DanaG: When AMD started to make their own north- and south-bridges, I read an interview where someone from AMD was asked, why their south bridges didn't have an AMD developed sound codec.
DanaG: The answer was something like: "The devil is on the analogue side, and we have no plans on waking him".
DanaG: Good description. Digital logic, I can do... analog stuff, I've disliked, in all my EE classes.
DanaG: oh yeah, and the 2.6.31 doesn't do KMS on R600... is it supposed to?
DanaG: What I'd really, really love to see somebody do in audio: raytraced audio, like Aureal did (or I think they did).
fireun: go talk to the Ogg guys, someone there has already got something going I'm sure
fireun: #xiph, or #vorbis
bridgman: DanaG; no 6xx KMS in 2.6.31 - the KMS code is there AFAIK, but it can't be enabled without 6xx/7xx memory management
fireun: goes a little stir crazy, not having a gaming outlet
fireun: and UT at 8fps is decidedly frustrating
hifi: my counter-strike addiction required me to run windows until the radeon driver has some moar fps in it
fireun: hifi: I |really| hate rebooting into windows, but yes, I have DoD and CS as well
fireun: hifi: lately, with the fglrx driver I was running before mistakenly upgrading to Jaunty, I was playing UrbanTerror, which is very CS like (not as refined, but still playable)
fireun: a Q3A derived gfx engine, which shows some.
hifi: I didn't want to abandon my ubuntu so I set up dual head with synergy
hifi: instead of real dual head I was using before
hifi: I can play cs and IRC at the same time :)
fireun: I think I was doing that a decade ago with x2vnc
fireun: but I like the mac integration
fireun: might have to try it.. if it supports old enough macosx
fireun: I'm gonna have to put linux on my powerbook before too long, 10.3.9 is getting long on the grey
hifi: win-lin synergy has a lot of bugs but is enough until I get more juice out of my radeon
hifi: haven't really done any programming as it would require me to reboot into linux
fireun: and I see no point in installing snow leopard on a pb 500
fireun: yeah, I prefer the linux
fireun: even if it does paint me into corners like this xpress200 catalyst 9.3 boondoggle
hifi: my R500 was dropped out of fglrx, I have no choice
hifi: and I don't really want to stop playing
fireun: well that was my problem, I upgraded to jaunty which doesnt work with the last catalyst version that supported my card
fireun: I should have stayed in intrepid, with a working version of fglrx
hifi: I actually knew that before buying this card but back then I only played Quake 2
hifi: which works perfectly with the radeon driver
fireun: I bet your card has support in an older fglrx catalyst release
fireun: not to talk you into a non-free solution
hifi: I'm running karmic so I'd have to downgrade a lot of stuff for fglrx
hifi: but it's only a matter of a few fps for counter-strike to be playable
hifi: I get a nice 100fps in small areas, big areas drop it down to 30 or so
hifi: though wine had some opengl memory management improvements lately so it might help
fireun: wait, you play CS thru wine?
suokko: fireun: I have tried to but freezing too often. HL1 works well for hours
fireun: slowly I imagine
fireun: wine->karmic->radeon, yes?
fireun: my radeon 3d performance is, unacceptable
fireun: I cant imagine adding wine on top of it
suokko: Last time I played HL I was using intrepid with radeon 9200
soreau: Here testing nexuiz on jaunty locks the machine as soon as there's lots of gpu intensive interaction
suokko: But r200 cards used to have the best open source drivers for long time (and I still think old dri1 driver is very good for that hw)
fireun: has a IGP xpress200m
fireun: I thought I was getting the best linux supported chip, at the time
suokko: fireun: Have you tried to run the games with less visual settings? Low fps might very well be result of game trying to use gl features that aren't supported well enough. (Example large textures might be problem because mesa doesn't include compression because of patents)
fireun: suokko: oh yeah, turned UrbanTerror down to the mins, 15fps standing still, 6 moving
fireun: and jerks
fireun: yeah, compressed and uncompressed textures, 16, 32, and "desktop" bitdepth
fireun: down to 640x480 res
fireun: low texture details... ect
fireun: I get about 3000-3500 frames in 5 sec as per glxgears
fireun: I think I was getting more like 4-4.5k with radeonrewrite, or maybe it was the xorg.conf options I had at the time, hard to say.
fireun: (tried too many things in too little time)
fireun: [and none of it worked well]
ickle: any advice for tracking down the cause of "[drm:radeon_ring_test] *ERROR* radeon: ring test failed (sracth(0x15E8)=0xCAFEDEAD)"
suokko: ickle: agp card?
suokko: You could try radeon.agpmode=-1 in kernel boot option
ickle: might be.. it's a rv280 (radeon 9200) on a g4 mac mini
suokko: Should be agp card
suokko: btw, there might be endian problems
suokko: looks the code
ickle: endian problems in the display I don't mind -- but hopefully though not whilst programming registers :)
suokko: ickle: Problem is that your CPU and GPU are using different endian so driver has to take care of conversion
ickle: fails miserably to workout the bus hierarchy for the device
ickle: even in the cmd stream?
suokko: I don't know
suokko: I'm just making wild guesses
suokko: What kernel are you using?
suokko: self compiled I would guess
ickle: yeah -- first time for that machine
ickle: I'm still finding my way around the ppc options
suokko: You have 2 options: You can wait for airlied or glisse to wake up. Or you could try to hack it your self. You could try to convert the scratch register to little endian format
ickle: that's at least what I wanted to hear -- that's its fixable :)
nanonyme: Almost everything is fixable given enough interest and time? ;)
ickle: that it is fixable. ENOCOFFEE
suokko: I would think it is. But did you try to use agpmode=-1?
suokko: It might also be problem in agpmode
nanonyme: suokko: I heard Apple has a bit different AGP implementation than PC's do anyway.
nanonyme: Might mean for better, might mean for worse.
