Nightwulf|work: hi all
agd5f: mcencora: http://people.freedesktop.org/~agd5f/0001-radeon-r200-r300-r600-add-check_blit-vtbl-function.patch
airlied: glisse: btw packages latest ddx for f12->14
airlied: agd5f: tried zaphod on m7 laptop and rs480 laptop under kms both failed to light up LVDS
airlied: will try and test next week
xmikos: Hello, I am running Arch Linux and using PKGBUILD for libdrm-git from AUR. Today I wanted to do upgrade but I see that author of PKGBUILD commented out "--enable-radeon-experimental-api" from it. Is there some reason like that this option isn't needed anymore or something like that? Or should I enable it?
BioTube: it was enabled by default
xmikos: BioTube: Great, thanks for quick answer
DemoOn: is smthing wrong with driver? http://pastebin.com/d393c2fae
edgecase: r100 nexuiz KMS makes me sad [drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation !
glisse: edgecase: yeah i am thinking on the best solution to address this issue
mcgregor: wine could be a really nice application if it had the capability to crash X all the time ...
DemoOn: Where can i post my problem with radeon driver?
edgecase: otoh, recent builds make openareana work stable. I wonder what could cause "mouse lag" tho? it's like there's a motion buffer delay
Neo_The_User: ARMA II says I have a 20XX texture size, Operation Flash Point Dragon Rising says i have a 1024 texture size, vdrift says 4096 so how do i find out what my actual texture size is?
Neo_The_User: using open source stuff....
maligor: don't think glxinfo showed it off hand
spstarr: agd5f: any patches for me to try?
spstarr: glisse: any patches for me to try?
maligor: Neo_The_User, do those actually work? (since it's in wine)
maligor: Neo_The_User, and the maximum texture size depends on the hardware
Neo_The_User: ARMA II and OFDR is in windows 7 and xp
Neo_The_User: im wondering how to do it in linux
Neo_The_User: i know my memory size is 1024.
maligor: I think new cards supported 8192x8192
Neo_The_User: i have a radeon pci e 4650
Neo_The_User: by HIS
maligor: HD3870 has maximum of 8192 on fglrx
maligor: since that's what I'm currently running
spstarr: looks at ml/and git trees
Neo_The_User: i wonder if texture size is actually allocated by the games themselves
spstarr: i see new patches from glisse
spstarr: [PATCH] drm/radeon/kms: free fence IB if it wasn't emited at IB free
spstarr: agd5f: when i tried dynclck option with KMS it never showed any power tables to use ?
spstarr: mplement setting active PCIE lanes on R600+
spstarr: new ddx
spstarr: airlied: are you rebasing mesa/ddx often from upstream right now basically?
edgecase: hey is atombios based on x86 code callbacks from OS, or just init data tables?
edgecase: we're talking about this in #coreboot
airlied: not sure what you mean x86 code callbacks
airlied: there is a virtual machine
airlied: and you execute the command tables in it
airlied: the vm has a few entry points into the OS to do register read/write, fb access etc
edgecase: ok but you don't need x86 emulator, to run x86 realmode code...
edgecase: ok that's all i need to hear
DemoOn: is this driver error? http://pastebin.com/d211a390f
DemoOn: how is called radeon driver if i want to find it ?
adamk: DemoOn: That does look like it might be a driver error. There are various components to the radeon driver.
adamk: There is radeon.ko which is the kernel module, radeon_dri.so (or r200_dri.so, r300_dri.so, r600_dri.so depending on the hardware) which is the Mesa 3D driver, and radeon_drv.so which is the Xorg 2D driver.
DemoOn: adamk: yah, and im usin the newest, ill try to somehow report it
adamk: Latest? You pulled it from git and your using one of airlied's testing branchs?
DemoOn: this isnt latest xf86-video-ati-6.12.4-r1?
adamk: That is probably not the component that is giving that error. That message is from the kernel.
