Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2008-11-26

Search This Log:

EruditeHermit: airlied: congrats
spstarr_work: looks at koji
spstarr_work: new DDX fixes
spstarr_work: new kernel radeon drm fixes
spstarr_work: the DDX change affects me a little, R300 stuff, but kernel change no
maleadt: Hi all.I'm trying F10 out to see KMS in action, but it seems that there are still some problems with KMS and my R520 card (AGP based). Is this a known issue?
adamk: maleadt, Yes.
adamk: maleadt, AGP + kms seems to be an issue, and airlied is certainly aware of it.
maleadt: adamk: okay :) is there a way to help debugging this, provide some info or smth else?
adamk: Not that I'm aware of. I think it's more a matter of the developers having the time to fix it.
maleadt: allright
adamk: maleadt, Though if you are familiar with the intenals of AGP chipsets and GPUs, you could always look at the code and see if you can track down the source of the problem :-)
maleadt: I wish I was... and it doesn't seem to be the most easiest problem either, so I think i'll wait for a better opportunity to get used with driver development :)
maleadt: another quick question though: the kms failure resulted in a "eq overflow" error message, which I have encountered previousely at ubuntu. Circumstances were completely different though: it happened while having worked for some hours with compiz and exa enabled. Could those two be related (as the ubuntu ticked is still open, and it seems like many people suffer from it)?
maleadt: with "those two", I mean the KMS + AGP failure now and the Ubuntu bug previousely
adamk: I'm not sure. My KMS failure didn't show any error messages, in fact. Xorg just didn't work properly (or at all, depending on the AGP setting).
adamk: I'd guess that they are really separate bugs, but manifesting themselves with the same error message.
maleadt: I see
maleadt: well thanks for the info, I'll definitely keep my eye on the irc logs etc for any improvements
MrCooper: yeah, that message just means 'something is stuck', it can be caused by an infinite number of different things
adamk: maleadt, Did you try different agp settings?
maleadt: not really, do you mean at bios-level or rather xorg.conf settings
maleadt: currently it's pretty basic, no agp fast writes, etc
adamk: No, in the kernel boot line.
adamk: You can set radeon.agpmode to -1 (force it to PCI) or 1,2,4,8 to set the AGP mode
adamk: None of them work properly for me, but it make the difference between X failing to display anything, and X being annoyingly distorted :-)
maleadt: allright, I'll give it a shot. See you in 10 minutes or so :)
maleadt: adamk: sadly, changing the AGP mode does nothing
maleadt: the corruption still occurs
maleadt: strange thing is AGP doesn't initialise at all when using modeset. No "(==) RADEON(0): Using AGP 4x" messages, but rather a complete silence, or when I manually force a mode: "(WW) RADEON(0): [agp] AGP not available"
spstarr_work: adamk: like this?
spstarr_work: http://www.sh0n.net/spstarr/agp-kms-corruption.ogg
spstarr_work: http://www.sh0n.net/spstarr/kms-nonagp-kde-corruption.ogg
spstarr_work: ?
spstarr_work: :)
spstarr_work: or this
spstarr_work: http://www.sh0n.net/spstarr/agp-corrupt2.ogg
adamk: LIke this:
adamk: LIke this:
spstarr_work: adamk: some yeah i see some of that
spstarr_work: airlied: when EXA was first spec'd was there the concept of GEM/TTM (kernel memory managment?)
spstarr_work: management
airlied: spstarr_work: we knew it was coming, just not what it would look like.
spstarr_work: oh
airlied: dmn
airlied: AGP r5xx.. /me checks how disabled they are.
ball: Aha!
ball: Is there an open source driver for Radeon 9200, or have docs not been released for that?
airlied: ball: its been supported for years
ball: airlied: great, thanks a lot!
ball: We had a chap in #xubuntu who installed closed drivers from ATI and wonders why it isn't working. I'm trying to gently steer him towards the open source driver.
airlied: therre isn't aclosed 9200 driver
airlied: anymore
ball: Is the 9200 a card, or just a chipset?
airlied: a card.
ball: He wants to use S-Video out
airlied: fglrx dropped 9200 about 2-3 yrs ago
airlied: that should work fine
ball: I'm not familiar with X.org. What would he have to do to get video out of his S-Video port?
ball: (does it mirror the VGA/DVI video?)
ball: hello anubis
anubis: Hello
anubis: I'm assuming the xorg.conf has to be edited manually?
ball: anubis: I would be surprised if you couldn't do it through something like xorgconfig
anubis: command not found.
