dmb__: spstarr, how hot?
spstarr: hot enough that you dont want your finger on the bottom very much
spstarr: not good
spstarr: in Windows Vista/7 it was much cooler with SL
spstarr: but i know the driver is not optimized/PM not opt yet
dmb__: spstarr, you can't get any temps?
dmb__: i think normally it will underclock itself
dmb__: but if it gets TOO hot, it will shut down completely
dmb__: spstarr, SL = secondlife?
spstarr: yes
dmb__: yeah, secondlife can warm up CPUs
dmb__: tried running it on an intel integrated GPU, the fans went crazy
spstarr_: ive gone back to intel GMA for now again
spstarr_: the GPU is waaay too hot..
spstarr_: that is worrying i dont want to fry the laptop ;p
spstarr_: even if the GPU would prevent such high temps
spstarr_: VBOs seem brokenin mesa 7.7
Pallokala: hi, was the dga-extension/support going to killed from radeon/xorg-server?
Pallokala: +be
hifi: it was killed
hifi: in dri2
Pallokala: cool, so I can then remove the dga use-flag :)
MostAwesomeDude: Please, yes.
ootput: hi guys. Just popped in to ask if there was antialias support atm for r7xx
ootput: I'm using compiz atm, and have noticed rough edges of window manager
MostAwesomeDude: Not yet.
MostAwesomeDude: I'm not sure if the docs cover AA. If not, we'll get it sooner or later.
ootput: is it being worked on, or is it impossible?
ootput: okay
MostAwesomeDude: It's not being currently worked on, but the hardware totally can do it.
hoo-hah: ootput: horrendous lag :)
hoo-hah: are you using tor?
ootput: yep
Guest46387: name fly_
Guest46387: lol oh yeah this isnt counterstriek
fly_: kdekorte: ping
fly_: or anyone for that matter i forgot what ./configure options to use to build the 32bit mesa
cockroach: hi. it seems with a git snapshot from last night i don't have a mouse pointer on the second screen (dual head DVI, R570, KMS) - should i file a bug report about that or does anyone have an idea?
fly_: this is all i could remember of that configure script "./configure --prefix=/usr --with-dri-drivers=r300,swrast --disable-gallium" whats the rest needed in order to compile for 32bit
fly_: ok i think ive got it
fly_: ./configure --build=i686-pc-linux-gnu --prefix=/usr --with-dri-drivers=r300,swrast --disable-gallium --enable-32bit CC="gcc -m32" LDFLAGS='-m32' --without-demos
fly_: does that look right?
sdsdsd: Hello guys, will radeon driver work if i try using compiz?
sdsdsd: (3D desktop)
biotube: should
sdsdsd: im using Debian Squeeze AMD64, with self compiled radeon and mesa (compiled 3 days ago)
sdsdsd: but when i try start compiz, it gives an error
sdsdsd: XGL not found ou something like that
biotube: did you set things up to load the new versions?
sdsdsd: yes... they are all working
chithead: compiz will try xgl and nvidia tfp first, then aiglx
chithead: so xgl not found is expected
sdsdsd: ill try reproduce the error again,
sdsdsd: Checking for Xgl: not present.
sdsdsd: xset q doesn't reveal the location of the log file. Using fallback /var/log/Xorg.0.log
sdsdsd: No whitelisted driver found
sdsdsd: aborting and using fallback: /usr/bin/metacity
sdsdsd: but there are no things about it on Xorg.0.log
sdsdsd: what do you think about that?
fly_: damnit! i hate this error mv: cannot stat `libGL.so.1.2': No such file or directory
fly_: doh
sdsdsd: "No whitelisted driver found" -> radeon driver is on whitelist
soreau: sdsdsd: Is it a laptop?
sdsdsd: yes
soreau: What is the output of glxinfo|grep renderer ?
sdsdsd: yes
sdsdsd: OpenGL renderer string: Mesa DRI R300 (RS690 791F) 20090101 NO-TCL
sdsdsd: \
soreau: I think I can safely guarantee that is not the output of the command I just posted
soreau: Ok, try SKIP_CHECKS=1 compiz-manager
sdsdsd: compiz-manager>?
sdsdsd: wait a minute
fly_: i think... i may... have gotten it
fly_: shh nobody disturb the compiler :x
fly_: gmake[4]: *** [libnurbs/interface/bezierEval.o] Error 127
fly_: WHAT?
sdsdsd: soreau, i havent found compiz-manager
soreau: sdsdsd: Debian, just use compiz as the command
soreau: SKIP_CHECKS=yes compiz &
sdsdsd: http://pastebin.com/m6806def3
soreau: looks
soreau: Appear it started ok
sdsdsd: about the last line?
soreau: mesa warning
soreau: harmless informational message
sdsdsd: humm
sdsdsd: so it must be working?
sdsdsd: about the "no whitelist driver found" ?
soreau: yup
soreau: Well if it is working, you should see some effects and stuff in ccsm should work
soreau: Most noticeably wobbly windows when you drag your windows.. give wobbly is enabled in ccsm
soreau: given*
sdsdsd: no one .. i have activated some effects, noone working
sdsdsd: on Compiz-fuzion icon, it appears : :Select Window manager > compiz (selected)
sdsdsd: but no efect found
soreau: You know, we should really continue this conversation in #compiz
soreau: see you there
sdsdsd: ok
fly_: grrr... now im stuck at libGLU.so.1 and i HAVE mesa-libGLU & devel i686 & x64
fly_: beats linux with a hose
fly_: still stuck
biotube: fly_: never had a problem, never used gentoo. the two might be related
fly_: im not on gentoo
biotube: which distro?
fly_: 'dora
biotube: ah
fly_: actually you might be able to tell what wrong if you dont mind looking at this output
biotube: sure
fly_: i started copying at some strange output im not sure its the cause of the failure but here it is http://www.pastebin.ca/1625133
biotube: it's trying to link amd64 object files as ia32, which seems to be the reason you keep getting that file-not-found error
biotube: try building just the 32-bit version
fly_: is that a ./configure parameter issue?
fly_: ./configure --build=i686-pc-linux-gnu --prefix=/usr --with-dri-drivers=r300,swrast --disable-gallium --enable-32bit CC="gcc -m32" LDFLAGS='-m32' --without-demos
biotube: I'd add --disable-64-bit
fly_: oh i missed a -
fly_: --enable-32-bit
fly_: redface
fly_: i think once i get this process down pat im gonna write a proper cross-compile HowTo
fly_: at least for fedora
fly_: probably take one more run through ill format one of my other drives and do a scratch build
fly_: omg it worked.. after like 12 dependency packages lol
fly_: brb
cockroach: hi. it seems with a git snapshot from last night i don't have a mouse pointer on my second screen (dual head DVI, R570, KMS) - should i file a bug report about that or does anyone have an idea?
Guest71653: Bwhaahha great success
Guest71653: i am now spying on my neighbors via google earth (allbeit a REALLY old image our backyard is filled with blackberry bushes which we got rid of 3 months ago)
Unggnu: hi all
Unggnu: Does anybody know when Powerplay support will be shipped with radeon for RV770?