DemoOn: ok, ill try 2.6.32-r5
DemoOn: now i have 2.6.32-r3
adamk: You might want to try the latest 2.6.32.* kernel or the latest 2.6.33-rc kernel.
adamk: Gotta go.
cire: I still get "You have old & broken userspace please consider updating mesa" (drm is radeon-testing, ddx&mesa are master).
cire: This always happens when running glxgears
BioTube: gears segfaults here
cire: then my situation is better than yours ;)
DemoOn: hmm dunno what to do
DemoOn: with my situation
BioTube: the interesting stuff runs for the most part, except the oblivion launcher and the bioshock updater crash
BioTube: which makes installing the latter impossible
mekius: I don't suppose any devs happen to use Evas gl engine?
chithead: mekius: if it is free enough and you can do benchmarks with it, consider submitting it for inclusion in phoronix test suite
chithead: then it will be on developers' radar
mekius: chithead: hmm, might do that cause atm it's not running on my r600 :/
mekius: with latest git stuff
mekius: at least as new as tuesday i think
chithead: and kms?
mekius: kms and dri2 working
mekius: i can do a lot of other things
mekius: like play games and use compiz
mekius: i normally use e17 though, compiz was just a test
DemoOn: if i can choose opengl from xorg11 and ati, whats the difference?
mekius: expedite is the benchmark program for evas, supports a software engine, xrender engine and gl engine
mekius: so could do cpu and 2d/3d accel tests, of course the gl is only doing 2d
soreau: DemoOn: On gentoo, xorg-x11 is for the open radeon driver and ati is for the proprietary fglrx driver
chithead: DemoOn: x11 is used together with the open source drivers, ati is used with fglrx
mekius: chithead: where would i submit this, just email someone?
DemoOn: seems for now that newer kernel helped resolve my situation
chithead: mekius: it would need a bit more than emailing someone. find out about the test suite on phoronix.com, ask whether the developers would consider for inclusion, then create a patch
mekius: chithead: ah, involved, i see :P
mekius: at least it's all coded in php which I know :)
mekius: something to put on my todo list
mekius: maybe once i'm done with work work :)
cjb: Hi, I see things like "[drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (9216000, 4, 4096, -12)" a lot on R100 -- known?
glisse: airlied, agd5f ok unmappable vram patchset sent
glisse: so far i haven't seen any issue but i barely tested the
glisse: unmappble path
glisse: patch i used to test the unmappable path on r7xx hw:
glisse: i set 40M as on my desktop with compiz i quickly hit that
agd5f: glisse: cool. reading over the patches now
glisse: cjb: you should open a bug and attach full kernel log
glisse: but it's kind of know issue
cjb: glisse: ok, thanks
cjb: airlied: Hm, I'm having trouble changing depth without modesetting, when going X->kdb->X. It looks like I'm restoring all of the changed base registers except gen_cntl_reg, which ends up different because its modified relative to its starting value.
cjb: will keep poking it, just wondering if you have any obvious ideas
cjb: oops, s/its/it's/
cjb: .. that's odd, the kernel thinks it's writing a new value to it, but radeontool regs shows it having the same value before and after
spstarr: glisse: hullo
spstarr: glisse: patches?
spstarr: oh testing memory sizes
glisse: spstarr: nothing that might remotely be helpfull to your case
glisse: next on my todo list if fence cleanup and improved ib recording
glisse: thus might prove more helpfull to your case
cjb: airlied: in particular, I get this image when I set_base away and then back, even though radeontool regs is the same: http://dev.laptop.org/~cjb/radeon-set-base.jpg
cjb: does that look like purely a color depth problem, or could it be something else?
airlied: cjb: thats the paleeete
airlied: palette even
airlied: the lut needs to be reloaded, and possibly stored across the set base stuff
maro: is it normal that R7xx 3D performance blows with KMS? or is it just that mesa 7.7 is too old?
maro: when not using KMS the performance is good, with KMS it looks like software rendering
BioTube: maro: the DRI2 codepath is newer and less optimized
maro: yeah, but we're not talking a 20 or even 50% performance hit, it's many times that
maro: which leads me to believe it's something with my setup
BioTube: how are you measuring performance?
