agd5f: airlied: http://www.botchco.com/alex/xorg/0001-drm-radeon-kms-fix-internal-tmds-setup-or-legacy-di.patch
agd5f: should do the trick
airlied: it doesn't fix me yet though ;-)
agd5f: might have to bust out the radeontool and compare kms to ums
airlied: yeah I'm not finding it
airlied: its wierd PLLs are different but looks sane in both cases
agd5f: so what's happening? no image? bad timing?
airlied: no image at all
airlied: works fine i I run it on crtc 0
agd5f: got a reg diff?
airlied: at the moment I've mostly got the same
dmb__: agd5f, sorry, kept disappearing (power outage), any updates on that bug?
agd5f: dmb__: sorry haven't had a chance to look at it much yet
dmb__: oh
airlied: agd5f: http://pastebin.com/m63ef3718
agd5f: airlied: ---kms +++ums?
airlied: agd5f: http://pastebin.com/m62b26376 is the notworking kms set
airlied: agd5f: < ums > kms
SmallCat_: MostAwesomeDude, las question, about those regs and temps, two theories, if hw context is used during validation with shader_api.h , is there some c way to mangle with the regs, fast methods, or couple of files looked almost like, while including hw context file, then structs are like hw ones also?
airlied: SmallCat_: please if you are going to keep coming back try and make sentences meaninful
airlied: maybe you could PM questions instead of flooding the channel
airlied: agd5f: any other regs I shuold dump?
agd5f: airlied: I usally do regmatch '*" and ignore the non-display regs
agd5f: airlied: RADEON_FP_H2_SYNC_STRT_WID and RADEON_FP_V2_SYNC_STRT_WID
agd5f: airlied: we aren't setting those regs at all it looks like
agd5f: airlied: http://www.botchco.com/alex/xorg/0001-drm-radeon-kms-fix-internal-tmds-setup-on-legacy-di.patch
airlied: agd5f: FP_H2 helps
airlied: agd5f: and the winner is you, thanks for that!!
agd5f: airlied: np
agd5f: I think there are some bugs related to that we can close
fly_: r300VertexProgUpdateParams:Params exhausted <-- what the heck does that mean
taiu: seems I fail at dri wiki captcha, anybody knows what it expects?
taiu: aa, got one correct!
fly_: r300VertexProgUpdateParams:Params exhausted <-- still havnt figured this one out
fly_: so the problem appears to be when using direct x 8.1 or higher on my r5xx
fly_: is there no support for higher than 7.0?
fly_: (this is direct x in wine so opengl
MostAwesomeDude: You'd have to ask the Wine people about Dx support.
fly_: they support up to 9
MostAwesomeDude: Heh, ish.
fly_: is
fly_: i only want 8.1
fly_: I was wondering if it was a problem on the driver side given the r300VertexProgUpdateParams:Params exhausted error
MostAwesomeDude: Hm. Which game causes that?
fly_: CounterStrike Source
MostAwesomeDude: Ah.
fly_: i cant find any relevant info on google
fly_: my google fu has failed me
MostAwesomeDude: Basically, it's trying to run a shader that's too big.
fly_: too big for what lol
MostAwesomeDude: For the HW.
fly_: for my x1950xtx?
fly_: i should hope not
MostAwesomeDude: It doesn't go in VRAM. It goes in a special part of the 3D engine.
MostAwesomeDude: Anyway, ISTR that other people have gotten Source games to work, so I'm not sure whether this is a recent problem.
MostAwesomeDude: Are you using Mesa from git?
fly_: no im on 'dora "rawhide"
fly_: so i have their repo
MostAwesomeDude: Close enough.
fly_: yeah
fly_: of course i am 64bit
fly_: let me check my mesa ver
fly_: for i686
MostAwesomeDude: Shouldn't matter too much; of course, your 32-bit Mesa could be ancient. :3
fly_: thats what im afraid of
MostAwesomeDude: But ISTR Fedora builds multilib all-at-once.
fly_: er no i had to get them seperately
fly_: there was a wine wiki for it
chithead: hi, I have a problem with xf86-video-ati 6.12.4, rv710 and an lcd monitor. when connected via dvi, the monitor image will suddenly go dark for a few seconds then reappear. this does not happen if I use radeonhd or the vga connector. Xorg.0.log: http://nopaste.com/p/aHraB0qzrb
MostAwesomeDude: Icky.
fly_: appears to be mesa 7.6
fly_: libdrm is 2.4.14
fudje: It's possible it's trying to use a program that's past the end of hardware. A lot of games programmers seem to think that because hard coding 75664636 in their shader works on their development machine that it will everywhere else too.
MostAwesomeDude: Mesa doesn't kick developers' asses enough when it comes to shader restrictions, unfortunately.
fudje: My favourite is when someone tries to insert a uniform at chunk 255 in an HLSL shader... hardware should support it, but Wine's already put a restriction on the amount available by inserting stuff in the very end to draw with.
fudje: I wrote a patch that would reshuffle such numbers so that they always fit, but you end up having to keep a list of pairs, and it gets messed up, and ends in a massive headache and apparent backwards progress
fudje: and then it turned out they were referencing these directly in the game's code as well and gave up.
MostAwesomeDude: Happens in open-source world too; see KDE4 and nVidia binary drivers.
MostAwesomeDude: Or, rather, see KDE4 and any driver that's not the nVidia binary driver.
fudje: I certainly get sick of the "It works on nVidia binary blob therefore we're done testing" attitude that seems to be about.
SmallCat_: git clone git://github.com/Cpasjuste/SDL-13.git heres the pandora port, maybe someone could find KMS suitable points there, it doesn't seem like that particular library is free though
SmallCat_: load like krh's egl instead and try target console
SmallCat_: moving to #android with that stuff
fly_: this is driving me crazy lol im so close i got into game and shot a few people in the face but still only dx 7.0
fly_: and i have to have 8.1 to compete because a popular plugin that enforces client settings will kick users with dx 7.0
twnqx: hi
twnqx: any idea why xrandr -q reports 1920x540 as avilable resolution on my displayport connector, even though the devices support only 1920x1080 ?
glisse: twnqx: display port ?
twnqx: (and 1920x1080 is not reportes, manually setting a 1920x1080@24 modeline and using it works however
glisse: unlikely we don't support display port
twnqx: displayport. the successor to hdmi.
glisse: hdmi ?
glisse: and the monitor is a display port enabled monitor ?
twnqx: no, i use a displayport => hdmi adapter (which is a level shifter)
twnqx: 1280x720 is correctly reported, but 1920x1080 turns into 1080x540
twnqx: 1920x1080, obviously
glisse: weird
twnqx: yes indeed
glisse: with kms ?
twnqx: no without
twnqx: i'm to clueless to set that up.
glisse: you should open a bug report with xorg log and monitor ref
glisse: can you connect the monitor through another output see if radeon get proper resolution then
twnqx: tried with two already
radian_29: hi
twnqx: also, is there a way to stop X from letting the kernel see ctrl+alt+del or Alt-F1 etc?
twnqx: i hate to accidently reboot my laptop when i want to lock something in a remote desktop
honk: Option "HandleSpecialKeys" "when"
honk: man xorg.conf
honk: maybe? ;p
radian_29: i own a RV370 [Radeon X300SE] and i'm testing kms on karmic beta with a kernel 2.6.32-rc4-149-ga3ccf63
radian_29: the card in only 32M, i use dual head (tv+vga) and i experience bad allocs on firefox and thunderbird
radian_29: is this a driver bug ?
radian_29: here my dmesg
radian_29: [ 660.449457] [TTM] Failed moving buffer. Proposed placement 0x00060004
radian_29: [ 660.449464] [TTM] Out of aperture space or DRM memory quota.
radian_29: [ 660.449469] [drm:radeon_object_list_validate] *ERROR* radeon: failed to validate.
radian_29: [ 660.449473] [drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation !
glisse: radian_29: agp ?
radian_29: pci exxpress
glisse: radian_29: got a dmesg ?
radian_29: yep above
glisse: full dmesg
glisse: pastebin somewhere
radian_29: okay
hoo-hah: what's the kernel option for KMS?
hoo-hah: I can't find it under graphics options
honk: staging
radian_29: here it is : http://pastebin.com/m3f2a904
twnqx: also: how do i activate the audio on the displayport? i guess there's some mixer on the graphics adapter that needs to be enabled, as the digital output seems to work otherwise
hoo-hah: thanks
hoo-hah: honk: thanks. Will I still be able to use compiz-fusion when disablign radeon-kms mode?
