EruditeHermit: hey, is there a DRI2 version of the radeon driver?
EruditeHermit: by that I mean is there an experimental repository with buildable code?
MrCooper: none that I know; so far DRI2 is tied to TTM
MrCooper: none that I know of even
wojak: how can i enable hdmi output on my radeon x1250?
syntropy: Happy SysAdmin Day!
nha: osiris_: replying here instead of #xorg-devel
nha: osiris_: one obvious problem is that for the texture instruction, your SrcReg.File should be PROGRAM_INPUT instead of PROGRAM_TEMPORARY
nha: I guess we should output a warning when instructions try to read from undefined values :}
nha: also, I'm not sure setting a register to (PROGRAM_INPUT, 1) is such a good idea (magic constants and so on)
spstarr_ols_: nick spstarr_ols_
nha: glisse: how uptodate is your user repository / agd5f's user repository wrt your work on lockups?
nha: glisse: in particular as far as debugging tools are concerned
agd5f: nha: http://people.freedesktop.org/~glisse/hacklockupfreedrm.tar.bz2
agd5f: although I think glisse has some new work beyond that
rindolf: Hi all.
Gnutoo: hello, is there someone there...my radeon x700 crashes when playing nexuiz...the driver is the xf86-video-ati-6.9.0
Gnutoo: it also sometimes block my computer when i switch from a real tty or to a real tty
airlied: michaellarabel: the first kernel atom user code is now out as well :)
michaellarabel: airlied: Yes, noticed already from your RSS :) Nice job.
michaellarabel: airlied: Is there any R500 hardware that's currently problematic with KMS?
Gnutoo: hello, will the r300 driver be improved in order not to crash when playing nexuiz and not to crash when swithching from tty to X or from X to tty?
airlied: michaellarabel: I've tested on rv530 laptop, r580 desktop.
airlied: thats it so far :), but it should work fine on vga/dvi/lvds.
michaellarabel: Okay, I'll give it a go then.
TobiasTheCommie: airlied: so, when you write kernel modesetting works for r500, does that include rv530?
airlied: TobiasTheCommie: yes all r5xx variarnts.
TobiasTheCommie: very nice
TobiasTheCommie: just, with the whole (support for rs690, but not for r600) i thought i would ask
TobiasTheCommie: good work :)
airlied: michaellarabel: not so sure the DDX code isn't full of bugs.. I've hardcoded some memory sizes...
TobiasTheCommie: just wondering, did amd release the specs for r100? or just r200-r400?
TobiasTheCommie: 400? i meant 600
agd5f: TobiasTheCommie: yeah, r128 and mach64 as well, but all of it was NDA
TobiasTheCommie: is it still NDA?
agd5f: r200-r400 stuff was NDA as well
TobiasTheCommie: i mean, iirc r200+ isn't NDA anymore...
agd5f: no time to review it with legal, etc. it for release
TobiasTheCommie: ah, oki
TobiasTheCommie: so, is it just the mach64, r100, r128, and r700 under NDA right now?
agd5f: we hope to get back to the old stuff after we get r6xx documentation out the door, but that's proving to be a nightmare
TobiasTheCommie: haha, oki.. sorry to hear that :)
TobiasTheCommie: i'm just really looking forward to the first amd card with specs for the codec acceleration(can't remember what the chip is called) to replace my rv530
michaellarabel: airlied: Do you have any timeframe for when you'd ideally hope the radeon KMS would go to master?
airlied: michaellarabel: plans is GEM first, then modesetting, for Intel, the radeon..
airlied: michaellarabel: I'm hoping to ship it in F10 beta also..
