amarks: airlied: rv790, drm detected vram 256MB
zhasha: dmb: gotta love old software "Christmas has been detected. Activate Christmas mode? (Y/N)"
DanaG: Even better is old software that uses busy loops for timing.
dmb: dos software did
DanaG: I once tried SimCopter on a modern computer... to get it to run, I had to use MS VPC (S3 Trio at least does the required DirectX 3(!)), and ODCM throttle the host down to like 500 MHz.
dmb: but there was nothing else
DanaG: That was a Win95 game.
dmb: DanaG, dude, i have been trying to get that to work for AGES!
DanaG: Oh, and it needed guest to be Win98.
zhasha: my rear weapon in tyrian has changed to shooting the texture "NOEL"
dmb: for some reason, graphics are always screwy
DanaG: hmm, is that in Wine, or what?
dmb: DanaG, it won't work in virtualbox or vmware?
DanaG: No Win9x driver for VBox, and VMware won't do DirectX, will it?
DanaG: Unless newer vmware has fixed that.
DanaG: It needs DirectX 3, at the very least.
dmb: DanaG, why can't it just do software directx?
DanaG: Hmm, you could try it in one of those things, with win9x.
DanaG: But the default VGA driver didn't work for it.
DanaG: Oh, and the thing that's affected by busy loops is the idle wobble on the helicopter.
DanaG: Instead of lazily wobbling a little bit, it goes WooOoOOobbBBbBBlLLlleEE all over the place like crazy.
dmb: DanaG, it was working a little i think, but fonts were missing, and textures were all screwy, and it kept crashing
dmb: this was with win9x
airlied: glisse: http://fpaste.org/Biwh/
dmb: what a poorly coded game
DanaG: It did still often crash for me, though.
airlied: glisse: still happens even after object rework, but hard to reproduce
airlied: I suspect memory corruption somewhere else
DanaG: It even crashed with native winxp.
airlied: topic dudes please
dmb: its so hard on irc to keep topic :)
MostAwesomeDude: I gave up on /topic for now; went back to paid code. :T
airlied: not if i kick you off, this isn't #efnet
DanaG: doesn't know #efnet. =þ
DanaG: Anyway, that topic is done now. =þ
MostAwesomeDude: I think we're miles better than efnet. At any rate...
zhasha: MostAwesomeDude: what do you get paid to code? also, how's shatter?
DanaG: oh yeah, how did that conversation even start? oh... pci bandwidth -> old software "christmas mode" -> old software -> busy loops. Right.
MostAwesomeDude: zhasha: I get paid to work on Pydra, for OSUOSL.
DanaG: Gotta' love the way the human brain works.
MostAwesomeDude: zhasha: Shatter will get moar love over winter break.
DanaG: googles pydra.
MostAwesomeDude: DanaG: http://pydra-project.osuosl.org/
MostAwesomeDude: It's a distributed computing setup.
zhasha: winter break is in several months
MostAwesomeDude: zhasha: My winter break is technically started, although I won't get full-bore into coding mode until next week.
zhasha: oh I confuse myself. In Denmark we have Christmas break which starts around the 22nd and lasts for about 2 weeks, and winter break which is usually in week 7
MostAwesomeDude: We have spring break around that time.
zhasha: that would make sense except it's winter till like half way through March here
zhasha: and after that it's monsoon season, then straight back to winter
MostAwesomeDude: It's traditional in the USA for people to go down to Mexico for that break.
MostAwesomeDude: So they'll go to Cancun and Tiajuana and get wasted, and then come back.
zhasha: but anyway, shatter, the ultimate goal is to kill the giga root pixmap?
zhasha: and how far off is that from now?
MostAwesomeDude: A ways.
zhasha: but supposedly you are making EXA work on a per-screen scale?
MostAwesomeDude: Well, just making EXA split rendering into shards.
zhasha: will it have any benefits except the obvious memory reduction?
MostAwesomeDude: Paves the way for getting rid of Virtual.
zhasha: oh right, I completely forgot that retardation was still in there
zhasha: I just haven't had to worry about it ever since I started using KMS because I can't render >8kx8k
Nightwulf|work: hi all
glisse: airlied: what do you do to trigger it ? Use little ram and force swapping ?
airlied: glisse: just use my PC ;-)
airlied: I have 6GB of RAM
airlied: so I'm not sure why it happens, but firefox scrolling around on a few big web pages + emails + lots of terminals seems to trigger it
airlied: I was thinking of adding a debugfs hook to just force the shrinker to be called
airlied: so I could try and trigger it more often
glisse: definitly need some way to easily test such path
glisse: airlied: i don't see what could be wrong as radeon_bo ptr should be valid otherwise the fault would be in ttm or in radeon_bo_check_tiling
glisse: it seems that this is the callback clear_surface_reg which is wrong
glisse: at least if am doing proper math with oops message
airlied: glisse: yeah its like the asic structure is corrupted or something
airlied: or rdev is invalid
airlied: which if the ttm is in a radeon_bo container might happen
airlied: not sure how that wuold happen
glisse: sounds like we need ugly debug code to check if bo is one we know and to compare rdev ptr against staticaly stored one
airlied: stick a magic number in the radeon bo
airlied: and just check against it
airlied: for debugging purposes
glisse: yeah was thinking to that too
glisse: also we are not passing gem file for swap storage
airlied: yeah we should probably hook that up as well
laumonier: someone could help me?
laumonier: i have a problems with my mobility radeon x1700 on linux i do not find the right driver
laumonier: which version of catalyst must i download?
cornair: I thin k you should ask another channel for that use fglrx
cornair: I mean #ati
laumonier: oki thx
kdekorte: Is the r700TryDrawPrims line 965 warning known?
kdekorte: Sorry line 955
gnuman: Hello, please is basic power management already in KMS or do I have to still use UMS for power saving (ForceLowPowerMode, DynamicPM and ClockGating)?
adamk: gnuman: You still need UMS.
gnuman: adamk: Is there some timeframe where one can expect it? This is important issue for people with laptops...