honk: ofc
honk: you dont need to recompile your kernel to disable kms though
honk: just add radeon.modeset=0 to your boot line
hoo-hah: cool
hoo-hah: thanks
hoo-hah: I'm guessing this will switch it back to DRI1? I'm after faster compiz performance
honk: just try it
honk: stop asking questions which you can answer by just doing it anyway
hoo-hah: yeah, kernel is building atm
honk:
honk: ^^
hoo-hah: I upgraded to 32.rc4, and accidentally used a vanilla config. I had to reconfigure anyway to enable KMS for radeon and some other hardware drivers
honk: you could've just copied the old config ;p
hoo-hah: my old config was pretty much building every module under the sun
hoo-hah: So i decided to properly strip it of unnecessary stuff
hoo-hah: - to reduce compile times
hoo-hah: - starting with a bare config was a better option
honk: I see :)
zeier: following http://xorg.freedesktop.org/wiki/radeonBuildHowTo, I get http://www.pastebin.ca/1621905 when compiling mesa for R600, which I was told was an header problem. I downloadedairlied/drm-2.6.git
zeier:
zeier: but now I do not know how to tell mesa to compile against drm-2.6
hoo-hah: twnqx: curiously, I never figured out how audio-through-hdmi worked. So instead of running one dvi cable to lcd, and one cable to audio box, you only use one hdmi cable to the LCD? Does this mean the LCD has speakers to use?
twnqx: yes
twnqx: it's "mixed" into the video bitstream
hoo-hah: okay
twnqx: that's why current high-end home cinema amps have HDMI inputs... they extract the audio stream from it and hand the video only over to your display (TFT, projector, whatever)
hoo-hah: digital audio from that differ from spdif at all?
twnqx: higher bitrate, yes
twnqx: like 7.1 LPCM (~6mbit instead of the 1.5mbit allowed on spdif)
hoo-hah: ah okay
twnqx: or the "enhanced" dts versions (dts-hs, dts-ma) or dolby true HD
fly_: seems like everything I try stupid things like glxgears still hard freezes my system. Is this most likely driver related?
adamk_: fly_, Yes.
fly_: Ok. and how would I fix this issue lol
adamk_: First, what video card do you have?
fly_: x1950xtx (r5xx)
adamk_: Are you using the drivers that come with your distribution?
fly_: er my mesa?
adamk_: No, I mean if you are using Ubuntu 9.04, for example, are you using the drivers that came with it, or have you updated them manually at some point?
fly_: im using the drivers that downloaded with fedora "rawhide"
adamk_: Do you know if KMS is enabled?
fly_: i dont know how to check that
adamk_: What is the output of 'cat /proc/fb' ?
fly_: nothing
adamk_: fly_, Have you tried accessing the machine from another box when it locks up?
fly_: no i dont have another machine handy
fly_: i know its still running because glxgears continues producing results
fly_: but i have no control and have to hard reset
soreau: Maybe you can look in /var/log/messages to find some clues
adamk_: Interesting.. Well, there are a few things I'd suggest... First, you could try updating your drivers. rawhide is probably pretty close to what is currently available.
adamk_: If you can borrow another machine, it'd probably be very helpful in diagnosing the problem if you could log in and copy off the output from 'dmesg' and your /var/log/Xorg.0.log file.
soreau: The worst problems I have are cool rainbow colored corruption in my favorite 3D game on the point sprites :)
soreau: (well it might not be point sprites, but it that's what it would look similar to)
fly_: thats one of the problems lol i have a feeling if this werent locking up it might allow me the support i need to run my favorite 32bit game better
fly_: doesnt /var/log/Xorg.0.log only write at launch of X
soreau: What is the output of 'glxinfo|grep renderer'? does it show DRI2?
fly_: OpenGL renderer string: Mesa DRI R300 (R580 7240) 20090101 TCL
kdekorte: fly_, might try adding pcie_aspm=off to your kernel command line if you are on Fedora rawhide
fly_: not until i know what it does heh
kdekorte: if that works can you report back
kdekorte: it is power management, affects some cards
fly_: k
fly_: ill give her a try
fly_: brb
soreau: It looks like he's using DRI1
adamk_: I'm surprised that KMS isn't enabled in rawhide.
fly_: ok well that stopped it from freezing
fly_: but it also made the fps go from 5000 to 275
Bibber_: Hi
Bibber_: damn I love the OSS drivers. Very stable 3d etc
Bibber_: But is there any plans of optimizing the drivers?
adamk_: fly_, Is the output of 'glxinfo | grep renderer' the same as it was before?
Bibber_: I get ~40 fps in counter-strike, and 99 fps with fglrx
fly_: OpenGL renderer string: Mesa DRI R300 (R580 7240) 20090101 TCL
adamk_: Interesting...
fly_: indeed
adamk_: fly_, I'd check to see if there is a bug report open for this and, if not, start one...
glisse: .win 32
fly_: huh? for why turning off power management makes frames go lower?
adamk_: It may be running at a lower clock speed by default.
fly_: i believe it is
adamk_: Usually glxgears framerate is pretty irrelevant, but that's a pretty significant drop.
fly_: i thought i saw something in the Xorg.0.log about clock being 94
adamk_: I'd check other applications to see how they perform, too.
fly_: also i havnt heard my fan kick on ever (is this normal)
fly_: in windows my video card sounds like a 747 taking off
kdekorte: fly_, hum, I didn;t see the frame rate drop when I did that... also looks like you are not running in KMS mode, you don't happen to have nomodset do you?
fly_: how to check for nomodset
glisse: Bibber_: optimizing is not yet top priority
glisse: fixing bugs is
fly_: <-- isnt good with linux video only servers
kdekorte: fly_ you can check the kernel line in grub, can you paste bin this "dmesg | grep drm"
fly_: http://www.pastebin.ca/1621999
fly_: nomodeset
fly_: its in there
Bibber_: glisse: Thats understandable. But how much do you think, we will get in extra fps's with opmitizations?
kdekorte: fly_, remove nomodeset, but leave pcie_aspm=off and see how it works
glisse: Bibber_: hard to say but i think we can get in the 70 - 110% range of fglrx depending on application
fly_: ok.. brb
radian_29: glisse : have you seen my dmesg ? http://pastebin.com/m3f2a904
glisse: radian_29: yes, did you try with lastest ddx & libdrm ?
fly_: k now i get consistent 300 fps
radian_29: not yet, i'll give a try when i have a moment
fly_: i had one hard freeze on the reboot after i made the change
fly_: being as this is a testing distro and it happened on the login i doubt it has much to do with video tho
kdekorte: fly_ when you type 'glxinfo | grep renderer' does it say anything about DRI2 now?
fly_: DRI2?
kdekorte: yes, it should
fly_: OpenGL renderer string: Mesa DRI R300 (R580 7240) 20090101 TCL
kdekorte: that is odd, did you get the fancy bootup screens?
fly_: what fancy bootup screen? i had to disable phymouth
fly_: plymouth
kdekorte: oh interesting... I get this OpenGL renderer string: Mesa DRI R300 (RV530 71C4) 20090101 x86/MMX/SSE2 TCL DRI2
fly_: what? no fair
fly_: gimme
adamk_: fly_, What is the output of 'cat /proc/fb' now?
kdekorte: any my glxgears is around 1400... so your card should be able to do better, although I don't have pcie_aspm=off on this machine
fly_: adamk_ nothing
adamk_: So you don't have KMS working.
fly_: i did roll my own kernel
fly_: which might explain that
adamk_: Well you might have mentioned that earlier :-)
kdekorte: fly_, then you probably missed something
fly_: missed what per say
kdekorte: drm-next
fly_: drm-next?
kdekorte: see topic
kdekorte: read the wiki, it will tell you how to build a correct kernel
kdekorte: but you should be able to use a fedora kernel
fly_: im using a modded one anyway
fly_: 2.6.31.4
fly_: cuting edge
fly_: :D
kdekorte: not really
kdekorte: 2.6.32-rc4 is cutting edge
soreau: drm-next even more so
fly_: Linux Combaticus 2.6.31.4-rt14 #1 SMP PREEMPT RT
soreau: What kernel does the version of fedora you're using supply?
zeier: I still have this issue: http://www.pastebin.ca/1621905. Could someone give me a hint? Somebody told me that this might be a header issue, but I don't know how to tell mesa where my headers (kernel sources) are located...
El_Angelo: Linux camelot 2.6.32-rc4-git4 #1 SMP PREEMPT Thu Oct 15 13:59:15 CEST 2009 i686 athlon-4 i386 GNU/Linux
fly_: 2.6.31.1-56.fc12.x86_64
fly_: close.. but the 31.1 has known issues
El_Angelo: aw
fly_: its a special child
El_Angelo: my glxgears frame rates went down :(
El_Angelo: do you need an xorg.conf ?