TobiasTheCommie: though i will admit that the intel g45 is appealing as well.. but, since that won't come on a pcie card... *shrug*
airlied: michaellarabel: but I'd hope to push radeon modesetting upstream with intel modesetting
airlied: or close enough behind it
airlied: also my modesetting branch needs intels GEM kernel patches to build.
rx__: airlied working on radeon modesetting?
rx__: reads scrollback
TobiasTheCommie: rx__: read the blog as well
bridgman: Gnutoo: just ask, if there's nobody on then someone will probably see the backlog
Gnutoo: bridgman, ok thanks...
bridgman: worst case you asked a question in an empty room :)
Gnutoo: bridgman, i have 2 problems: 1) switching from X to tty sometimes make my computer crash...2) i can't play to nexuiz because it will crash at a moment or at another
bridgman: OK, #1 is always tricky because it depends a bit on the other (tty) driver as well, but if you describe the distro and any changes/options you have picked that affect the console driver someone might have an idea or ask you to file a bug ticket
bridgman: for #2, how long does it typically run before crashing ?
bridgman: and what GPU & drivers are you running ?
bridgman: actually your #1 question is interesting too -- my experience has been that VT switch (X to tty) either always works or always hangs, so if you're seeing "sometimes works" that might be a clue
bridgman: can you give me more information on #2 first ? Which card, which drivers ?
Gnutoo: bridgman, i run the x700 GPU
Gnutoo: and the r300 driver
bridgman: ok. Did you build the driver yourself or are you using the drivers that came with your Linux distro ?
Gnutoo: i used the driver that came with gentoo...
bridgman: OK, good. There are two other drivers needed for 3d - "drm" and "mesa" -- do you happen to know versions for those ?
Gnutoo: ok i'll look
Gnutoo: mesa 7.0.3
Gnutoo: and the drm of the kernel: 2.6.26
groo: hi/2 all :)
groo: eia airlied, agd5f and ati hackers
bridgman: Gnutoo: thanks; give me a minute - how long does it normally run before crashing ?
Gnutoo: bridgman, don't know...i could check...
Gnutoo: bridgman, but not now...
groo: eia bridgman
groo: im the lucky bastard with the rs485 with the garbled video when suspending to ram :D
Gnutoo: bridgman, but it can be something like 15 min...
Gnutoo: or even more
groo: anyone have any ideas or similar suspend problems with rs485?
bridgman: groo: I think agd5f's laptop has an RS480 or RS485 so don't give up hope ;)
bridgman: Gnutoo: agp bus ?
groo: mine is an acer 5102
groo: its the only problem i have with the open source driver.. resume works but tottaly garbled..
bridgman: I think we just need to get him another PC so he can run Linux on his laptop ;)
Gnutoo: bridgman, no pci express
groo: and no one else apear to have the same rpoblem
bridgman: groo: that's how people become open source developers :)
groo: bridgman: you dont know what you are proposing LOL!!!
rx__: they solve their own problems?
bridgman: groo: probably...
groo: bridgman: i already oppened the bug report, tested patches from dave and alex... tried every single config i could think off
bridgman: rx__: yeah, one of the big surprises for me was that the best way to get open source support for something is to give a laptop with that chip to a developer
groo: bridgman: besides calling a priest for a possession session, i dont now what else i can do
bridgman: As long as they need a laptop, you just wait a couple of weeks and miraculously everything works
rx__: sounds reasonable
groo: bridgman: LOL
rx__: free swag for support
groo: bridgman: if i had that kind of money i would buy one with an nvidia card ;)
groo: rx__: also called pimping :D
groo: bridgman: dont get me wrong.. im very impressed with the open source ati driver..