suokko: But agp driver should probably handle the difference
nanonyme: Probably. :)
nanonyme: I'm gonna try that too when I get home.
airlied: ickle: we have some reports of ppc working on a macbook.
airlied: ickle: but I haven't gotten my G5 with an x86 card in it to run yet
airlied: same error you saw there
ickle: at least I'm in good company ;-)
airlied: its on the list of things to investigate, just not on top :)
ickle: figures out yaboot
ickle: after appending radeon.agpmode=-1, drm appears to be successfully enabled :-)
airlied: ah I think MrCooper said he hasn't gotten agp working on his box yet
airlied: mine is PCIE
ickle: hmm, GTT 512M but I only have 256M of RAM?
airlied: GTT is the AGP aperture
airlied: or in PCI case
airlied: PCIGART aperture
airlied: we don't actually fill it with anything
ickle: ok, I was thinking i915
hifi: hm, when I boot 2.6.31 I think KMS is used and my screen goes blank
ickle: ok, like Sarvatt the fb is a subdued yellow
airlied: ickle: yeah there are fixes from Michel on the mailing list
airlied: he messes with the palette.
airlied: I'd prefer we used the hw swapping if we can
airlied: hopefully I can solve the PCIE/ppc issues this week
airlied: though it assumes I can get my ass into the office
ickle: cairo-drm bootstrapped onto the r200, wayland here we come! ;-)
airlied: ickle: jeez you pick the crappy cards :)
hifi: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch.
hifi: [dri] radeon.o kernel module version is 2.0.0 but version 1.17.0 or newer is needed.
hifi: what package is outdated, libdrm?
airlied: you really aren't doing any favours using shaderless hw :)
ickle: works with what he has lying around ;-)
airlied: hifi: libdrm with --enable-radeon-experimental-api and DDX
airlied: ickle: I'm sure someone can send you a cheap r600 :)
airlied: if you can get cairo-drm working on that I'd be actually interested in it :)
hifi: airlied: thanks
ickle: gah, I was getting my hopes up because of r200_fragshader.c
hifi: ubuntu karmic now has KMS enabled 2.6.31
airlied: ickle: it has a very limited frag shader
hifi: but the console stays blank after KMS is initialized
airlied: ickle: http://oss.sgi.com/projects/ogl-sample/registry/ATI/fragment_shader.txt
hifi: I had to blindly login and startx to get any picgture out
airlied: its not even up to i915 level
airlied: though it does have a vertex shader
airlied: but again its rather limited
hifi: airlied: do I need drm-next for that?
airlied: hifi: no, all stuff is in master in thoery
airlied: though its all a bit unstable until I fix some more bugs
hifi: ah, forgot to reset my git repo to origin
ickle: hmm, the vertex shader might be enough -- we only need rsqrt and friends for computing radial texture coordinates
hifi: airlied: also master DDX is now ok?
ickle: hmm, might want to change *ERROR* unimplemented radeon_gem_pread_ioctl to WARN_ONCE ;-)
phoenix64: is there any program available to show the temperature of a r500?
phoenix64: (wait, and is it "a r500" or "an r500"? :D)
_moep_: i the say the chip/the graficcard :D also male :D
nanonyme: phoenix64: The latter sounds more grammatically correct to me.
nanonyme: Not exactly sure though. :) *shrug*
nanonyme: (as in, while r is most definitely a consonant as a first character in a word, it'd sound to me as if it has a vowel in front of it if you say it separately)
phoenix64: and any answer to the real question? :D
phoenix64: <- no native speaker :D
nanonyme: Me neither.
nanonyme: And yeah, obviously I don't know if there is such a utility. ;)
twoerner_: agd5f: ping
suokko: hifi: do you load fbcon module?
suokko: And are you sure you don't load vesafb or radeonfb modules?
osiris_: does anyone know a fix for [drm:r300_cs_track_check] *ERROR* [drm] Buffer too small for color buffer 0 (need 1048576 have 524288)?
AndrewR: just want to say what my "tvtime hang machine under KMS" bug was fixed with latest xf86-video-ati. Thanks !
lesshaste: hi all
lesshaste: what does open /dev/fb0: No such file or directory mean?
lesshaste: it's the only serious looking line in my X log
suokko: lesshaste: That means you don't have KMS enabled in kernel but your xserver is trying to use it (if I'm not mistaken)
lesshaste: suokko, kms is hard to google :) what is that?
suokko: kernel mode setting
_Groo_: hi/2 all
_Groo_: im compiling the ddx fixes from today and all the new codebase, lets see if it mitigates my horrid bug
_Groo_: any dev alive?
_Groo_: MostAwesomeDude: ping?
suokko: _Groo_: It is generally better if you just say what you want to say and then wait that someone responds
suokko: Then you might find someone knowing the answer who you didn't except to know it
_Groo_: suokko: i just wanted to chat :D since MostAwesomeDude already knows what im talking about,..
_Groo_: suokko: but any radeon dev can do the trick ;)
_Groo_: reboot x, brb
osiris_: nha: ping
nha: osiris_: pong
osiris_: nha: could you take a look at 0de50a531d9b61880dedfb9c87e33965fe1a9fbc on shaders_cleanup branch (my private repo). there's a bug somewhere and I spend over 12 hours trying to find it, but failed
nha: what's the anon git remote for your repository?
nha: uh oh
nha: I'm afraid that's probably too much for my post-mountain hiking-tiredness state :}
osiris_: ok :)
osiris_: anyway, I'm going to push this branch to master when I fix the bug
nanonyme: airlied: Any idea if the KMS+mm stuff in 2.6.31 should be compatible with Fedora 11 userland?
nanonyme: Just wondering how much stuff I should expect to have to upgrade to newest from git. ;)
fir3_: how complete/fast is the radeon driver with an x800 atm? is it possible to play games like nexuiz/ut2k4?