El_Angelo: with git ?
soreau: zeier: Maybe you need latest libdrm?
zeier: I'll check that. thanks
fly_: so which portion of this wiki am i reading and how much am i going to hate you guys when im done
fly_: lol
fly_: and what is up with this gallium ive been hearing so much about (my interest is peaked)
soreau: gallium isn't ready yet
fly_: butbutbut.. aww
fly_: oh g-d it its in staging drivers
fly_: i can even remember passing over that stupidly
twnqx: is an RV635 R500 and older oder R600 and newe?
kdekorte: rv635 is an r6xx type..
fly_: so will KMS make my r5xx good again?
kdekorte: fly_, I'm using the -56 kernel, what is wrong with it?
twoerner_: airlied: ping
fly_: kdekorte, what?
fly_: kdekorte, when did i say that
biotube: [07:52]
fly_: i was talking in regards to the fedora kernel
adamk_: fly_, It's hard to say if it will make it work again. But it's certainly an option to try.
fly_: more importantly will it maybe fix r300VertexProgUpdateParams:Params exhausted
fly_: plzplzplz
soreau: fly_: You're wasting time by asking questions. We can't predict the future. Just try it
fly_: but the title say "We can predict the.. " oh wait no it doesnt
dileX: hi
dileX: in VT-1 (booted with nomodeset, loaded radeon k-m with modeset=1) I can not use the full resolution
dileX: Console: switching to colour frame buffer device 160x50
dileX: approx. from mid below and the right side seems to be limited - no cut-n-paste possible
dileX: this is with linus-tree and drm-next on top
dileX: mesa latest GIT
biotube: after you change the tty resolution, you need to restart gpm
dileX: biotube: sure?
biotube: yes
dileX: I will try - sound funny
dileX: but anyway
dileX: hm
dileX: OK
dileX: next boot
dileX: switching from VT-7 (Xorg) to VT-2 -> no problems
fly_: 3000 fps glxgears
El_Angelo: on what gpu?
fly_: x1950xtx
El_Angelo: hmmm
El_Angelo: only 700 on this rs780
zeier: I do not want to overwrite my debian drm-files, so I added a prefix when compiling. How do I tell my system to use /opt/drm/include/drm... instead of /usr/include/drm?
fly_: hmmmm thats interesting i booted up wine and got black lines
El_Angelo: er,
El_Angelo: no
El_Angelo: 180fps :)
El_Angelo: horrible
El_Angelo: with a 100% cpu load
fly_: cat /proc/fb \\ 0 radeondrmfb
fly_: yeah the game crashes now and i get those weird lines
fly_: and was that supposed to change my DRI?
fly_: OpenGL renderer string: Mesa DRI R300 (R580 7240) 20090101 TCL DRI2
fly_: it did
fly_: well now the question becomes.. what happened to my game lol
fly_: thinks KMS kicked my dog (i love the staging description)
radian29: glisse : last git ddx and libdrm -> same bad alloc
glisse: radian29: dual screen ?
glisse: resolution ?
radian29: yes
radian29: tv + vga
glisse: 800x600 + 1280x1024
radian29: yes
El_Angelo: oh
El_Angelo: is that a known issue?
El_Angelo: i have the same on my 1920x1200+1280x1024
glisse: radian29: Compiz ?
radian29: nope i
radian29: compiz white (i suspect mesa)
dmb__: hehehe, doing cat /dev/urandom > /dev/fb0 does some funny colors :)
El_Angelo: rofl
glisse: El_Angelo: how much vram ?
El_Angelo: you mean on the graphic card?
glisse: yes
dmb__: vram=video memory
El_Angelo: i think 512
dmb__: might have bad memory on the card :)
dmb__: which i have had before
El_Angelo: Sapphire HD4830 512mb DDR3 PCI-express
El_Angelo: hmmm
dmb__: you see artifacts sometimes?
El_Angelo: a mouse pointer that gets multiplied 5 times is something i see quite frequently
El_Angelo: other than that nothing really i noticed
Bibber_: glisse: Wau, up to 110%? That sounds sweet
glisse: Bibber_: depending on application
Bibber_: Yeah ofcourse
El_Angelo: dmb__: is that something that classifies as artifact?
radian29: i get garble screen at gnome launch, but after it's ok
Bibber_: I guess fglrx has alot of optimization popular games
glisse: but i don't expect we will outperform fglrx in many or any app
dmb__: El_Angelo, i'll let the experts deal with it, i honestly have no idea what i'm talking about other then past experiences :)
dmb__: glisse = expert
El_Angelo: it's actually possible cause i did physically damage the card
glisse: radian29: you should open a bug and attach full dmesg + full xorg log
El_Angelo: i soldered everything back
El_Angelo: and it is kind of working
glisse: you have enough vram so it should work
radian29: okay will do
glisse: it's definitly a bug in our side
dmb__: so these video card distributors, asus, sapphire, etc, does AMD just sell them the chipset and they solder everything together and put there own twist into it and sell it?
adamk_: Pretty much.
glisse: dmb__: it's bit more complex than what it seems, they choose which ram chipset they will use and they have to choose various other component on the board
glisse: quality of the card is highly dependent on the quality of this various others components
dmb__: oh
dmb__: interesting :)
glisse: if you buy the cheapest there is likely a reason for it to being cheap
dmb__: does ATI sell there own, or a reference implementation?
dmb__: looks at his $20 video card :P
glisse: ati provide reference implementation AFAIK but reference implementation doesn't mean it gives specific ram chipset
dmb__: oh
glisse: there is many ram chipset manufacturer it's like kingston, corsair, ... for the ram of your cpu
glisse: i think other component on the board are pretty similar as they are not that much complex
Bibber_: El_Angelo: Maybe you got the same bug as me
dmb__: oh
Bibber_: El_Angelo: The ddx driver is apparently giving a bogus pointer to vram
Bibber_: Because you got > 256 mb ram
Bibber_: You have to patch the ddx driver, and everything will be fine
Bibber_: El_Angelo: Look at http://bugs.freedesktop.org/show_bug.cgi?id=24301
lab: Hi. I have installed the radeon driver for a x1200 card. unfortunately I can only set the 640x480 resolution. what could I do ?
lab: here's xorg.0.log: http://sprunge.us/DgDR and xrandr output: http://sprunge.us/fiPG
lab: (on xubuntu jaunty 9.04)
jcristau: hmm
jcristau: (II) RADEON(0): Query for AtomBIOS Get Panel EDID: failed
jcristau: (II) RADEON(0): Added native panel mode: 640x480
lab: jcristau: yes but... what do you suggest?
lab: :)
jcristau: try a newer driver, and if that doesn't work better, file a bug?
lab: a newer radeon driver ?
jcristau: yes
lab: I installed this version from a deb package
lab: jcristau: is this link good for amd too? http://wiki.x.org/wiki/radeonBuildHowTo I see the string xf86....
fly_: K... so I am crashing glxgears still it seems. and my game wont run. so i decided to experiment and make a new xorg.conf by issuing (Xorg -configure) to which it made one but also made a point of saying "(II)[KMS] No DRICreatePCIBusID symbol, no kernel modesetting"
jcristau: lab: should be.
lab: ok, let's try
lab: jcristau: I really don't understand which part of this tutorial refers to the installation of the drivr
jcristau: lab: it should be enough to download the xf86-video-ati 6.12.4 tarball from ftp.x.org, run apt-get build-dep xserver-xorg-video-ati, then in the unpacked source tree ./configure --prefix=/usr && make && make install
Bibber_: glisse: Will you wait for galium3d, before you start with optimising?
lab: let's try jcristau. thanks
glisse: Bibber_: it's more complex than gallium/classic
Bibber_: Well it isn't in mesa/galium3d one would optimise?
glisse: i think biggest bottleneck are in ddx & kernel right now
Bibber_: Wauu. I really didn't think that
Bibber_: I though all the optimizing work, laid in galium3d or mesa
lab: jcristau: in which folder is xf86etc. ?
