DanaG: yay, got my good laptop back!
spstarr: DanaG: and mine needs repair.. the LCD inverter is dying
spstarr: I have to keep KMS off, keep switching from text mode to X
spstarr: then it resets the inverter briefly.. sometimes for 10-15 mins sometimes for < 1 minute
DanaG: oh yeah, and HP didn't put the lid back together exactly correctly... it was bulging, until I took my ID card and poked under the bezel a bit.
DanaG: So, anything new besides that DynamicPM and ForceLowPower?
spstarr: we'll see, if the cost is worth fixing vs buying a new laptop
DanaG: They also didn't replace the fingerprint reader. :(
DanaG: oh yeah, so, I wouldn't quite say "don't buy them", but now I do agree with the people calling their support mediocre or crappy.
DanaG: Still an awesome product, though.
dileX_: airlied: enjoy your holidays :-)
DanaG: Now even the consumer ones are using R600 and R700.
spstarr: seems if i lower brightness to lowest, the LCD backlight doesn't turn off
DanaG: reboots to try radeon, and random faster boot tweaks.
DanaG: hmm, which mesa branch for r600_dri?
Iowahc: good morning
Iowahc: quick question, is there HDMI support in radeon driver?
hifi: Iowahc: if you mean sound, no
Iowahc: no, actually video
hifi: should work as far as I know
Iowahc: should it be listed with xrandr?
hifi: I think so
Iowahc: there it is
Iowahc: never saw it
ajavid: hi, I'm on debian and using radeon in debian unstable on x1900xt
ajavid: I have a -dpi 100 in my xargs
ajavid: however xpyinfo reports 89xsoemthing and my fonts look horrible
ajavid: >xdpyinfo | grep -B1 dot dimensions: 1680x1050 pixels (480x270 millimeters) resolution: 89x99 dots per inch
ajavid: grep DPI /var/log/Xorg.0.log
ajavid: (**) RADEON(0): DPI set to (90, 140)
ajavid: what the hell is going on?
ajavid: I do startx -- -dpi 100 and still nothing
ajavid: I try -xpi 96 and no change
ajavid: er.. dpi*
nanonyme: Are you sure you should be allowed to give it as a command line switch?
nanonyme: X.org calculates DPI based on resolution, I think it might be overridable via xorg.conf.
MrCooper: command line should override xorg.conf
ajavid: root 17629 12.3 1.7 220212 146768 tty7 Ss+ 01:44 0:04 /usr/bin/X -br -nolisten tcp -dpi 100 :0 vt7 -auth /var/run/xauth/A:0-X4GwGF
ajavid: (++) RADEON(0): DPI set to (100, 100)
ajavid: now its ok
ajavid: but my fonts still look like they are not right
ajavid: I ahve anti aliasing enabled but I think that maybe out of the domian of this channel
ajavid: I was trying to figure out if it was a bad dpi problem
ajavid: so is my dpi calculated properly?
ajavid: if I don't specify a dpi switch i get the 02:41 < ajavid> (**) RADEON(0): DPI set to (90, 140)
ajavid: if I do, I get 100x100
nanonyme: Depends on whether you are using radeon driver from the beginning of the year with EXA on. ;)
nanonyme: (And with r6xx-r7xx)
nanonyme: It used to break antialiasing *bad*.
ajavid: debian sid
nanonyme: Right, irrelated then.
ajavid: i would use windows xp
ajavid: but my wifi driver doesn't work in xp 64
ajavid: and I don't want to use x86 cuz i have 16gb ram
ajavid: so life kinda sucks for me right now
ajavid: im having to look at these ugly fucking fonts.
ajavid: curses more
ajavid: does it look like i have anti alised fonts or not?
MostAwesomeDude: Your fonts look fine.
MostAwesomeDude: Perhaps your subpixel hinting is incorrectly set?