terracon: r300 barely playes nexuiz/ut2k4 heh
soreau: Um, wrong
soreau: fir3_: I've been playing nexiuz on the open radeon driver here with my radeon 9600 and it supports your x800 too afaik
fir3_: nice, did you also try other games?
soreau: No, I don't know where to get any of the UT games
fir3_: http://www.atomicgamer.com/file.php?id=34669 ;)
osiris_: fir3_: yes ut2k4 is playable
fir3_: osiris_: with similar detail settings like on win?
osiris_: never played ut2k4 on win
_Groo_: hi/2 all
_Groo_: latest ddx commit appears to have fixed bug 22606
_Groo_: gonna do some more testing before closing it :)
_Groo_: could any dev take a look at why kwin3d stopped working since 2 weeks ago? :(
DanaG: hmm... is R600 KMS just not going to be in 2.6.31?
nanonyme: DanaG: That's a trick question as far as I've understood... The r600 KMS as of now isn't really very useful without the memory manager. ^^
DanaG: Ah. Without memory manager... KMS prevents you from using acceleration.. was that it?
_Groo_: freeorion still crashes dri, oh well [ 4146.043397] <6>freeorion: segfault at 54 ip b58a1ef1 sp bfbe7c90 error 4 in r300_dri.so[b584f000+23b000]
fir3_: osiris_: did you try any other games?
osiris_: warcraft3, frozen throne, world of goo
_Groo_: world of goo works very well with dri2/kms :) i just closed it using rs485
nanonyme: Wonder when the merge window for 2.6.31 closes/closed anyway.
_Groo_: i cant make wow work yet, still :(
fir3_: osiris_: all worked fine? did you use some dev version of the driver or the one with your distro?
osiris_: fir3_: I use current git version of mesa
osiris_: fir3_: yes, all worked fine
nanonyme: Safe to assume devs are using git unless otherwise mentioned? :3
fir3_: thanks :)
fir3_: guess i'll get a used x800 then
osiris_: I also played Painkiller, but recent wine is broken
nanonyme: osiris_: If broken means "not fully compatible with Windows programs", it's probably the nature of it... I've heard it's not exactly trivial to implement WinAPI even though the API docs are public.
osiris_: _Groo_: I tried playing trial wow - it starts, but then trying to login it complains about problems with wow version identification
nanonyme: (often have to still reverse-engineer to find out whether the outcome is the same as on real Windows)
DanaG: X800GTO(squared): http://www.techpowerup.com/articles/overclocking/vidcard/127
osiris_: nanonyme: I meant the game worked great with earlier versions of wine
nanonyme: I suppose someone with enough time and interest should do git bisecting then. (do such people exist?)
_Groo_: osiris_: use a private server, that appears to be a wrong realmlist
_Groo_: osiris_: i cant make wow to start here witn dri2/kms and latest wine
_Groo_: osiris_: it crashes and takes r300 dri with it
osiris_: _Groo_: try forcing backbuffer for offscreen rendering mode
_Groo_: osiris_: how?
nanonyme: Shouldn't that be default?
osiris_: nanonyme: FBOs are default (if available) because they are faster
nanonyme: osiris_: That's changed then.
_Groo_: osiris_: opengl or dxd9 mode?
osiris_: _Groo_: 3d with kms is known to be slower than with non kms
osiris_: _Groo_: opengl
nanonyme: osiris_: It used to be backbuffer still at least last year. I guess they finally got FBO bugs mostly squished.
_Groo_: osiris_: and fbos are working fine (at least with mesa tests)
nanonyme: _Groo_: Well, it can break at more than one point anyway.
osiris_: _Groo_: you haven't tested it enough ;)
nanonyme: That is, can have bugs with driver or Wine with FBO's.
_Groo_: oh btw, i see that xv for dri2/kms now has brightness/contrast (as per xvinfo), but its not working.. maybe the code isnt complete yet?
_Groo_: osiris_: yes i didnt, it was fixed just this week :)
osiris_: it's still broken. run fbo_firecube and try changing texture formats
bridgman: DanaG; I don't think anyone believes KMS/MM for 6xx/7xx is going to get into 2.6.31
fir3_: osiris_: do i need a specific kernel for the driver?
_Groo_: osiris_: let me check
bridgman: I haven't given up hope for 6xx/7xx non-mm 3D support though; it's a lot smaller and much less invasive
DanaG: What's weird is that, in one version, I DID have KMS on my R600. I just couldn't use acceleration on top of it.
_Groo_: osiris_: could that be the reason kwin stopped working? hes trying to use fbos?
osiris_: I don't know
bridgman: the KMS code itself has been around for a long time, it's just not much use without mm and all the other associated bits
nanonyme: Useful as in you might consider it lucky if you get it to start X at all. ;)
_Groo_: osiris_: yeah it crashed
DanaG: What's weird, is that it DID start fine for me. =P
bridgman: the problem is that there's no reason to test and fix it yet
bridgman: but the KMS code itself (excluding mm) should be very similar to 5xx
_Groo_: bridgman: hi john
bridgman: hi Groo
nanonyme: Yeah, X might start. Just not a good idea to rely on that it will. :)
_Groo_: bridgman: is there a workaround to enable kwin3d?
osiris_: _Groo_: kwin3d works for me pretty good with Xrender accel
_Groo_: bridgman: btw i have strange color bugs with firefox and some websites (like my own blog, lol)... and only shows using cairo, if i use qt/webkit it sghows fine.. i know its a radeon bug, because it triggered the bug 22606 i opened
_Groo_: osiris_: opengl mode
_Groo_: osiris_: it was working fine (always did) till 2 weeks ago
osiris_: _Groo_: yeah, opengl mode doesn't work
_Groo_: osiris_: it used to.. compiz still works tough
osiris_: _Groo_: try bisecting it
_Groo_: osiris_: arrrghhh not that again! i cant!! i tried.. i cant pinpoint wheres the error... i though it was mesa.. then i tried drm, and finally ddx... but its a strange combination of the three
_Groo_: osiris_: im just not skilled enough to find that particular bug... i suspect its more deep then just one commit.. might be a implementation change
MostAwesomeDude: _Groo_: Use git-bisect.