Bibber_: Well, keep up the good work. I hope you know that you outperform fglrx in everything but 3d and powermanagement :)
lab: I can't find the tarball
lab: found it.
lab: (I have just found it)
lab: jcristau: how can I check if I really have installed the tarball version? I mean... I cofingured, made, made install etc. but...the screen is always 640x480
fly_: .. sigh
lab: I also rebooted
jcristau: lab: look for ati_drv.so or radeon_drv.so in the X log, the line below should have 'module version = 6.12.4' if it's the new one
jcristau: instead of 6.12.1
lab: jcristau: yes, it appears as 6.12.4
lab: but the problem is still the same
jcristau: lab: sadness.
jcristau: lab: is this on a laptop? if so what's the panel's resolution supposed to be?
lab: it's an industrial pc. with a 1200x600 (or something similar) monitor
fly_: great now im just randomly freezing
jcristau: what's an industrial pc? :)
jcristau: lab: the external monitor seems to be detected as 1680x1050
lab: yes, it's correct
lab: jcristau: it's mini-pc intended for a factory:)
lab: jcristau: 1680x1050 is the right monitor res
jcristau: lab: ok. so it has a mobile chip, and the driver expects a laptop panel. it doesn't find its resolution (because the panel isn't there), so it assumes 640x480.
jcristau: lab: as a workaround, a Monitor section in xorg.conf with Identifier "LVDS" and Option "Ignore" should get you going
lab: jcristau: unfortunately xorg.conf is empty
jcristau: lab: so add 4 lines to it :)
lab: jcristau: easy to say:)
jcristau: barring bonghits, 'Section "Monitor"\nIdentifier "LVDS"\nOption "Ignore"\nEndSection\n' should work
taiu: seems r600SetTexOffset does not set the bo size to right value
taiu: fixing that might fix the compiz dri1 corruption
lab: jcristau: I had to remove those lines. The screen was not visible at all
lab: sadness
jcristau: lab: without the xorg.conf, does 'xrandr --output LVDS --off; xrandr --output DVI-0 --auto' get things to normal?
fly_: hmmm i think ill give the new rc a try as well
fly_: might have better staging
fly_: also my 1000hz real time preemption might be partly causing these hard locks
lab: jcristau: do you mean that I should type those two commands ?
jcristau: lab: yeah
lab: let's try (anyway... what a nightmare :) )
agd5f: taiu: shouldn't matter. the dri bo is faked for tfp. it's just a pointer to a X pixmap. see radeon_legacy_bo_alloc_fake()
lab: jcristau: black screen :( :(
taiu: agd5f: but it's used as a texture when rendering?
taiu: taiu: and when size is wrong Syncsurf does't sync it
taiu: ok highlighting myself :)
agd5f: taiu: the mesa texture object should have that info already
agd5f: oh right
agd5f: the sync. good call
taiu: agd5f: I patched it a bit and cannot reproduce currently
agd5f: taiu: excellent!
kdekorte: taiu is this for the texture reuse in clutter?
taiu: kdekorte: nope :(
kdekorte: no worries...
taiu: agd5f: I don't know it I can rely on t fields there, so I would like smb else to verify/patch this, my attempt http://dpaste.org/GZaP/
taiu: or maybe there's some other way...
taiu: bbl
agd5f: taiu: looks good. probably want to align to hw pitch requirements.
agd5f: taiu: http://dpaste.org/ixpS/
lab: anyway.... I shrug
fly_: plays "rawhide" while he "rolls" a kernel
Plouj-: do _you_ guys dual-boot rawhide with stable versions of Fedora?
lab: jcristau: thanks for all your help
jcristau: lab: np. i have no further idea, so maybe file a bug with a description of the problem and your logs, and agd5f or someone can have a look.
fly_: great success
fly_: no more glx freezing
fly_: i actually gained 100fps on glxgears
fly_: added bonus
fly_: HOLY HELL :D im getting 85fps (my monitor refresh rate) in the stress test on counterstrike source
fly_: its jittering like it REALLY wants to go higher
fly_: but i still get - r300VertexProgUpdateParams:Params exhausted - when i try to go dx 8.1
kdekorte: fly_ so what combination are you using now?
Bibber_: fly_: Are you serious? What gfx card and with what gfx stack?
rnoland: agd5f: corruption fixed?
agd5f: rnoland: yup
rnoland: awesome...
rnoland: looking for commits now...
rnoland: did my patches look ok, btw?
agd5f: rnoland: I pushed to the 7.6 branch
agd5f: your patches too
rnoland: ah, cool
rnoland: has been screwing with zfs bootcode today.... so i'm a bit behind
fly_: bibber_ x1950xtx
Bibber_: fly_: kms?
fly_: kdekorte, trying the 2.6.32-rc4
fly_: yup kms
Bibber_: Would I get the same boost with r600?
mokoloko: I WANTS DAT FOR MI 4850 NAO!
kdekorte: fly_ is on an r500 card... the r6xx is not quite that fast yet
bridgman: agd5f; ping
agd5f: bridgman: pong
bridgman: I was talking to "lab" over on #ati; looks like the bios is simply wrong for the hardware; it says there's an LCD panel but there isn't one, just an external DVI-attached display
agd5f: bridgman: ok
bridgman: we ran into another user whose panel had died and was trying to make the driver ignore the BIOS and just use the external display but I don't remember details or whether there was success -- do you remember ?
agd5f: bridgman: there's a randr option to ignore outputs
bridgman: adamk was helping him earlier, let me check if that was tried
lab: bridgman: I tried to make the driver ignoring it, configuring xorg.con and then trying with xrand
agd5f: http://wiki.debian.org/XStrikeForce/HowToRandR12
lab: but without success
lab: agd5f: this was exactly what jcristau suggested and what I tried
agd5f: Option "Ignore" "true" in the monitor section for LVDS
lab: the result was: black screen
agd5f: lab: pastebin your xorg log and config
bridgman: looks like the suggestion was Option "Ignore" without the "true"
jcristau: bridgman: the parser handles that just fine
lab: here's xorg.0.log: http://sprunge.us/DgDR and xrandr output: http://sprunge.us/fiPG
fly_: omg .... someone plz help me figure out this params error.. i can get into a game but as soon as i try to join a team it crashes
agd5f: also make sure the monitor section is associated with the LVDS output
fly_: i can almost use dx level 80
fly_: 8.0
agd5f: so it doesn't get applied to another output
kdekorte: fly_ are you using mesa from git?
agd5f: fly_: the wine translation is making gl program that is too big for the card
bridgman: anyways, taui / agd5f congrats on the fix
fly_: so theres no way to fix that?
bridgman: lowers the cone of silence and goes back behind the office firewall
agd5f: fly_: need to fix the wine dx->GL translator to be more efficient or make the driver compiler more efficient
zeier: I successfully compiled mesa, xf86-video-ati and kernel with KMS. Now I have the (known) problem that AIGLX doesn't recognize my /opt/xorg/lib/xorg path (set in corg.conf). To get HW 3D to work, what should I enable? (where can AIGLX be disabled, and sth other enabled)
fly_: so in otherwords thers nothing "I" can do
agd5f: fly_: not unless you want to hack on some code
kdekorte: zeier, use a symlink?
adamk: lab: Did you get/understand that suggestion from agd5f about associating the monitor section with the LVDS output?
osiris_: fly_: I have a patch in my local repo for the parameter exhaustion
zeier: kdekorte: great idea... Something the problem seems so complicated and is so easy to fix...
fly_: hmm? plz dont be joking
zeier: s/Something/Sometimes
fly_: osiris_ whats this about a patch?
osiris_: fly_: http://pastebin.com/m29ede25f
fly_: run as a LD_PRELOAD?
osiris_: if you don't install compiled mesa, then yes
kdekorte: Actually there is a LIBGL_PATH variable I think.. might be LIBGL_DRI_PATH... would need to look it up
fly_: now this is the 32bit mesa i need to edit correct?
kdekorte: yes
soreau: kdekorte: LIBGL_DRIVERS_DRI ?
soreau: Err.. LIBGL_DRIVERS_DIR
taiu: ok, the first one correct gets the prize ;)
soreau: LIBGL_DRIVERS_DIR=$PREFIX/xorg/lib/dri IIRC
taiu: nope
taiu: hmm, actually i may eat my words
fly_: wait im confused here.. this cant be complete code
agd5f: taiu: I pushed a slightly modified patch. thanks for tracking that down!
fly_: no includes no end bracket it doesnt make sense
soreau: fly_: git apply /path/to/file.patch
fly_: if im compiling from source
taiu: ok
fly_: ok... so -eyerub- remove mesa i686 and rebuild it from git Si?
fly_: then git apply
fly_: i mean apply before building
fly_: ssorry im very tired
rnoland: waits for the merge to master...
fly_: osiris_ sorry I must be a pain right now but isnt this going to configure mesa x86_64
fly_: and i think im missing how exactly to apply this
fly_: ah i see i can specify the arch in the config
osiris_: airlied: there are still some problems with tex bo accounting on r300. I'm getting "radeon_cs_gem.c:121: cs_gem_write_reloc: Assertion `bo->space_accounted' failed." when running wine d3d9 test, and some games
TCW_: xserver-xorg-video-radeon 1:6.12.3-1 X.Org X server -- ATI Radeon display driver <- how stable may thet driver be for a rv530 chip with 3D / xv /whatever other features a user would expect to work like a charm?
