Zajec: i can see commit to r6xx-rewrite: "Fix r6 code bugs."... does it change "WIP" to something already usable?
MostAwesomeDude: Probably not.
bridgman: I didn't see an "it runs!!" email from Richard so probably still WIP but much closer
bridgman: Richard's estimate was "this weekend", I asked him to push early so Jerome et al could take a look at the cs bits
Zajec: bridgman: great :)
Zajec: thanks for explaining
bridgman: just don't ask me to "twitter"
MarkusKrogh: Hi, is there currently any way to affect the powermanagement of a radeon x1600 card?
bridgman: if you're running the latest radeon ddx there are a couple of options - ForceLowPowerMode puts the chip into a lower power state, and the other option (whose name eludes me) lowers power further when display is blanked by DPMS
bridgman: info on agd5f's blog : http://www.botchco.com/agd5f/?p=41
bridgman: or check latest manpage
MarkusKrogh: ok thanks
bridgman: I don't remember if Clock Gating is on by default or not, check that as well
bridgman: Clock Gating uses some internal chip hardware to automatically stop clocks on unused hardware blocks; the newer options add engine clock and PCIE lane width controls as well
MarkusKrogh: bridgman: How do I check the Clock Gatering
bridgman: read the man page to see if it is turned on by default; if it isn't, turn it on
max_r: is current r6xx-rewrite supposed to somehow work?
rah: bridgman: wouldn't it be more appropriate to have just one "Low power/performance mode" option that affects these things rather than individual options?
rah: bridgman: otherwise it kind of turns the driver into just a HAL layer?
bridgman: this PM stuff is all temporary anyways until we transition to KMS; power management really needs to be in the kernel if we're going to make it nice and automatica
bridgman: in the meantime these options are new and might cause problems, so they were released as independent options for now
bridgman: max_r; don't think so, I expect Richard would have said something if the ported code was running ;)
bridgman: probably pretty close though...
bridgman: he found the big problem, now just working through the problems that were hiding behind it ;)
bridgman: rah; I expect that some of these options will default to "on" after a while, but they'll probably be kept individually controllable for troubleshooting
rah: bridgman: I see
bridgman: there are just so many different configurations out there, and if the drivers don't have some tools to allow user troubleshooting the chances of supporting all those configurations is pretty much zero
bridgman: for Windows all the troubleshooting happens with the OEM before the product ships
bridgman: for Linux it mostly happens six months after the product ships
mcgreg: bridgman: hehe as a normal user you have simply no idea :)
bridgman: although that is starting to change
mcgreg: no idea how complex and difficult it probably all is
bridgman: being a user, or driver development ?
mcgreg: *user* :)
bridgman: yeah, probably ;)
bridgman: I tend to be a boring user; I install the distro and I don't muck with it
mcgreg: being a user .. really. as a user all you see is : working - or not wokring
bridgman: always seems to work well for me
bridgman: but if you're not interested in the guts then you probably do need to choose a distro that takes care of this stuff for you
bridgman: and hardly any distros have the resources to do that
mcgreg: actually I've used debian ... and I love it :)
mcgreg: I've tested gentoo too... it was interesting... but very much time consuming
bridgman: so do devs at some other distros, but I won't name names ;)
mcgreg: you dont need name names... the fact speak for itself :)
phercek: I already reported here a bug that when I pull out second unused vga card (RV100 PCI) the first one (RV670Pro AGP) stops to work. The image is corrupted. When both VGA cards are in (or when dri is off) and only the AGP one is used it works correctly. Today I tried with latest radeon from git and the behavior changed. I do not get any image, screen is black. Only mouse cursor is rendered. You can find log file and version info here http://www.hck.sk/users/peter
bridgman: can you update the bug in Bugzilla pls ?
phercek: I can post a new one; before I did not create one; since I did not have drm drivers from 2.6.30 rc?; is it ok to report a bug even with drm from git://git.freedesktop.org/git/mesa/drm; branch r6xx-r7xx-support origin/r6xx-r7xx-support ?
bridgman: I think so; we probably would have merged that branch code to master if it weren't also tied up with other projects (KMS/MM) going on in parallel
phercek: Ok, I'll create a bug report in few minutes.