BioTube: gnuman: there's a patch, but it needs editing before it applies
adamk: No idea. I'm not a developer. There are still a number of items that need to get fixed in KMS before it can replace UMS, frankly.
gnuman: I know there were some patches in september (just basic infrastructure, but I have hoped that they will evolve). They are not in Git yet?
gnuman: Ah, I have found it: http://www.phoronix.com/scan.php?page=news_item&px=NzUyOA (In-Kernel Power Management For ATI KMS).
gnuman: So these patches hasn't been accepted?
BioTube: gnuman: no
gnuman: BioTube: And you mean these patches by your previous comment or are there any other?
BioTube: gnuman: well, the basic infrastructure got accepted, but not the code to actually implement it
gnuman: BioTube: So where I can find that patch you have talked about before?
BioTube: but, like I said, it won't apply cleanly without editing
rah: agd5f: https://bugs.freedesktop.org/show_bug.cgi?id=25520
adamk: rnoland: What r6xx/r7xx card(s) do you have?
rnoland: 3650, 3850, 4650 iirc.
adamk: Have you benchmarked openarena with the anholt demo on on the 3650?
rnoland: adamk: no, i have the 4650 in now...
rnoland: i don't have anholts demo...
adamk: Alright, I'm going to test something quickly... I'm seeing about 20 fps difference between FreeBSD and Linux on that demo.
adamk: I am using KMS in Linux, though, so I'm going to reboot with KMS disabled and see if that explains any of the difference (which is roughly 50 fps on FreeBSD and 70 on Linux).
adamk: I haven't noticed this at home, so I'm thinking this might just be something with my setup.
adamk: Alright, so KMS does seem to make the difference. Maybe I don't see it at home because the 4350 is faster than the 3450 that it's not noticeable.
rah: using drm-radeon-next or drm-radeon-testing with KMS locks my machine while loading a firmware file
rah: is there a missing file, or is this a bug?
BioTube: the file's only on the mailing list(something about licensing)
adamk: One would think that the machine shouldn't lock just because a firmware file is missing.
DanaG: Yeah, it should either fall back to no-interrupts, or at least refuse to finish loading the module.
rah: BioTube: which mailing list? what's the subject of the email that contains it?
BioTube: rah: the dri-devel list; I forget the subject
DanaG: google for R600_rlc.bin
DanaG: or R700_rlc.bin
rah: DanaG: thanks
rah: this is really bad, btw
rah: this shouldn't happen
rah: looks at agd5f
DanaG: Looks like it needs a different failure mode.
adamk: agd5f: Thanks for the patch. I should be able to test it this evening.
hifi: so DGA doesn't work with KMS/DRI2, right?
hifi: (the input bits, not video)
adamk: Last time I tried it, it did not work. Not sure if there's any plan to change that.
cxo: brrr il fait cold... today was the first semi-proper snow day
hifi: still waiting for the final snow that stays
cxo: Can you believe it, 1 week into december and we aren't covered with the stuff yet
hifi: global warming!
cxo: yeah its not too bad living in Canada, but it must be a nightmare for people on the equator
agd5f: hifi: correct. DGA assumes a fixed front buffer pointer
agd5f: with kms the front buffer may move
hifi: agd5f: so is it completely killed for KMS/DRI2, forever?
agd5f: hifi: you can enable it, but it's dangerous
hifi: what about the input part only?
rah: I still get a hang with the _rlc.bin files in /lib/firmware/radeon
agd5f: hifi: they are coupled together
rah: do they need to be there during compilation?
hifi: my understanding is that DGA is only used for input these days
agd5f: rah: no, but you'll need to add them to your initrd is youa re using one
hifi: and not really for graphics access
rah: ah ok
agd5f: hifi: DGA supports both, most apps don't use the direct framebuffer stuff anymore
hifi: would it be a problem to have it enabled and working without the framebuffer stuff, by default?
hifi: if the framebuffer stuff is the dangerous part
agd5f: hifi: yes, the framebuffer stuff is the dangerous part
agd5f: hifi: you'd have to change the DGA code in the xserver
hifi: will the DGA framebuffer be ever supported on KMS? I spitted out on Wine appdb that KMS/DRI2 killed DGA for good with radeon, and intel dropped DGA support too if the xf86-video-intel changelog is correct
glisse: airlied: with which GPU did you produce the oops with swap ?
glisse: i have been trying all day to trigger it or find out anythings wrong in the code and nothings popup
mjr: dga is pretty much in the past, as far as I can see; if you're interested in input, xinput2 is the way one should be looking at for funkier-than-generic X input nowadays
glisse: i will update a patch to fdo that you can use, it will mostly dump info in front of a ill radeon_bo structure
hifi: mjr: there is a little debate on Wine bug 6971 about how to solve some of the mouse issues and DGA was promoted for a quick fix, but as it turns out it's going to die along with KMS
hifi: agd5f: other source tells me the graphics part was dropped from Xorg completely some months ago?
agd5f: hifi: not that I know of
hifi: in september
hifi: agd5f: http://lists.freedesktop.org/archives/xorg-announce/2009-September/001049.html
hifi: Perhaps the most intrusive change is the removal of framebuffer support from
hifi: airlied should know about it, he's one of the ones talking about in the relevant post
agd5f: hifi: not quite. it just moves responsibility from the driver to the server
agd5f: responsibility for calling dgainit
glisse: hifi: anyway i think bottomline is DGA is dying and it should not be use anymore for whatever reason
glisse: xinput2 should allow you to do what you want
hifi: but it's not complete yet? or is it?
Zajec: glisse: could you check once again thread about PM/modesetting, please?
Zajec: glisse: i don't 100% get your idea about that functions you mentioned
Zajec: glisse: not sure about placing them (well, just check my mail :) )
Zajec: watching Gallium videos :)
glisse: hifi: i think it's complet, Peter posted tutorial on his blog
hifi: oh, nice
glisse: airlied: http://people.freedesktop.org/~glisse/0001-drm-radeon-debug-swapout-V2-without-TTMCHANGe.patch
spstarr: looks at git and ml
Jonet: Hey guys, I got a small problem
Jonet: My two monitors isn't separated over VGA-0 and DVI-0
Jonet: When doing xrandr --output DVI-0 --left-of VGA-0
Jonet: Instead of my right monitor extending to my left, both monitors get's the same offset and extends a monitor on the right that doesn't exists
Jonet: If you understand my problem
adamk: Jonet: Pastebin the output of 'xrandr --verbose'
adamk: I had that problem when both outputs used the same CRTC.