Unggnu: 4870 are pretty noisy without Powerplay
biotube: loves his passive cooler
Unggnu: Does it make sense to create a bug report if suspend freezes with KMS?
biotube: Unggnu: power management isn't support with KMS yet
Unggnu: biotube, no option if you want to play modern games
Unggnu: biotube, even without it isn't supported
Unggnu: only FGLRX has support for it
Unggnu: Or do you mean suspend with power management?
biotube: I meant power management period
biotube: but the suspend bug seems to have something to do with aspm
biotube: try booting with pcie_aspm=off
fly_: my fan has never turned on..
fly_: :'(
fly_: suspend bug? i think i may have had that happen earlier
fly_: i couldnt click anything but a few buttons in my tray
fly_: then i recompiled my kernel ;) havnt seen anything of that problem since : Linux Combaticus 2.6.32-rc5 #1 SMP PREEMPT
Unggnu: biotube, powerplay (changing clock and shader) is power management for me too
Unggnu: The suspend problem isn't so important since it works without KMS. But the noise is annoying and FGLRX is no real option
Unggnu: fly_, which Radeon do you have?
fly_: x1950xtx
Unggnu: Is there a road map for fglrx?
Unggnu: *powerplay
Unggnu: /radeon :)
fly_: fglrx is pretty outdated for us legacy folk
Unggnu: XV is still tearing with Fglrx, performance is low with 3D and it crashes regularly.
fly_: sounds like a good time to switch to open drivers
fly_: ;)
Unggnu: yes, I use them all the time but the fan is noisy. I have already patched the firmware so temperature isn't so high but it is still more noisy than with powerplay
fly_: passive cooler
fly_: :D
fly_: hey be happy your fan turns on
fly_: im so worried im frying my gpu
fly_: actually its not even that warm i tested yesterday
Unggnu: haha, yes, that's why I normally would prefer an active fan if there is a working power management
fly_: id still love to be able to set my gpu fan to 50% (which isnt very loud) and underclock it to 500mhz (if it were even running at 500mhz im not sure)
fly_: the freq on my card is something like 666mhz (yes i know lol funny number)
fly_: but i am getting 959fps in fullscreen glxgears
fly_: (i know its not a bench)
marvin24: OT: am I the only one who gets only 10 KiB/s pulling mesa?
fly_: mmm yup i downloaded at 3.2mbps
fly_: not even oh 3 hours ago?
fly_: Does anyone know if rovclock works with open source drivers?
marvin24: fly_: propably related to dsl lines as from my university account I also get mbps... - strange
fly_: people at your uni are downloading lots of pron
biotube: fly_: don't you know? they're downloading music. RIAA said so.
fly_: that too but a LOT of porn
fly_: lol
fly_: mmm lunchable
marvin24: I don't think that the high speed network is made (or primary) used for this - at least not here
fly_: well what is your uni running 1 OC3
marvin24: what is it?
fly_: what kind of connection do they run
fly_: speaking of porn .. my desktop is so sexy
MostAwesomeDude: And we're officially offtopic.
marvin24: back to the original OT: git seems to have problems with all freedesktop repositories on my dsl line
fly_: Yeah sorry bout that
fly_: wipes drool from face. "Oh Hey 'dude, do you know if rovclock works with the Open Drivers?"
biotube: marvin24: the git protocol could be throttled; try http instead
fly_: ugh why would they throttle git.. thats just racist.
marvin24: as I said, it works fine from another (high speed) account.
marvin24: an I can download with > 200 KiB/s on my dsl line, so the problem is on the fdo side
marvin24: are there some operators who I can contact?
fly_: its not fdo i assure you
MostAwesomeDude: marvin24: It's not our problem.
fly_: go to the open source lab and do a git from there i bet is FLYS
fly_: it*
MostAwesomeDude: fly_: Misuse of "racist;" perhaps you meant "infantile."
marvin24: ok - was OT anyway
fly_: MostAwesomeDude, heh I meant to joke. But you never answered my question'
MostAwesomeDude: fly_: Dunno if rovclock works, probably not.
fly_: hmm any way to gague temp?
fly_: outside of me using a external thermo
MostAwesomeDude: Not yet. Nobody's written the code to talk to the thermal sensors.
soreau: someone was talking about it the other day iirc
fly_: it would be a great help in achieving better performance
fly_: thermo readout and fan control
MostAwesomeDude: Actually, it wouldn't help performance at all.
MostAwesomeDude: It would help with power-saving though.
soreau: Since the hw is capable and many people are interested in it, we should see fan/temp stuff sooner or later I imagine
fly_: well not directly no but itll give us the tools to improve the performance without frying our gpus
MostAwesomeDude: As usual, the problem is that nobody's written the code, not that we don't know how to do it.
fly_: I believe (dont quote me on this) that most of the cards covered by the r300 drivers can handle a core temp of 90*C (at least i know my x1950 can)
rillian: fly_: actually, you still have to wait for compression and delta resolve, so being on a fast network doesn't really make a git clone instant
fly_: rillian: oh of course but we were talking about his 10kb transfer
rillian: ooch, that is bad
soreau: Will it ever be possible to use the directX portion of cards in linux? or is dx exclusively proprietary?
fly_: exclusive
fly_: MS has their dirty paws all up in that cookie jar
rillian: isn't it implemented on top of the same capabilities the opengl code is?
rillian: why would it be harder to implement directx than opengl?
biotube: there are critical differences that make a 1:1 translation impossible
biotube: plus, I'm pretty sure the docs don't talk about the DX interface
rillian: they don't talk about the gl interface either?
MostAwesomeDude: The cards have switches to use Dx rasterizer rules instead of GL.
MostAwesomeDude: And they're documented.
biotube: ahh
fly_: I really think OpenGL still has a lot of unlocked potential
fly_: er untapped*
soreau: MostAwesomeDude: Is that to say it is in fact possible?
otaylor: Well, really, that there is no "DX part of the cards"
rillian: hi otaylor
otaylor: at least up to r500 (I haven't looked at the r6xx docs in detail) there are just a few switches that can be flipped to control some behavior differences between DX and OpenGL
otaylor: Hey rillian
otaylor: soreau: so it, might conceivably be useful for Wine to be able to toggle those switches to get better compat without workarounds, but there's no "hidden power"
fly_: it would be nice i dont know if theres any legality issues behind that
soreau: Thanks MostAwesomeDude, otaylor
fly_: or i would assume they would have already
otaylor: fly_: would have what?
fly_: toggling the DX switches
biotube: reverse engineering for compatibility is protected under IP law, at least in the US so long as you don't break any DRM
otaylor: fly_: probably more that there's a big bunch of stuff between the radeon mesa driver and the wine translation code
biotube: you'd need to write a DX interface and have a way to check if it's available
otaylor: fly_: So the interest in getting things working a bit faster with radeon cards may not equal spec'ing out OpenGL extensions, implementing them in Mesa, etc. And you'd still need the fallback compat code.