BioTube: KMS currently forces VSYNC on if you;ve got an IRQ-supporting kernel
BioTube: UMS doens't support it
WhiteRabbit56: to anyone: is it possible with the current radeon drivers to find the clock speeds of the gpu and graphics memory
BioTube: debug/dri/0/radeon_pm_info has that
BioTube: though for me it reads "Hz" instead of "KHz:
maro: BioTube: if by IRQ supporting kernel you mean the special firmware, then yes, I do
spstarr: BioTube: thats only displayed if you enable dynpm=1 no?
maro: my measure is the lovely glxgears but also the ioquake3 MENU is too slow to use (3 fps or so)
BioTube: spstarr: I don't know; I'd presume it would still be there
spstarr: BioTube: DebugFS?
spstarr: or sysfs
BioTube: DebugFS, yes
spstarr: yeah not available bufs/clients/name/queues/vm/vma only
maro: I'm using pretty recent bits though, 2.6.33-rc8 + ati git
maro: only mesa is at 7.7
maro: I guess KMS-shipping distros use some newer git snapshot of mesa
BioTube: well, mesa puts together the command streams, so it's probably going to be the biggest speed boost
maro: guess I'll try again when 7.8 is ready
maro: thanks :)
AndrewR: sorry, trying to hunt down strange hang/freeze ... Can someone with non-integrated radeon boot his/her machine with mem=96m and try KMS (with light X session)?
AndrewR: i was about to run mozilla, and got my freeze right after browser startup. Now trying wih same 96M but with UMS
AndrewR: i'm not sure, but _may be_ same bug affecting my system, even without KDE .. http://www.phoronix.com/forums/showpost.php?p=113530&postcount=9 - backtrace looks like some reference problems inside kernel/ttm?
AndrewR: http://bugzilla.kernel.org/show_bug.cgi?id=15276 - kernel bugzilla
DanaG: great, one of my friends is griping about how he "hates ati" because the "open driver doesn't work" -- just because he refuses to build anything (and is on Debian with no "enable-radeon-experimental-api").
DanaG: Oh, and do we have support for 3 heads yet?
airlied: unles you have one of the new 5xxx, 3 heads generally means two cards
airlied: or a matrox convertor box
AndrewR: "-24 C" outside .... good weather for some hacking. Anyone know any radeon-specific ttm stress test?
AndrewR: my kernel currently - mainline git, up to f8b55f251012e104093e105483c45c5d85ad3040 ("Orphan DECnet")
Neo_The_User: does anybody know what is inside the context of the ucode by agd5f? like if they are instructions or what not.
Neo_The_User: my dad (L84Supper) was planning on designing a motherboard with a PCI E radeon card with an arm based processor.
Neo_The_User: so basically im wondering if the ucode is x86 specific
BioTube: since it's ucode for the various controllers on the card, I doubt it ever gets touched by the CPU after upload
DanaG: hmm, does ARM even DO pcie?
Neo_The_User: well there is no mb that has arm and pci e
DanaG: If you could get arm and PCIe SLOT, even, that would be a winner.
Neo_The_User: yeah hes doing that
agd5f: Neo_The_User: the code runs on the GPU, not the host processor
Neo_The_User: is mesa r600 and and libdrm_radeon CPU independant as well?
agd5f: Neo_The_User: yes
DanaG: Neo_The_User: I'd buy such a board, if I could get one.
agd5f: Neo_The_User: lots of people run radeons in powerpc macs which are big endian like arm. I'm not sure anyone has tried it in an arm, but in theory it should work
Neo_The_User: wonderful! :D thank you agd5f and BioTube for the input. :D
spstarr: building git master kernel
L84Supper: 2GHz ARM Cortex SOC http://www.marvell.com/products/processors/embedded/kirkwood/88f6282-3_pb.pdf attached to an ATI HD5x00. The PCIe is only x1 vs x4 or x16 so that would a choke point until we do PCIe x16
L84Supper: good to hear that Linux drivers won't be a problem for ARM
Neo_The_User: AndrewR: did you still want someone to test to see if mem=256m to see if it would work?
AndrewR: hm drm/tests seems to have drmstat utility ... but it gives me only "No driver available"
AndrewR: Neo_The_User, try with 96m, if possible
Neo_The_User: oh ok
AndrewR: Neo_The_User, i have 256m currently (installed in machine). so i can only go down
Neo_The_User: with or without KMS?