osiris_: fly_: patch -p1 < patch.file
fly_: oh my god duh
fly_: i do that all the time for kernels -bang head on wall-
osiris_: airlied: here's a patch for wine that narrows down the problem http://pastebin.com/m545392d6
fly_: patch unexpectedly ends in middle of line \\ patch: **** Only garbage was found in the patch input.
fly_: thought it looked incomplete
osiris_: airlied: and here's the backtrace http://pastebin.com/m1ab39e78
osiris_: fly_: copy the patch text from the textarea, not the colored version
fly_: i did
fly_: im supposed to be in the mesa directory (that git created) correct?
osiris_: yes
soreau: fly_: Pastebin ate some characters I bet, maybe @@
agd5f: osiris_: might be related to bug 24131
fly_: im pretty sure pastebin did eat some chars
fly_: and there were some missing @@'s
osiris_: agd5f: I don't think so. it looks to me like a call to radeon_cs_space_add_persistent_bo is missing somewhere
soreau: fly_: Seems pastebin uses @@ at the beginning of a line to highlight it
fly_: yeah that seems probable as all those lines are highlited :)
fly_: what an oversight
devilshawk: fly_: you where on gentoo before? because if you where i got working 32bit libs. i get direct rendering context for both glxgears and wine apps
fly_: i switched
devilshawk: *were
fly_: gentoo was being a pain in my rear as everything required my taking a nap
devilshawk: fly_: you nuked the other system?
fly_: no i have it on my other harddrive'
fly_: has way too many harddrives for his own good lol
fly_: ive got a debian drive a gentoo drive a freebsd drive a windows drive and now a fedora drive
fly_: @_@ it gets kind of confusing some times lol
devilshawk: fly_: so if you want i can but the binaries on server (actually just a updated 32bit mesa and libdrm where needed)
devilshawk: fly_: should work with the current emul-linux pkgs in gentoo
fly_: hmm share the wealth with the gentoo community :) if this works i might make gentoo my "project" drive because i believe it has the potential to be great but its soo cumbersome
osiris_: fly_: does the patch work?
fly_: osiris_ it patched im just trying to figure out how to install the bugger
fly_: make linux-dri-x86 gives me make[3]: makedepend: Command not found
fly_: and im like no... its make depend
fly_: i need imake ;)
fly_: mv: cannot stat `libGL.so.1.2': No such file or directory
fly_: next error on the chopping block
osiris_: fly_: ./configure --prefix=/usr --with-dri-drivers=r300,swrast --disable-gallium --with-demos="" && make -j3
fly_: howd you know i run a dual core ;)
fly_: but theres no configure in there
fly_: thats the first thing i tried
osiris_: then run autogen.sh first
fly_: me beats libdrm with a spoon *Requested 'libdrm >= 2.4.15' but version of libdrm is 2.4.14
fly_: yet another thing to update
kdekorte: fly_ you can find it in koji
fly_: whats on koji?
fly_: looks for his cat
kdekorte: you are on fedora right?
fly_: yeah
kdekorte: koji.fedoraproject.org
kdekorte: all the latest builds
fly_: i thought you were talkin about my cat lol
fly_: its name is koji
kdekorte: think of it as rawhide+
kdekorte: sometimes things are broken, but it can be useful
kdekorte: search for libdrm and you should find the 2.4.15 version of live drm
spstarr: r6xx + kwin = win
kdekorte: there is also a command line took named 'koji', use it to download builds ' koji download-build --arch=i686 [build #]
kdekorte: koji is a good cat name
fly_: kojis fat
fly_: i got it and installed it
fly_: i think sutogen.sh is looking at my x86_64
fly_: autogen*
kdekorte: fly_ I have a command for building mesa in 32bit mode
kdekorte: ./configure --build=i686-pc-linux-gnu --prefix=/usr --with-dri-drivers=r600,swrast CC="gcc -m32" LDFLAGS="-m32" --disable-gallium --without-demos --enable-32-bit
fly_: i cant ./configure until i auto
fly_: theres no configure
kdekorte: you can feed most of that to autogen.sh as well
dileX: @r300g users: is openarena broken with latest mesa GIT?
fly_: Requested 'libdrm >= 2.4.15' but version of libdrm is 2.4.14
mokoloko: tried opensuse 11.2 rc1 and this time computer only freezes when I enable kwin desktop effects :)
mokoloko: I bet the freezing on fedora and mandriva was related to KMS
mokoloko: since opensuse doesn't use it
agd5f: fly_: you can remove that check, it's not really required for radeon
soreau: dileX: Has is ever worked? 0.o
soreau: s/is/it
fly_: how would i remove that lol
dileX: soreau: it has - only the logo looked strange. and yes, a bit slower than classic-mesa. means you are shot a bit earlier. but I am no good gambler.
dileX: (the intro logo)
agd5f: fly_: edit configure.ac and change LIBDRM_REQUIRED=
osiris_: fly_: git revert 859828cc4fb989bc5b67d26991a090a9f37e7c05
agd5f: to whatever version you have
fly_: osiris_ whats that about
osiris_: removes the 2.4.15 check
fly_: oh . but the fact its still looking at 2.4.14 a good indicator that its not trying to build a 32bit?
fly_: isnt*
agd5f: fly_: nothing to do with that
osiris_: fly_: why are you building 32bit mesa anyway?
fly_: for my 32bit game that has the param error
osiris_: aah
fly_: http://pastebin.com/d5c000d42
fly_: thats the configure output
fly_: doesnt look very arch descriptive
osiris_: pass the --with-dri-drivers=r300,swrast --disable-gallium options to make it compile faster
osiris_: and --without-demo
osiris_: *demos
osiris_: got to go
kdekorte: fly_, why are you not using the configure I gave you?
fly_: i am now :p
kdekorte: run autogen.sh and then rerun the configure and it should compile properly
dileX: Is there a need to rerun configure (after autogen.sh)?
jcristau: autogen runs configure
fly_: there we go now it looks official
soreau: dileX: Well I have r300g from nha mesa repo installed to /opt and testing on a per-app basis with LIBGL_DRIVERS_DIR=$PREFIX/xorg/lib/dri and I get only black screen with running openarena
fly_: btw kdekorte im compiling for r300 :P
fly_: thx tho
dileX: soreau: which r300g are you testing? here: dri/st
fly_: omg... mesa does not like me at all - mv: cannot stat `libGL.so.1.2': No such file or directory
fly_: pissy kittens everywhere
soreau: fly_: You were told how to configure it to get around that..
soreau:
fly_: i did configure it
fly_: this is gmake
kdekorte: you missed --libdir
soreau: --with-demos="" is key
fly_: so now gmake wants to holler at me :p
kdekorte: ack I don't have it in mine
kdekorte: nope, libdir should be defaulted fine
kdekorte: soreau, why not just "--without-demos"
fly_: can i use those same tags on gmake? lol
kdekorte: you just use "make"
kdekorte: no options
kdekorte: and then sudo make install to install it
soreau: dileX: I'm compiling --with-state-trackers=dri,glx and cp'ing radeon_dri.so -> r300_dri.so
fly_: make install clean :p
kdekorte: fly_, no that is wrong
soreau: dileX: I don't have dri/st :P
kdekorte: make clean or make install not together
kdekorte: and do make and make install separately
dileX: soreau: you are using dri/st though
soreau: dileX: Is that what that means?
dileX: yupp
soreau: dileX: Well I'm using rv350
fly_: make && make install clean (thats how i compile most things)
dileX: interesting will be xorg/st (modesetting xorg driver)
kdekorte: fly_, probably wrong
kdekorte: you only need clean when you want to erase all the compiled code
soreau: dileX: The only thing that runs with it is gears and a handful of demos I've tested from mesa most all have corruption except gears in default size window
fly_: nopw still giving me - mv: cannot stat `libGL.so.1.2': No such file or directory
soreau: dileX: What card does it work on for OA?
kdekorte: fly_ probably due to the clean
fly_: i didnt do the clean thank you
kdekorte: run make, and then make install
dileX: soreau: I forgot I am on r300g for two weeks
fly_: i just did
fly_: lol
kdekorte: you screwed something up then
fly_: i used your command
kdekorte: run make clean, run the configure, run make and then finally make install
soreau: dileX: Are you using ~nh/mesa ? and which card?
fly_: that is EXACTLY what i did
kdekorte: I just did that here myself and it worked fine using the git tree...
fly_: im using a patched git tree on a 64bit
dileX: soreau: OA runs here on rv515 (on mesa master) but very slow (as non-gambler I say not really playable) - no colour corruption in the demo scenes
kdekorte: I have 64bit and 32 bit git trees of mesa on my machine
kdekorte: so I know it can be done
soreau: fly_: Can you should the tail end of configure where it shows you the configure options it will use?
