AStorm: hello
Zajec: AStorm: this channel? ;)
AStorm: how is the dri2 rewrite going wrt RadeonHD?
AStorm: I see the "old" driver now has more EXA accel
AStorm: (also, where's that radeon-rewrite tree?)
AStorm: ah, my bad, it's in mesa
eschat: hi, i have a problem with xorg, radeon mobility (thinkpad) and freebsd7.2 again. the last hint fixed the problem only for one time
AStorm: btw, how is that r7xx drm going with the new code available? anything usable now/soon?
AStorm: or even... is anyone working on that?
eschat: google has no answer on my problem
chithead: AStorm: r6xx-r7xx drm (for exa and xv acceleration) is in kernel 2.6.30. if you want the experimental 3d support you need to fetch drm from agd5f's r6xx-r7xx-3d branch
glisse: eschat: dri disabled ? and what is your problem ?
AStorm: chithead, this. but what about KMS? :)
AStorm: (I know it doesn't have 3D yet)
chithead: afaik glisse's kms code works only up to r5xx cards (correct me if I am wrong)
AStorm: wouldn't the about only thing needed be a similar API conversion?
AStorm: chithead, yes, it doesn't do r6/7xx
eschat: dri is disabled. my problem is that the complete system hangs if i start x. i have tryed to downgrade the radeon driver, i have tryed anything in xorg.conf. i have tryed to disable kernel modules for radeon cards, i have tryed to downgrade the last changes in kernel radeon code via cvs
eschat: sometimes it works. knoppix works too
AStorm: chithead, I suspect I need agd's mesa r6xx-rewrite branch too?
AStorm: (looks fairly ancient)
chithead: mesa is only for 3d and use the mesa/mesa git repository. either r6xx-rewrite or r6xx-r7xx-support. but don't expect it to work with games or compiz
AStorm: or rather, mesa/mesa branch r6xx-rewrite
AStorm: heh. but I do expect it to still work with 2D
AStorm: and Xv
chithead: if you only need 2d and xv, use kernel 2.6.30. it comes with all necessary support
AStorm: actually, I'd prefer KMS, but oh well
AStorm: I've the drm set up correctly right now anyway
AStorm: unless there have been fixes this month
glisse: eschat: try with option noaccel in xorg conf see man radeon
eschat: i have tryed this several times
glisse: eschat: i wouldn't understand lockup with noaccel option are yoou sure you properly set it
glisse: and the xorg log reported to disable all accel
glisse: also this should lead to very very very slow X
AStorm: unless it's a modesetting failure
AStorm: 2009-04-17 Initial 3D driver for R6xx/R7xx chips Richard Li 50 -5/+21663 - I love such huge commits... :-/
eschat: glisse: i try
AStorm: does the r6xx-rewrite even work? (if it does, isn't that KMS to test?)
max_r: AStorm: r6xx-rewrite allows glxinfo to show renderer string
AStorm: hmh :)
max_r: AStorm: + there are claims that it allows redbook hello program to run
max_r: for me it only shows window with nothing inside
AStorm: no glxgears yet, I see
AStorm: it doesn't do triangles
AStorm: anyway - for KMS, I'd need drm w/ TTS/GEM?
AStorm: and proper xorg driver?
eschat: glisse: it works this time, but why? i have just commented out the noaccel line
chithead: yes, kernel memory manager is necessary for kms
max_r: but there is no kms/mm/... for r6xx atm
eschat: glisse: this time i have busmode pci too. this is new
AStorm: max_r, you say there's no DRI2 rewrite of the r6/7xx command pipeline?
AStorm: ouch
AStorm: looking for a non-existent thing is indeed hard :)
max_r: AStorm: don't know, r6xx-rewrite name suggests that it should work both using classic mesa system and kms/dri2/etc, but there is no kms/dri2/ttm/drm/... for r6xx so it is impossible to test it
chithead: rewrite is underway, but dri2 does not imply kms
AStorm: doesn't? uhm, it doesn't move all the modesetting into the kernel?
AStorm: interesting
AStorm: I'm wondering now since there's drm-next-radeon - that's DRI2 for r5xx actually, right?
AStorm: or is it DRI2 + KMS?
glisse: chithead: dri2 imply memory manager, for radeon we decided that memory manager imply kms, so for radeon dri2 imply kms
AStorm: hm
AStorm: so, there's no working DRI2 for r7xx yet? (as in, 2D-working, KMS possibly?)
glisse: dri2 is for 3d
AStorm: nah, that needs mesa support
AStorm: :)
AStorm: hmmhmm then again, it does imply some measure of 3D to work
AStorm: what I meant was - DRI2 r7xx driver with at least basics working for 2D support
AStorm: or did I just confuse myself?
glisse: talking about dri2 regarding 2d accel means nothings
dileX: for 2D you dont need dri2
glisse: 2d accel is totaly unaware of dri/dri2 ....
dileX: hi glisse
AStorm: glisse, uhm, Xv?