Gnutoo: bridgman, sould i pastebin my xorg.conf?
groo: bridgman: what im not impressed is with ati in general, altough thei have changed. big time
dmb: groo, in my opinion, they changed for the good
bridgman: Gnutoo: I'm just looking through recent fixes to see if there's anything applicable
groo: dmb: agreed
Gnutoo: bridgman, ok thanks a lot
dmb: nvidia has been nothing but crap for me
dmb: especially on linux
dmb: in fact, their card literly burned up
rx__: groo; every time you mention nvidia, bridgman kills a kitten
dmb: that i was using
RTFM_FTW: several of us do :P
dmb: i will never use nvidia again :P
groo: dmb: i agree, i was talking all this years of frustration with fglrx, but this late year as been mind boggling .. just to see the ati radeon git is like watching a child grow
groo: rx__: bridgman: i know he works for AMD.. i just couldnt help myself :D bridgeman thinking.. note to self.. no more drivers for groo
bridgman: note to self... find out where groo lives, offer troubleshooting, park car around back...
groo: bridgman: and get is only notebook :D leave him alive and without .. now THATS punishment
groo: bridgman: besides i leave in brasil.. im used to being robbed !!!
bridgman: dang, that's a long drive
groo: bridgman: now, seriously, any ideas? radeontools pll?
groo: bridgman: eheheh
bridgman: groo: for rs4xx I ask agd5f and airlied
groo: airlied is so sick of me bugging him he is hiding every time i go online :D
groo: agd5f is alex right? i tried is lvms patch.. didnt work either :(
bridgman: groo: yes
groo: see https://bugs.freedesktop.org/show_bug.cgi?id=16537
rx__: has suspending ever worked?
groo: like i said, besides calling a priest i dont know what else i can do besides crying or learning C + linux X driver dev
rx__: older versions?
groo: rx__: just with fglrx
bridgman: rx_: seems to work real nice for some, not for others; ditto for radeonhd but they usually work well on different systems
groo: rx__: suspend/rsume WORKS.. the problem is that the screen comes back totally garbled..
rx__: well i mean nongarbled then :)
groo: rx__: and STAYS garbled no matter what i do..
bridgman: groo: once we get a bit further with 6xx we'll probably go back and spend more time on the older chips... hold on, it stays garbled even after the windows start getting updated again ? How about when screensaver kicks in ?
groo: rx__: it used to work, but last time i tested radeon open source besides post 6.9.0 was in.. hmm dapper i believe
rx__: that's a start
bridgman: Gnutoo: I'm not seeing any mesa changes since 7.0.3 which jump out as addressing crashes but there sure are a lot of fixes for 300/400 in general; which X server version are you running ?
groo: bridgman: it stays garbled in console, X, with opengl, wihout, with drm, without. with one monitor.. dual head also
bridgman: I'm asking because if you are running a new X server (which seems likely) then upgrading mesa should be pretty easy
Gnutoo: xorg 7.3 with xserver 1.4.2
groo: im using packages from xorg-edgers, X 1.4.9xxx, radeon from git, mesa 7.1 rc1
rx__: it's possible to do a git bisect and find the offending commit
bridgman: groo: at some point someone who knows more than me will ask you to start dumping registers before and after the suspend/resume when they have time to look at the results ;)
bridgman: can you borrow another machine and get in via network or serial port ?
groo: bridgman: dave airlied already did that.. we spent a good couple of hours.. me rebooting and sending.. him scratching is head
groo: bridgman: he couldnt find anything out of the ordinary
bridgman: hmmm, feel like I'm walking down a well trodden path ;(
bridgman: have you tried shaking it ?
groo: bridgman: but not stud?
bridgman: Gnutoo: OK, that would need a new X server as well which is a bit of a pain
groo: bridgman: besides throwing it into a wall.. actually yes
bridgman: ok, guess they've tried all the usual things...