_Groo_: MostAwesomeDude: i did thta
_Groo_: MostAwesomeDude: i spent yesterday afternoon trying to pinpoint the bug
_Groo_: MostAwesomeDude: and i failed
_Groo_: MostAwesomeDude: first i thought it was mesa... but after a few compiles , ive seen that drm influences kwin3d behaviour also... so i went back bisecting drm, but i couldnt found exactly where is broken
_Groo_: MostAwesomeDude: i just found one build from 2 weeks ago that works, besides that.. nill..
_Groo_: hmm the cairo rendering might be associated with an old libpixman... compiling latest xorg and xproto libs as of now
nanonyme: Right, this time the 2.6.31 might actually compile in a sensible time. :P Scrapped several dozen extra modules out. :)
_Groo_: nanonyme: by any chance do you have an atheros wireless card?
_Groo_: nanonyme: ah ok
nanonyme: Broadcom one.
_Groo_: nanonyme: with 2.6.31, ath5k now starts in off mode, and it always started in on mode... so if i forget to switch it on via the flip i dont have wireless when i enter kde (until i turn it on that is).
nanonyme: Isn't that good?
_Groo_: nanonyme: nope.. the default was always on
nanonyme: That is, imo WiFi should by default be in powersaving mode unless you a) want to find a hotspot b) want to use a hotspot.
_Groo_: nanonyme: and i cant make the stupid rfkills to work... and it worked fine for .30 and below
nanonyme: Having WiFi on by default in a laptop sounds like wasting battery.
_Groo_: nanonyme: if it was always on before.. leave it on i say
nanonyme: Old incorrect behaviour shouldn't be fixed just because everyone's used to it? ;)
_Groo_: nanonyme: exactly :D
_Groo_: nanonyme: btw does brightness/contrast works for you in dri2/kms with xv?
nanonyme: _Groo_: I haven't yet gotten a system configured where KMS succesfully works.
_Groo_: nanonyme: chip above r500?
nanonyme: rv350 on PPC and rv670 on amd64.
nanonyme: Both combinations aren't really the best supported ones.
groo_: nanonyme: i have an rs485 (the evil one, the most shittye one)
groo_: nanonyme: you just need: master (drm, mesa and ddx) and latest linus tree.. there... kms/dri2 :)
nanonyme: groo_: Unless you've a ppc CPU, it's not that bad. :p
groo_: nanonyme: oh.. old apple?
nanonyme: As I said, the *combinations* are troublesome.
groo_: nanonyme: yeah, that sucks :)
groo_: nanonyme: but arent you appletars (just kidding), suposed to only used steve jobs approved goods and sex machines?
nanonyme: You make compromises when you've lost install discs and don't want to pay for maintenance.
groo_: nanonyme: lol :D but thats the apple way.. pay through the nose!
nanonyme: Well, it's actually better than with all other computers. ;)
nanonyme: Usually once warrantee is over, you're screwed.
nanonyme: With Apple it just means you have to pay.
nanonyme: And you can still get components for Apple machines sold five years ago.
groo_: nanonyme: im just kidding :) apple is fine, i would never buy one, but its ok if you like to pay extra
groo_: nanonyme: and dont want to have the trouble.. thats why apple charges more... all proprietary all the way, but assle free
nanonyme: The main thing though isn't the proprietariness but the limited variety in hardware.
nanonyme: Easier to support like ten components *really* well than hundreds of components in a satisfiable manner, really.
fir3_: groo_: i manage to make every OS unstable, including macosx(freezed more than once) ;)
nanonyme: fir3_: You're not the only one.
groo_: fir3_: same here...
groo_: my linux is always a work in progress :D
nanonyme: OSX isn't any miracle and the package manager is highly unstable.
fir3_: and the mouse acceleration in osx is horrible (you can't even turn it off)
nanonyme: Then again, package management in a safe way isn't being done in that many systems.
DanaG: fir3_: damn straight.
nanonyme: Try pulling the power from your Linux while doing a distro version upgrade and you know what I mean. ;)
DanaG: It's like it's stuck in a layer of molasses.
groo_: nanonyme: apple is good in locking everything up and making their products talk well with each other (logical since they control both hw and sw).. but besides that, its all proprietary shit
DanaG: Ironically, to get my mom's printer and scanner to work in OS X, I had to install gutenprint, foomatic-rip, hpijs, and SANE and sane-TWAIN bridge.
nanonyme: The recovery operation might require so much effort it's less effort to reinstall.
DanaG: All are things that are mainly for Linux.
nanonyme: If the power was cut in the middle of some particurarly important package.
DanaG: er, s/gutenprint/ghostscript/
groo_: DanaG: cups was bought by apple some time ago
DanaG: Yeah, but those OTHER things are not Apple-owned.
Plouj: who here works at AMD?
groo_: DanaG: yep.. but macosx is pratically a bsd with some stupid win manager on top
nanonyme: groo_: Stupid enough that its new and kinky features get ported to Windows and Linux? ;)
AStorm: darwin, not bsd
groo_: nanonyme: dont start with me... lots of 3d stuff appeared first in sun project and compiz... yes, apple inovates with a few features but thats it.. and windows copies from both linux and mac... so no news there
nanonyme: groo_: The 3D stuff is irrelevant.
groo_: AStorm: bsd.. darwin is the graphics layer... the kernel is bsd
groo_: nanonyme: irrelevant? is 99% of the apple "value" lol
AStorm: the kernel is Mach
AStorm: not any of BSD
nanonyme: groo_: The search bar and tray were the bigger innovations.
nanonyme: 3D isn't a big issue in OSX.
groo_: nanonyme: so, where did linux copied from? we had all that even in windowmaker.. its just wasnt that "fancy"...
groo_: nanonyme: apple pimped it up and sold it as "inovation".. which is balmers favorite word for everything...