AStorm: and the vertex pipeline is used for 2d accel in the current driver it seems
nanonyme: AStorm: Yeah, I don't think KMS + EXA worked together on r6xx/r7xx the last I checked (which was quite long ago) on rawhide kernel, haven't tried glisse's tree.
TheBrayn: howdy folks
otaylor: agd5f: any idea why access to evicted buffers would become dead-slow after suspend/resume? (Tried repeating radeon_modeset_agp_init() - didn't seem to help, though conceivably something subtly wrong about it)
agd5f: otaylor: interrrupts not enabled properly?
otaylor: agd5f: how would that affect things?
agd5f: timeouts waiting for stuff
agd5f: there was a problem with the old drm a while ago, where rendering would be slow after resume and it was due to interrupts not being re-enabled on resume
otaylor: agd5f: Hmmm, but drawing is still fast as long as the buffers are in vram
agd5f: ok, probably something else then
agd5f: otaylor: AGP or pci/e
agd5f: ?
otaylor: agd5f: AGP
agd5f: otaylor: maybe try and compare gart setup in the pre/post resume cases
otaylor: agd5f: by code inspection or elsewise?
agd5f: although, I'm not really familiar with the agpgart drivers
agd5f: I would suggest dumping regs
otaylor: card registers?
agd5f: well, in the agp case, you really just point the card to the agp aperture
agd5f: the actual gart is handled by the agp driver
agd5f: otaylor: does pci mode do the same thing?
otaylor: agd5f: hmmm, so you are wondering if it goes from dead slow, to deader slower?
otaylor: agd5f: I should check that
agd5f: otaylor: also, does it work after resume if you unload the driver and reload it?
agd5f: that would point to bad init
otaylor: agd5f: is that possible with kms and fb console?
agd5f: otaylor: yes
agd5f: but you'll need remote access
agd5f: just unbind the console from the fb
agd5f: you have to echo 0 to something in /proc I think to unbind. let me see if I can find the path
agd5f: otaylor: echo 0 > sys/class/vtconsole/vtcon1/bind
otaylor: tries
otaylor: agd5f: unload/reload gets things back to full speed
agd5f: otaylor: probably some missing init then
otaylor: agd5f: yeah
wltjr: anyone got any experience with dri and xpress200m? seems to hang/crash x when I attempt dri :( np without just slow....
otaylor: agd5f: Maybe some of the stuff in radeon_gart_init(), though if the card didn't have the right agp location, I'd expect complete non-function not slow
agd5f: yeah. strange
wltjr: for anyone interested here is my info, x/driver versions, xorg.conf, and xorg.0.log, with dri x crashes, and never get a working display, but have a mouse....
wltjr: http://pastebin.com/m6bda1153
mib_u7xwy53s: i was thinking of getting the 4850 but since the 4770 is on sale friday should i wait for the 4770?
icewaterman: mib_u7xwy53s, neither is supported by the radeon driver.
yangman: that's not true
rektide: agd5f:
rektide: mjg59:
rektide: http://www.phoronix.com/forums/showthread.php?p=76341#post76341
rektide: my request to AMD to document power savings features / aid in d3
mib_u7xwy53s: but which is the better card?
yangman: mib_u7xwy53s: IIRC there's still issues with bringing up a 4770 properly. 4850 should work with both radeon and radeonhd
agd5f: rektide: you could just shutdown your PC in teh sort term :)
rektide: its my fileserver!
rektide: its my everything
rektide: world be damned. ;)
rektide: i promise i'll contribute if they actually deliver some documentation
rektide: gotta put this computer engineering degree to some kind of use
rektide: might as well be _saving the planet_
icewaterman: uhm, the titel still says no 3d, and pretty much any graphics card can do 2d as good as the high-end cards.
hax0r1337: few care about the planet these days
agd5f: rektide: you can save a signifigant amount of power using the existing power options already available
agd5f: such as turning down the engine and memory clocks and pcie lanes
agd5f: the information for that is available
rektide: i spent quite a while reviewing docs this weekend, not sure how i missed that
agd5f: rektide: it's not in the docs, it's in the code
yamato: on load a 4770 uses 40 odd watts less than a 4850/70, right?