Gnutoo: bridgman, ok thanks i'll update my xserver to 1.4.99.05
groo: bridgman: you are starting to understand why im getting nuts, an airlied is getting hidden :D
bridgman: Gnutoo: that will let you pick up the latest Mesa which should definitely be better for gaming even if it doesn't fix your crash
Gnutoo: bridgman, ok thanks...
bridgman: I'm assuming the drm in 2.6.26 is pretty new but will check that next; glisse has been doing some interesting work dealing with GPU lockups but I think you should already have most of those fixes
bridgman: groo: yeah... I wonder if there's some kind of easy brute force solution like reinitializing the whole GPU on resume
bridgman: ... or you could do like I do and not suspend, works every time ;(
groo: btw, anyone here tryied vmware 6.5 beta 2? the new guest 3d accel doesnt work with open source radeon :( says the opengl is < 1.4, i openned a but report with vmware
Gnutoo: bridgman, ok i''l use mesa 7.1_rc3
groo: bridgman: tryed post to bios.. hang up the entire system.. see bug report
bridgman: Gnutoo: that's the good stuff ;)
bridgman: groo: OK, will take a look at the bug
bridgman: Gnutoo: just checking through drm fixes now, give me a minute
groo: bridgman: it has a nice video of the problem also :)
groo: ppl, gotta go before my wife solves the video problem throwing my note by the window..
TobiasTheCommie: groo: there is an x11 git overlay through layman
groo: bridgman: if you could think of something i would really apreciate
bridgman: groo: good luck ;)
TobiasTheCommie: groo: so, if you want, yo can do like me, and just git it all...
groo: TobiasTheCommie: link?
bridgman: Gnutoo: is there a driver date for drm showing up in the X log or anywhere else ? Probably either 20080528 or 20080613
TobiasTheCommie: groo: http://dberkholz.wordpress.com/2008/03/30/hot-new-x-leetness/ <- that is the closest thing to a guide i found... and it isn't really that close
Gnutoo: bridgman, ok i'll look
TobiasTheCommie: groo: just remember to install layman with subversion flag on(else you won't see the x11)
bridgman: I just realized I don't know how to map between kernel versions and drm commits in fdo... 'cause Dave might cherry pick fixes from mesa/drm master for the kernel tree
bridgman: anyone ?
groo: TobiasTheCommie: thanks, i will give it a spin
TobiasTheCommie: in any case, i have to sleep now.. but feel free to ask me later if you can't figure it out :)
TobiasTheCommie: everything is hardmasked, so you will have to unmask it in /etc/portage/
TobiasTheCommie: and now i go sleepies
groo: TobiasTheCommie: im using ubuntu. usually i do backports of the latest git tree.. X included
groo: night tobias..
Gnutoo: bridgman, i've compiled the drm inside the kernel(not as module...) i'll pastebin the log
groo: ppl, gotta go.. thanks for all the help, eventually this bug will be crushed and my note will be 100% ati complete :)
groo: good night/2 all
TobiasTheCommie: wait what...
TobiasTheCommie: why did i think groo used gentoo..
bridgman: Tobias: because Gnutoo used gentoo and they have a lot of letters in common
TobiasTheCommie: just figured it out
TobiasTheCommie: when checking the back log
TobiasTheCommie: stupid me
bridgman: Gnutoo: I'm wondering if going to a slightly older kernel might be a good first thing to try (if you aren't already updating x server & mesa)
Gnutoo: bridgman,here's Xorg log: http://www.pastebin.ca/1083093
TobiasTheCommie: Gnutoo: well, then what i just said to groo might be of interest to you...
TobiasTheCommie: that was confusing
TobiasTheCommie: and now i sleep.
Gnutoo: bridgman, now i can update Xorg and mesa...
damentz: bridgman: btw, you're open source drivers for everything r300 and less
damentz: are the best out there right now i think
bridgman: Gnutoo: I'm being stupid... if you're running Gentoo then you have all the source code there, right ?
damentz: with the opengl support
Gnutoo: bridgman, yes i have
damentz: i was playing warsow on my radeon 9600
damentz: with mesa 7.1 rc3
damentz: 60fps at 1024x768
damentz: everything was rendered correctly
damentz: with anistropic filtering
damentz: that's cool
bridgman: that is good; there are some really good people working on the drivers
bridgman: I still shake my head some nights when I read the backlog
bridgman: Gentoo: can you look in drm src/shared-core/radeon-drv.h ?