fly_: hmm oh i wasnt suggesting.. i have no clue why wine hasnt just assumed it was legality
fly_: it would be very nice though
biotube: wine's also just an API project
biotube: going DX-OGL is less complex
otaylor: fly_: I tihnk people are more interested in DirectX on top of Gallium, though I'm not sure how much work has actually been done on that
soreau: is interested in gallium, namely fbos and glsl
soreau: and ultimately gl2.1 at least
otaylor: soreau: fbo's work pretty much OK in the non-gallium driver
soreau: Then whatever will make compiz alpha blur work with hw fastness :)
fly_: yeah im interested in the bo handling fix that i need to produce a bug report for (downloading the game now to do just that)
fly_: that'll hopefully! be the last thing needed to allow the r300 drivers to do wine with DX > 7
fly_: I can at least modify my games quality settings to produce the better framerate
fly_: especially the Valve games because they have so much client control over rendering options
fly_: wtf.... its not crashing now
fly_: @_@
fly_: im so confused
fly_: new updates perhaps?
soreau: fly_: You may have finally did it right
fly_: uhh nope aparantly not.. even though i have dxlevel set to 9.0 and the game reads it as such i still get kicked from games with zblock
fly_: it says im still using DX 7
fly_: ok so it will run dxlevels higher than 7 but i am missing textures EVERYWHERE
fly_: and the framerate is unplayable
tzaeru: I wonder if it's still feasible to switch between ATI binary drivers for R300 and these open source ones
fly_: maybe i dont need to build that parameter fix into the mesa build anymore... i dont know if that would effect the textures any
fly_: or the framerate
fly_: OH i also havnt set my video card options in xorg lol
fly_: hahahaha
tzaeru: should those actually be specified?
fly_: its supposed to autodetect now...
tzaeru: I only have few features put on, nothing else ;3
tzaeru: yep, I'm a bit doubtful of the autodetect.
fly_: but ive seen differences changing options
tzaeru: especially when I can't seem to see what it has detected anywhere.
fly_: 100's of fps differences
tzaeru: enabling pageflip, I managed to jump FPS with glxgears from 2700 to 3600 :-3
tzaeru: though, with these drivers, playing many games that worked well on ATI drivers is impossible due to FPS issues.
fly_: i dont even think i have option "DRI" in the video card section
fly_: come to think of it is there an Option "DRI2" ?
WhiteRabbit56: thought: is thee a way to control/set AA AF?
fly_: theres a load for it so i would assume there would be an option "DRI2" "bool"
fly_: aw man thunderstorm... i better not lose my power
fly_: that would piss me off
fly_: actually if anyones got a list of current xorg options preferably with descriptions itd be awesome
fly_: oh man theres the lightning
fly_: rubs his surge protector for good luck
fly_: brb
fly: :/ 12fps and missing textures still
Guest10210: i wonder if that parameter fix osiris gave me doesnt need to be added anymore..
fly_: gonna test compiling mesa without the patch
zhasha: agd5f: hello
fly_: Nope i aparantly still need the parameter fix -sigh- i dont know how i can possibly fix these textures or the framerate at this point
MostAwesomeDude: Arg.
MostAwesomeDude: There's no legal reason why Dx can't be implemented.
MostAwesomeDude: Just look at Wine.
zhasha: MostAwesomeDude: MS doesn't like WINE
fly_: MS doesnt like freedom
zhasha: they WILL go out of their way to break it
MostAwesomeDude: zhasha: But they haven't sued the project or issued takedowns.
zhasha: MostAwesomeDude: no, they just continually make it as hard as possible to maintain compatibility
mjt: they can't
stikonas: but the resources of m$ are not infinite
fly_: im assuming if msoft hasnt thusfar they cant.. theyve got plenty of bored lawyers
mjt: they can't break compatibility for other and their own programs
zhasha: I'd love a D3D state tracker, but it'll probably awaken the wrath of microsoft
fly_: ^ indeed
MostAwesomeDude: zhasha: VMWare can't confirm or deny it, but they have a D3D st.
fly_: but i wonder what the legality is with freely distributed software.
zhasha: they probably have a basic D3D8 st
fly_: i mean if you arent turning a profit im not sure ms has a case
zhasha: which in all honesty, isn't all that much work
zhasha: fly_: the implementation is what they own, the API is free for all to implement AFAIK
adamk_: I've always wondered about vmware's interested in gallium3d... Can someone explain it? Is it just to support 3D acceleration on their products?
zhasha: but an implementation of DX that rivals their own, they won't be happy about that
zhasha: adamk_: I imagine they want to provide superior 3D acceleration in virtual machines
fly_: oh nonono that would bring down the wrath of the paperclip
adamk_: Does anyone know how many folks vmware has working on this? I know Brian is a vmware employee.. Anyone else?
fly_: i just had someone call me a communist for using linux
fly_: lmao
zhasha:
zhasha:
MostAwesomeDude: adamk_: VMWare bought TG.
fly_: almost pushing 1000fps in glxgears fullscreen hmm what else can i tweak
osiris: fly_: if you have r300 or r400 enable ColorTiling
fly_: did it
fly_: is there a full list somewhere?
adamk_: Ahhhh... Interesting. I didn't realize the entire company went over to vmware.
adamk_: Well, I hope that's a good thing.
osiris: color tiling and pageflip are the only that can give you some performance boost
mattst88: IIRC it's been about a year since VMware bought TG
fly_: osiris, OH, by the way after a fresh install of all my graphical components i no longer have the bo handling error but i am missing textures all over the place
fly_: and my fps is 12ish
fly_: in DX8+
osiris: fly_: are there any errors in the console?
fly_: gimme a second
fly_: errr i dont believe so... -tests again-
osiris: pastebin whole output
spstarr: hmm
spstarr: AMD folks, is it alarming you that the GPUs are getting quite toasty with the r6xx driver?
fly_: alright but i think i get some wine related errors (other than that not much output at all
fly_: Console output: http://pastebin.ca/1625604 Screenshot: http://img266.imageshack.us/img266/615/testhardware0001.jpg
fly_: that happens in DX8+ but not DX7
MostAwesomeDude: Have you asked the Wine guys?
terracon: I think wine only really cares about proprietary nvidia. Everything else. it may work, If it doesn't... oh well
fly_: yeah they dont tend to do more than say "oh it must be a driver issue"
dmb: agd5f, ping
dmb: anyone know who wrote avivotool?
osiris: fly_: remove the WINEDEBUG=-all
terracon: well most of the windows stuff these days is going to use every possible option out of direct x and open source opengl doesn't implement alot of these advanced features. So windows games in wine will run like shit if they even load at all
dmb: oh wait, has to be run by root or it will segfault
dmb: brbs
fly_: osiris, k
fly_: http://pastebin.ca/1625627
fly_: err:d3d_surface:fb_copy_to_texture_direct Texture filtering not supported in direct blit
fly_: lots of those
zhasha: should really just implement a D3D state tracker
zhasha: D3D9c
fly_: well thatd be fine with me :X my cards max support is 9c
airlied: the hard bit is integrating with the rest of Window drivers
zhasha: DX10 sucks ass anyway :P
airlied: though a D3D( state tracker for Linux/Wine might be sensible
stikonas: Wine people will not care anout D3D state tracker until it is implemented
Dr_Jakob: zhasha: actualy its the complete opsite. D3D9c is just a big pile of steaming shite, while they actulay got DX10 right.
stikonas: at least it seems to me that Wine people will not help to write it
fly_: as soon as i can verify this all works and i can actually play in the higher dx levels I want to write a detailed HowTo for r300
fly_: at least for Fedora
osiris: wine hackers aren't going to implement D3D state tracker on top of gallium, because they need to preserve current infrastructure for closed source drivers and writing maintaining another one would be to much of a problem
zhasha: Dr_Jakob: then there's the matter of which performs better. Even under Vista DX9 outperforms DX10
terracon: I find some wine apps go to software rendering when you load them. You have to try a bunch of times before hardware accel kicks in
Dr_Jakob: zhasha: I find that very hard to belive. Sure DX9 is faster if it doesn't do the _same_ things under DX10.
fly_: but if you guys have any experimental patches you wanna offload on me im your guinea pig :D
fly_: x1950xtx is the epitome of compatibility issue
fly_: or so it would seem
MostAwesomeDude: fly_: Seriously, r500 is not that bad off. I'll gladly blame Wine over GL any day of the week.