AndrewR: radeon survived test=1 with only 64m, so built-in testing not enough
AndrewR: Neo_The_User, with KMS
L84Supper: another issue with ATI HD5x00 on ARM is the smaller cache with ARM, 256K L2 and 32KB L1's
Neo_The_User: AndrewR: alright. compiling KMS code
DanaG: OOOh, that marvell thingy... sweeeeeeeeeet
DanaG: Single core, or dual?
L84Supper: haven't decided yet
DanaG: Having a swappable slot would be even better than an on-board gpu.
L84Supper: we're designing a laptop with a modular used replaceable mainboard
L84Supper: it's x86, but I'm looking at other options using ARM for other markets
DanaG: ARM + radeon could be really cool. Bonus points if we could get a special, low-power ATI.
DanaG: I wonder how much wattage it would take to get, say, an old Radeon 9800 Pro sort of performance, on a modern process.
L84Supper: they offer single core 2GHz or dual core 1GHz ARM
L84Supper: or being able to throttle the GPU clocks for battery and AC powered modes
DanaG: hmm, and something standard, like MXM, would be cool to have.
L84Supper: not sure of any gotchyahs yet for the combination
Neo_The_User: mem=* and test=* goes where? kernel cli?
L84Supper: we don't need vgabios
DanaG: Wait, where does AtomBios come into play?
L84Supper: do the FOSS drivers make any calls to BIOS?
L84Supper: how does the PowerPC with Radeon get around AtomBios or does it just work?
DanaG: Does it still have video-bios?
AndrewR: and ttmtest failed to compile for me: http://pastebin.ca/1803324
AndrewR: Neo_The_User, mem=64 - kernel command line
AndrewR: *64m, not just 64
spstarr: hullo bridgman
AndrewR: Neo_The_User, memtest=4 if memtest compiledin just in case
Neo_The_User: memtest=4 is the kernel config option right?
AndrewR: Neo_The_User, test=1 was module parameter for radeon, testing AGP (in my case)
Neo_The_User: and im compiling my own kernel with memtest off.
bridgman: trips and falls over the registration requirement again
L84Supper: DanaG : my understanding is that the Radeon cards work with PowerPC as-is
AndrewR: Neo_The_User, i just enabled "memtest" in kernel config, may be there was some additional default config i missed
bridgman: L84supper; the radeon and KMS drivers call AtomBIOS for 5xx and up, plus some 4xx
Neo_The_User: oh i need that on. thanks
Neo_The_User: recompiles kernel yet again
bridgman: radeonhd calls for 7xx and up, not sure what the current status is for 5xx and 6xx
L84Supper: agd5f: is there anything special to get radeon working with PowerPC? or do they just work?
L84Supper: "the atom parser works on powerpc"
bridgman: probably the best statement would be that it has worked, and frequently works and sometimes needs a tweak to keep it working
DanaG: ah, so on ARM, you'd still need an atombios image?
L84Supper: I'll have to see what it does with ARM
DanaG: That'd make video bios more of a challenge... you'd have to have an EEPROM user-accessible for people to put their own vbios into it.
L84Supper: DanaG : not sure, I heard AtomBios has more than just a table
bridgman: copying a PC BIOS wouldn't quite work since the parser in the bios is compiled down to x86
L84Supper: bridgeman: that's what I'm trying to fund out
bridgman: you would probably want to compile the atombios parser into your firmware and use that instead of the parser in the bios for initial POST
L84Supper: x86 binary BIOS won't work with ARM
bridgman: 98% of the BIOS is portable bytecode
bridgman: it's just the bytecode interpreter that's compiled to x86
cjb: recalls Alpha machines that could run video BIOS code in emulation
bridgman: the rest is interpreted on the fly when the driver makes calls
bridgman: the syntax is basically
bridgman: push parameters
bridgman: point to bytecode routine
bridgman: call interpreter (written in C)
L84Supper: how often and when is the other 2% called?
bridgman: the other 2% is the interpreter ;)
bridgman: it's probably more like 4-5% really, since the PC calling conventions require some x86 assembly code wrapping the interpreter
L84Supper: would you have to port the interpreter to ARM or could we emulate x86 for those times on the ARM?
bridgman: either... the drivers have complete interpreters in C so once the driver is up there's no problem anyways
bridgman: it's just if you want something on the screen before the driver is initialized
L84Supper: oh ok
bridgman: as long as you're running KMS the driver comes up very early... there's an atombios interpreter in the kernel driver already
Neo_The_User: bridgman: not just in DDX?