bridgman: damentz: 9600 not 9700 ? That *is* good
bridgman: Gnutoo: don't see anything odd in the log other than what seem like a lot of resume messages
bridgman: try drivers/gpu/drm/radeon/radeon_drv.c
Gnutoo: in /usr/src/linux/drivers/char/drm/radeon_drv.h there is #define DRIVER_DATE "20080528"
Gnutoo: bridgman, i have previous logs where it didn't completely freeze my computer as it does now if you want
bridgman: OK, so that is before Glisse's lockup fixes; sounds like it's not the final 2.6.26 either but I could be wrong
bridgman: Gnutoo: was anything different when computer did not freeze, or same configuration different day ?
Gnutoo: bridgman, but theses logs were done with older Xorg,drivers etc...
Gnutoo: bridgman, yes older drivers...
bridgman: Ahh, ok. Yes please on the log
Gnutoo: bridgman, but that's valid only for nexuiz...
Gnutoo: ok i'll pastebin the logs...
Gnutoo: as valid for nexuiz i meant: the log, but also the fact that it didn't crash with driver that are a lot older
Gnutoo: i think it was xorg 7.2
bridgman: OK, guess the choices are to go back to older drivers and find where the problem started (which will be a pain because it doesn't crash immediately) or go to a newer kernel with glisse's lockup fix and see if that helps
bridgman: I don't know much about gentoo & safest way to upgrade pieces of it... are you comfortable picking up a newer kernel ?
bridgman: I guess we should make sure that there is a newer kernel with Jerome's fix in the drm, or you'll need to just pick up a new drm instead
Gnutoo: bridgman, i can patch my kernel if you want
bridgman: I haven't built and installed drm myself so I'm thinking we might want to wait until someone with a bit more experience shows up
bridgman: I don't think a single patch is going to help; looks like there was a fix, then a fix to the fix, then a fix to that fix etc...
Gnutoo: bridgman, ok there are newer masked drm in portage...
bridgman: Is there a date on the latest one ?
bridgman: OK, that seems pretty good... it picks up pretty much all the fixes which seem useful, just one missing
bridgman: I would give that a try
Gnutoo: bridgman, ok thanks...
MostAwesomeDude: So I hear we have new codes.
bridgman: yep, all kinds of neat things happening
bridgman: two sets of new accel code in radeonhd, new atombios parser, initial kms & memory management code...
bridgman: .. and we still can't get a damn triangle out of the 6xx or 7xx
bridgman: we have a test program that works fine; we put EXACTLY the same code in the driver and it doesn't work
MostAwesomeDude: bridgman: Could be worse; you haven't heard of my bicubic adventures yet?
bridgman: I was watching them for a while; you're pretty much at the same point aren't you ? Everything nice, everything finished, great code, doesn't actually work ?
bridgman: Yeah, that sucks.
MostAwesomeDude: I'm doing exactly the same stuff as the EXA code, which works perfectly, but I'm hanging the card.
bridgman: Not like I can help, but what are the symptoms ?
MostAwesomeDude: Well, right now, it's just a soft-lockup whenever I try the Xv.
bridgman: Well, we're all learning a lot more than we ever wanted about lockups so I imagine we'll get past all these problems pretty soon
MostAwesomeDude: I know the second tex is getting to the card; my problem is just getting the card to USE it for something without locking up.
bridgman: Dumb question -- do we support multi-textured rendering in opengl today ? I assume the answer is yes but don't know for sure
MostAwesomeDude: bridgman: Oh, yeah, definitely.
MostAwesomeDude: ...Are you, by chance, talking about the Flash GPU requirements?
bridgman: Please repeat the question
bridgman: I didn't understand the first time ;)
bridgman: Flash like crappy youtube videos ?
bridgman: and recently announced gpu acceleration ?