AStorm: windowmaker is copied from NeXT
AStorm: which guess who invented
groo_: AStorm: true, but next was founded when jobs was KICKED from apple, so its not truly a apple product..
nanonyme: groo_: And I don't mean the Gnome-like tray, I mean the tray you see in the bottom part of OSX screen.
groo_: AStorm: it was suposed to be apple next big thing.. as usual jobs was right :)
AStorm: except the hardware, which wasn't
groo_: nanonyme: thats the one i was talking about.. old IBM workstations had one of those, it just wasnt that "fancy"
groo_: nanonyme: by fancy i mean 3d cutsy utsy..
AStorm: hey, it was more than 15 years ago
groo_: nanonyme: im old enough to know most if not all the "inovations" from major players (read, ms and apple) really existed in some form for more then 20 years!
nanonyme: groo_: Hmm, combined shortcut/program in use bar? Good for them.
nanonyme: It got to Windows 7 too finally, btw.
groo_: nanonyme: did you used windowmaker?
groo_: nanonyme: or OS/2?
AStorm: I find that bar funny with 1-button mouse
AStorm: and unwieldy if I start many instances of an app
nanonyme: groo_: Only vague recollections about OS/2.
groo_: nanonyme: the start bar wasnt even invented by MS lol.. it was a os/2 3.0 invention
nanonyme: groo_: Sure, of course not.
groo_: nanonyme: win95 that come out a year later used it.. and os/2 3.0 DIDNT had ms code anymore... the last one from both ibm and ms was 2.0
nanonyme: The start bar design sucks, I don't mean that part of it.
groo_: nanonyme: im not saying it sucks of not.. im saying the perception ppl have of both MS and APPLE "inovations" are highly misconceptions at best
nanonyme: I mean the being able to crop running programs as shortcuts into the bar so that a shortcut and program in use really *are* the same.
bridgman: yeah, given that pretty much everyone fed off the work done at Xeroc PARC
AStorm: they are not, really
groo_: nanonyme: now.. apple is good in USABILITY!!... bot not inovation. see iphone or ipod.. incredible machineries with lots of usability.. BUT they DIDNT invent anything
AStorm: a running session vs new session
AStorm: it's a broken design
nanonyme: AStorm: Yeah, well. But UI-wise.
AStorm: it looks nice, but is unusable
AStorm: if they split it into 2 parts, then maybe
nanonyme: Err, why?
AStorm: because it adds a problem when switching sessions of an app
nanonyme: (plus who the heck uses a Mac with a one-button mouse anyway, you can plug in a regular mouse into one)
AStorm: the same reason why I dislike that nested menu for "same app" selector in windows
nanonyme: (actually the UI seems to be designed for a two-button mouse)
DanaG: Win7 is spiffy: try setting it with small-icons and group-only-when-full.
DanaG: Oh yeah, Sun makes VM additions for OS/2 .... beats me why.
nanonyme: groo_: And yeah, true, iPhone seemed to have gotten tons of influence from eg Palm.
groo_: DanaG: thats because innotek was an os/2 company
groo_: DanaG: they did a lot of os/2 developemnt back in the day.. i should know.. i used os/2 for a decade...
nanonyme: still recalls the Gnome 3.0 visions of removing dropdown boxes and replacing them with a wheel ;D
groo_: nanonyme: try kde 4.3 ;)
nanonyme: If it's anything like I just mentioned, I probably don't want to. I consider the idea pretty much ridiculous. :)
suokko: btw, Why none is using cyclic context menu? Maya has been only useful application that I have seen using that kind of menus. (I have seen a few games using them)
suokko: I guess pie menu is the right name
nanonyme: suokko: It's probably exactly what I was talking about.
nanonyme: Been suggested for Gnome 3.0 for pretty much everything. (then again, Gnome 3.0 is in so much future that who knows what it'll have)
groo_: nanonyme: probably it will be another e17
groo_: nanonyme: or duke nuke forever :D
groo_: nanonyme: my favorite vaporware :)
Guest52147: I have a 3200hd
Guest52147: on a debian testing. can somebody help me to get xv and dualscreen working?
bridgman: do you know what versions of kernel, radeon driver and mesa driver you have now ?
Guest52147: a self compiled 2.6.30
nanonyme: Stock Debian testing seems to mean 6.9.0 of radeon and 7.0.3 of Mesa.
bridgman: ok, that should work; modular drm is strongly recommended, did you build the kernel that way ?
Guest52147: not at the moment
Guest52147: when i start x
bridgman: I don't know if you absolutely have to but it is strongly recommended
Guest52147: i get something like xorg deaktivated
bridgman: with the driver versions nanonyme posted you don't have a hope of it working properly
bridgman: pick up current radeon and mesa drivers
MostAwesomeDude: Shouldn't need Mesa for Xv and dualscreen.
nanonyme: Also X server 1.4.2.
bridgman: if you have already updated the radeon driver then pastebin an X log
bridgman: if you're running something old like 6.9.0 then get a new radeon driver first
bridgman: in the meantime you could pastebin dmesg output to see if the kernel driver is coming up ok
Guest52147: ok for that I need a little reboot
Guest52147: while my kernel is building drm module
Guest52147: where do i get the radeon drivers
suokko: Guest52147: You could new one from debian sid
Guest52147: ok and without confusing aptitude?
suokko: Guest52147: There is tips how you can keep mixed system in debian documentation
suokko: And you can download the packages and manually install
suokko: But if prefer compiling you can download source from http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/
Guest52147: I tried the compiling way, but with an error http://pastebin.ca/1484918 .
nanonyme: Guest52147: Try installing xutils-dev.