MostAwesomeDude: If they think it's our fault, it'd be nice if they could provide unit tests or samples of failing code.
fly_: dude, yeah me too but thank god they are getting a little better
zhasha: we need a D3D state tracker. throw out the stale wine
fly_: im just obsessing because i wont be able to scrimmage under dx7 because of that damn plugin -.-
osiris: fly_: I believe these errors are actually indicating some real problems. anyway r300 doesn't have real (fully accelerated) support for FBOs so the game would be slow anyway
mokoloko: could one play half life 2 with r600+ cards on dx7 or even 8.x mode?
mokoloko: in theory
mokoloko: through wine of course
osiris: mokoloko: it will be slow
fly_: I dont know but it certainly plays under windows heh im not sure how much fbo effects it there
bridgman: spstarr; are you running 6xx with kms ?
zhasha: fly_: everything uses DX surfaces (FBOs)
bridgman: agd5f added power saving code for non-kms, but the code hasn't gone across to kms yet AFAIK
osiris: fly_: I meant that r300 driver doesn't have some things implemented yet, that's why it's slow for FBOs
fly_: Oh
adamk_: played HL2 on r300 with the open source drivers with cedega.
osiris: the HW is fully capable of FBOs
fly_: i thought you meant hardware support and i was like hmm?
fly_: lol
zhasha: bridgman: you accidentally the whole SU:SU_TEX_WRAP_PS3
fly_: wishes he was a better coder, hed love to help any way possible
fly_: talks in 3rd person a lot too so its even better
dmb_: i wish a knew a bit more about video cards, and i would contribute :)
dmb_: but honestly, after coding the whole day at a company, coding in your free time sounds bad :)
osiris: dmb_: just ask questions
osiris: dmb_: agreed. that's why I contribute to r300 so little now
fly_: shakes fist at osiris, j/k <3
dmb_: yes, why i'm considering changing my career to IT kind of stuff and just coding in my free time
dmb_: IT is sooo much less stressful
fly_: honestly I SHOULD upgrade my card but with its HW specs I just dont see that being necessary for another 5 years
fly_: save for DirectX support
bridgman: zhasha: ??
zhasha: bridgman: there are now 2 entries for SU:SU_TEX_WRAP_PS3 in R5xx Accel Guide (1.4)
dmb_: i'm considering buying a 5770 card
dmb_: any of you have one yet?
bridgman: duplicated info ?
zhasha: yep
zhasha: page 214
fly_: well honestly my next card will most likely be an Nvidia -ducks- unless ati does something amazing
dmb_: anybody see whats wrong with my registers http://bugs.freedesktop.org/attachment.cgi?id=30505 :)
bridgman: fly_; what would you consider amazing ?
dmb_: is there a way i can resolve the names of the registers?
dmb_: 00007ba8 doesn't mean much to me
fly_: i dont know... lol i havnt seen it yet
fly_: the 5870 does look pretty though
yangman: fly_: http://www.youtube.com/watch?v=N6Vf8R_gOec
dmb_: too much $$
fly_: yeah
yangman: >.>
fly_: reaches into pockets and pulls out moths
fly_: heya yang
yangman: heh
yangman: I'm totally getting a Evergreen once I can afford 2 more 24" monitors >.>
fly_: wishes beard trimmings were a form of currency
dmb_: is the 5770 that bad, or is it really worth paying like a couple hundred extra for the 5870?
bridgman: dmb, what are we looking at ?
fly_: 5870 is pretty much overkill
fly_: <-- likes overkill
bridgman: in your link ?
dmb_: 12270*5160 i would love that resolution
dmb_: bridgman, DVI doesn't work with my card on the radeon or radeonhd driver
dmb_: but works with fglrx and VESA
dmb_: strangly enough, VGA output works fine
bridgman: so this is a register dump ?
dmb_: bridgman, yeah, avivotool
fly_: WELL THAT WAS INTERESTING
fly_: clicked on a webpage and x died
dmb_: thats the radeon driver's dump
dmb_: fly_, maybe flash
fly_: maybe but i have flash
bridgman: what do you get with radeon, ie do you get sync but no video ?
dmb_: bridgman, nope, get no signal
dmb_: i can switch back to a vt, and video works again
dmb_: but once back in x, no signal
bridgman: and presumably no errors in xorg log ?
dmb_: nope
bridgman: wierd... what GPU ?
dmb_: all the details are http://bugs.freedesktop.org/show_bug.cgi?id=24313
dmb_: http://bugs.freedesktop.org/show_bug.cgi?id=24313
dmb_: erm
dmb_: Gigabyte Radeon HD 4350
bridgman: ok, tx; that's better than me asking a million questions :)
fly_: the gtx 295 has a 896-bit Memory Interface
fly_: 5870 = 256bit
dmb_: it was a cheapo card, going to replace it, but i figure others might run into the same issues as I did, so I want to help fix it if at all possible
fly_: but the 295 is about $100 more
dmb_: fly_, no need for a 896-bit memory interface
dmb_: even 128 is fine
fly_: BUT it also has nearly 2gb ram
dmb_: and what kind of textures are you going to be loading that total to over 2gb of vram?
fly_: ALTHOUGH.. its only GDDR3 and the 5870 is GDDR5
dmb_: GDDR5 is a huge improvement
bridgman: dmb, with monitor attached to DVI do you get an image during bootup ?
dmb_: bridgman, yes
dmb_: bridgman, able to access the vts fine also
dmb_: bridgman, also, whatever radeon is doing wrong, radeonhd is also doing wrong
bridgman: pity; it makes life easier when the drivers behave differently ;)
dmb_: yes
fly_: doing wrong? FBO's?
fly_: -.- that was a spiteful comment
bridgman: not to mention confusing
bridgman: or maybe it's just me ;)
dmb_: heh
dmb_: FBO?
bridgman: I think we have conversation bleed-through happening
dmb_: oh
fly_: http://en.wikipedia.org/wiki/Framebuffer_Object
bridgman: fly_: that's not helping me understand the relationship between FBOs and dmb's issue with DVI
fly_: oh no i was just injecting my 2 cents
fly_: about what radeon hd and radeon are doing wrong lol
bridgman: but... but... (scratches head)... FBOs are implemented in mesa & drm, not radeon & radeonhd
yangman: hahaha
bridgman: needs more coffee
Bibber_: Is there issues with audio over hdmi?