Jonet: Oi, it's you again! :D
Jonet: Btw, adam. Thanks for last time, radeon drivers works flawless and with decent GLX acceleration! :)
adamk: Jonet: Yeah, they appear to both be using CRTC 0
adamk: No problem!
adamk: I have no memory of our discussion, but I'm glad I could help before :-)
Jonet: How do I change that?
Jonet: It was on #ati, i had issues with fglrx on a r420 card. Wich apperantly was unsupported
adamk: Try 'xrandr --output DVI-0 --crtc 1 --left-of VGA-0'
adamk: I *think* that will work.
Jonet: Nope, DVI-0 was turned of instead
Jonet: Is there a way in xorg.conf to set this?
adamk: Alright, let's swap... 'xrandr --output DVI-0 --crtc 0 --left-of VGA-0 --output VGA-0 --crtc 1'
Jonet: I noticed in the configuration that only one monitor was identified
spstarr: looks at the r6xx changes in DRI
adamk: Yes, you can specify both monitors in the xorg.conf and place them right-of or left-of each other. Not sure if you can force a crtc, though.
Jonet: Now it turned of my right monitor instead
Jonet: What exactly is the crct?
adamk: Jonet: I may be in over my head on this one, then. I'm not actually sure what the acronym means.
DanaG: crt controller?
Jonet: Well, I don't need to know the acronym. Just what it does basicly
adamk: Jonet: However, here's the xorg.conf file that I use to define two monitors and place VGA-0 on the right: http://pastebin.com/mc6f52fd
DanaG: Why's it still called that even when it's used on DVI? =þ
adamk: Jonet: As I understand it, it's the connector on the video card that talks to the monitor(s).
adamk: Jonet: Nearly all readeons have two CRTCs, even if they have more than two outputs.
DanaG: You have the stuff that creates images and the stuff that pushes them out to the displays... crtcs are the latter.
Jonet: Ah right, I get it. So basicly crtc1 is the DVI connector or so :)
Jonet: I get it
DanaG: Not quite... the connector is a physical thing, and you can route CRTCs around to different connectors arbitrarily. Some cards have fixed CRTC to connector mappings, right?
adamk: DanaG: I believe so, yes.
adamk: Jonet: You could try the xrandr command I gave above, but swapping the 0 and 1 for each port.
adamk: Though I'm not convinced that will do anything other than turn off the other monitor
DanaG: I sometimes wonder why my old 9800 Pro had a "(Secondary)" output in Windows, yet my modern card does not.
adamk: DanaG: That was necessary for dualhead support on WIndows 95/98/ME, as I undersand it.
adamk: DanaG: Your modern card doesn't care if it works in those OSes anymore :-)
Jonet: My GP has Primary and Secondary, atleast on Windows 7 :P
DanaG: Aaah. And W2K did things differently from WXP.
adamk: Jonet: Yeah, but the card doesn't have two entries on the PCI Bus. That's the difference.
adamk: It may show up in the Device manager as primary and secondary, but not on the PCI Bus.
Jonet: I changed my xorg.conf and added a new Monitor Section. Since both screens on 24" are identical the settings are the same.
Jonet: I'l try some stuff with xrandr and see what i get =)
Jonet: Thing is that DVI-0 on xrandr --verbose says "CRTs 0 1"
Jonet: While as VGA-0 says "CRT 1 CRTCs 0 1"
adamk: DVI-0 shows "CRTC: 0" "CRTCs: 0 1" just as VGA-0 does... At least from the xrandr you pastebin'ed.
Jonet: No no, after i set VGA-0 to crtc 1
adamk: Yeah, this is definitely over my head.
Jonet: That's the output after your command there
adamk: Yeah, no CRTC for VGA-0.
airlied: glisse: 3600 of some sort, it might be due to my 6GB RAM machine (I'll run memtest on it ;-)
Jonet: Well i did --crtc 0 on DVI-0 and --crtc 1 on VGA-0
Jonet: So it should be different CRTC
Jonet: Thing is it doesn't seem like crtc 1 is getting any output : (
DanaG: GP? What's GP?
Zajec: glisse: thanks
DanaG: (10:59:23 AM) Jonet: My GP has Primary and Secondary, atleast on Windows 7 :P
Jonet: GP should've been GC :P
Jonet: Graphic Card =)
edgecase: ok i haven't researched this yes, but i have some drawing glitches on R100 with Ubuntu Karmic, anyone heard of that? KMS is disabled, should I update and turn it on?
airlied: adamk: alex fixed coherent mode so thats one less thing kms needs for you ;-)
DanaG: what card?
Jonet: Mine? It's a Radeon X850 Platinum
DanaG: I have a Mobility HD3650 (current card), a GeForce Go 7600 (never much liked nvidia drivers), a 9800 Pro (oldie but goodie), and a Radeon 7500.
adamk: airlied: Yeah, I have the patch. I'll give it a shot this evening and close the bug. Now if only I could disable it with an option to the kernel/module :-)
edgecase: anyone tested on 7500 mobile? am i crazy to try KMS on that?
edgecase: put another way, what is the *best* card to try KMS on?
adamk: edgecase: I've used it on various r1xx, r3xx, r4xx, r5xx, r6xx, and r7xx cards.
adamk: edgecase: I've had various problems, some serious, but most have been resolved by this point.
edgecase: ok so there's a chance it will work. I think i'm too lazy to take it on today, i'll dry disabling XAA stuff and see if my glitches go away
lordheavy: running drm-radeon-next (or drm-next) but it looklike fine here
spstarr: builds mesa git master
jonrafkind: How can I get an opengl2.0 implementatino? does mesa not support it yet?
stikonas: jonrafkind: there is very experimental opengl2.0 implementation for R600/R700 cards
jonrafkind: oh, i have r300
stikonas: then you will have to wait untill gallium3d driver matures
jonrafkind: ok, any idea when that is? 6 months / 5 years?