fly_: http://pastebin.com/d50fe9b85
kdekorte: looks ok
fly_: then i make
kdekorte: find the error when it is compiling libGL.. it is near the top...
kdekorte: probably something silly
kdekorte: btw, when you installed libdrm, did you install libdrm-devel from koji?
fly_: yup
fly_: wait no just libdrm
fly_: there is no devel it says
kdekorte: you need that, if you are going to compile mesa
kdekorte: fly_ install this http://kojipkgs.fedoraproject.org/packages/libdrm/2.4.15/1.fc12/i686/libdrm-devel-2.4.15-1.fc12.i686.rpm
kdekorte: Assuming you have already installed http://kojipkgs.fedoraproject.org/packages/libdrm/2.4.15/1.fc12/i686/libdrm-2.4.15-1.fc12.i686.rpm
fly_: where the hell was it hiding
kdekorte: http://koji.fedoraproject.org/koji/buildinfo?buildID=135882 in plain sight
fly_: well i searched and it didnt pop up
fly_: brb speaking of koji cat needs food... well she doesnt need food because shes really fat but shes meowing
kdekorte: this will also fix the problem of mesa looking for libdrm 2.4.15
rillian: I have a radeon hd 4850 card (manufactured by asus) which I purchased in a moment of foolishness
rillian: it is, at this point, about the loudest thing in my computer
rillian: I gather that the proprietary driver can adjust the fan speed, at least one windows
rillian: is there any way to get temperature readouts and control the fan with free tools?
mokoloko: rillian: I feel you. I solved it by buying a passive cooler for it
rillian: mokoloko: I feared that might be next. How do you know it's working if you can't read the temperature sensor?
mokoloko: Though I think there was something wrong with my stock cooler since it sounded like a turbine and still temperatures were over 90c
fly_: im in the same boat but the other way around it bugs me that my fan isnt making noise
fly_: in windows in a game my x1950xtx sounds like a 747 taking off
mokoloko: yup i had that too before
rillian: mokoloko: it's certainly generating an impressive amount of heat considering it's only running in 2d mode
rillian: my previous machine had an nvidia and the fan was quiet unless I ran an opengl application, which was more or less acceptable.
mokoloko: I bet mine had lack of paste between the cooler and the core
rillian: it seems like the radeon fan starts out at 'noticeable' and climbs to 'loud' over the first few minutes, then stays there
octe: fly_: you running counter-strike source fine on a x1950xtx?
fly_: mmm depends on what you mean by fine
octe: does it crash?
octe: err, freeze
fly_: in DX7 yeah it ran about 60fps
octe: freezes sometimes for me, even with dx7
octe: i have an x1900xt
octe: just got your Params exhausted error in civ 4 though :)
octe: never gotten that before, though civ has crashed before.
fly_: i was freezing a lot then i upgraded to the 2.6.32-rc4 kernel.. seems to run a lot better now
octe: ah
fly_: using kms
octe: i haven't tried it since i used the rc2
octe: maybe it gets better
fly_: kdekorte, ok so i build a brand new git. patched it..
rillian: anyway, apparently aticonfig can read the temp through the proprietary driver, aside from a discussion about a year ago about airlied and mjg knowing how to approach the problem there's not currently any support
rillian: is this something I could help with? are there docs for the i2c interface?
kdekorte: fly_ so does it work?
fly_: im changing the configs to build for x86
fly_: then i will autogen.sh
kdekorte: why do you need to run autogen after you run configure
kdekorte: run autogen.sh
kdekorte: and then run the configure command I gave you, except change r600 to r300
kdekorte: then run make
fly_: because i bet its going to give me the error
fly_: ill do it anyway
fly_: checking for LIBDRM... configure: error: Package requirements (libdrm >= 2.4.15) were not met: \\ No package 'libdrm' found
kdekorte: if you did have libdrm-devel installed before it probably tried to link the 32bit code with the 64bit libdrm
kdekorte: did you install libdrm-devel from above
fly_: yup
kdekorte: pkg-config --cflags libdrm gives you?
kdekorte: please do "rpm -qa libdrm\*" and post the output
fly_: Package libdrm was not found in the pkg-config search path. \\ Perhaps you should add the directory containing `libdrm.pc' \\ to the PKG_CONFIG_PATH environment variable \\ No package 'libdrm' found
fly_: libdrm-devel-2.4.15-1.fc12.i686 \\ libdrm-2.4.14-1.fc12.x86_64 \\ libdrm-2.4.15-1.fc12.i686
kdekorte: something is screwed up if you have libdrm-devel installed but pkg-config can't find it
fly_: heh,
kdekorte: might want to update the 64 bit one to 2.4.15 and install the libdrm-devel for 64 bit as well
fly_: they will conflict
kdekorte: no they won;t
fly_: Transaction Check Error: \\ package libdrm-devel-2.4.15-1.fc12.i686 (which is newer than libdrm-devel-2.4.14-1.fc12.x86_64) is already installed
fly_: i could rpm it i guess
kdekorte: yeah use the command line to do that
kdekorte: I have them all installed without conflicts, so I know it works
kdekorte: I just upgraded them this morning
spstarr: hmm
fly_: conflicts
kdekorte: error please
fly_: and if i uninstall it it takes out 56 dependency packages with it
kdekorte: rpm -Uvh libdrm-2.4.15-1.fc12.x86_64.rpm libdrm-devel-2.4.15-1.fc12.x86_64.rpm
fly_: ok that got it
fly_: hates rpm
fly_: ok so they should be linked eh?
fly_: mv: cannot stat `libGL.so.1.2': No such file or directory
fly_: i ran the configure just as you posted it
fly_: then make
fly_: make gives me that
kdekorte: what is the error message above it
kdekorte: there should be a message as to why libGL can't be compiled
fly_: /usr/bin/ld: skipping incompatible /usr/lib64/libXdamage.so when searching for -lXdamage \\ /usr/bin/ld: cannot find -lXdamage \\ collect2: ld returned 1 exit status \\ mklib: Installing libGL.so.1.2 libGL.so.1 libGL.so in ../../../lib
jcristau: so you need a 32bit libXdamage
fly_: im on it
kdekorte: yum install libXDamage-devel.i686
kdekorte: might be a little D
kdekorte: yeah libXdamage-devel.i686
fly_: hmm? yeah it was little d
kdekorte: might need to run configure again
kdekorte: you should have got both the -devel and the non -devel package using yum
fly_: i also had to get libXfixes
spstarr: actually
spstarr: SL might be 'GOLD" for r6xx except framerates are poor?
spstarr: or is framerates not included in the http://xorg.freedesktop.org/wiki/RadeonProgram
kdekorte: spstarr, yeah SL runs pretty close to how it ran on fglrx
spstarr: kdekorte: slow frames?
spstarr: i cant get more than 5-10 FPS
kdekorte: spstarr, how do you show frame rates?
kdekorte: mine feels faster than that
spstarr: it shows in debug on console
spstarr: i see lots of chugging
spstarr: it is still better than the intel GMA however
kdekorte: I'll look in a minute
fly_: omg i got it done
fly_: i also needed libXmu
fly_: soooo aparantly im done?
fly_: r300 successfully patched.. i hope
fly_: ill brb with results
fly_: wellll... glxgears is the same, I can run dx 8 now havnt tried higher but my cpu usage in glx gears is 40% and 100% in CS:S
fly_: so the patch succeeded but there appears to be a cpu usage problem now
kdekorte: are you using the 32bit glxgears?
spstarr: airlied: i had some problem with KMS
spstarr: er non-KMS
kdekorte: go into the mesa directory/progs/xdemos
kdekorte: run make
kdekorte: and then ./glxgears
spstarr: VT switch shows blank console but.. the goal is to use KMS so i wont worry about it
kdekorte: file on ./glxgears should say it is 32bit
kdekorte: spstarr, I have to get the new secondlife... so it will be a few mins before I can test
kdekorte: I wish snowglobe was in Fedora...
fly_: 74.2 0.2 0:25.45 glxgears
spstarr: ok
spstarr: kdekorte: build it from source
spstarr: although trunk is broken right now
spstarr: the LL people are horrible at coordinating an open source project
fly_: and my game now has unplayable fps even in DX7
spstarr: 2009-10-15T19:26:09Z INFO: display_stats: FPS: 11.30
spstarr: 2009-10-15T19:26:39Z INFO: display_stats: FPS: 6.50
spstarr: kdekorte: it varies of course more primitives (avatars) to render slows things down
spstarr: but even still, shouldn't we get 20 FPS even? i know the driver is not optimized yet etc
fly_: its as if the video card isnt doing any of the work its passing it all off to the cpu
fly_: OpenGL renderer string: Mesa DRI R300 (R580 7240) 20090101 TCL DRI2
kdekorte: fly_ is that the 64bit glxinfo?