fly_: eh
dmb_: just had some hazelnut coffee, good stuff
fly_: im just obsessing.. where have you been
bridgman: in an RV with three generations of family driving around Cape Breton, NS
bridgman: it's nice to be back at work ;)
fly_: also just had to pick up a half a bag of fritos in under 10 seconds
soreau: gives bridgman a ton of ground coffee beans
bridgman: goes off to make coffee
fly_: i got it all in 9
yangman: bridgman: speaking of which, I'm just going to point people at this everytime the "but it's in Mesa" answer needs to come up ;) http://yangman.ca/blog/2009/10/15/linux-graphics-driver-stack-explained/
fly_: osiris, so I dont suppose you have any "experimental" patches for FBO huh? :3
bridgman: yangman; looks good
bridgman: there should be headlines saying "radeonhd will never have 3D' all over the internet within a week ;)
dmb_: heh
Sclytrack: wut no 3D?
bridgman: ;)
zhasha: why do you keep updating the R5xx guide by the way?
fly_: No 3D? omg -runs to tell others-
dmb_: bridgman, so, at ati, do you get to have one of http://www.youtube.com/watch?v=N6Vf8R_gOec at your desk? :)?
bridgman: during the course of driver development we find missing stuff
bridgman: dmb; I'm on 26.4Kbps dial-up, can you describe the video to me ?
dmb_: bridgman, 24 monitor display
bridgman: zhasha; either a question comes up and we find the docs weren't sufficient, or we know something was missing but couldn't find internal documentation at the time the doc was prepared
bridgman: dmb; no, but I saw the rig set up in the Toronto office before the show
yangman: hahaha
bridgman: main impression was "it's *big*" ;)
bridgman: we didn't have all 24 displays running 1 app when I saw it, so I only saw 6 screens per app
spstarr: hullo bridgman
bridgman: hi spstarr
spstarr: bridgman: I have an interview Wednesday
dmb_: oh
spstarr: not development job but linux stuff
spstarr: bridgman: I realllly hope I get this job
spstarr: bridgman: question about GPU temp
yangman: spstarr: congrats
spstarr: bridgman: the r6xx is getting VERY HOT right now with the linux driver
dmb_: spstarr, just don't get nervous
dmb_: because they can tell
spstarr: bridgman: the GPU will prevent itself from overheating?
dmb_: sound confident in yourself
spstarr: dmb_: yes, i know. I'm going to focus on the questions not the person interviewer.
spstarr: (s)
dmb_: i always go to interviews assuming i'm not going to get the job so i don't get nervous
bridgman: spstarr; there is a thermal shutdown but I don't know if it has to be set up by driver/bios or not
spstarr: eep
bridgman: are you using kms or user modesetting ?
spstarr: it got very hot enough that touching underneath the case of laptop for long.. would be quite uncomfortable
osiris: fly_: nope. implementing it would require major changes in the driver, and currently I think no one would like to spend too much time on r300 classic driver
yangman: thermal shutdown is usually in the high 90C or low 100s
spstarr: bridgman: no KMS that locks up still
spstarr: bridgman: I have switched back to Intel GMA for moment
bridgman: ok, if you're not using KMS then there is a ForceLowPowerMode option
fly_: osiris, is/will it be in gallium?
bridgman: laptops usually get hot, so that on its own is not a concern
spstarr: bridgman: well i can't see the GPU temp right now since we dont expose this to sysfs yet
osiris: fly_: in gallium you get it for free
fly_: I mean not to sound selfish or anything but im going to have to install windows again if i cant run at LEAST DX8
yangman: you can theoretically cook an egg on the bottom of a laptop and the hardware will be just fine
spstarr: !!!
fly_: doesnt want to do that
spstarr: yangman: still concerning it gets that hot
yangman: although, if can actually bring water to a rolling boil, be worried
yangman: ;)
bridgman: yep; we routinely run the chips up to ~90C and cut off at just over 100C; eggs are normally poached between 90-100C
bridgman: yangman; that's not totally true
bridgman: the egg yolk stains the plastic
fly_: osiris, So I guess my question is would you like me to run gallium and help testing? (seeing as i have no life )
yangman: bridgman: it adds character ;)
bridgman: has too much character already
fly_: im already sacrificing my home pc for server development (as of 2 months ago) so its no trouble to run experimental drivers while i do that
osiris: fly_: currently r300g doesn't need testing, what it needs is a manpower to implement missing pieces, and fix all major bugs (most of them are already known)
fly_: oh, uh i dont think i can help there :P unless you need me to code a calculator or something lol
spstarr: bridgman: so the GPU will protect itself
fly_: btw whats different about the compiling/setup process to run gallium? I think im gonna
dmb_: probably hard to compile/setup
dmb_: someone should make a guide
fly_: er i dont think its hard
osiris: fly_: it's simple
fly_: so ill have a r300 and a r300g?
osiris: run ./configure with additional parameters: --disable-gallium-intel --enable-gallium-radeon --with-state-trackers=glx,dri
fly_: ok thats close to what i expected
osiris: then the library will be in lib/gallium/radeon_dri.so (you need to rename it to r300_dri.so)
fly_: then my xorg driver is gallium?
osiris: noe
osiris: *no
osiris: it's only a dri driver
fly_: oh ok so still radeon then
fly_: just curious.. how will i know its running
bridgman: spstarr; are you running with ForceLowPowerMode ?
spstarr: dont have that set, no
spstarr: I should however
bridgman: yep, give that a try
bridgman: it will slow down 3D performance but make everything run cooler
spstarr: but is the GPU heating also due to driver not being optimized yet?
osiris: glxinfo will report different OpenGL renderer string
spstarr: in Windows I tried Second Life and the GPU was not that hot
bridgman: there's no dynamic power management in the driver, so it runs at the same speed all the time
bridgman: once dynamic PM is implemented the clocks & voltages can go up and down depending on what you are doing
bridgman: so average power is quite a bit lower
spstarr: ahhhh
spstarr: bridgman: that explains it
bridgman: that said, the cooling solution on the laptop should be able to handle continuous full speed operation eg gaming
spstarr: bridgman: that will be exposed via sysfs?
bridgman: unless it's full of dust ;)
bridgman: mostly invisible; the issue is that dynamic pm needs access to both modesetting and acceleration activity information
bridgman: so it really needs to be built into the drm on top of kms
bridgman: you can't downclock memory unless you know the engines are all inactive, so we can't do that in the ddx driver (ddx doesn't know about 3d)
bridgman: we can't do it in drm without kms because drm doesn't know about display modes
bridgman: so for the moment you have a "go slow" switch, until someone has time to add dynamic PM to the KMS drm
bridgman: same with reading temp, setting fan etc
bridgman: on all but the newest chips that's handled by an external (non-AMD) chip connected via i2c, but the registers are also used by modesetting
bridgman: so if you implement temp/fan support in the kernel (where it needs to be) it conflicts with modesetting in usermode accessing the same registers
bridgman: in fglrx we added a special protocol between ddx and drm to handle this stuff, but with the open drivers it makes more sense to build on kms
bridgman: the newest GPUs have on-chip fan controllers (and temp reading I think) but we haven't dug up that info yet
yangman: bridgman: is that for Evergreen, or includes r7xx?