MostAwesomeDude: As in, http://blogs.adobe.com/penguin.swf/2008/05/flash_uses_the_gpu.html
bridgman: OK. No.
bridgman: I was just thinking it might be a good place to look for "known working multiple texture" code
MostAwesomeDude: Yeah, I've looked through the stuff in r300_tex, and in the EXA masking code, which is pretty basic multitex stuff.
MostAwesomeDude: I figure I'll be figuring it out sometime next week.
MostAwesomeDude: And then I'll bash my head against the wall a few times.
MostAwesomeDude: Hm, I've gotta try the KMS stuff sometime soon.
bridgman: Assume you've picked through the status bits re: which blocks are busy ?
MostAwesomeDude: Busy blocks? Status bits?
MostAwesomeDude: What sorcery is this?
bridgman: The best kind... documented spells
MostAwesomeDude: Ooh, I see.
MostAwesomeDude: No, I haven't tried poking those at all.
bridgman: section 9.1.9 is the starting point I think
bridgman: 5xx right ?
bridgman: GA_IDLE register
bridgman: Not sure if you can do all that via debugger or if you need to have driver figure out when it's screwed and dump out the data; Matthias has been doing a lot of this on 6xx and might have ideas
bridgman: Then again if you ask Matthias about hung blocks he might scream and run away and we wouldn't want that
MostAwesomeDude: I would probably just do all of that at the end of my 3D register writing, and then get all of the status stuff dumped.
bridgman: Only catch is that docco says "first make sure the HW is hung by doing...."
MostAwesomeDude: Yeah, not useful for general-purpose stuff right now.
MostAwesomeDude: But for dumping info at the end of a known-to-hang process? Sure.
bridgman: Oh, does it hang that quickly ? Yeah, guess that makes sense
MostAwesomeDude: Well, it hangs before drawing the first triangle.
MostAwesomeDude: So either it's hanging in the compositing stage, or it's hanging in the rendering stage.
MostAwesomeDude: And if it's hanging in the compositing stage, then it's gonna be hard to find.
MostAwesomeDude: So I'm just gonna assume that it's me that's messing up, and go from there.
MostAwesomeDude: Hmm, I wonder if I can get anybody behind my crazy idea.
MostAwesomeDude: If we get RGB Xv support into the Radeon and Intel drivers, then Flash could use those.
bridgman: Good point. RGB should be easier than bicubic though, shouldn't it ? Just a different texture format ?
MostAwesomeDude: Oh, yeah, I've already written the patch. :3
MostAwesomeDude: But it's hard to find any ways to test it.
MostAwesomeDude: The mplayer devs should be writing me back soon, on how to force mplayer into RGB playback.
bridgman: www.youtube.com ?
bridgman: Ah ;)
MostAwesomeDude: It's sorta frustrating, because really only a few drivers support the RGB color formats right now. Nouveau, neomagic, trident, and maybe tdfx.
MostAwesomeDude: So vendors see that YUV is the only way to go.
MostAwesomeDude: And then applications like Gnash and Snes9x do RGB->YUV conversion for Xv.
MostAwesomeDude: And on-card, the texture engine does YUV->RGB.
MostAwesomeDude: It's sorta silly.
bridgman: It sells CPUs...
MostAwesomeDude: Yeah, I guess so.
MostAwesomeDude: eyes Intel's raytracing advertisements...
bridgman: Hold on, the article you linked implies that Flash uses OpenGL not Xv and no discussion of using RGB Xv even if it was available
bridgman: gotta run... ttyl
MostAwesomeDude: Aw, didn't get to finish the thought; the reason they're going with OGL is because most drivers don't provide RGB Xv...
dmb: whats the branch for the r500 modesetting support?
rx__: modesetting gem?
dmb: is there a livecd that exists so we can test it easily?
rx__: i wish :)
rx__: beg airlied for a fedora livecd
dmb: begs airlied
dmb: rx__, whats modesetting-101?
rx__: older stuff