs480
EruditeHermit: oh
airlied: just an intel cpu instead of amd one
EruditeHermit: is that chip specific issue you debugged/
EruditeHermit: is that a*
airlied: not sure it might affect other rs480s
EruditeHermit: or do you think it will help an rv350
airlied: does s/r not work on that?
airlied: I thought it was just LVDS not coming back
airlied: don't think we need to do the same on rv350
EruditeHermit: ah well its the LVDS not coming back
EruditeHermit: you're right
EruditeHermit: but i lumped that in with s/r
eosie: does anybody know what is R300_SC_SCREENDOOR?
hyc: ok, a slightly off-the-wall question... I recently tried to replace the 1680x1050 screen in my HP dv5z notebook with a 1920x1200 screen
hyc: the image on the new screen was quite unstable
airlied: hyc: the internal panel?
hyc: I'v read in various places that maybe the EDID needs to be reprogrammed
hyc: yes
airlied: it could also be in the bios tables
hyc: I was under the impression that the screen itself provided its timing parameters and such
hyc: no?
airlied: not always
airlied: for laptop panels it ships in the bios usually as well
AndrewR: airlied, hello ...i tried drm-next on my rs780 and it hangs machine hard in KMS mode ...
hyc: also, this notebook has an HD3450. is there any known limitation in the card itself driving this rez over LVDS?
airlied: hyc: no
airlied: AndrewR: did it used to work?
hyc: airlied: thanks
airlied: hyc: you should be able to see the EDID in the x logs
AndrewR: airlied, yes, it works at least with drm in 2.6.32-rc5
airlied: AndrewR: don't suppose you'd like to bisect?
airlied: AndrewR: oh did you pull -next into -rc5?
hyc: airlied: yes, I saw a fair bit of stuff and a bunch of modelines that looked like they should be fine
AndrewR: and it was working in -next, too ... but unfortunately i can't remember exact good commit
AndrewR: airlied, , no separate trees ..
hyc: using fbcon, the screen looked normal actually, for plain text.
AndrewR: airlied, one moment will try bisect
airlied: AndrewR: if you pull -next into -rc5 tree
airlied: you should be able to bisect it quite easilyt
hyc: but my shell prompt also uses boldface, and any boldface text was distorted.
hyc: and the blinking cursor was also distorted. horizontal tearing in both cases.
hyc: would this be something I should be able to fix with manually configured modelines? or is there more in the BIOS tables
hyc: that this would not address?
hyc: I spent several hours tweaking modelines and never getting anything rock solid.
airlied: I suspect the bios tables are being used for some timing, you would have to hack around it in the driver I suspect
hyc: ok
AndrewR: airlied, last time i tried this - it ends up with conflicts i was unable to solve
airlied: AndrewR: that isn't always the state
airlied: AndrewR: it depends on what stuff is in Linus tree and my tree at the time
airlied: it should merge fine at the moment
AndrewR_rs780: not sure if remote X is right solution for copy/paste between machines ... but http://pastebin.com/mf84caa7 (/var/log/messages ends up here for 'bad' drm). It goes past this point with 'good' drm.
AndrewR_rs780: airlied, bisect in progress .. already found good commit
airlied: AndrewR_rs780: excellent