bridgman: 7xx and some of the 6xx
bridgman: just 670 I think but maybe also 620/635
bridgman: somewhere in there
fly_: osiris, and i need not apply the param patch right?
osiris: nope
yangman: ah, good to know
sgcb: osiris: would --with-state-trackers=xorg do anything for me on an r7xx
bridgman: we'll know more after the docs are written ;)
bridgman: needs to go cut up some trees
osiris: sgcb: no, there's no gallium driver for r600 gpus
sgcb: oh, drat
fly_: hows this look
fly_: ./configure --build=i686-pc-linux-gnu --prefix=/usr --with-dri-drivers=r300,swrast --enable-32-bit CC="gcc -m32" LDFLAGS='-m32' --without-demos --disable-gallium-intel --enable-gallium-radeon --with-state-trackers=glx,dri
osiris: looks fine
fly_: kk
dmb_: would i notice 2d performance difference between Radeon 43xx and the Radeon 5770?
bridgman: when scrolling big windows, I imagine; that's probably the place where 3D engine performance and bandwidth make the biggest difference
bridgman: possibly video playback under a compositor; not sure how close to the bw limit we are on a 4350
dmb_: oh
mjt: scrolling become significantly slower when switching to kms ;)
dmb_: yes
mjt: dunno what was the cause or the real change but it _is_ slow :)
dmb_: well, i'm guessing with the r8xx, its going to be a while before I will be worrying about KMS :)
bridgman: my guess is that the slowdown comes from resending full state with each command packet under DRI2
bridgman: since overlapping blits have to be broken down into a bunch of small regions
dmb_: last time I was playing with KMS on r515, i had the fragmentation problem
dmb_: got slower and slower over time
dmb_: until i rebooted
bridgman: yeah, video memory managers are some of the hardest, touchiest code around
bridgman: we had to pretty much put armed guards around the code base for the first couple of years when we transitioned the proprietary drivers onto a common memory manager
dmb_: lol
bridgman: the nice thing is that once they work they tend to keep working
dmb_: i'm not sure if any of the memory manager stuff has changed in the last couple of months to fix that
dmb_: yes
dmb_: question is, when do you do a pass for degraging?
dmb_: whenever you alloc new memory?
dmb_: defragging*
bridgman: not sure what the current code does, but the most common approach seems to be "defrag when an allocation fails"
dmb_: ugh
bridgman: that tends to give you periodic pauses though, so you tend to end up with a combination of checking for low-hanging fruit on every call plus defrag on fail
dmb_: i guess it fails when there is no more consecutive slots?
dmb_: actually, i have no idea how vram is allocated
dmb_: i'm guessing no paging system like main memory?
mjt: and by default with linux's overcommit allocation never fails :)
bridgman: yeah, it depends on the size of the buffer you are trying to allocate
bridgman: no paging yet
dmb_: would paging be useful in vram?
dmb_: it also makes things 100x more complicated
bridgman: I used to play on Lisp machines; they would run for about 24 hours before having to garbage collect, and would pretty much shut down for 2-3 hours until garbage collection finished
dmb_: wait, there were lisp machines?
bridgman: paging is one of those "useful in theory, not so much in practice" things
dmb_: i only knew it as a programming language
dmb_: annoying one at that
dmb_: bridgman, seems to work well for main memory in most operating systems
bridgman: yep, lisp used to be incredibly slow on general purpose processors, so special machines were built with tagged memory and explicit support for list operations
dmb_: oh
bridgman: it was arguably the Symbolics vs LMI war that prompted Stallman to dedicate his life to software freedom
dmb_: lol
bridgman: http://www.donhopkins.com/drupal/node/109
dmb_: hahahah
bridgman: they were roughly the same level of coolness as smalltalk machines, but lisp machines happened within driving distance of where I lived while smalltalk machines were on the other side of the continent
bridgman: I assume you've seen : http://xkcd.com/224/
fly_: osiris, uhh theres no /lib/gallium or /usr/lib/gallium .. but there is /usr/lib/dri/radeon_dri.so
dmb_: oh
osiris: fly_: I meant $(mesa dir)/lib/gallium
fly_: oh so i pull the made radeon_dri.so and move it?
osiris: or copy
fly_: ok -moves it so he can change its anme at same time ;o
fly_: name8
fly_: alright -poof-
fly_: OpenGL renderer string: Mesa DRI R300 (R580 7240) 20090101 TCL
fly_: 0_o
fly_: i dont think it changed...
dmb_: bridgman, anything you can suggest I test?
dmb_: bridgman, or should i just wait for agd5f
soreau: has quit (Remote closed his connection)
fly_: wellll... maybe it did work.. my water is no longer discolored. just missing and the colors look right
fly_: still 10fps
fly_: hmmm Im not sure aboot this.. before i was fine in the menu now it lags. (badly) cpu usage at 98%(of one core) in menu
mokoloko: could one play quakelive with r600+ cards nao?
fly_: oh and that was DX7
fly_: i think im not actually running with HW acceleration unless the drivers are THAT unfinished
fly_: osiris, Im really not sure I did it right
amarsh04: my rv280 complained when trying to run etracer under kernel 2.6.31 with otherwise stock debian unstable
amarsh04: %%% etracer warning: Attempt to bind to Texture unloaded texture: `b-herring _run_icon'
amarsh04: drmRadeonCmdBuffer: -12. Kernel failed to parse or rejected command stream.
amarsh04: etracer: page allocation failure. order:4, mode:0xd0
bridgman: dmb; let's wait for agd5f, he's probably already done some work on it
fly_: is so confused
[Enrico]: mhm i gento RADEONDRIGetVersion failed because of a version mismatch..... i'm using archlinux with kernel .31. kms is enabled and at boot is works, i can see it changing the screen resolution during kernel loading
airlied: [Enrico]: the DDX isn't build with kms support
[Enrico]: doh!