rektide: ahhhhhhh ffffffffffff
agd5f: see radeon_pm.c in xf86-video-ati
agd5f: rektide: there are already some options that might help
agd5f: DynamicPM for example
agd5f: reduces your engine clock to 1/4 and pcie lanes to 1
rektide: superb, i'll do that.
agd5f: when dpms kicks in
rektide: excellent
agd5f: the info is there if you want to try other things
rektide: was that code from AMD or ? how did it get there ?
ckybam69: but the 4770 seems to have more overclocking room and once drivers that are good thats put out wont the 4770 be better?
yamato: 4770 is the first 40nm chip out the door, so there is good power savings to be had
yamato: the yields over at tsmc are shot though
yamato: if you want power efficiency, I heard about a 4670 using a mere 3W idle over at silentpcreview
ckybam69: power efficency is good but i want a good match between that and fps
yamato: well, 4770 is a strong card in that court as well
ckybam69: i think i will wait till friday and get the 4770
ckybam69: its new so drivers down the road will make it better and it seems to have room for overclocking
agd5f: rektide: yes from amd.
agd5f: I wrote it :)
tball: Hello everybody. How long are we from a partial working radeon-rewrite driver, which work with r6xx and r7xx? :-)
max_r_: tball: define "partial working"? able to run glxgears?
tball: yeah, or basic 3D.
tball: Maybe compiz
max_r_: I think you shouldn't wait fully stable working compiz till the end of the summer, but it is just my observation how fast things are going
tball: Okay. I guess the devs are working on other thing in paralel
max_r: yeah, that drives me mad :) there is work going on to bring support for older cards first, but afaik some devs are working on r6xx-rewrite
tball: But older cards does have support . Cmon :)
max_r: tball: apparently we have to wait till that support goes to master branches. It makes sense though, because if they will focus on support for older cards now they can make it into 2.6.31 kernel release
MostAwesomeDude: max_r, tball : I'm not even doing radeon work right now; I'm getting paid to do other stuff.
MostAwesomeDude: You'll have to let the AMD guys pick it up.
tball: Damn... :P
max_r: and one of amd guys (richard), who did the initial work for r6xx 3d is on vacation now, so we have to wait till he returns
max_r: according to bridgman his vacation is two weeks and by looking at git logs I can tell that it probably started about a week ago, so he hopefully will return at the begining of may
max_r: s/may/june
spstarr: :)
tball: MostAwesomeDude: Are you getting paid now? Didn't you start for yourself reading about driver development? :)
max_r: and accourding to this: http://airlied.livejournal.com/66515.html, another radeon developer will join as at the beginig of june, so I hope for burst :)
tball: max_r: spstarr: What do you say? Should we write the opengl part ourself? I will soon be finished with all the exams
otaylor: agd5f: So, it turns out that the agp location/memory registers on the card did need to be set on resume
spstarr: tball: I'm not a driver dude, right now
otaylor: agd5f: I have no real theory about how it was working at all without that
max_r: tball: do you have experience in graphics driver development?
otaylor: agd5f: maybe it was automatically falling back to pcigart in that case
tball: max_r: Not displaydrivers, and not for Linux.
tball: But for embedded system drivers, which is completely different, I know.
agd5f: otaylor: weird.
MostAwesomeDude: tball: I've never been paid for my Radeon work.
tball: I just thought that of your statement before: " I'm getting paid to do other stuff.". Sorry if I misunderstood you. I appreaciate your work.
tball: and all you other developers
otaylor: throws
otaylor: http://fishsoup.net/tmp/kernel-2.6.29-radeon-agpresume.patch onto his mental 'waiting for airlied to get back' queue
otaylor: agd5f: thanks for your help with that
nanonyme: vacations++
spstarr: hire me :/
nanonyme: spstarr: Who?
spstarr: just hire me ;p
nanonyme: I mean, that's kinda why you send applications: you probably want some specific company to hire you. :)
spstarr: waiting for replies :)
nanonyme: Although I suspect vacancies in companies might still be a bit scarce. Might get better in a year or two.
agd5f: otaylor: np
max_r: agd5f: is it expected that redbook hello program will actually show something with current r6xx-rewrite?
spstarr: nanonyme: i hope not
nanonyme: You hope vacancies aren't scarce or you hope things will not get better in a year or two? :p
spstarr: i hope things get better *NOW*
yangman: heh. you and I both
nanonyme: Well, things get better for *you* immediately if you figure out some new software idea that you can sell. ;)
spstarr: no thanks, i need money now not later
nanonyme: The whole IT sector probably will take some time to recover. Heck, it still hadn't recovered from the IT bubble bursting towards the beginning of the millennia, then we had this economic crisis...