kdekorte: you need to use the 32bit glxinfo
fly_: the one in mesa
fly_: the git
fly_: because i ran that one they both give me horrendus cpu usage
kdekorte: spstarr, I'm getting between 8 and 18 FPS in secondlife
kdekorte: switching to 800x600 with low settings I get about 20fps
kdekorte: fly_ you might need to reboot to load the new libdrm into X... but unsure of that
kdekorte: also you might need to install a 64bit version of the mesa code
kdekorte: so that everything is on the same level
spstarr: kdekorte: did
spstarr: im using newest
kdekorte: did 64bit performance change?
spstarr: im using a 64bit laptop :)
fly_: i did a reboot
kdekorte: fly_ can you check the date on /usr/lib/dri/r300_dri.so
fly_: 1sec
fly_: i went ahead and did the process again it appears to have worked now
fly_: odd
fly_: though i still cant seem to do dx 8
fly_: big fat dump that i suppose i have to give to wine
airlied: damn osiris left
fly_: yea
fly_: he left a while ago :/ i think i farted or something
kuaera: Is occasional crashing still to be expected on xorg-edgers versions of the driver on a R580 [ ATi Radeon X1950XTX ]?
fly_: but at least he game me this nifty patch that gets rid of an error and aparantly creates a whole new one
fly_: gave*
fly_: kuaera ive got a x1950xtx
fly_: i was experiencing random crashes with kms in kernel 2.6.31.2
kuaera: Not sure which kernel version is on that PC...but it's rather recent
kuaera: Should I try to gather some crash information to forward to some developers?
kuaera: Or should I just wait-and-see, since the driver is known to be incomplete?
fly_: well i can say i moved up to 2.6.32-rc4 and i have no more random crashes
fly_: who knows what other bugs ill find though
fly_: i seem to have a nack for finding bugs lol
fly_: So, from the crash dump i got out of wine it appears the new problem im having is also driver related... i think the patch osiris gave me may have caused a big issue in the driver
MrSunshine: http://paste.cplusplus.se/paste.php?id=24 humm ? :)
osiris_: airlied: but I'm back
airlied: osiris_: I suspect some sort of flushing going wrong usually when I see those errors
airlied: its most likely dma bo related of course
osiris_: airlied: no it isn't
osiris_: look at the backtrace
osiris_: the rendering action is initiated by the app
osiris_: fly_: can you pastebin the crash dump?
airlied: osiris_: need to figure out why that tex is getting emitted there and not add as a psersistent bo earlier
rnoland: wishes brianp would go ahead and merge 7.6 > master already....
gsedej: omg. Nexuiz 2.5.2 is really slow on R350 :D
stikonas: gsedej: Nexuiz needs insane amount of GPU :)
stikonas: even on R700 cards it has decent FPS only with Normal quality of effects, setting them to High or Ultra makes Nexuiz lag :)
mokoloko: :O with free drivers?
gsedej: stikonas: 2.5.0 is working quite well
stikonas: of course with free drivers
gsedej: but 2.5.2 is really slow
gsedej: anyway... Nexuiz is really masterpiece on nevidia:P
gsedej: I meant Nexuiz 2.4.2 is quite fast:)
mokoloko: stikonas: what card?
stikonas: RV730
stikonas: OpenGL renderer string: Mesa DRI R600 (RV730 9480) 20090101 TCL DRI2
stikonas: mokoloko: ^^
mokoloko: well maybe some day my 4850 will work :)
stikonas: why isn't it working?
mokoloko: latest fedora snapshot still freezes :'(
hnsr: yay they found balloon boy
amarsh04: in the garage?
hnsr: in the attic yeah
hnsr: ah well, it was a good and entertaining mystery
amarsh04: I was fearing death by misadventure
hnsr: yeah
hnsr: especially right before they found him they were saying some sheriff deputy saw a box falling from it
amarsh04: I fully understand why some .au government people were concerned about a teenage girl attempting a solo round the world sailing trip
biotube: did she build the boat herself?
hnsr: wasn't that the dutch goverment? 13 year old girl
MostAwesomeDude: Well, this patch makes readPixSanity work: http://paste.pocoo.org/show/145202/
MostAwesomeDude: :3
hnsr: though some australian girl already did
amarsh04: if someone does all their homework and is a prepared as possible and know the risks, then fine
MostAwesomeDude: So it is rendering errors, after all.
biotube: wonders who'll be the first kid ot solo to the moon
WhiteRabbit56: Where can I find the documentation from ati
biotube: http://www.x.org/docs/AMD/
WhiteRabbit56: biotube:Thanks
spstarr: do we expose the GPU temp in /sysfs yet?
spstarr: is quite curious the temperature cause it just went up and its quite hot
spstarr: in laptop
airlied: no not yet
spstarr: ok
rillian: airlied: is there anything new on that? I found a discussion about temp/fan support from a year ago on lm_sensors
rillian: but nothing since
airlied: rillian: what needs to be done hasn't changed
airlied: just nobodys done it yet
airlied: the atom tables explain what temp chips are connected and to what i2c bus
WhiteRabbit56: airlied: Is there power management implemented on r600/r700?
airlied: just need to hook lm sensors driver to those buses
airlied: WhiteRabbit56: not under kms yet
rillian: ah, ok, it's in the atom tables
airlied: I looked at it briefly on my r580 but it needed an lm64
airlied: which lm sensors doesn't support yet
rillian: are the requisite bits in the published docs?
airlied: that atom tables document it
airlied: i2c is already written
airlied: the temp sensor chips aren't AMDs
MostAwesomeDude: airlied: http://paste.pocoo.org/show/145202/ makes readPixSanity work. Guess I need to go back and figure out why I'm misrendering things.
airlied: MostAwesomeDude: yes I noticed arbocclude misrenders even without occulison not working
airlied: MostAwesomeDude: also that texture routing thing needs ifxing
spstarr: is happy i can use the r6xx now
rillian: airlied: ok. I was wondering if it was something I could help with. at least I have a pointer to start with now.
spstarr: the heat however, is concerning
airlied: rillian: decode the table on your GPU and see what chip it has
spstarr: but i do not know if this is due to optimizations for power management while using 3D are supported yet or not, or if this is normal temp
rillian: gosh but 'atom' is a hard tool name to search for
airlied: rillian: its not a tool
airlied: though we do have AtomDis tool to help with decoding it
airlied: its mainly the ATOM_POWERPLAY_INFO tables you'd be worried abouyt
rillian: how does one dump the tables?
airlied: well you can either add code to read them to the driver, or clone git://people.freedesktop.org/~mhopf/AtomDis
airlied: throw in an atombios.h header file from somewhere and get your bios
rillian: thanks. I wasn't clear that would actually ask the driver
MostAwesomeDude: airlied: Which test were you using for multiple textures?
rillian: ah, it doesn't dump the bios. still trying to figure out how to do that.
airlied: MostAwesomeDude: multitexarray or texrect
airlied: rillian: cd /sys/bus/pci/devices/
airlied: rillian: echo 1 > rom; cat rom > tmp/myrom ; echo 0 > rpm
rillian: airlied: thanks!
MostAwesomeDude: rillian: $ wget http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/plain/src/AtomBios/includes/atombios.h
MostAwesomeDude: $ make CFLAGS="-DATOM_BIG_ENDIAN=0"
MostAwesomeDude: (Fix appropriately if on BE.)
Dr_Jakob: MostAwesomeDude, airlied, agd5f: Can any of you guys test that patch I just sent to xorg-devel on radeon?
dmb: agd5f, if the video works on DVI before radeon driver is loaded, should i even bother with that first patch?
dmb: tried the coherent mode, no good
rillian: MostAwesomeDude: thanks. I was using the one from the radeonhd driver, but except for the endian changes there's not much difference
rillian: what am I looking for the in the table. there's nothing especially string-like
rillian: http://pastebin.ca/1623010
rillian: also found an mjg blog post about there being a powersaving mode the radeon drivers aren't using which would probably also help with the fan noise issue
MostAwesomeDude: Dr_Jakob: I don't have a patchable xserver built at the moment, sorry. Looks sound from the EXA side, though.
airlied: Dr_Jakob: maybe later, (where's MrCooper when you need him ;-)
airlied: rillian: look at atombios.h
airlied: ATOM_PP_OVERDRIVE_THERMALCONTROLLER_*
airlied: there is 8-bits for the controller and i2c line
airlied: has another 8-bits
Dr_Jakob: airlied: in Japan apperently :)
rillian: so I have to ask the sensor what it is over i2c?