[Enrico]: airlied: so i have to disable kms i guess
[Enrico]: or wait archlinux devs to enable it
airlied: pretty much
[Enrico]: mhm that's why gentoo is better
[Enrico]: [but that pc is too old to compile gentoo :'(]
[Enrico]: i think i will reinstall fedora :D
[Enrico]: [but last time i got some problem with composite, bad rendered windows]
[Enrico]: eheh damned r200 :D
spstarr: bridgman: i will try the Xorg config option for forcing power low
m03sizlak: just reinstalled f12 alpha from DVD
m03sizlak: sill get random lockups using radeon
m03sizlak: when can i expect the next major release that might actually fix something for me?
m03sizlak: lotsa glitches too
[Enrico]: airlied: oh sorry before i forgotten to thank you, with kms disabled dri is working :D
[Enrico]: i really must switch to fedora i think
fly_: OK so either I did something horribly wrong or the gallium driver just doesnt work for me. I tried compiling it for 64bit using "./configure --prefix=/usr --with-dri-drivers=r300,swrast --without-demos --disable-gallium-intel --enable-gallium-radeon --with-state-trackers=glx,dri" then i did "make" followed by "cd ~/mesa/lib/gallium/ && mv ./radeon_dri.so /usr/lib64/dri/r300_dri.so" overwrote then X wouldnt start
fly_: [Enrico] help test rawhide (its actually pretty stable)
[Enrico]: fly_: well normally i prefer a stable system, but i can pick up radeon and mesa from rawhide, thanks
m03sizlak: [Enrico] i wouldnt suggest that
MostAwesomeDude: fly_: Again, you can't use the r300g for AIGLX, so don't install it over your system DRI.
m03sizlak: fedora 12 rawhide for me is less stable than a one-legged drunken hooker
m03sizlak: on crack
fly_: ah, well that would make sense
[Enrico]: m03sizlak: eheheh, well before i was using f11, but i had problem with composite, it worked but with half rendered windows :'(
m03sizlak: yeah well glitches are the least of the problems im seeing in f12
fly_: try updating kernel
m03sizlak: everything is up2date
[Enrico]: note glitches, not rendered windows. i mean just few parts of windows were rendered
fly_: Linux Combaticus 2.6.32-rc5 #1 SMP PREEMPT
m03sizlak: radeon locks up more often than a bank of america branch
fly_: maybe you dont have kms set up
mokoloko: :D
m03sizlak: yeah i do
m03sizlak: its f12 installed from alpha DVD and updated
m03sizlak: on r300 hardware
fly_: well i have had 0 lockups on my r300
soreau: m03sizlak: Works fine here on gentoo with rv350
fly_: :p
m03sizlak: agp or pci?
fly_: pci-e
soreau: agp
m03sizlak: well thats prlly why...i have agp
fly_: kernelver?
dmb_: do they sell agp cards?
m03sizlak: what kernel is that fly_ ?
m03sizlak: its a laptop
fly_: post uname -a
[Enrico]: i have a r200 AGP
m03sizlak: roooofl
m03sizlak: [paul@megaton ~]$ uname -v
m03sizlak: #1 SMP Tue Sep 29 16:32:02 EDT 2009
fly_: a
dmb_: #38-Ubuntu SMP Fri Oct 2 11:06:40 UTC 2009
m03sizlak: oh yeah
m03sizlak: 2.6.31.1-56.fc12.i686
fly_: move up to some rc of 32 you wont regret it
fly_: ;_)
m03sizlak: how to do that w/ yum in f12?
fly_: you dont
m03sizlak: why not
m03sizlak: is there a rpm
m03sizlak: cus im not trying to compile from source
fly_: well if you wanna do it right you gotta compile from source
fly_: :)
fly_: check it (this will take a second)
m03sizlak: the f12 kernel branch is very different from the mainline kernel
m03sizlak: especiually in KMS
fly_: mkdir /usr/src/
fly_: takes deep breath
m03sizlak: yeah i know how to compile a kernel thx
fly_: well its easy why would you just trust fedora to do it for you
m03sizlak: but the fedora kernel is completely different from a kernel.org kernel
fly_: yeah mine doesnt crash
fly_: lol
m03sizlak: and btw you can just download -rc5 instead of patching some earlier kernel
fly_: i grabbed the wrong link so sue me
m03sizlak: lol
fly_: been up for 12 hours already :P
m03sizlak: thats not very long
fly_: it is when its only 3:50pm
m03sizlak: time is relative
fly_: to sleep
fly_: anyways you dont have to remove your fedora kernel.. give my method a try
m03sizlak: yeah i used to use kernel.org kernels w/ fedora
m03sizlak: but the fedora kernel is vastly different (newer) in terms of radeon driver
m03sizlak: and kms
m03sizlak: the kms in kernel.org kernel is OLD
fly_: have you tried .31?
fly_: 2*
m03sizlak: no
fly_: pretty sure its up to date
m03sizlak: its not
m03sizlak: airlied, how different is kernel.org .32 from fedora kernel in terms of radeon & kms ??
fly_: id like to know too
fly_: ill roll a fedora kernel from source if its better
fly_: seeing as how theres a *.32
amarsh04: is there any reason why extremetuxracer would die on kernel 2.6.31 with radeon driver like "drmRadeonCmdBuffer: -12. Kernel failed to parse or rejected command stream."?
m03sizlak: yes
fly_: m03sizlak, did you mayybe try compiling a kernel with custom settings before flaming the fedora-qa cannel?
m03sizlak: why
amarsh04: are there other bits that have to be compiled to be compatible with kernel 2.6.31?
fly_: because you CANT expect fedora to know exactly whats best for your particular system
fly_: thats just foolish
MostAwesomeDude: Sure I can.
MostAwesomeDude: Or, rather...
MostAwesomeDude: I can expect the two or three packages that *need* to know my CPU, to have CPU runtime detection.
MostAwesomeDude: *runtime CPU detection, even.
fly_: :)
amarsh04: radeonhd and the guide at http://www.x.org/wiki/radeonhd:experimental_3D worked for my RS780, just wondered why the RV280 had problems
fly_: premade packages are the lazy way out theres a whole WORLD of settings :)
MostAwesomeDude: I'm not lazy; I just think that my CPU time and personal time is too valuable to constantly waste on compiles.
fly_: oh well it takes me like 5 mins now
fly_: lol ive compiled many a kernel
fly_: the "fedora way" looks a little daunting and cumbersome tho
fly_: but hey i might just roll one (customized of course) to see if my results are better
fly_: I still doubt ill see over 10fps in game
fly_: after the obligatory cigarette of course
fly_: (i suppose the drivers would be newer as airlied works directly with them)
fly_: MostAwesomeDude, feel free to confirm or disprove the driver version difference between kernel.org and fedora kernel (if you feel so inclined) as performance is an obsession for me id really like to know
MostAwesomeDude: fly_: Sure. Fedora's kernel is newer than kernel.org.
m03sizlak: tolf you :P
m03sizlak: *told
zhasha: ahrem, graphics driver wise
MostAwesomeDude: Understand that the kernel doesn't incrementally improve WRT DRM; it goes in big jumps and leaps.
zhasha: MostAwesomeDude: so I installed fedora on my spare 20GB, on an extended partition (sda4), giving it its own boot partition (sda5) and installing the system itself onto sda6
fly_: (told you :p ) is something a 12yr old would say, especially after I admitted that there is a distinct possibility due to airlied involvement in the project
zhasha: fedora rawhide that is. well the funny part is grub says Illegal device specified when I try to chainload from sda5 (hd0,4)
zhasha: despite having installed grub onto that partitions boot sector several times :(
fly_: but i can guarantee you my custom compiled kern wont crash ;)
zhasha: fly_: cat /dev/random > /dev/mem
fly_: huh
fly_: makes a new folder "combatadora"
zhasha: MostAwesomeDude: phoronix proved useful for once (I updated the xml file with the newly added registers)
fly_: phoronix was useful?
