mikkoc: hi, im not able to compile http://cgit.freedesktop.org/mesa/drm/log/?h=r500-support with a 2.6.25 kernel
mikkoc: 2.6.24 is fine tho
mikkoc: more details: http://rafb.net/p/8fUbyI71.html
damian: Hi. In this config file: http://dpaste.com/44454/ I changed the radeon driver to vesa because with the radeon driver Xorg doesn't work properly. The Xorg starts but: 1) background is purple (shouldn't be) 2) I see a mouse pointer, it's running but I cannot click anything (for example icons in XFce). Is my configuration file wrong? (It's generated with Xorg -configure)
arekm: damian: better try to solve radeon driver problem than switch to vesa
damian: I'm trying... vesa is just to be able work...
damian: arekm: Where should I begin? I have no idea how to work out my problem...
tilman: what card are you using? try getting the latest code from git and see if it works better than the release
damian: RV370 5B60 [Radeon X300 (PCIE)] - Does it say something about my card? (It's not my computer, I don't know where is documetation..) OK. I'll try to get lates code... If doesn't help, I'll back...;)
tormod: damian: try without DRI and AGP. See also https://wiki.ubuntu.com/Bugs/AtiDriver for some common issues and tips.
arekm: well, without dri can cause problems since almost no one tests such setup
dennisb: I have no luck with the textured xv output on a x1550, no errors but just random triangles outputed
partymola: last git version for the driver makes my computer crash when X start
partymola: computer hangs completely :(
MrCooper: partymola: if you have a known working version, please try and isolate it with git bisect
partymola: i will have to go back step by step
partymola: until i find the stable version
partymola: but i am sure it was some of the changes
partymola: introduced on April 10th
partymola: oh, wait
partymola: i found what the problem is
partymola: my card keeps putting in the Xorg log
partymola: (EE) RADEON(0): Idle timed out, resetting engine...
partymola: i have like 1200+ lines of those in my log
partymola: I have a ATI Radeon XPRESS 200 5974 (PCIE)
partymola: i am using EXA
dennisb: and I do get xv working when using EXA, not without
partymola: MrCooper: no clues about the "resetting engine" ?
dennisb: with some tearing effects however, which is the main reason a wanted xv in the first place.
MrCooper: partymola: probably means the GPU locked up
partymola: well, i am curious about why it can log that, considering that the system gets competely locked. even ssh stops working
partymola: how can i get older revisions from git?
partymola: i don't know how to use git actually...
partymola: i just make git pull from time to time and compile it...
partymola: i think i get how to use bisect now
partymola: ok, it seems the bad one is 79c8d4 by Alex Deucher tagged as "RADEON: cleanup"
partymola: gotta confirm it
dennisb: partymola: look like a change that only remove comments in the code
partymola: weird :/
dennisb: yes. That's all there is in the diffs here: http://gitweb.freedesktop.org/?p=xorg/driver/xf86-video-ati.git;a=commit;h=79c8d4ca36a1c3e5fe759d4ccc379c36af8f1676
partymola: ok, this is weird... now even the last commit works :/
partymola: given that i always compile after a make clean, this makes no sense
partymola: how can i be sure that i am using the last git driver version? is it logged somewhere?
eboettcher: git fetch right before make?
partymola: git pull
partymola: oh, by the way
partymola: when compiling (on FreeBSD)
partymola: i get an error
partymola: that PACKAGE_VERSION_MAJOR, PACKAGE_VERSION_MINOR, and PACKAGE_VERSION_PATCHLEVEL are not defined
partymola: i just put them on config.h
partymola: and i can compile the driver
partymola: ok, last version is working now, so i guess it was some problem on my end
arekm: agd5f: http://carme.pld-linux.org/~arekm/MVI_0213.AVI
arekm: agd5f: I was talking about this problem with bright colors and tv-out on r300
edgecase: ok these DVI -> component dongles work with X300 in windows Catalyst, but not from the BIOS at bootup, which is a bit disapointing
edgecase: but it is progress
fbond: Hi; I have an iBook G4 and my 3D performance is not very good. glxgears starts off at about 9-10 fps, but then gets much better after I move the window. Sound familiar?
Magnade: fbond: move it once or continuing it to move it causes it to stay fast?
fbond: Magnade: moving it once.
fbond: Performance in openarena is not very good, ever, from what I can tell (but I can't figure out how to move the window since openarena captures the mouse).
Magnade: are you using some 3d window manager?
fbond: Magnade: no
fbond: Magnade: I'm just using metacity with no compositing or anything.
Magnade: you search the bugzilla for mesa?
fbond: Magnade: No, I'm not sure what the best search terms are. I've Googled several times.