eosie__: jonrafkind: OpenGL 2.1 is supported in the R300 Gallium3D driver, but the driver is far from complete
stikonas: 6 months looks reasonable
jonrafkind: oh? ok cool
MostAwesomeDude: stikonas: Oh, cool, you're proposing deadlines. I take it that you're going to contribute some patches? :3
stikonas: I'm not saying that it will be done in 6 months
eosie: jonrafkind: it depends on how much time devs will devote to it
jonrafkind: of course
eosie: MostAwesomeDude: do you have any clue about what might be the cause of weird colors on the framebuffer background? (e.g. glean/teapot) the state tracker uses clear_with_quad here
glisse: airlied: yeah likely somethings like that :) anyway at least it leads me to spot a bug on my ttm change :)
glisse: i did allow for the ring buffer to be evicted
glisse: funny things happen when you evict the ring buffer
coucouf: Hi guys, I pulled a radeon version supportive of my RV740 yesterday and it just rocks ! So I wanted to drop a big THANKS for your work. :)
eosie: MostAwesomeDude: it's basically one interleaved vertex buffer with two vertex elements and the second one appears to be always busted
eosie: MostAwesomeDude: it's also interesting that gen_mipmaps takes the same path and it appears to be busted in exactly the same way
WhiteRabbit56: has ARB_texture_non_power_of_two been implemented in the r600 dri code yet? or is that a non-standard(as far as i could tell its supported by OpenGL) function?
airlied: its in r600 dri
WhiteRabbit56: do you know since when?
airlied: since I added it last week I think
spstarr: WINTER STORM WARNING FOR:
spstarr: =NEW= CITY OF TORONTO
WhiteRabbit56: (im running the fedora 12 dri-experimental)
WhiteRabbit56: thank airlied thats the info i wanted
spstarr: brigman: get out yer shovels for 2009's first snowstorm :)
airlied: maybe 2 weeks ago
spstarr: bridgman even
WhiteRabbit56: btw... to all the devs here: You guys are doing a great job!!!! keep up the great work!
WhiteRabbit56: @spstarr it's been snowing here for the past 3-4 days
spstarr: in Iowa?
spstarr: thats coming here to Toronto
WhiteRabbit56: Is it possible for me to build the newest dri without having to rebuild my kernel & mesa & everything else?
airlied: no DRI is 3 pieces
airlied: latset implise building all 3
WhiteRabbit56: alrighty then, so sounds like im gonna be building everything from source then (thanks airlied)
airlied: win 18
adamk_: Hmmm... agd5f's patch doesn't apply cleanly here.
adamk_: The only chunk that fails deals with displayports, though. I *think* it's safe to continue :-)
airlied: yeah shuold be fine
dileX: whats the recommended drm-2.6 GIT-branch to use against 2.6.32?
adamk_: rebooting... Wish me luck,.
spstarr: is running on new mesa, but no interrupts for now
spstarr: airlied: no response yet from LKML on high interrupt rescheduling issue yet
adamk_: agd5f, airlied, You guys are great. That patch worked exactly as advertised.
adamk_: Now to find out why compiz is so painfully slow.
Ghworg: dileX: drm-next or drm-radeon-testing depending on how bleeding edge you want
spreeuw: please please make an interrupt patch for 2.6.32 ;D
spreeuw: once it works
adamk_: Hmmm... GL compositing via kwin is fine, so it must be a compiz thing.
rah: agd5f: I'm using mkinitramfs to generate initramfs files but it doesn't seem to be picking up the new .bin file; do you know if it should?
airlied: rah: its distro specific, so need to ask them
dileX: Ghworg: hmm drm-next should be fine. just looking into the other GIT branches. drm-linus-stage seems to be the super-bleeding-edge-st
rah: airlied: why is it that other radeon .bin files are built in to the kernel but the _rlc.bin files aren't?
airlied: rah: license
airlied: kernel doesn't accept new firmwares either
airlied: the old ones got in before the deadline
rah: airlied: I'm confused there are .bin files in /lib/firmware/radeon that the kernel reports it's using
rah: airlied: is it getting those files from /lib/firmware/radeon, or are they included in the kernel sources?
stikonas: airlied: have you managed to try kdm on R700 hardware and see if X server starts?
airlied: rah: the kernel installs them in /lib/firmware/radeon
adamk_: Is it possible to get openarena not to sync-to-vblank with KMS? I understand that vblank_mode does not work with KMS any more, correct?
airlied: stikonas: not yet
airlied: stikonas: can you file a bug if you haven't
airlied: adamk_: you have to hack the DDX
adamk_: I don't care enough to do that :-)
stikonas: airlied: I haven't filed it yet
stikonas: will do so
airlied: adamk_: its not a major cause of slowdowns
adamk_: stikonas, What issue are you having? kdm from KDE 4.2.4 works fine here with KMS.
airlied: I suspect doing faster clears in mesa will bring us back within reach
adamk_: airlied, Yeah, didn't think so. Just curious how the x1900 compared to an HD4350.
stikonas: adamk_: I cannot start X with kdm 4.3.4 unless 88a50a is reverted in DDX
airlied: adamk_: r600 should vblank as well
airlied: under kms
maderat: maybe in r600_textstate.c /* YUV422 TODO conversion */ /* X, Y, Z, ONE, G8R8_G8B8 */ supposed to be Z, Y, X, like in r300?
rah: airlied: the kernel makefiles install them?
airlied: rah: yes
rah: that would explain it
rah: (as I'm using make-kpkg)
agd5f: maderat: r600 doesn't support yuv->rgb in the texture engine. you have to use a pixel shader to convert
adamk_: Hmmm... This is interesting. On FreeBSD, I have to not only disable coherent_mode, I have to change the resolution for it to actually do anything. I can, however, immediately set it back to 1280x1024 and it continues to work fine.
airlied: adamk_: not sure the DDX does a modeset on propery change always
adamk_: Not a big deal. I just realized that I had never tried it before in FreeBSD, only Linux. On Linux, changing coherent_mode (with UMS) or coherent (with KMS) changes it immediately.
pendrachken: Any one else using the debian testing/sid Xorg 7.4+4 getting a complete kernel lock when init launches kdm/gdm with the radeon driver? On my machine the kernel locks so hard even magic sysreq commands don't work, but ONLY when kdm/gdm is launched from init. Root kdm launches and startx with an ~/.xinitrc start normally. No errors or warnings in dmesg/syslog/Xorg.o.log either
airlied: adamk_: btw what model is your moniotr?
airlied: I'd like ot make coherent mode quirks someday
adamk_: Westinghouse something or other.