airlied: ball: http://lolworks.se/2008/11/xrandr-script-force-svideo-detection-on-debian-with-radeon-x600/
airlied: play with xrandr
ball: I don't know much about X.org yet, but on Xfree86 there was a program called xf86config that asked you what you had and how you wanted it configured.
airlied: you can do it dynamically now
ball: airlied: thanks
airlied: at least to test.
airlied: http://wiki.debian.org/XStrikeForce/HowToRandR12
airlied: also might be useful
ball: anubis: definitely check out that first link, but be aware that you may need to fix your driver config first
anubis: This is a Radeon 9200
airlied: anubis: the same command work for all the cards.
ball: has yet to try a Radeon card
ball: I've been on mach64 for years
ball: mach32 and mach8 before that
ball: Very happy with the 2D performance
ball: 3D has been a non-issue
anubis: I just want to use the s-video out.
ball: video's okay with Xvid
ball: anubis: that's discussed in the first link airlied gave us
ball: ...but like I said, fix your driver.
anubis: Dammit, what kind of stupid error is this. To edit the xorg.conf can't open to write.
anubis: UGH
ball: anubis: you're probably not root
ball: don't edit it by hand though
ball: that's just asking for trouble
anubis: Ok
airlied: ah I'd go edit it by hand using the instructions, but only if you get it working with xrandr first
airlied: if xrandr can turn on the TV then you know its possible
anubis: Holy crap, that little script thingy worked.
ball: there you go.
anubis: Well, the 2nd display is a copy of the first, but at an unsupported resolution.
anubis: Ok... Now the trick is to make the tv display it's own display, and not mirrored.
anubis: Like it's own desktop.
ball: Whether that's possible or not will depend on your hardware
anubis: It did it under windows. LOL
airlied: you need to add a virtual line as per the debian paeg
airlied: then use --right-of
ball: anubis: that's a good sign then.
anubis: It says that already : xrandr --output S-video --set load_detection 1 --right-of LVDS
airlied: yeah with the virtual line in xorg.conf that should work to put it on its own
anubis: Ok, so edit xorg.conf manually, right?
airlied: yup just follow the instructions on the Debian wiki page
anubis: Do you know how to make the TV an indepenent display?
airlied: anubis: yes adding the virutal line will do it
airlied: Section II.5
airlied: with that the --right-of will work properly
ball: I'm starting to wish I had a fancy graphics card :-)
airlied: less cahnce t-vout will work :)
ball: airlied: I wonder whether any of the cards I have laying around at home even have TV-out
ball: I know my main boxen don't
ball: Mrs. ball is due for a new computer, but I want to make sure it has DVI out.
ball: ...since presumably we won't be using CRTs forever
ball: (her monitor is on its last legs)
ball: ...think that count out the Asus Eeebox
anubis: I gotta run guys, thanks for all of your help. I'll keep playing with this.
ball: good luck
ball: For those cards with motion video hardware, is Xvid likely to take advantage of it?
airlied: no not yet.
ball: (those cards with open source radeon driver)
ball: ah, okay.
airlied: AMD might release some info later
ball: Okay. Has the 2D hardware changed much over the years?
airlied: the cards evolve over the years, various bits change in each release
ball: I didn't know whether 2D hit a plateau when they went all out on the 3D hardware
ball: I suppose 2D may inherit some benefit from 3D work
airlied: there is no 2D hw on newer chips
airlied: you just use 3D engine to do 2D operations
ball: Interesting.
airlied: even on older chips we use the 3D engine for desktop renderings
ball: I'm not even sure my card had 3D hardware
ball: perhaps it did, and just lacked motion video.
ball: breaks out google
ball: Is Radeon very different from (say) Rage Pro?
RTFM_FTW: considerably different
airlied: mach64 is grandfather of radeon.
airlied: but they don't share much.
ball: I know I used mach64 drives with my Rage XL boards.
ball: Radeon's a different animal then?
RTFM_FTW: very much so
ball: Is that because they went back to the drawing board?
airlied: thry just evolved bits for newer featuresets.
RTFM_FTW: Radeon also marks the age of the ArtX team at ATI (concerning HW design)
airlied: the mach64->r128->r100->r200->r300 is pretty much piece by piece changes.
ball: Is it possible to say one design (or driver family) is "cleaner" than the other?
airlied: not really... the newer chips have a cleaner design in some parts
airlied: because they stop evolving it and rewrote it.
RTFM_FTW: heh none of the older stuff can be considered "clean" IMHO
ball: Is r300 current?