Magnade: or for radeon driver itself i guess
Magnade: well i think id suggest looking at all open bugs for your card
Magnade: or the other route would be to just updating mesa and the radeon driver to latest
Magnade: and see if the problem still exists in the first place
fbond: Magnade: thanks, I'll see what I can do.
fbond: Magnade: I might just upgrade from Ubuntu 7.10 to 8.04.
mattmatteh: anyone have a banding problem with black and dark gray using xv ?
Magnade: fbond: might help there is some packages you might try tho hold a sec
fbond: Magnade: Okay, thanks.
Magnade: fbond: ^^
Magnade: fbond: if nothing else will help with 8.04
fbond: Magnade: Thanks.
bridgman: edgecase: I think we had a special dongle ("SmartDongle") which was recognized by BIOS
bridgman: at bootup but not sure how to lay hands on them. IIRC they had a dip switch which
bridgman: you used to pick a component video mode (eg 720p) then the BIOS knew what to do
bridgman: It might have only been for older chips, not sure. Do you think your component vid/dvi
bridgman: dongle is able to get edid information into the card ? If not, BIOS will go thpptth!
edgecase: bridgman, i have one of those dongles, someone here told me that they use GPIOs not DDC
edgecase: linux/Xorg doesn't find anything connected to DDC/i2c, but Catalyst under windows detects it
edgecase: i think the bios in x300 dropped support for it
bridgman: I don't think the smart dongle uses gpio *or* DDC -- there's a special i2c address
edgecase: i'm waiting to get back my 7500 DVI out to try it at bootup
bridgman: that gets read. I'll ask around and see if I can find out about current support
edgecase: ic, so an address scan of i2c bus then
bridgman: my recollection from late 07 was that smart dongle was supported on 5xx parts
edgecase: i didn't see any support in radeon driver,
bridgman: yeah, not even an address scan as just a read from a specific slave
bridgman: the sad reality is that getting tv working right seems to be about as much work
bridgman: as getting 3d working right, and 3d probably makes more users happy ;)
edgecase: oh ok i'd be nice to know, the website and manuals are a bit spoty on supported cards
bridgman: edgecase, which card did your smart dongle come with ?
edgecase: i bought chinese knockoff dongles off eBay ;<
edgecase: i have a range of cards here mach64 pci thru x300
edgecase: i think there are some TVs/projectors that are still viable, but don't have HDMI inputs, hence my interest in Component out
bridgman: interesting -- didn't know anyone had copied the smart dongle. Does your dongle
bridgman: definitely have the little dip switches for video mode ?
edgecase: i suspect, as with many chinese copies, that the outsourced mfg of branded parts had the specs leaked
edgecase: yes, w2k catalyst drivers on x300, shows a panel with checkboxes that correspond to the dip switch settings
bridgman: just to be clear, are there dip switches on the dongle itself or only in the ccc window ?
edgecase: dongle itsself has dip switches, 480i, 480p, 540p, 720p, 1080i, 16:9
edgecase: i have found that 7500 DVI card will bootup to panel's native mode
bridgman: do you think that is using edid or the dip switches to get the right mode ?
edgecase: and x300 will bootup with s-video, so i'm hoping to find a card that will bootup in component out mode
edgecase: well i think the dip switches are detected by catalyst, i only tried 480i since that's the only TV I have
edgecase: i tried Xorg but it didn't find edid, and someone here said they used gpio instead
edgecase: which i guess is possible, there's a lot of extra pins in DVI connector
edgecase: but there are VGA -> component dongles also, with dipswitches
bridgman: yeah, there are an awful lot of different ways to get video to a display ;(
edgecase: and lots of ways to wreck it
bridgman: nice for users, not nice for driver/bios developers
bridgman: well I have another question to add to the list, but it's not like I was anywhere
bridgman: close to running out of questions ;)
edgecase: what's your role in all this? former ATI developer? i gather some of the people must have been
bridgman: Yeah, I'm from the ATI side -- about 10 years. Engineering management for a while,
bridgman: then process engineering, then IP management, now I get to do something fun again ;)
edgecase: the other thing i can't figure out, is the 7-pin mini-din, supposedly for component out, but i can get nay a card to do other than s-video on it
bridgman: We've got at least one ex-ATI driver engineer about to join the team as well,
bridgman: along with Alex and one or two of our Shanghai driver devs
rx__: what ex-ati driver engineer? :P
bridgman: I *think* we had a smart dongle-type thing for the 7 pin as well. The main point is
edgecase: it seems Xorg/linux is on the verge of some critical restructuring, kms architecture, exciting times
bridgman: that component video doesn't have any signalling info other than analog load detection
bridgman: and that's not enough to know how to drive the display
edgecase: well, it might be,
bridgman: wait til' he starts and then we'll give out his name ;)
bridgman: yeah, there are very cool things happening in the x world right now. I haven't
bridgman: seen this level of change and collaboration for maybe 10 years
edgecase: if you consider, composite = 1 pin loaded, s-video= 2 pins loaded, component = 3 pins loaded?