adamk_: It shows up in Xorg log as: Manufacturer: WDE Model: 5140 Serial#: 0
airlied: can you pb the whole section in the log?
adamk_: Monitor name: L1928NV
adamk_: Yeah, give me a second.
adamk_: I have this tendency to end up with odd, quirky hardware.
airlied: well I won't be using the year for a test then, 2007 is prerty new
kdekorte: airlied, is the warning from r700TryDrawPrims line 955 a known issue with mesa git?
airlied: it is now
kdekorte: File r700_render.c function r700TryDrawPrims line 955
kdekorte: Rendering was 3 commands larger than predicted size.
kdekorte: showed up today with my rv635
airlied: probably should file abug
kdekorte: will do
airlied: the predictor needs fixing
agd5f: already is one: 25514
airlied: ah cool
|moe|: after all the help you gave and all the wisdom you have shared i broke my pc and am now having an intel gpu. so thank you very much for doing right the mess i have done wrong. you guys rock! bye
|moe|: many years of help and gathered knowledge down the drain...
|moe|: c u around
adamk_: airlied, How about an xorg.conf option in the mean time?
adamk_: airlied, Don't get me wrong, not having the console expectedly blank when KMS is enabled would be great, but simply having X start up properly without having to run xrandr commands would be good to :-)
adamk_: airlied, Also, it doesn't just seem to be a matter of the monitor. This monitor works fine with an x850, HD3450, HD4350.
adamk_: (Without messing with coherent)
airlied: adamk_: oh so its only the x1950, wierd
adamk_: Well, to be fair, I've never checked to see what coherent is set at with those other cards, but I've never had to adjust it with them, one way or the other.
adamk_: And, the x1900 works fine with two LCDs at work.
adamk_: So it's just this specific combination :-)
adamk_: Hmm... It may be time to upgrade all my workstations to r6xx/r7xx cards now that the drivers are doing so well.
soreau: So is there any reason I would want to disable aiglx? It seems I have the functionality with it on but without I just can't use indirect rendering or dri1
maligor: well.. it's not surprising you can't use indirect rendering without aiglx, since that's what it is
RiotingPacifist: GLXinfo says i have no-tcl, can i build my own radeon driver or recompile my kernel to get TCL on "ATI Technologies Inc RS482 [Radeon Xpress 200M]", and how can i find this sort of thing out without crawling to IRC everytime?
adamk_: RiotingPacifist, The Xpress 200 doesn't have TCL.
adamk_: It's a hardware limitation.
RiotingPacifist: thank you for the info and the quick reply
adamk_: Sorry the answer isn't what you want. It's a pretty low-end GPU :-)
RiotingPacifist: i know but it came on my motherboard not much i can do short of getting a new laptop or learning to solder,buying a broken high end laptop, breaking everything.
MostAwesomeDude: No, you're stuck with it. You can't really switch to another GPU.
bridgman: on the other hand, I'm pretty sure the XPress200M was the fastest IGP around when it was introduced ;)
RiotingPacifist: I bought this laptop a couple of years ago :'( but it was damn cheap and i needed a laptop fast due to a bug with the screen on my previous one
twnqx: but how do i switch between the embedded intel and amd cards? :X
bridgman: intel IGP and ATI/AMD discrete graphics ? should be a BIOS option - integrated, discrete, or switchable
spstarr: ah yes that :)
twnqx: yes, switchable :P
twnqx: but i wonder... how to switch
spstarr: bridgman: still no word from IP if they can give us the spec on that?
spstarr: twnqx: there is some voodoo that makes it 'switch' d
spstarr: other than reverse engineering the windows vista/7 driver to do it
twnqx: so i have to sacrifice kittens if i want to do that on linux :X
bridgman: you don't want switchable right now, no driver support - go for integrated (intel) or discrete (amd)
spstarr: twnqx: Lenovo laptop?
bridgman: and honestly, before we mess around with dynamic graphics switching I would really like to see things like dynamic power management first
spstarr: I toggle between the GPUs if the ati one gets unstable during trunk tests
bridgman: dynamic pm is maybe 1/10th as much work
twnqx: spstarr: yes
spstarr: twnqx: W500? :)
RiotingPacifist: dont both ATI and Nvidia both offer IGP+discrete setups, who are they scared is going to copy them and steal their customers :S
twnqx: bridgman: the problem is... the ATI eats up battery, and i only need it to drive the displayport connector
bridgman: that's not the point - the IP for each system was implemented under three way agreements, so getting agreement to release is 3^3 times as hard
bridgman: there are other things we need to get out first
spstarr: yes, PM support :)
bridgman: twnqx, that's why I think power management should be a priority
spstarr: bridgman: I tried secondlife with windows 7 amd driver... we're still far from that point, but getting there
bridgman: good - any luck with that LLVertex thing ?
spstarr: bridgman: cant find any mesa path yet
spstarr: bridgman: gdb isn't showing me anything in the stacktraces
spstarr: bridgman: do we have any tests for VBO?
spstarr: anything in mesa demos ?
bridgman: I guess either LL devs are going to have to figure out what's not implemented or we'll just have to wait until enough stuff gets implemented for it to start working
spstarr: bridgman: thats likely how it'll end up
spstarr: bridgman: might be missing impl in the DRI driver
bridgman: yeah, the question is "of what" ;)
spstarr: bridgman: get your shovel out, blast of winter is coming tonight
DanaG: It'd be nice to at least have a gpu-switch after zapping xorg.