Guest52147: no that did not work
suokko: Guest52147: apt-get build-deb xorg-video-ati ?
nanonyme: suokko: Git radeon might have dependencies in X.org stack that are only available in sid.
suokko: It is also possible that debian testing has too old version of some dependecies
nanonyme: That's actually what I said.
nanonyme: Specifically probably xutils-dev is too old.
Guest52147: ok so a newer xorg is needed
Guest52147: xutils-dev 1:7.4+4
nanonyme: Well, that version number sadly says nothing to me since that's technically just X.org version number.
nanonyme: X.org macros 1.1.5 are required according to the bug report.
nanonyme: (and xutils-dev seems to provide them in Debian)
DanaG: Look at xserver-xorg-core
Guest52147: ok thanks for help. I'll try the sid packages and hopefully XOrg will come up again
cbmuser: hi, anyone knows whether it's sufficient to pull the current xorg-driver for radeon from git to get both DRI and kms working with ubuntu/9.10 with 2.6.31 ?
bridgman: depending on GPU, you shouldn't need to pull anything
bridgman: sorry, I missed kms, ignore previous
suokko: cbmuser: https://launchpad.net/~xorg-edgers/+archive/radeon-kms
suokko: That is all you need if your card is supported
cbmuser: I was talking about this issue:
cbmuser: [dri] radeon.o kernel module version is 2.0.0 but version 1.17.0 or newer is needed.
cbmuser: [dri] Disabling DRI.
cbmuser: suokko: I was using this repo up to a few days agao
cbmuser: it didn't work back then
cbmuser: so, I should give it another try ?
suokko: cbmuser: your libdrm is not build with --enable-experimental-radeon-api
cbmuser: ah, ok
suokko: One in raden-kms should be configured with that so it shoudl work from repo
suokko: If you installed everything from there
cbmuser: ... in progress ;)
cbmuser: ok, let me just reboot, brb :)
tormod: latest -ati commit has brought some font rendering corruption (RV410)
cbmuser: suokko: works, thanks
cbmuser: but there is some problem with the MTRRs, a friend discovered
cbmuser: the cpu caches 64MB of the graphics mem only instead of 128MB
suokko: cbmuser: You can thanks tormod for providing the packages ;)
cbmuser: is that known
cbmuser: when I switch back and forth to console, its somewhat slow
cbmuser: I can fix it with this however:
cbmuser: echo "base=0xc0000000 size=0x8000000 type=write-combining" > /proc/mtrr
suokko: Are you sure that card include 128M vram?
cbmuser: tormod: thanks for the packages
cbmuser: yes, it does
cbmuser: and the line I pasted speeds things up
tormod: cbmuser, you're welcome, but rather thanks Sarvatt for keeping it up to date :)
agd5f: groo_: the contrast/brightness/etc. Xv attributes only work when bicubic filtering is disabled.
agd5f: xvattr -a XV_BICUBIC -v 0
suokko: I think it would be worth making bug report to http://bugs.freedesktop.org/
cbmuser: Sarvatt: thank you as well ;)
cbmuser: suokko: you mean the mtrr-issue ?
cbmuser: in any case I will have to talk back with my friend who fixed it for me
suokko: yes. Unless agd5f knows about it inmidiatly
cbmuser: he is more into kernel-hacking thats why he discovered so fast
cbmuser: I showed him my problem and he immediately said "Yeah, that looks like an MTRR-problem, let me check."
cbmuser: and, *bamm* it was fast again ;)
suokko: cbmuser: I know the feeling :)
cbmuser: there is one kernel-version which consists of just his patch
suokko: cbmuser: Is that mtrr problem only with KMS or also with old driver?
cbmuser: suokko: I think it came with 2.6.31
cbmuser: I have to check again
suokko: But you should check if your friend has already reported it ;) He might have done it
cbmuser: it appeared after updating my Ubuntu 9.10 one day
cbmuser: suokko: nah, he didn't
cbmuser: he doesn't even have a radeon
cbmuser: he just found out by inspecting it
cbmuser: I will ask him to join once he is back on
Veerappan: tormod: still wondering about the font rendering issue?
Veerappan: I had the same issue until I updated mesa earlier
tormod: Veerappan, yes, I though maybe it was "some rendering seems to get dropped on the floor sometimes" from the commit
Veerappan: All I know is that between the major commits this morning to -ati and the mesa update, things are working pretty well for me now
Veerappan: watching a DVD right now to test to make sure that Xv is functioning correctly.
tormod: I am on mesa 86248807... I miss one r300 commit
Veerappan: my current: http://cgit.freedesktop.org/mesa/mesa/commit/?id=6fff62ee3fdbfe7d8ba15d3ad001f9afd120c307
Veerappan: that one seemed to fix it for me, but I also updated libdrm/kernel around the same time, so those might have fixed it as well
tormod: Veerappan, yes I am just missing that one, 6fff62ee...
Veerappan: ahh, try that out, and see if it makes it all better :)
tormod: I'll wait for Sarvatt to update the xorg-edgers packages :)
Veerappan: or that
Veerappan: umm... just rest easy knowing there's hope :)
tormod: Veerappan, thanks :)
Sarvatt: oh, did i forget to upload it? lol
tormod: Sarvatt, I was worried, no update in more than 8 hours ;)
Sarvatt: they dont update things on weekends usually so i stop checking :D
Sarvatt: uploaded now
tormod: thanks, I'll update again tomorrow, good night
fireun: jaunty bites - my printer stopped working too
cbmuser: suokko: still there ?
twice11: suokko: cbmuser sent me here, regarding the MTRR stuff.
suokko: I was just about to go sleep :)
twice11: I am a bit confused why the size of the MTRR range is the aperture size.