L84Supper: maybe libv's textmode for coreboot could be done for ARM?
cjb: Neo_The_User: the kernel driver can modeset, therefore the kernel driver speaks atombios.
L84Supper: we've been talking about Radeon with ARM for almost 10 years
L84Supper: now the ARM's are fast enough to try it out
DanaG: I was wondering about marvell's thingy...
DanaG: 1080p and 3d... at the same time.
Neo_The_User: AndrewR: booting up in 64m
cjb: DanaG: which one in particular?
DanaG: but no mention of whether that means 1080p WITHIN 3d.
L84Supper: many ARM soc's have hardware codecs, I have a ARM 11 mid that does 1080p playback just fine
DanaG: The new one, that was introduced as a prototype named "ebox".
DanaG: By "within 3d", I mean compiz, of course. =þ
cjb: sure, you'd expect that to work
L84Supper: it will be interesting to test and see where it bogs down
cjb: the 1080p will be coming over Xv or something
DanaG: I've heard that the beagleboard, on the other hand (omap3 and powervr) can't do compiz -- it's GL ES, not GLX.
bridgman: AFAIK things like snapdragon are basically radeon + arm
DanaG: wishes somebody would make an architecture called LEG
L84Supper: yeah the beagleboard OMAP has PowerVR for GPU
DanaG: Or a GPU.
DanaG: Then you could have.... yeah, a pun.
cjb: DanaG: oh, I'm pretty sure that Marvell part is GLES too
bridgman: so LEG would be graphics ? so you could have an ARM and a LEG ?
L84Supper: bridgeman: the radeon's for ARM were pretty trimmed down
bridgman: Low power Embedded Graphics ?
cjb: DanaG: I don't see why compiz would be so hard to run on GLES, though, probably just needs some work?
AndrewR: Neo_The_User, welcome back
Neo_The_User: hi there. got my KMS kernel running
spstarr: airlied: drm-linus / drm-radeon-testing conflicts
Neo_The_User: mem=64m is set
spstarr: im guessing stick to drm-linus at the moment?
Neo_The_User: just clone drm-radeon-testing
DanaG: bridgman: yup, you got the pun. =þ
DanaG: And added a viable name, too!
L84Supper: snapdragon - Improved 3D graphics - up to 80M triangles/sec and 500M+ 3D pixels/sec
cjb: DanaG: I guess the big question is whether we can implement GL_TFP inside ES
L84Supper: that about where PowerVR is
bridgman: we like backronyms at ATI, yes we do
spstarr: i'll stick to drm-linus until the two sync up
DanaG: You might need that old thing called "XGL".
spstarr: XGL == fail
spstarr: AIGLX == future
Neo_The_User: AIGLX is stable now IIRC
AndrewR: Neo_The_User, so, so far it works ok? For me just launching mozilla wwas enough .... or on 2nd test - just one Eterm.
spstarr: has been for over 2 years AFAIC
DanaG: hmm, but how would that jive with GL-ES?
Neo_The_User: im still in cli
DanaG: I'm not really up on how ES is different.
Neo_The_User: i need to finish building mesa and DDX
Neo_The_User: KMS isrunning very slow though. takes it 30 secs to read what im typing
L84Supper: the down side to Marvell with a FOSS project is their secret policies on docs, but devs would not need Marvell docs since PCIe and the ARM core is well documented
AndrewR: Neo_The_User, ok. Thanks for testing, anyway ... hopefully gcc 4.4. will work ok with 64 m, for me there was troubles, too ... with swap of cource
Neo_The_User: im suing gcc 4.4.3
Neo_The_User: sorry im making somany typoes. i cant read what im saying :(
AndrewR: Neo_The_User, 4.4.3 here too, slackware -current, so glibc-2.11 and not-so-recent binutils
L84Supper: how has ATI been with open docs for Radeon recently?