DanaG: Even if we had to log out to switch GPUs, it'd be better than nothing.
spstarr: bridgman: how much of the DRI driver is implemented? 80%? or its scattered
DanaG: Better yet to just save energy with the Radeon, though.
bridgman: implemented relative to what ? GL 1.5 ? GL2 ? GL3 ? GL 3.2 ?
spstarr: GL/GLSL, core
bridgman: GL* plus a bunch of other vendors extensions ?
spstarr: that also
bridgman: core what ? 1.5 ? 2 ? 2.1 ? 3 ? etc ?
spstarr: right now we have no GLSL (Shader language) support yet
spstarr: bridgman: core driver essentials for the GPU
bridgman: what the heck does that mean ?
spstarr: bridgman: power management, and other AMD specific stuff
bridgman: sorry, weren't we talking about 3D ?
bridgman: so do you mean "relative to fglrx" ?
spstarr: bridgman: well, all the bits needed to make a DRI driver function fully
spstarr: bridgman: yes
bridgman: it does function fully
spstarr: bridgman: fglrx or catalyst driver
bridgman: but again it depends on what you mean by fully
bridgman: fglrx goes way past any DRI driver today, and bypasses chunks of DRI in order to do that
bridgman: the "reference DRI driver" is probably swrast, followed by Intel
bridgman: but again, DRI != driver stack
bridgman: DRI doesn't handle power managent or any of the other things we're talking about here
spstarr: bridgman: the drm will do this (when in KMS mode later)
spstarr: bridgman: but in terms of stuff like:
bridgman: quivers in antici.......
spstarr: basic shaders
spstarr: Atmospheric shaders, etc
spstarr: GLSL stuff
bridgman: what the heck are atmospheric shaders ?
bridgman: arb_fp/vp implemented, GLSL wip
spstarr: in second life im stuck using 'low' mode vs ultra which i can use in win7
bridgman: hold your thumb and forefinger together then stare at it for a second
bridgman: the distance between them is what I know about second life ;)
spstarr: bridgman: well, let's look at it in terms of 3D support
spstarr: how much is done now roughly in %
bridgman: relative to fglrx on 6xx/7xx, maybe 30%
bridgman: relative to what you probably need for sl, maybe 80%
spstarr: well, excluding SL itself
bridgman: relative to windows, maybe 25%
spstarr: SL really wants GLSL
spstarr: bridgman: well considering it's been about 6 months things are going good.
bridgman: right, and glsl just needs some additional Mesa IL instructions implemented
bridgman: most of the code is in mesa, not the driver
spstarr: so GLSL is generic for all GPUs?
bridgman: most of the code is - are you familiar with the Mesa IL stuff ?
lindi-: any idea if controlling the fan of hd 3450 (RV620 according to wikipedia) is supported? Ideally the whole card should power down when I turn my monitor off but that is probably not possible?
spstarr: bridgman: that is good, I was not sure if one needed to implement GLSL differently for each GPU company
spstarr: (ie if GLSL was part of OpenGL spec)
stikonas: GLSL is part of OpenGL spec
bridgman: spstarr : http://cgit.freedesktop.org/mesa/mesa/tree/src/mesa/shader/prog_instruction.h
bridgman: scroll down to "typedef enum prog_opcode"
bridgman: lindi; the BIOS is supposed to load a temp vs speed program into the fan controller at startup, so lowering power should lead to lowering temp which should in turn lead to lower fan speed
bridgman: there seems to be a lot of variation between vendors in terms of how good the initial program is
bridgman: we don't have external fan control code in the open source drivers yet
lindi-: bridgman: interesting, and there's no free software that I could load to the controller?
bridgman: not yet
bridgman: getting to kms is really a pre-requisite for all that stuff, and that's what the devs are focusing on right now
bridgman: once kms is picked up by everyone then they can start plugging in power & fan control code
Wizzup: Hm... is it just me or is XGetImage with KMS really slow?
Wizzup: I'm not sure how it was without KMS
lindi-: bridgman: 'sudo rovclock -i' dies with "Floating point exception" in debian stable. trying unstable now -- is this the right command to lower power?
bridgman: didn't think rovclock supported the newer chips
maligor: the bios fan control stuff was scary on my card, it pretty much got to 90C easily
bridgman: not sure though
lindi-: bridgman: does HD 3450 count as a new chip?
bridgman: do "man radeon" and see the xorg.conf options
bridgman: new relative to when most of those utilities were written, yes
bridgman: you've got ForceLowPowerMode which statically reduces the engine clock to 50% of default, and DynamicPM which reduces it to 1/4 of default when your display is blanked by DPMS
bridgman: you need to restart X for them to take effect
bridgman: the code also uses fewer PCIE lanes which causes problems on a few systems, that's why they aren't on by default
bridgman: also ClockGating, but not sure if that does anything on 6xx/7xx, IIRC it doesn
bridgman: note that the power options have been implemented for UMS but not for KMS
maligor: I thought the pcie lanes were meant to be varia..err..switchable, seen some hardware hacks on that
lindi-: debian unstable's "man radeon" lists Option "DynamicClocks" "boolean" but not those, I guess I need some even newer version
bridgman: sounds like it
bridgman: maligor, they are in theory but there seem to be chipset-specific issues
bridgman: both sides of the bus need to negotiate
Sinuvoid: Where's the SVN again
maligor: debian experimental seems to have 6.12.4
maligor: Sinuvoid, what svn? you mean git repo?
jcristau: 6.12.3 and 6.12.4 are pretty much the same thing
AndrewR: hey all. latest xserver changes broke my screen a bit ()http://img687.imageshack.us/i/xserverbug9122009.png/
AndrewR: ops. some fonts are unreadable
Sinuvoid: Hey, how do I check what version of radeon card I have?
Sinuvoid: Also, could someone give me the link to the svn/git?
bridgman: lspci should tell you the card/GPU
Sinuvoid: What does ls mean by the way
Sinuvoid: (off-topic question but I see people use it all the time)
Sinuvoid: Thank you :)
Sinuvoid: so would ls pcm work too?
bridgman: build instructions, source code location etc... are at :
bridgman: scroll down for link to build instructions etc..
maligor: heh, I guess what you'd want is 'aplay -L' or 'aplay -l'
Sinuvoid: Don't worry, I've built it before :D
Sinuvoid: bridgman, I just want to see the svn for updates
Sinuvoid: If there isnt much then I wont bother.
bridgman: at some point you might want to consider referring to git rather than svn... don't know if xorg ever used subversion, did it ?