wltjr: spstarr: probably need it now and later :) always need $$$ :)
spstarr: wltjr: cause i own a condo, and bills dont stop.
wltjr: spstarr: I am self employed.... :)
wltjr: had to fire my employee due to economy, and I am very busy, damn credit markets, etc....
spstarr: :/
wltjr: also bailed from contributing to Gentoo just before economy tanked, that kept me in business :( though not really any FOSS contributions :(
wltjr: while I am making noise anyone familiar with dri and xpress200m, done with fglrx, impressed with radeon, and close with dri... :)
mjg59: Haha xpress200
wltjr: yes everyone hates it
wltjr: it worked across the board with fglrx for a brief period of time, can't recall combo of driver, x, etc
wltjr: but mostly no xvideo for days and that's been very annoying, now I have that but 2d performance sucks :(
wltjr: likely due to composites, but I like my mim eye candy via metacity, not beryl
wltjr: next will be my hd2600 pro, that's still fglrx but very unstable
nanonyme: thought to say he doesn't really care much of eye candy, then realized he's running Windows 7
wltjr: my eye candy is just true transparent terminals and shadows, that's it
wltjr: oh and advant :)
wltjr: er avant
wltjr: damn advanta on the mind, jerks!!!
rah: SDL seems to be having problems displaying fullscreen
yangman: mjg59: is the work you've done so far for radeon temperature sensor available to look at anywhere?
mjg59: yangman: Afraid not
yangman: anything you're willing to put up somewhere?
mjg59: It's on a machine that's off at the moment
mjg59: I can push it later on?
spstarr: mjg59: that will be hooked into sysfs?
yangman: np
yangman: thanks
mjg59: spstarr: Yes
spstarr: nice
nha: I've decided to make the scripts that I'm using during X.Org development available
nha: http://cgit.freedesktop.org/~nh/dxo/
spstarr: :)
nha: they're very specific to my setup, so won't be directly useful to anybody, but anybody who wants to steal is welcome
rah: DanaG: ✈ is that an aeroplane or a dart? :)
nanonyme: Definitely not a dart with this font. :)
chithead: ✈ U+2708 Name: AIRPLANE
AStorm: ✈ you mean? :)
AStorm: yeah, that one
rah: "Running with scissors. And throwing airplanes[sic] at stuff. I love Unicode!"
AStorm: neat
nanonyme: AStorm: Weren't those the same?
rah: that doesn't make sense
AStorm: nanonyme, yeah, but typed from keyboard
nanonyme: Maybe he's referring to 9/11? I dunno. :)
AStorm: ctrl+shitf+u 2708 = win
rah: heh
rah: control, shit and fuck you 2708 = win
rah: you ain't wrong
AStorm: ;P
AStorm: ٦ - the devil's character - coincidentally hebrew? :p
AStorm: ah btw, some person asked me if RV740 is supported (HD 4770)
AStorm: is it?
mjr: not for 3d yet, no
AStorm: I know not for 3d
AStorm: but modesetting and 2d
mjr: oh yeah, it was quite new for r700 series too, then I dunno
fat_chris: AStorm: :o
AStorm: fat_chris, yeah you
fat_chris: as far as i can tell its not supported yet, but im not _that_ bothered about it
spstarr: well there is '3D' for the r6xx/r7xx..... the 2D is using the 3D engine ;)
spstarr: but not OpenGL 3D
AStorm: I suspect one of the guys here could code the modesetting in some minutes given debug output...
AStorm: :)
fat_chris: heh well if i knew how to do that id do it ;)
nha: blah, I should limit the number of visuals that I'm running the Glean tests on
nha: this is taking ages
Kano: hi, could somebody test a hd 2600 agp card
Kano: i have the report that it does not show a pic
benh: ouch
benh: /home/benh/linux-powerpc-test/drivers/gpu/drm/radeon/r600_cp.c: In function ‘r600_do_init_cp’:
benh: /home/benh/linux-powerpc-test/drivers/gpu/drm/radeon/r600_cp.c:1954: warning: cast to pointer from integer of different size
benh: /home/benh/linux-powerpc-test/drivers/gpu/drm/radeon/r600_cp.c:1956: warning: cast to pointer from integer of different size
benh: /home/benh/linux-powerpc-test/drivers/gpu/drm/radeon/r600_cp.c:1958: warning: cast to pointer from integer of different size
benh: might be harmless... oh well, I'll have to toy with kms first anyway