bridgman: It's not just linux either; Solaris is now using almost the same X framework as Linux
edgecase: sure, the player OS that run Xorg, need to keep in sync or loose it
bridgman: re: # pins loaded, good point, something else to ask about
edgecase: how's X on novell these days :)
bridgman: my guess is that we probably only do load detection on one line, ie green
edgecase: yeah i can see in the load detection code, I sense there are circuits there the code just needs extending
edgecase: like, the .h files allude to it
bridgman: Novell has Dave Reveman (compiz) so they have lots of options. We've all been
edgecase: 3 constants defined in numeric order, only 2 used in load detection
bridgman: a bit driver-focused for the first months but starting to try to tie together
edgecase: i mean X on novell netware
bridgman: work at all levels of the graphics stack
edgecase: for comedic value :)
bridgman: Ahh, sorry, I was thinking SuSE. Is there still a netware product ?
edgecase: dunno, gcc can make netware object files etc, i used netware once and saw X in action
arekm: wonders if these disorders are tv thing? I get similar one when I connect canon camera (not just only tv-out on r300), weird http://carme.pld-linux.org/~arekm/MVI_0213.AVI
bridgman: Don't be askin' me to look at videos on my 21kbps dial-up line now ;)
edgecase: bridgman, it would be awesome to use the AVIVO cards with the smart dongle, the trend in HTPC being somewhat polytheistic in terms of source material codecs
arekm: the bright areas look like old-tv hardware without signal
bridgman: I'll see what we can find out. I'm surprised nobody is making dongles with
edgecase: having an ultra low-power HTPC based on mpeg2 hardware, that chokes on xvid, isn't going to gain much traction
bridgman: programmable edid blocks, so you just program in the EDID information you
edgecase: so either fast CPU + non-xvmc overlay that 90% of cards can do,
bridgman: want the card to see. I'm sure they exist but haven't seen them
edgecase: or *maybe* AVIVO is flexible enough?
edgecase: well any respectable hardware hacker would just pull the board out of on junk monitor, and cut the write-inhibit pin on the EEPROM
bridgman: A lot of monitors don't seem to have write inhibit in the first place; there is
bridgman: at least one dead monitor lying around the SuSE lab from early I2C testing ;)
edgecase: that's one avenue i want to explore for making RGB outputs "plug and play" for fixed frequency non-ddc monitors,
bridgman: I don't think mpeg2 is going to be a problem; I figure we can probably open
bridgman: up the IDCT/MC hardware on current chips pretty easily. The challenge is
bridgman: going to be h.264/vc1 -- there's more drm IP tied in there so still wait and see
bridgman: that said, I think our IDCT/MC hw can handle the odd block sizes from h264/vc1 so
edgecase: what about non-mpeg2 formats, wouldn't we loose the hardware assist ?
bridgman: we might get enough performance without uvd
bridgman: We can get quite a bit of hw assist anyways. Motion comp is mostly texture blending
edgecase: ok, sounds neat, i'm still considering hedging my bets by giving HTPCs beefy CPUs tho
bridgman: The nice thing is that you can get pretty beefy low power CPUs these days
bridgman: We should know a lot more in a few months anyways, so the other option would be
arekm: bought popcorn hour instead of trying to make own HTPC
bridgman: to make sure you have a socket that will take a more powerful CPU if needed
bridgman: sorry, jumping threads too much. Was trying to say "we'll know a lot more in a
bridgman: few months so other option is to wait a bit and see"
tigerchen: hmm, after killing my X-server with radeon i get the following in my logfile and x won't respond anymore:
tigerchen: Output 23 enable success
tigerchen: (EE) RADEON(0): Idle timed out, resetting engine...
tigerchen: [mi] EQ overflowing. The server is probably stuck in an infinite loop.
tigerchen: [mi] mieqEnequeue: out-of-order valuator event; dropping.
bridgman: tigerchen: this seems to have just showed up in the last day or two. Someone else
bridgman: did a bisect and found the offending change, but the change only seemed to affect
bridgman: comments so last time I looked heads were being scratched. Not sure if anyone
bridgman: has an answer yet
bridgman: Ahh, here we go ;)
tigerchen: ah, well, whatever, and i got no working gears anyway atm :( although they should be working
tigerchen: libGL warning: 3D driver claims to not support visual 0x54
tigerchen: Error: couldn't get an RGB, Double-buffered visual
edgecase: hmm popcorn time looks pretty good
bridgman: tigerchen, what is your card ?