airlied: I thought we used to set it to the vram size, it might be wrong in the new codebase
twice11: As I understand it, the AGP aperture is a virtual memory area the card can busmaster to that can be remapped by the host.
twice11: It is a kind of IOMMU.
airlied: its not the AGP aperture
twice11: Ah, something different :)
airlied: its just a PCI MMIO address space
airlied: the VRAM is just made available to the CPU via it
twice11: Ah, I see. You have rdev->mc.aper_size and rdev->ddev->apg->agp_info.aper_size
suokko: airlied: Did you read earlied report where only half of vram was set to wc mode so performance was horrible? twice11 did found the problem and fixed it with manual mtrr change for cbmuser's computer
airlied: suokko: it sounds like the code misdetected the vram size then
twice11: cbmuser: What codenumber does you chip have? (I mean rv770 or the like)
nanonyme: airlied: Care to comment on the API compatibility question I had? (as in, should F11 userspace run Linux 2.6.31)
otaylor: nanonyme: worked fine for me (with the current rawhide kernel)
airlied: nanonyme: if you get the latest f11 -ati from koji it might.
twice11: Oh, doesnt matter. It always is "rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0);"
airlied: nanonyme: the F12 one definitely does
suokko: twice11: so I was hoping that you could write a bug report to http://bugs.freedesktop.org/ (DRI category) unless airlied can solve it now
airlied: the API is moving a bit so I'm not always keeping up with the compat
otaylor: nanonyme: err, actually, wasn't referring to the radeon driver, but to general compat in other areas. Haven't tested the radeon driver
twice11: OK. I will write or help cbmuser write the report.
cbmuser: twice11: M22/X300
airlied: otaylor: btw did you test that F11 kernel I added those intel patches to?
airlied: cbmuser: what kernel where you running?
suokko: thanks. I will go to sleep. good night
cbmuser: glaubitz@oslo:~> uname -a
cbmuser: Linux oslo 2.6.31-1-generic #14-Ubuntu SMP Thu Jul 2 16:03:01 UTC 2009 i686 GNU/Linux
cbmuser: Ubuntu 9.10
otaylor: airlied: no, off at guadec now, only my r300 with me.
airlied: so is that rc1 or previous?
airlied: otaylor: no worries
nanonyme: airlied: Right, sounds good. Then I might get something to actually work as soon as I figure out how to compile a kernel on this build target. :)
airlied: cbmuser: can you pastebin a dmesg from kms on it?
cbmuser: airlied: ok
cbmuser: airlied: d #radeon
cbmuser: 01:23 < cbmuser> airlied: ok
cbmuser: sorry, wrong paste
cbmuser: airlied: at the end of the dmesg you'll see that I also get the errors with the "command stream"
cbmuser: a few weeks ago, these messages usually meant I would have to reboot
twice11: Can you check what lspci -v says about yout graphics card and paste that too?
cbmuser: because X slowed down dramatically, always happened when surfing with firefox
twice11: It should output a some memory area lines. I think I checked the VRAM was 128M in it.
twice11: Should look like this line from my computer:
twice11: Region 2: Memory at e0000000 (32-bit, prefetchable) [size=256M]
cbmuser: Memory at c0000000 (32-bit, prefetchable) [size=128M]
twice11: Yeah. That's 128M
twice11: Looks like airlied is right. If radeon_vram_info called by radeon_device.c:550 returns only 64M, the used aperture size is limited to 64M, and thus the MTRR.
airlied: its wierd though the driver claims only a 64MB bar size as well
airlied: as only 64MB RAM
twice11: No. You overwrite the bar size in line 555
twice11: If the bar size was bigger than the VRAM size.
airlied: ah I need to check that, I suspect its just as wrong as the previous change
twice11: I think that the problem is that radeon_vram_info returns the wrong size for cbmuser's card.
airlied: cbmuser: do you have a log from X without KMS?
cbmuser: should have, wait a second
twice11: OK. The M22 is called rv370
airlied: my m22 only has 64MB of RAM
airlied: but you can get them with 128MB
cbmuser: airlied: http://pastebin.com/m3e824c1f
cbmuser: oh, wait
cbmuser: think its wrong
cbmuser: this one's with kms already
Veerappan: not sure if it helps any here, but my lspci -v for my 128MB x300 shows: Memory at d0000000 (32-bit, prefetchable) [size=128M]
cbmuser: airlied: I would need to reboot w/o kms
cbmuser: can you wait like 15 minutes ?
airlied: thats only the aperture, not the actual VRAM
airlied: no worries I'll still be here.
Veerappan: I do have a 64MB x300 mobility in a box in the closet I can swap into my laptop, but I'd prefer to avoid it... but if really needed, I can troubleshoot
airlied: my x300 is just sitting on the floor behind me.
twice11: Ah. Could it be that the X server does use the second half of the aperture even if there is only 64MB of video RAM?
airlied: with kms it doesn't get to see the aperture
airlied: thats what I don't understand
airlied: if we report 64MB VRAM we sohuld only try and use the first 64MB of it
twice11: cbmuser: When you experienced the slowness: Was the X-Server in kms mode or in user-level-modesetting-mode?
airlied: though I suspect safest thing is to set the mtrr across the whole aperture
airlied: thats what my old kernel did
cbmuser: twice11: I think kms only
cbmuser: I will check in a few minutes
cbmuser: ok, brb w/o kms ;)
cbmuser: airlied: xorg.log w/o kms: http://pastebin.com/m364992ab
cbmuser: twice11: switching is as usual when not having kms, i.e. the usual delay
cbmuser: but not slow redraw like with kms when switching to X
twice11: But image appears at once, not from top to bottom?