Sinuvoid: bridgman, should I get master branch or "r6xxx/r7xxx support" branch?
lindi-: maligor: it's probably not enough to just fetch the radeon driver and use it with stable's Xorg even though X is now "modular"?
bridgman: master - those were just feature branches
bridgman: merged back to master now
Sinuvoid: I want r6xxx support though
maligor: lindi-, it should be
maligor: lindi-, you might need libdrm also
bridgman: Sinuvoid; that was the branch used for initial development; it's obsolete now
Sinuvoid: Ah ha
Sinuvoid: Master it is :P
maligor: lindi-, since I tried that on debian unstable a few days ago
lindi-: maligor: at least with xserver-xorg-video-glamo some #ifdef's needed to be added to make it backwards compatible
lindi-: ok, i'm first building linux 2.6.32 anyway to check if I can control the motherboard fans (F8000 superio)
bridgman: spstarr; just tying a 20x28 tarp over the woodpile ;)
Sinuvoid: bridgman, how do I 'update' the current git clone I already have
lindi-: maligor: ok thanks, I can run X from debian unstable chroot
chithead: git pull
DanaG: yay for fancontrol.
maligor: lindi-, I just made it into a package
lindi-: maligor: thanks, can you put the .dsc somewhere?
Sinuvoid: chithead, isnt it git push? (from what im reading here)
maligor: lindi-, dsc?
chithead: maybe you want to make yourself familiar with git first
lindi-: maligor: it's what dpkg-source expects. http://www.debian.org/doc/maint-guide/ch-build.en.html
Wizzup: Sinuvoid: pull pulls, push pushes. :)
lindi-: maligor: I'm assuming you don't have a full-blown repository for your package yet :)
maligor: lindi-, no, and it was a hack job anyway
Sinuvoid: Wizzup, makes sense :)
lindi-: maligor: seems I need xserver-xorg-dev from experimental as well
lindi-: maligor: (or maybe the dependencies are just too strict)
maligor: lindi-, I haven't a clue what you mean :P
lindi-: maligor: sorry, i'm quite tired and did not explain this well. xserver-xorg-video-radeon in debian experimental has Build-Depends set to xserver-xorg-dev from experimental
maligor: oh, ok, I used the unstable as a base
lindi-: maligor: ah and just fetched the newer version from git?
maligor: yes, and changed the version
jcristau: lindi-: or you can ignore that bit from the build-depends, it's just there to ensure the version in experimental gets built against that server, it's not strictly necessary
lindi-: maligor: and you are using debian unstable without any chroot tricks?
maligor: lindi-, yes
Sinuvoid: ok, this is not working
Sinuvoid: git pull git://anongit.freedesktop.org/xorg/driver/xf86-video-ati
maligor: I'm not using it at the moment since I need the evil features of fglrx libGL tho
moosipurk: could someone help me with this:
moosipurk: File r300_render.c function r300Fallback line 371
moosipurk: Software fallback:ctx->RenderMode != GL_RENDER
chithead: just go to the directory where you clones
chithead: moosipurk: use a pastebin
moosipurk: oh, sorry
chithead: moosipurk: you can use driconf to disable low-impact fallback
moosipurk: ok :)
Sinuvoid: ok, this is not working
Sinuvoid: git pull git://anongit.freedesktop.org/xorg/driver/xf86-video-ati
MostAwesomeDude: git clone, not git pull.
Sinuvoid: I want to update the current clone I have here though
bridgman: the clone should already know where it came from; just cd into that repo then do git pull origin master, or stomething like that (origin master is probably optional assuming previous clone was from master as well)
lindi-: DanaG: seems "pwmconfig" fails with "manual control mode not supported" :(
cxo: dang on it
agd5f: lindi-: don't use rovclock. it doesn't do the right thing even on supposedly supported chips
agd5f: it changes the feedback divider without taking into account the post dividers
DanaG: I'll be glad when that KMS powersavings code gets into the git drm modules.
eosie: zhasha: hey buddy, a month ago you asked me whether I know why mipmaps don't work... now I have the answer and even the fix ;)
zhasha: eosie: oh please do tell me because I've broken my brain trying to figure it out
eosie: and also the weird colors on the background are gone
zhasha: eosie: do you have commit on mesa yet?
eosie: zhasha: if the state tracker wanted TCL to be bypassed, routing of vertex streams did not work at all, only position was routed correctly, but not colors nor texcoords
eosie: zhasha: no, I don't, I finished the patch about an hour ago
zhasha: I meant whether or not you have commit access on fd.o
eosie: I see.. no, I don't have an account yet
spstarr: bridgman: im making a pizza
eosie: and what's even more interesting is that piglit/fbo-generatemipmap works on r300g but not on r300c ;)
soreau: eosie: Does the mipmap fix happen to fix fisheye mode at all?
airlied: glisse: looks like it wasn't bad RAM, time to debug then
eosie: not sure, I believe the fisheye mode dumps CS
soreau: eosie: Who has been pushing your patches? MAD?
zhasha: and once in a blue moon, I'll push them. but seriously, get an account
soreau: Well I hope someone hurries up and pushes them because I want to test it ;)
soreau: eosie: Any reason you don't have an account yet?
zhasha: it took me 2 months to get one
soreau: That's a good and valid reason
Nezmer: stupid question. Is supporting power management in KMS a priority for r700?
eosie: in 2 months I'll be very busy on an internship in France
[Enrico]: Nezmer: there is fair work to provide it afaik , but i ignore the current status
Nezmer: [Enrico]: Do we expect it to be ready for 2.6.33?
[Enrico]: Nezmer: as i said i don't know the current status
Nezmer: [Enrico]: ok. thanks for replying
Sinuvoid: checking if xorg-macros used to generate configure is at least 1.3... configure: error: configure built with too old of a version of xorg-macros.m4 - requires version 1.2.0 or newer
Sinuvoid: How do I fix this?
MostAwesomeDude: eosie: Patch? Link?
eosie: MostAwesomeDude: I am working on it
Sinuvoid: Can no one help me out
MostAwesomeDude: Sinuvoid: Update xorg-macros and run autoreconf.
bjaglin: Sinuvoid: http://www.x.org/wiki/CompileXserverManually#head-6a888084850e6c731fec78ffdc1774e6653fd5d6
soreau: Sinuvoid: Which distro are you using anyway?
Sinuvoid: soreau, fedora 11
AndrewR: it was 955b9f23a34cc79a5cd9676b45b3df4ffcc7302b in gitxserver (master). reverting it fixes corruption ... ( EXA: ModifyPixmapHeader_mixed fixes.)
Sinuvoid: checking if xorg-macros used to generate configure is at least 1.3... configure: error: configure built with too old of a version of xorg-macros.m4 - requires version 1.2.0 or newer
Sinuvoid: I keep getting this error!