fly_: man every time i have a problem and phoronix shows up in my google search it always ends up being a waste of time
[Enrico]: fly_: i think he means the test suite
[Enrico]: not the internet site
fly_: oh well the site sux too
fly_: lol
fly_: hi rhodan
m03sizlak: look at this font corruption:
m03sizlak: http://i38.tinypic.com/2vimck9.png
rvalles: hi hi
rvalles: been using KMS for a while on r700
rvalles: just updated the kernel to rc5 from rc3, and now colors are funky (paletted)
rvalles: what's wrong: the kernel (a bug in the newest kernel) or else (api change, need to build super up to date radeon driver)?
biotube: rvalles: staging drivers aren't affected by normal kernel freeze rules IIRC
biotube: you might've just found a bug
rvalles: yeah, might
rvalles: I guess I'll go back to rc3 for now
rvalles: as I need to get some stuff done tonight
rvalles: forget that, biotube
rvalles: I rebuilt
rvalles: libdrm
rvalles: x11-drm
rvalles: xf86-video-ati
rvalles: and did "kill X"
rvalles: I instantly got a proper-looking kdm.
biotube: interesting
biotube: well, the interface is still unstable
hnsr: i noticed my colors were off a bit after trying out the most recent libdrm/xf86-video-ati/mesa
hnsr: then I rebooted to windows and they were still off :(
hnsr: not sure if I'm just imagining it
rvalles: hnsr: were off a bit -> in my case, totally palletized LSD-25 experience.
rvalles: hnsr: if it's not what you've seen, then it's not the same problem.
hnsr: ha, yeah that's not what I was having
hnsr: in fact im pretty sure it probably is my monitor giving me different colors depending on what input I use (analog or dvi)
hnsr: since I did chance my monitor setup recently
ajami: I am starting to attempt some r600 code for the gallium driver. I just want to make sure that I need the Mesa Master, libdrm master, and a kms kernel. Am I missing anything?
biotube: a straightjacket?
hnsr: :D
ajami: but then i couldn't type...
ajami: Wait, I have my toungue!
ajami: so, mesa and drm, build drm, then mesa? Is there a link for this stuff...? (for working on gallium r600)
ajami: Build libdrm (./configure --experimental-radeon-api?), build Mesa (./configure --????)
MostAwesomeDude: ajami: From the topic: http://wiki.x.org/wiki/radeonBuildHowTo
biotube: give mesa --enable-gallium-radeon --disable-gallium-intel
ajami: Except that is for no gallium whereas I want to work on gallium stuff MostAwesomeDude
MostAwesomeDude: ajami: Does --enable-gallium-radeon not work for you?
ajami: MostAwesomeDude: I did not say that, but the instructions you sent me do not include such an option, but rather a --disable-gllium
ajami: biotube: libdrm first?
ajami: or does it matter?
biotube: ajami: yes
biotube: without a KMS-enabled libdrm, Mesa won't build with it
MostAwesomeDude: ajami: Hm. Looks like somebody changed the page. I'll fix it later; trying to fix r300 PSC right now.
ajami: ic
ajami: PSC?
biotube: pixel shader code?
MostAwesomeDude: Programmable Stream Control.
MostAwesomeDude: It's used to swizzle verts on their way into the shader.
ajami: now that my drm is buit with --enable-radeon-experimental-api, how do I tell the mesa build to use that drM?
ajami: drm*
MostAwesomeDude: I use PKG_CONFIG_PATH=/usr/local/lib/pkgconfig as part of my configure.
ajami: what's that?
MostAwesomeDude: Well, it forces pkg-config to look for libdrm.pc in /usr/local.
MostAwesomeDude: Assuming that's where you installed it.
ajami: with the mesa config do I need to add --with-dri-drivers=swrast,r600
ajami: Is this the config line: PKG_CONFIG_PATH=/usr/local/lib/pkgconfig ./autogen.sh --with-dri-drivers=swrast,r600 --enable-gallium-radeon --disable-gallium-intel
MostAwesomeDude: You could, if you wanted to build classic Mesa drivers.
ajami: ah
MostAwesomeDude: Here's mine: $ ./configure PKG_CONFIG_PATH=/usr/local/lib/pkgconfig --enable-debug --with-driver=dri --with-dri-drivers=swrast --disable-gallium-intel --enable-gallium-radeon --with-state-trackers=dri,egl --disable-glw --disable-glut --without-demos
MostAwesomeDude: I didn't want r300 classic built, but otherwise, this works for me.
mattst88: MostAwesomeDude, have you tested out the xorg st?
MostAwesomeDude: mattst88: Not really, no. Trying to iron out more bugs in dri st first.
MostAwesomeDude: xorg st's way more complex.
rnoland_: MostAwesomeDude: what did you just merge?
rnoland_: is still waiting for someone to merge 7.6 -> master
MostAwesomeDude: rnoland_: TBH I'm really not sure.
MostAwesomeDude: I can't find any stray commits from it.
rnoland_: MostAwesomeDude: it looks like you merged master to master
rnoland_: Mesa (master): Merge branch 'master' of ssh://git.freedesktop.org/git/mesa/ mesa
MostAwesomeDude: rnoland_: Yay me? :T
zhasha: MostAwesomeDude: I had an interesting experience earlier after getting rawhide running.
zhasha: one of the mipmap tests showed its texture embedded in itself
MostAwesomeDude: Honestly, I forgot to use $ git pull --rebase when I pulled, but I didn't have any outstanding patches...
MostAwesomeDude: zhasha: r500?
zhasha: the lower left corner of the texture was the texture itself, and it nested
zhasha: yes, r500
zhasha: I still don't have an r300
MostAwesomeDude: Hm. I think it's r500-specific; I'm stuck on an r300 and not seeing this at all.
zhasha: but the whole misrendering thing was a kernel/libdrm mismatch issue
zhasha: this is something new
zhasha: it looked like the mipmap levels were drawn onto the texture and not beside it (memory wise)
MostAwesomeDude: Hm.
MostAwesomeDude: It *would* explain a lot.
mattst88: MostAwesomeDude, do you have all the graphics hardware you need for r300g testing/development?
MostAwesomeDude: mattst88: At the moment, yes.
ajami: MostAwesomeDude: do you have r500 hw?
MostAwesomeDude: ajami: Not until Tuesday.
ajami: now that I have a skeleton for the gallium r600 driver, how do I start this softpipe stuff you were talkinga bout MostAwesomeDude ?
MostAwesomeDude: ajami: RADEON_SOFTPIPE=1
MostAwesomeDude: For more info, read src/gallium/winsys/drm/radeon
rnoland_: sigh... ok, i figured out the problem with hte merge...
rnoland_: lex.yy.c is a generated file....
MostAwesomeDude: Urgh.
MostAwesomeDude: Damn. Did I break master?
rnoland_: MostAwesomeDude: no...
MostAwesomeDude: Oh, well then I didn't *completely* fail.
MostAwesomeDude: I still feel bad though. Do I need to do anything to fix it?
rnoland_: but osiris was going to merge 7.6 to master for me yesterday, since brian hasn't done it...
rnoland_: MostAwesomeDude: i have no idea... i don't think it hurt anything other than just a rogue empty commit
MostAwesomeDude: rnoland_: That's good.
rnoland_: there, it took me two branches to get everything in the right place, but i finally have master+7.6+my local commits+uncommitted wip