RTFM_FTW: the newer stuff is better
airlied: ball: r600 is current
airlied: * office
ball: were ther 400 and 500?
RTFM_FTW: then again the newer stuff is considerably more complex as well
airlied: ball: yes, there are r700s now as well
edgecase: what's the newest that's "fully" supported by OS driver?
RTFM_FTW: R600+ is fully supported by the closed source driver
RTFM_FTW: err R6xx, R7xx that is
ball: what about open source?
ball: likes open source
RTFM_FTW: probably R5xx I'd imagine
chithead: r300 is fully supported by open source, r500 minus tv-out
edgecase: heh i guess i could have read the topic
RTFM_FTW: to what extent I wouldn't know
chithead: edgecase: http://wiki.x.org/wiki/RadeonFeature
ball: bookmarked! What does the (RHD) signify?
chithead: radeonhd
ball: HDTV ?
ball: Not easy to pick a sweet spot from that table. I suppose it depends what a person is using the card for.
RTFM_FTW: I'd imagine
chithead: r500+ actually
logari81: for a RV250 is the following right?
logari81: OpenGL vendor string: Tungsten Graphics, Inc.
logari81: for use of the radeon driver I mean
adamk: logari81, Yes, I believe that is correct.
logari81: is there any documentation about what to expect as OpenGL vendor string for diverse cards?
chithead: logari81: the vendor string comes from the driver, not the card
logari81: on my X700 I see DRI R300 Project
logari81: I also have seen "Mesa Project" in other cases and I m trying to understand this differences
adamk: logari81, It all depends on who wrote the driver :-)
ball: Does R100 have significantly better OpenGL support than Rage 128?
adamk: ball, Well, I don't know how significant it is, but r100 is capable of running compiz. I don't believe rage 128 can.
adamk: That may have more to do with the card than the opengl support in the drivers.
logari81: I refer only to cases regarding the opensource driver, I think I am missing something
adamk: logari81, It still depends on who wrote the driver.
adamk: For example, the r200 driver was written by Tungsten Graphics.
ball: Is there much difference in terms of power/heat between the various Radeon chipsets?
adamk: The r300 driver was not.
ball: (any to avoid?)
chithead: ball: avoid rs480
logari81: adamk: ok I understand now, thank you
ball: Oh, important question: which can run fanless?
adamk: logari81, Basically, I would not take the vendor string to mean much.
chithead: ball: all but the most energy hungry cards exist in fanless variants. depends on how large a heatsink you can live with
ball: chithead: any idea which are the coolest-running?
chithead: ball: the slowest and (usually cheapest). also very low power are the igp rs690 and rs740
logari81: adamk: I use it as a criterium to check for leftovers of the fglrx
ball: brb
ball: What's this ES1000 thing?
ball: Oh well. I have to go home now.
RTFM_FTW: mmm R1xx hardware is considerably better for GL than R128 is
RTFM_FTW: one classic example of that is in the realm of NPOT (RECT) support
RTFM_FTW: since R128 only supports a very limited number of cases in the EXT_texture_rectangle specification
chowmeined: on r300 im having multiple issues with opengl, in some apps the whole screen turns white, in another it makes the app crash in certain situations
chowmeined: id use the fglrx driver, but that makes lots of apps segfault also
chowmeined: slightly different but still unusable
spstarr: chowmeined: AGP or PCIe?
chowmeined: spstarr, AGP
spstarr: which distro
chowmeined: spstarr, ubuntu 8.10, ATI Technologies Inc Radeon R350 [Radeon 9800 Pro] (Secondary)
Amaeth: ´╗┐Hello; i need to see movies at tv-out with a radeon 9250..... any idea?
spstarr: chowmeined: no idea on 'buntu
chowmeined: spstarr, i dont really see what difference that makes
Amaeth: why? and another gnu/linux?
spstarr: chowmeined: big differences in distros
spstarr: kms vs no kms support, different patches not upstream yet
chowmeined: what does kms have to do with opengl being borked
spstarr: kms = Kernel Mode Setting
spstarr: i dont think 'buntu has kms in their kernel or a ddx that is kms aware
chowmeined: yes, it doesnt
spstarr: your mesa is 7.2?
chowmeined: spstarr, OpenGL version string: 1.3 Mesa 7.2
spstarr: what sort of crash do you get
chowmeined: the program segfaults
chowmeined: in the other app, (it goes full screen), instead of displaying, the whole screen goes white
spstarr: what output pastebin
chowmeined: but sound continues to play
chowmeined: k one sec
chowmeined: spstarr, well i cant get it to do it
chowmeined: :S