Sinuvoid: Nope, same error.
bjaglin: Sinuvoid: did you do http://www.x.org/wiki/CompileXserverManually#head-a20e40009e1d4bd0286a4bcb0944fdeb8e64f9d2? i built X for the first time yesterday too... it took some time :)
Sinuvoid: I did that
Sinuvoid: (the macro part)
MostAwesomeDude: Did you autoreconf?
Sinuvoid: doesnt say to
flyback: seeing stars wtf
MostAwesomeDude: It also doesn't say what to do with that error.
flyback: seeing little black holes all over the place
MostAwesomeDude: And you should run autoreconf.
Sinuvoid: MostAwesomeDude, just like that?
Sinuvoid: ./autoreconf ?
MostAwesomeDude: $ autoreconf
Sinuvoid: aclocal: couldn't open directory `/opt/xorg/share/aclocal': No such file or directory
Sinuvoid: autoreconf: aclocal failed with exit status: 1
Sinuvoid: I used this btw:
Sinuvoid: $ export ACLOCAL="aclocal -I /opt/xorg/share/aclocal"
Sinuvoid: Which I think broke something
MostAwesomeDude: $ unset $ACLOCAL
MostAwesomeDude: Have you considered just upgrading to F12?
Sinuvoid: That would be hell.
Sinuvoid: F10 to F11 took like a week
MostAwesomeDude: Suit yourself.
Sinuvoid: MostAwesomeDude, I'm sticking with F11 for awhile now :)
MostAwesomeDude: Sinuvoid: And I'll keep on testing my code on F12 only. :3
Sinuvoid: Ohhhhh damn you :P
MostAwesomeDude: Took me one night to get three boxes updated to F12, and I didn't use preupgrade.
flyback: sinks his teeth deep into Sinuvoid's leg meat
MostAwesomeDude: Supposedly it's even easier with preupgrade.
Sinuvoid: flyback whoaaaa
flyback: has never and never will do drugs of any kind, this is 100% all nature brain damage
flyback: Sinuvoid, you damn eh-hole
flyback: that's for the rapper "snow" and other canadian acts of terrorism!
flyback: such as the sci fi series - lexx
MostAwesomeDude: I love how the kick reason is always "airlied."
flyback: I was joking :(
airlied: flyback: it wasn't funny
airlied: otherwise I'd have said haha instead of /kick flyback
Sinuvoid: no one was laughinh
soreau: flyback: You are distracting while people are trying to get real support for the radeon driver. You are completely off topic
airlied: you seem to think saying the word canuck is funny, when clearly it isn't
flyback: yeah well 12+ yrs
flyback: I WILL NEVER LEARN
flyback: but I will keep it out of this channel :)
flyback: http://www.allaboutvision.com/conditions/ocular-migraine.htm <-- that field picture
flyback: oh airlied I talked to you before
flyback: I think you wanted to know how the video project was coming
flyback: I did find out one thing, you need at least a 800x480 res portable dvd player lcd or pc fonts aren't readable well on post
cxo: finds it hilarious, cos most of the time, he isn't even talking to a Canadian
Sinuvoid: Lol he was lucky this time
DanaG: always reads "cos" as "cosine".
DanaG: weirdest screen I've ever used: a 42" plasma, 1024x768 stretched to widescreen.
cxo: beat 1386x1050 (18.5inch acer, was on sale ) for weird
DanaG: Pixels would be 0.036 x 0.027 inches.
DanaG: BIg and rectangular. =þ
eosie: MostAwesomeDude: I've sent a little patch series on your email
MostAwesomeDude: eosie: Sweet, I'll get to them when I get home from work.
MostAwesomeDude: eosie: Also, I'll sponsor you getting commit access.
eosie: ok :)
soreau: eosie: MostAwesomeDude: Did I ever mention how much you guys rock? =D
soreau: my rv350 loves you too
DanaG: that's 9600 series?
flyback: DanaG, you know a cosine?
DanaG: Oh yeah, do R350 chips (not R360) have any sort of mechanism to shut off before they can overheat and die?
DanaG: My old 9800 Pro doesn't have a fan on it -- just a big honkin' pair of heatsinks with heatpipes. So, I'd want to make sure it won't die, at least.
soreau: DanaG: I have no idea.
soreau: This one was marketed as a 9600 Pro, it's AGP and has a fan. I don't usually care too much about monitoring temps but I always make sure my machines are in a cool, well ventilated area
soreau: I've never had any hardware problems with it over the 5 or more years since I've had it
DanaG: I'm just extra sensitive to fan noise.
flyback: didn't fry a portable dvd player after all :)
spstarr: temperature is critical
spstarr: +1C 10 mins from me -1C 2 mins from me
airlied: glisse: no idea how your system is running with your current tree ;-)
airlied: with the TTM ERESTART patches X was dying for me here, I sent a patch to fix it to the list
airlied: please review/ack it
airlied: your other patch for cleaning up the warning crashes stuff, so I've just cleaned it up simply here
airlied: glisse: btw radeon_object.h still checks for ERESTART
spstarr: airlied: where is glisse working now @ RH? still in .fr?
airlied: spstarr: yes/yes
DanaG: - if (unlikely(r != 0)) + if (unlikely(r < 0))
DanaG: interesting... "unlikely" function.
DanaG: Now that has me curious.
airlied: its a branch hint to the compiler
airlied: in that case probably fairly pointless
hax0r1: does radeon support anything higher than glsl 1.20?
cxo: didnt know you could add if-statements
airlied: hax0r1: it barely does that
DanaG: that's awesome.
hax0r1: airlied: cool, we will see if nv50 will break that barrier, that will probably require so changes to the gallium code
DanaG: though, it makes me think, "if it's unlikely that something is true".
airlied: hax0r1: http://dri.freedesktop.org/wiki/MissingFunctionality
airlied: where that says piles and piles of work it really means it
cxo: switching over to Thunderbird 3.0!
airlied: glisse: also seeing [TTM] Failed to find memory space for buffer 0xf750d02c eviction.
airlied: radeon 0000:06:00.0: object_init failed for (10833920, 0x00060004)
airlied: in my logs
airlied: [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (10833920, 4, 4096)
airlied: with latest TTM/radeon code
spstarr: airlied: thanks for that url
airlied: and now my desktop died