MostAwesomeDude: But seriously, that screenshot looks just fine over here.
dileX: ajavid: I had same display problems - selecting the right ddx helped (1/12 pixel precision and latest drm-next-radeon)
nanonyme: Yeah, looks normal to me.
dileX: my experiences were - you could read what was displayed - but font was fine-granulated misdisplayed. as it was not drawn exactly. updating ddx helped.
ajavid: sorry i had to go eat an apple
ajavid: ok so they look fine
ajavid: ok java applications however they are not aliased as system bitmap fonts are specified
ajavid: if i look in the /etc/sun-java6-bin/fonts.properties it only contains the old non ttf fonts like how you specify in al ong line
ajavid: like --x--blahblah--more.dashes909 blah
ajavid: i know this has nothing to do with radeon but if anyone know how to change fonts in jre apps that'd be nice, been trying to figure this out for days now
ajavid: i just wanna make all apps using this jre to use ttf fonts via the file
rehabdoll: radeon doesnt support hdmi-audio, right?
rehabdoll: only radeonhd
adamk: airlied, Any further thoughts about that AGP x1300 issue with KMS that we were discussing yesterday?
airlied: adamk: hmm didn't get much time I asked agd5f to see if he can help out
adamk: Cool. I'll be around should anything come up. I have another problem now, though :-) With a PCIe on an x1900, hooked up to a Westinghouse LCD, the screen constantly flickers on and off: http://www.npark.com/kms-xorg-x1900.mpeg
adamk: As I recall, I had a problem initially with this LCD under Xorg where I had to specify a specify hex number to xrandr to get just the right refresh rate, or something like that. Unfortunately, with KMS there's no way to specify a mode when booting up, is there?
adamk: And the final issue that I seem to have is when I switch from Xorg to console, and back to Xorg (with KMS enabled), one of my monitors loses its signal. Sometimes forcing them off via dpms and then waking them up helps. Sometimes using xrandr to reposition the monitors works. Sometimes neither works.
airlied: adamk: if you change the mode from X does it help?
airlied: for the refresh issue
adamk: Not consistently. Sometimes, in fact, the monitor just stays blank. I tried to work out a pattern to this yesterday evening, but couldn't come up with one. The very next time I rebooted, the monitor went blank as soon as KMS kicked in and stayed that way. I was unable to get it to turn on by changing the resolution via xrandr.
Kano: hi agd5f , airlied
suokko: Intresting problem: I have installed modesetting-gem drm modules and when I try to use modesetting I have problem that agp is first acquired by fbcon and later when xstarted ddx driver can't acquire agp
zhasha: suokko, which ddx are you using, whose kernel and whose modesetting-gem?
suokko: Ubuntu jaunty kernel and ddx is currently master
zhasha: does jaunty have KMS?
Kano: 2.6.30 has it when you add one option
zhasha: it sounds like you need a compatible ddx
suokko: nope. But I have managed to install modules to support KMS
suokko: I guess I have to go back try to make Jesse's ddx/drm work then :( They even had problems to boot at all
suokko: r200 agp card seems problematic
nanonyme: suokko: AGP is problematic, period. ;)
suokko: seems like that
suokko: But I don't understand why fbcon or ddx should acquire agp for exclusive use.
DanaG: hmm, for (even just basic) 3D on R600... which branch of mesa should I build?
dileX: DanaG: see here
DanaG: hmm, I'm curious when it'll move back into the more "normal" branch of the drm modules.
DanaG: "It's probably going to be a week or so before the devs get back to troubleshooting bugs. Right now the priority is getting the new 3D code running on the radeon-rewrite base, which involves different paths for the bottom part of the 3D driver and different paths in the DRM IIRC.
DanaG: The code where the problems are likely to be is the code that's going to be replaced in a week
DanaG: oopsie, didn't realize paste would do that in finch.
DanaG: I'll be most interested in jumping on it approximately whenever that happens.
mroconnor_: doing a fresh install using an older kernel and now git repos right now
mroconnor_: and I have to say you guys have done a lot of great work in a short amount of time
mroconnor_: just comparing the older and the newer ;)
pkt: has the suspend / resume problem with AGP cards been fixed?
DanaG: Oh, is it possible to have KMS and R600 DRI together?
nanonyme: Since when?
nanonyme: DanaG: Where
nanonyme: Where's you read that?