bridgman: we're wonderful ;)
L84Supper: we have been waiting for the AMD 780 coreboot patch for over a year
bridgman: actually those patches ended up on my desk last Friday
chithead: legal review started sometime around summer 2009 iirc
bridgman: that sounds right
bridgman: I think all the CPU stuff is done, then it bumped over to me for graphics
L84Supper: my crystal ball last year said that they would not be released until the 800 series ships
Neo_The_User: does memset=64 change the GPU ram usage or actual SDRAM usage? mesa is still on main/*
L84Supper: and nobody would want them anymore
L84Supper: we need 800 series docs now
chithead: yes, I listened to a talk of a coreboot developer who expressed concern that the code would not be released before the parts were eol
L84Supper: chithead: were you at FOSDEM?
chithead: no 26c3
L84Supper: probably Peter
L84Supper: we skipped a AMD 78x version fro a laptop since the port never arrived, was forced to use Intel :(
bridgman: yeah, I expected the 780 to ship a bit earlier than it did
AndrewR: Neo_The_User, mem= limit available memory, so you run like you have only 64 Mb of RAM (not gpu)
L84Supper: http://vr-zone.com/articles/amd-800-series-chipsets-launch-schedule/7577.html only a few weeks and the 7xx series is too old
chithead: well the existing stocks will not disappear overnight
AndrewR: Neo_The_User, main ide - kernel has less memory to play with - so bugs related to bad memory management will surface and bite you sooner
cjb: Neo_The_User: mem=64M makes the kernel pretend that it can only detect 64M of RAM in total. not much to do with GPUs.
DanaG: I do wish AMD had better stuff for laptops... if I were to have to make a recommendation for somebody, I'd be torn between wanting ATI/AMD for attitude reasons, versus wanting Intel for battery life.
bridgman: I think people are working on battery life...
bridgman: the worst part is that we're about the same efficiency when we're actually doing stuff
bridgman: the Intel CPUs just do *nothing* more efficiently ;(
bridgman: but modern CPUs do nothing quite a bit of the time so it matters
L84Supper: bridgman: any ideas on the best way to get 800 series docs into the open?
bridgman: yeah, we have to figure out how to make the damn thing work and then write the docs ;)
bridgman: modesetting code is out, and that has all the reg info
bridgman: shader instruction set is out
bridgman: I think Richard has the shader compiler changed to handle attribute interpolation in the shader rather than in the rasterizer
yao_ziyuan: i have a computer with a radeon 9200 card.
yao_ziyuan: opensuse 11.2, fedora 12 and ubuntu 9.10's open source radeon drivers don't support it well (can't play OpenGL games)
bridgman: we just don't have the pipe configured properly so we're still hanging on draw commands
yao_ziyuan: i wonder if the open source driver will support this old card again
yao_ziyuan: do open source driver developers focus on supporting new cards or old cards as well?
bridgman: both; until recently a lot of people said that the 9xxx was the best supported on open source
bridgman: did the open source drivers work better for you in the past ?
yao_ziyuan: opensuse 11.1's works perfectly
yao_ziyuan: later versions can't play OpenGL games due to a dmaXXXXX() function error
bridgman: ok, some people with 2xx cards reported problems with the fall distro releases; during the transition to kms a couple of things might have gotten broken
AndrewR: bridgman, i guess r200 (inc. my rv280) not very well supported in KMS mode ... i know few bugs, just not sure how to fix them ;) but "generic" instability with latest kernel worries me a bit more ....
bridgman: is there a bug report in that reflects the problem you are having ?
yao_ziyuan: i didn't file one
yao_ziyuan: i'm currently use opensuse 11.1
bridgman: AndrewR; I imagine airlied/agd5f know about the problems you are seeing ?