Dr_Jakob: ok thanks MAD/Dave.
WhiteRabbit56: using atombios can we currently change the clock speed?
soreau: Dr_Jakob: Curious, what was wrong with component alpha fall back?
WhiteRabbit56: because i did notice some things in atombios.h that implied it
airlied: Dr_Jakob: I'm going there in a day or two ;-)
airlied: rillian: no the table tells you the sensor you have to talk to it over i2c at the address the table should say
airlied: i2c doesn't really have a useful identification mechanism
MostAwesomeDude: For rectangles, is there any place I have to actually worry about the normalized coords, besides the frag shader?
airlied: not that I know off
airlied: texrect already works
airlied: apart from the multi-texture bit of it
Dr_Jakob: soreau: for operation over you can emulate component alpha with two draw calls, problem is that it doesn't turn component alpha off for those two drawcalls causing it to not be handled anyways.
Dr_Jakob: Since the driver can't deal with it.
Dr_Jakob: soreau: http://anholt.livejournal.com/32058.html <-- the math behind it.
MostAwesomeDude: airlied: Wait, it does? For you? It doesn't work on my r300.
airlied: I get one texture rendered
MostAwesomeDude: I get neither.
airlied: fail.
MostAwesomeDude: I do get a couple of fun shader error messages.
WhiteRabbit56: 00000000000000000000
soreau: Dr_Jakob: Neat. The math is a bit over my head but what is a use-case scenario?
Dr_Jakob: soreau: subpixel smoothing for text.
MostAwesomeDude: The math's actually genius, really.
MostAwesomeDude: airlied: Yeah, just double-checked. The texture/sampler setup is identical on r300/r500, so it must be shader compiler bugs.
airlied: gets laptop out my bag
rillian: airlied: ok, that makes more sense. I was confused because my ucOverdriveThermalController is only defined in the ATOM_POWERPLAY_INFO struct and my PowerPlayInfo table doesn't appear to have one of those
MostAwesomeDude: airlied: Just pushed the logical ID for samplers, multitexarray works fine here now.
airlied: MostAwesomeDude: nice
rillian: it says it's format rev 4 and has an array of ATOM_POWERINDEX_INFO_V4 and ATOM_POWERUNKNOWN_INFO_V4 tables
MostAwesomeDude: Gonna hafta dig deeep into r300compiler for the texrect. TBH I'll probably spend time getting gallium-blitter branch finished for r600g.
airlied: rillian: ah maybe agd5f can advice further on the v4 stuff
MostAwesomeDude: $ mv ~/corbin osu/lug-meeting
MostAwesomeDude: Later.
rillian: agd5f: you around? I'm trying to work out how to talk to the thermal controller on an hd4850
AndrewR: some random becnhmark: OpenArena 0.7.0 smp + RS780 + more-or-less current Mesa gives me 25 fps in 'anholt' demo, 1024x768x32, all max (it would be nice if someone restores download link to it). Something like 1/3 from my refresh rate (75 hz)
AndrewR: *benchmark
airlied: MostAwesomeDude: you broke them for me :-)
rillian: airlied: anyway, thanks for your help.
MostAwesomeDude: airlied: Wait, what. It broke?
airlied: MostAwesomeDude: wierdly yes, I only see the second texture now
airlied: or rather
airlied: in texrect I see texture 0 where I should see texture 1
MostAwesomeDude: airlied: Probably need to overhaul the texture enables by actually checking which ones are enabled and using a bitfield.
airlied: I wonder if we need to do the hw_tmu abstraction we have in classic as well
MostAwesomeDude: I think just counting should be sufficient.
MostAwesomeDude: And not doing the cheap thing I've got in r300_flush_textures.
rillian: http://www.botchco.com/agd5f/?p=45 also points at ATOM_GPIO_I2C_INFO
rillian: and I do have four device entries there
airlied: rillian: the powerplay table gives an index into the i2c table
rillian: airlied: or would, if that part of ATOM_POWERPLAY_INFO_V4 were documented?
agd5f: rillian: documentation for v4 is on my todo list
_Groo_: ji/2 all
_Groo_: agd5f: ping
_Groo_: airlied: ping
agd5f: _Groo_: pong
_Groo_: agd5f: hey ...
_Groo_: agd5f: the person i wanted to talk to
_Groo_: agd5f: did you had the time to add the patch to disable my fake vga (the one thats screwing kms/xrandr)?
agd5f: _Groo_: you never sent me your vbios or lspci
_Groo_: agd5f: like we talked 2 days ago (if my memory serves me)
_Groo_: agd5f: can i upload it to paste.bin?
agd5f: and I thought it was that your laptop had no tv-out port
_Groo_: agd5f: and i pasted lspci, i never pasted bios, thats for sure
agd5f: vbios is a binary file
_Groo_: agd5f: yeah, tahats it
_Groo_: paste the commands and ill send to you right now
agd5f: as root
agd5f: cd /sys/bus/pci/devices/
_Groo_: agd5f: by battery is screwed but i have enoguh juice to send it to you
agd5f: replace
_Groo_: done
_Groo_: lspci -nv ?
agd5f: yes
_Groo_: agd5f: email?
agd5f: alexdeucher AT gmail DOT com
_Groo_: sent :)
_Groo_: the patch goes in the kernel kms?
_Groo_: or the libdrm?
agd5f: kernel
_Groo_: agd5f: ok, can you send the diff to me (using latest linus tree).
agd5f: _Groo_: yeah
_Groo_: agd5f: thanks a lot..
_Groo_: agd5f: the out of sync problem i still have when returning from suspend, could it be related? that is, kms trying to set a resolution for a vga screen into my vlds?
_Groo_: lvds
agd5f: _Groo_: that's not the problem in either case
_Groo_: agd5f: well it still there, i was hopeful it would go away with newest kms code.. but... nope
agd5f: _Groo_: what model laptop is this?
_Groo_: http://pastebin.ca/1623240 this is what happens to x more times that not, with the scrwed kms bug
_Groo_: agd5f: acer 5102 with a evil rs485
_Groo_: agd5f: i believe you have the freedesk bug, if not i can tell it to you
_Groo_: when X starts the kdm background becomes a 800x600 square, and the new background stays behind it with 1280x800 size, then while loading X crashes with the pasted bugtrace
rillian: agd5f: ok, so it's not otherwise public, but one of those fields in an index into the GPIO_I2C_Info table?
_Groo_: if i restart it, after a few tries i can login, but if i do any xrandr stuff it crashes again..
_Groo_: robably because kms says one rez and X wants another
_Groo_: or the vga display tries to take over, dont know
agd5f: rillian: r6xx and r7xx are different
agd5f: not sure if they have i2c entries or not
agd5f: completely new tables
rillian: ok. so what's currently there might be nonesense?
_Groo_: agd5f: my bat will go out soon, since now you have my email, we can talk further by email (during work hours i dont have irc/Im acess)
agd5f: _Groo_: ok
rillian: I didn't trace through the code to see how it was interpreting the v4 PowerPlay_Info
agd5f: rillian: what code?
_Groo_: agd5f: thanks again for your work, you guys rock, even with these few annoyances, radeon works like a charm (when t Do work)
rillian: AtomDis
rillian: (lspci reports the card as ATI Technologies Inc RV770 [Radeon HD 4850])
agd5f: it has some reversed engineered support for newer tables
_Groo_: seeya ppl
agd5f: but I'm not sure how correct it is
rillian: ok, it doesn't report anything about thermal management
rillian: but there are quite a few undescribed fields
rillian: http://pastebin.ca/1623010 if you're curious
dmb: agd5f, i shouldn't have to apply the first patch you gave me if it works fine before radeon is loaded right?
WhiteRabbit56: rillian: did you try dumping with F as the command?
agd5f: dmb: yeah
rillian: WhiteRabbit56: I did, but I didn't read the whole thing, any search suggestions?
WhiteRabbit56: rillian:i2c showed up in mine quite a few times... I'm not really familiar with i2c but you could see if you can find anything usefull with that
dmb: agd5f, i'll update the bug, but yeah, coherent mode = 0 did not help
rillian: WhiteRabbit56: nothing obvious, but I don't really know what I'm doing
dmb: agd5f, would it help if i tested the radeonhd driver?
dmb: and see if that blanks out
agd5f: dmb: sure
rillian: ucVoltageControlI2cLine, SetupHWAssistedI2CStatus, ReadEDIDFromHWAssistedI2C, ATOM_I2C_ID_CONFIG_ACCESS?
rillian: 'UCHAR bfI2C_LineMux:4' shows up a couple of times