cbmuser: I'll check mtrr now
cbmuser: got ya:
cbmuser: glaubitz@oslo:~> cat /proc/mtrr
cbmuser: reg00: base=0x000000000 ( 0MB), size= 2048MB, count=1: write-back
cbmuser: reg01: base=0x07ff00000 ( 2047MB), size= 1MB, count=1: uncachable
cbmuser: reg02: base=0x0c0000000 ( 3072MB), size= 128MB, count=1: write-combining
cbmuser: without kms, the mtrr-size is correct
twice11: Yeah. The X server says 64MB RAM, but 128MB address range.
airlied: its wierd that the mtrr not covering the whole aperture is messing up
airlied: since we never report back anything from the second half
twice11: I can testify that I've seen it on cbmuser's machine. Switching from console 1 (drmfb) to console 7 (X server) took around 0.05 to 0.1 seconds to redraw the X screen with MTRR=64MB, but was fast with MTRR=128MB.
airlied: wonders if the fb mapping is doing something wrong
twice11: Switching from X to fb was fast
airlied: but I can believe we need to fix it so it works like previous driver and like my F11 kernel
airlied: since its the path of least surprise
airlied: ah I see some other bugs in there as well I think
DanaG: yargh, xorg can be irritating... when is Xorg going to be able to handle keycodes > 255?
EruditeHermit: anyone have ubuntu working with kms from edgers repository?
EruditeHermit: also on non KMS when I run glxgears, I get the message rrb is 0x94826b8 0 300x300
EruditeHermit: a million times in the console
airlied: oops I left some debugging in there
airlied: hmm it should be under a debug block
mpyne: would anyone happen to know what the following DRM error is about? [drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(2).
mpyne: with the occasional "[drm:radeon_bo_move] *ERROR* CP is not ready use memcpy."?
EruditeHermit: airlied: did you fix it in master?
EruditeHermit: err will you fix it in master
airlied: EruditeHermit: I can't see it in master anymore
airlied: ah MrCooper already fixes
EruditeHermit: 11cf8a28 still had it
EruditeHermit: which is the last commit I see
airlied: my master is at 645bfa
airlied: sounds like a local issue
EruditeHermit: oh I was looking at cgit.freedesktop.org
EruditeHermit: is that not updated yet?
airlied: it can get stuck sometimes
EruditeHermit: well I'll check back in a few hours
EruditeHermit: but if its just debug info, its not a big deal anyhow
airlied: EruditeHermit: the bug is in nemsa
EruditeHermit: is MrCooper Maciej Cencora
EruditeHermit: the last commit I see is from MAD
airlied: the fix was a while ago
airlied: like 2-3 days
EruditeHermit: guard debugging output?
EruditeHermit: well let me update mesa
EruditeHermit: sorry for wasting your time
EruditeHermit: airlied: one more question before i go test
EruditeHermit: is suspend/resume broken in kms or is it more a problem with distribution scripts having not been adapted to a kms system yet
airlied: EruditeHermit: it could be either
airlied: it works on my laptop at least, so distro scripts do need fixing
airlied: in theory pm-suspend --quirk-none should bypass distro hacks
EruditeHermit: i'll try that if I can get my kms system up
EruditeHermit: it was working and then it started segfaulting on X server startup recently
EruditeHermit: and when it does work, the icons in the minimize maximize, close buttons in the window manager top bar don't appear
EruditeHermit: gtk decoration or whatever its called
EruditeHermit: window decoration
EruditeHermit: ok off to test brb
mpyne: oddly enough I also had issues with window decoration icons, but with KWin...
EruditeHermit: the messages from running glxgears went away atleast
EruditeHermit: well my kms system won't even boot anymore
EruditeHermit: kms kernel
mpyne: I haven't updated it since I got it working. ;)
mpyne: I have checked drm-linus just to make sure
mpyne: but I haven't seen any kernel fixes since before like a week ago
mpyne: I can't actually use KMS though, as DRI2 doesn't work since sometime last week for me.
mpyne: CS fixes recently got the DRM error message to change at least so now I get the same error for root or non-root
EruditeHermit: oh when kms doesn't segfault X, DRI2 was working
mpyne: but still no love
mpyne: KMS is teh hotness
EruditeHermit: well good news is that its changing on a daily basis
mpyne: but when it was working my desktop was uber-slow, even with compositing disabled
EruditeHermit: teh hotness would be gallium 3d
EruditeHermit: I want openGL 2.1 =p
mpyne: I disabled DRI/DRI2 on a hunch since I had DRM errors after my next-to-last update caused X server b0rkage
mpyne: and everything was magically like 10x faster, in addition to booting at all
EruditeHermit: oh what X borkage?
EruditeHermit: did X fail to start?
EruditeHermit: for me thats the problem right now
mpyne: X would freeze hard
mpyne: only sysrq-magic keys worked (which was at least nice)
mpyne: then X would work but I couldn't see anything that was going on, I guess once the CS buffer-majig filled up?
airlied: I'm working on fixing the CS buffer at the moment
mpyne: oddly KDE 4 managed to last much longer (~ 5 s) than fluxbox (instant update moratorium, didn't even get a background)
airlied: KMS support in master is still in development
airlied: kms-support is much more likely to succeed
mpyne: airlied: thanks, I for one appreciate it (assuming that's the cause of my problem)
mpyne: ok, that's when the error happened too, was when I switched from kms-support to master since it looked (to me at least) that further development was happening in master
mpyne: I'm content since I have a working desktop again to just wait and test improvements as they happen
EruditeHermit: but master stuff works better for my non kms kernel which I normally use
EruditeHermit: its good to see it all in master branches
airlied: further dev is happeningi n master
mpyne: getting 3d someday would be cool too but I guess we'll see. :)
EruditeHermit: we can see the light at the end of the tunnel
airlied: it just not the same codebase
airlied: so it needs a lot more further dev
miickee: I have debian 5.0 (lenny) installed, and I want to update my ati drivers for a radeon HD 3450, would this debian wiki and packages described within (http://wiki.debian.org/ATIProprietary) be suitable?
airlied: #ati for fglrx q