yao_ziyuan: google ["opensuse 11.2" "radeon 9200" supertuxkart]
bridgman: yao_ziyuan; I don't remember seeing a mention of that dmaXXXXX error, might be worth at least searching the bugzilla to see if there's one already entered and add one if not
yao_ziyuan: the first result is a bug report
yao_ziyuan: bridgman: http://lists.opensuse.org/opensuse-bugs/2009-11/msg07456.html
libv: L84Supper: no it cannot be done on arm
libv: L84Supper: first of all, it's in/out, secondly, it's VGA
libv: real, native IBM style VGA
AndrewR: bridgman, yes, bugs filled, but they have enough bugs now .. so i'm trying to understand things enough to fix at least something .... not sucessful with this yet
AndrewR: yao_ziyuan, "drm/radeon: Multipart buffer for parser to avoid memory allocation failures (bug 23993)" - something like this on dri-devel ...
yao_ziyuan: bridgman: if you tell me all the dmaXXXXX() functions i may recall which one it is
bridgman: I've never seen a dma
yao_ziyuan: it's a function
yao_ziyuan: a traceback said this function call failed.
yao_ziyuan: and crashed the game
yao_ziyuan: it's like dmaRadeonXXXXX()
spstarr: bridgman: I moved back to Fedora :)
AndrewR: heh, new patch .... "[PATCH] drm/ttm: handle OOM in ttm_tt_swapout" may be i nedd to try it ....
bridgman: sorry, long week; I was thinking XXXXX being a number, not a text string
AndrewR: reboot for testing
spstarr: looks at dri-devel ML
bridgman: yao_ziyuan; couple of questions
bridgman: have you already gone through a troubleshooting exercise (logs etc.) with one of the devs here
bridgman: and have you already updated to newer code for all the components, especially drm and mesa ?
bridgman: the closest bug on freedesktop.org I have found is 25141, looking at that now
bridgman: it's an rv100 but talks about a performance drop in the same time period
bridgman: nope, the log for that one indicated they were running KMS, I bet that wasn't what they wanted
bridgman: hello ?
bridgman: apparently talking to himself again
spstarr: he left
spstarr: watches Mens' Skeleton final, Canada could win medal
gentoofu: [22:09:46] yao_ziyuan [~firstname.lastname@example.org] has left #radeon
DanaG: hmm, glad to see the DPI issue is now getting coherent discussion: http://bugs.freedesktop.org/show_bug.cgi?id=23705 -- now I can just sit back and watch; people have already presented the things I'd have to say.
bridgman: missed that... thanks
AndrewR: running with mem=96m. KMS active. Mozilla ok .. at least for blank page ... loks like ttm patch was good
AndrewR: can someone add note in this bug http://bugzilla.kernel.org/show_bug.cgi?id=15276 about http://article.gmane.org/gmane.comp.video.dri.devel/43363 ?
AndrewR: :/ hold on ... AGP init failed here .... need to re-test
spstarr: GOLD MEDAL
AndrewR: guess Canada folks go outside to celebrate .....
cjb: spstarr: hm, don't do that. I'm watching the olympics on DVR.
cjb: (but fortunately don't care much about men's skeleton.)
DanaG: "Given that we still haven’t purged the RVxxx naming from our vocabulary, I’m going to stick with RV870 for this story. But for those of you who have embraced the new nomenclature - RV870 = Cypress and at points I will use the two names interchangeably. The entire chip stack is called Evergreen. The replacement stack is called the Northern Islands."
AndrewR: compiling yet another kernel .... on mem=96m
cjb: AndrewR: fwiw, that doesn't sound likely to succeed to me.
cjb: (oh, but I suppose swap would get you through it, slowly.)
AndrewR: build failed. But machine still alive. let me reboot it and finish build normally ....
DanaG1: that's a hell of an article.... you could in fact make the story into a whole book.
DanaG: "Northern Islands"
DanaG: Northern what islands? Do you mean northmost islands of any continents... or some specific continent? =þ
virtuald: or northern ireland?
DanaG: And how do you rent an aircraft carrier? =þ
DanaG: oh yeah, go to page 1.
virtuald: first i'd like to know how to get the money to rent an aircraft carrier :)
DanaG: Coupled with the GT218 GPU on the G210 is 512MB of DDR2 RAM, using the customary 64bit memory bus. Interestingly, unlike most other entry-level products, the G210 only comes in 1 memory configuration: 512MB. Even more interesting are the DDR2 chips on our MSI card – 4 chips we can’t identify bearing the ATI logo.
virtuald: amd pwning nvidia?