DanaG: where what?
DanaG: sorry, I had a couple of topics mentioned.
nanonyme: 16:56 < DanaG> Oh, is it possible to have KMS and R600 DRI together?
DanaG: That's exactly my question. =þ
nanonyme: Well, afaik no.
nanonyme: I was just wondering if you have more current information.
nanonyme: Like some developer blog posting or whatever. :)
mroconnor_: that would be nice
DanaG: Interestingly enough, radeon still uses a more power than fglrx.
DanaG: s/ a //
nanonyme: Interestingly enough my computer had a hardlock with powersaving options on while I was in an exam.
nanonyme: DanaG: Where do you check your power consumption, btw?
chithead: if you run on battery, you can check /sys/class/power for how much current it delivers atm
nanonyme: DanaG: Doesn't sound reliable.
nanonyme: "PowerTOP analyzes the programs, device drivers, and kernel options running on a computer based on the Linux and Solaris operating systems, and estimates the power consumption resulting from their use."
otaylor: Anyone want to decipher this for me:
otaylor: File r300_render.c function r300Fallback line 444
otaylor: Software fallback:ctx->Stencil.Enabled && (ctx->Stencil.Ref != ctx->Stencil.Ref[back] || ctx->Stencil.ValueMask != ctx->Stencil.ValueMask[back] || ctx->Stencil.WriteMask != ctx->Stencil.WriteMask[back])
nanonyme: So it's not the real power consumption, just an estimate on what the running software maybe might consume.
mimikry: but it's more than a simple consumption, because if it was, then it would also work on computers which are not a laptop
nanonyme: Oh, right: no ACPI power usage estimate available
nanonyme: So it *can* use ACPI to get the real numbers.
nanonyme: I just get wakeup percentages.
jcristau: it gets power usage estimate when you're on battery
nanonyme: I wonder if you could optimize efficiency by some of those things it suggests, like say setting USB standby on and disabling CD drive polling on HAL.
nanonyme: Would make sense to me that if there's less stuff all in all, the stuff I actually want to do gets through more easily.
glisse: otaylor: it's for stencil
glisse: there is a limitation i need to look at gl extension
glisse: to refresh my memory
glisse: otaylor: up to r5xx we can't have different mask/value for back & front face
otaylor: glisse: Yeah, I've tracked it down a bit more
otaylor: Somehow the backface valuemask gets different from the frontface, but I think it's a bug somewhere in mesa - there's nothing apparent in the application using the two-faced stencils
glisse: otaylor: yeah there is very few apps with different valuemask/ref
glisse: so we should be accelerated on most case
sannes: glisse: noticed the oops mentioned earlier?
glisse: i also think we could handle this case by doing 2 pass rendering but this would need some infrastructure into mesa, gallium likely makes this easier
glisse: sannes: not really
nanonyme: glisse: Btw, what's the latest thing going on with KMS+memory management development, bugfixing with existing code?
glisse: nanonyme: i am tracking down a lockup on r3xx with scorched3d
otaylor: glisse: Found it - radeon_clear_tris() calls mesa_StencilFuncSeparate(GL_FRONT)
otaylor: Though, hmm, not sure that explains it, unless tsomething goes wrong with the save/restore in that function
kdekorte: Is KMS (F11) supposed to be working with the r635 yet? I still get lockups on boot with it
glisse: otaylor: does adding same line but with GL_BACK helps ?
otaylor: glisse: I was about to try changing FRONT to FRONT_AND_BACK
glisse: oh good too
otaylor: glisse: Though I donn't have a built mesa here, so "just" is relative
glisse: otaylor: i think airlied added this function to have clear with r100/r200
glisse: and he likely didn't test much with stenciled apps :)
otaylor: The use that clutter makes of stenciling is a llittle horrifying... the end of what we are actually wanting to do would be fine with a scissor, but clutter doesn't realize it's axis aligned, so it's doing all this stencil construction which must be killing the memory bandwidth
otaylor: "opportunities for optimization" :-)
glisse: what about clip plane ?