tigerchen: x1400 on a lenovo t60
tigerchen: with xorg x11-drm (r500-support branch) from git and airlieds mesa r500-test branch
agd5f: tigerchen: the r500 drm stuff was alreayd merged to drm master
mattmatteh: agd5f, hi
agd5f: mattmatteh: hi
tigerchen: agd5f: ok, then i kill that branch-line
mattmatteh: agd5f, i got a dvd calibration disc and was trying to calibrate the colors ( dark blacks or browns displayed as dark blue/green)
mattmatteh: agd5f, i had the disc in 2 windows side by side, xv, and x11, seems like there is a banding issue at the low black end
mattmatteh: agd5f, dont know if thats a driver issue or in the hardware
tigerchen: agd5f: and i noticed a garbled text-console whenever i switch
mattmatteh: agd5f, i tried to get fglrx to work with xv but no luck, i think i will try again, to see if the card is capable of showing movies
agd5f: tigerchen: using the 3d engine in the ddx seems to kill the console. not sure why yet
agd5f: console works fine with 2D only or no accel
tigerchen: agd5f: ok, and it happens with fb and without, but i think you know that already
tigerchen: agd5f: but i still don't get gears working
mattmatteh: agd5f, have you heard of banding problems with xv ?
mattmatteh: or am i the only one ?
agd5f: mattmatteh: I don't recall any offhand
mattmatteh: agd5f, hmm. ill keep trying. thanks for your help. you have been very helpfull
mikkoc: tigerchen: i think i got the 3d working
mikkoc: at least glxgears runs at 800-900 fps
tigerchen: oh cool
mikkoc: thx man
tigerchen: hmm, worked for you but not for me :/
mikkoc: well, im not sure if it's really working... how can i be sure?
tigerchen: glxinfo |head
mikkoc: direct rendering: No
mikkoc: server glx vendor string: Brian Paul
mikkoc: server glx version string: 1.4 Mesa 7.1
tigerchen: ok, no dri 4 u
mikkoc: i guess that's wrong
mikkoc: nvm then
tigerchen: perhaps you should also try the normal x11-drm ebuild (not mine)
mikkoc: i think i am already
mikkoc: yours didnt compile with 2.6.25 kernel
mikkoc: when i saw glxgears running smooth i thought i had 3d :P
mikkoc: Xorg.0.log doesn't show any relevant errors tho
tigerchen: hmm, that's strange, because i only added the branch-line
tigerchen: or did you mean the mesa-ebuild?
mikkoc: no the x11-drm
mikkoc: but i think alex said it was merged into master already
mikkoc: so it was useless
tigerchen: that's the reason i said you should try the x11-overlay one
mikkoc: which i did
tigerchen: but i don't get why you get indirect rendering and i get these strange errors
mikkoc: which errors?
tigerchen: do you get this line when running glxinfo: Warning, R520 detected, 3D HAHAHAHAHA!!.
tigerchen: libGL warning: 3D driver claims to not support visual 0x54
tigerchen: Error: couldn't get an RGB, Double-buffered visual
tigerchen: these errors
mikkoc: nothing like that
mikkoc: i think im missing something
tigerchen: looks like that
tigerchen: but the patches should work
mikkoc: any special option in xorg.conf needed?
tigerchen: perhaps that helps
mikkoc: i get this stuff into dmesg:
mikkoc: [drm] Initialized drm 1.1.0 20060810
mikkoc: [drm] Initialized radeon 1.28.0 20060524 on minor 0
tigerchen: same here
mikkoc: [drm] Setting GART location based on new memory map
mikkoc: [drm] Loading R500 Microcode
mikkoc: [drm] writeback test succeeded in 1 usecs
mikkoc: same for you?
mikkoc: ok, i had the radeon module loaded at boot
mikkoc: i dont have: Option "Accelmethod" "exa"
mikkoc: lemme try with that
mikkoc: that didn't help
mikkoc: tigerchen: do you have this in xorg.0.log?
mikkoc: (II) RADEON(0): [dri] Found DRI library version 1.3.0 and kernel module version 1.28.0
mikkoc: also got this:
mikkoc: (II) RADEON(0): Acceleration enabled
tigerchen: I'll look
tigerchen: yes, both
mikkoc: ok, im gone for today.. cya
mattmatteh: agd5f, i just got xv to work with fglrx and it looks alot better. not perfect, it still has some banding ( if thats the right term ) and the banding is different. also the options are different with xvattr.
mattmatteh: i am not sure how xv is set up or works in the drivers, but it seems something is different. is there a way to grab the state of the card and compare ?
rx__: there is radeontool to dump registers
rx__: not sure if that's what you need though
mattmatteh: rx__, well, i am not a device driver developer, but would like to fix and help fix this problem
mattmatteh: rx__, are you familiar with the radeon driver ?
rx__: i'm not that familiar sorry :)