glisse: i am not sure what exactly clutter wants to do thought
otaylor: glisse: it uses clip planes unless there are nested clips
otaylor: glisse: Basically every actor in clutter can specify a clip box in its user space coordinates, and those nest heirarchically
otaylor: (or that's my understanding of it)
otaylor: So, in the normal case where no rotations are involved, the composite clip is still an axis aligned rectangle, but you could have more complicated clip volumes
otaylor: (Not sure the clutter code actually deals with nested 3-D rotated clips)
otaylor: (but even in 2D case, you get a clip "polygon")
glisse: otaylor: i think what clutter wants to do can be done faster in 2 pass rendering first pass would build the stencil buffer, second would do the actual rendering and use stencil to discard pixel
glisse: and no need for 2 sided stencil here
glisse: which would enable hw accel on older hw or even maybe on embedded hw which might lack this stencil ext
glisse: everythings seems to isolated in cogl :)
glisse: nice design
otaylor: glisse: there's no actual use of two sided stencils, right?
spstarr: glisse: s/r doesn't work still :)
spstarr: glisse: but my LCD inverter is dying so im gonna be off laptop for afew days
nanonyme: Whoa. I never realised the people ATi has a license from that binds them to that DRM garbage was from Intel's subsidiary. :o
sannes: glisse: http://pastebin.com/m785f9c80, bo->vm_node == null, introduced by commit 7b3ccf290581d9478db480e71490f02bd5d8b0fb (stikonas who bisected it)
nanonyme: Hrm, that was a very unclear sentence. :)
nanonyme: (As in, didn't know that Intel's subsidiary technically forces ATi not to go open specs :P)
sannes: glisse: and to reproduce, just modprobe modeset=1 and try to display image with fbv
glisse: otaylor: haven't looked at code was looking at doc
glisse: sannes: haven't played with that code path yet
glisse: will take a look once i am done tracking lockup on r3xx
sannes: glisse: thanks, then atleast I know you know, .. you know :P (I'll continue testing though when the tree is updated)
otaylor: glisse: ah, OK. I think there's probably optimization to be done on the cogl use of stencils, but my first run at optimization would be to try to detect the case when Union(outer_clip, inner_clip) == inner_clip, and use clip planes rather than stencils in that case
glisse: otaylor: yeah according to grep there is no use of double sided stencil
xpk: Hello. Please help me with enabling direct rendering for Radeon 9600XT on Debian Sid. What exactly do I add to xorg.conf? I see next line in my Xorg.0.log: (WW) Radeon(0): Direct rendering disabled
glisse: xpk: could be kernel issue, pastebin dmesg | grep drm
agd5f: xpk: you need to install the linux-firmware package
hifi: apt-get install linux-firmware
glisse: oh forgot about this debian things...
nanonyme: glisse: Better hope the infection doesn't spread. ;(
xpk: seems so. thanks a lot.
glisse: more over we exactly know what the firmware does
xpk: I guess so :)
hifi: is it impossible to write open firmare?
hifi: with current knowledge
glisse: we don't know the opcode of the cp
glisse: one would need to figures this out
hifi: moreover why the firmware needs to be loaded and not on the device itself?
glisse: yeah could be cost and also lot easier to fix bugs
glisse: than to ask millions of people to flasg their gpu bios
glisse: thought i don't think their ever was much bug in the firmware
xpk: should there be some special lines in xorg.conf? cause it still seems not working although dmesg shows no error messages now
hifi: how did you define it's not working?
glisse: xpk: you might need to reboot or reload radeon module
cheetopet: glisse: do you wanna hear about issues with kms from your drm tree?
glisse: cheetopet: yeah
xpk: the thing is I've just upgraded to sid, and it uses kde 4, which has the compositing enabled by default
xpk: and I can't get it run properly
zhasha: xpk, dmesg | grep drm
cheetopet: card is a - 01:00.0 VGA compatible controller: ATI Technologies Inc RV505 CE [Radeon X1550 64-bit]
zhasha: sorry, not what I thought it was :/
zhasha: oh hey, there are 2 of you! my bad
zhasha: cheetopet, your issue wasn't what I thought it was, xpk, I've no idea about yours