bridgman: although I don't think multi-card support is getting a lot of attention right now.. but that will change
bridgman: either when RandR is extended to work across multiple cards, or when the KMS/MM transition finishes and the devs aren't so tied up with that one big task
phercek: notice that I'm not using the PCI card in my xorg.conf; It just needs to be there so that the primary card works
bridgman: understood; I think a couple of similar bugs have been reported on the radeonhd list; seems to be some common thread but it's not obvious what that is
phercek: actually because of later problems with multicard support I got matrox tripehead2go, so that I can work with one card only; but I still need to have the pci card in (even when not used)
bridgman: is the VBIOS in your 670 card stock or flashed ?
bridgman: good, one less thing to worry about
phercek: ok; going to fill in the bug report
phercek: me too
phercek: bridgman: http://bugs.freedesktop.org/show_bug.cgi?id=21763
bridgman: looks good; the log file is with the PCI RV100 removed, right ?
bridgman: two "monitors" on the card, one real and one Matrix 3:1 thingy ?
phercek: do you one with the RV100 in too?
bridgman: only if the devs ask for it, this should be fine for now
bridgman: is the BT879 a separate capture card ?
phercek: yes a separate PCI card; some really old one
bridgman: have you ever tested with that card removed ?
adamk: Anyone have any ideas about this crash in rawhide: http://pastebin.com/m393ab2c6
phercek: bridgman: no I did not
bridgman: might be worth a try; I remember one case where an old capture card clung grimly to the old graphics card and wouldn't let go
phercek: bridgman: I want to keep the TV card in, it works; do you need a test without it?
bridgman: yeah, just for testing
phercek: ok; I'lll try; I'll be back in about 5 mins (this is the machine with the problem)
phercek: bridgman: pulling the TV card out does not help (the agp card still does not work when both TV and RV100 are out)
phercek: bridgman: do you want a test with RV100 in and TV card out ?
bridgman: nope, card out and RV100 out is enough
bridgman: I figure the devs would have been wondering; still a tiny chance the card's driver could be interfering but that's much less likely
bridgman: thanks for testing
kristian__: does this driver support tvout for an ATI 3450?
bridgman: more or less; might still need to be forced on, and last I heard was NTSC only
kristian__: I need PAL in Denmark, right?
bridgman: I think so
bridgman: main thing it's been missing is a developer in a PAL country with TV and time ;)
bridgman: most of the people working on it recently live in NTSC countries
megari: bridgman: how complicated is it to make the the card output PAL signal as opposed to NTSC?
bridgman: in theory it should work now
bridgman: it just doesn't ;(
bridgman: it took some tinkering to get it working on NTSC, needs the same for PAL
megari: bridgman: I see. My question was just technical curiosity.
bridgman: yeah, in theory everything is being programmed correctly based on what we could get from internal docs and looking at fglrx
bridgman: but we need a tinkerer in a PAL country
bridgman: all the PAL and SECAM folks are working on other stuff ;)
lucxx: is the radeon r6xx rewrite in mesa branch functional?
bridgman: don't think so, close though...
lucxx: for test
lucxx: or to test it
lucxx: i ned agd5f drm branch?
bridgman: it's still really for development, prob not ready for test yet
bridgman: yes, agd5f's 6xx drm branch
bridgman: trying to get it running over a mesa-to-drm API that will work for KMS/MM as well
bridgman: once that is done we can merge into radeon-rewrite (or master if radeon-rewrite has merged) and do all the work in one place
lucxx: and you are close?
bridgman: drm will probably have to stay separate for a while longer until the main GEM/TTM code goes upstream
bridgman: yeah, Richard figured he would have it working this weekend some time
bridgman: you never know until it works though ;)
Gnutoo: hi I've Software fallback:ctx->Line.SmoothFlag
Gnutoo: (File r300_render.c function r300Fallback line 392
Gnutoo: is there a workarround because it gives horibles perf
bridgman: I thought the flag to disable those fallbacks was on by default these days
bridgman: are you running fairly recent Mesa code ?
Gnutoo: it's a big problem because I can't run the following softwares because of that:
Gnutoo: *smc(super maryo cronicles)
Gnutoo: *gnash compiled with opengl
bridgman: do you have driconf installed ?
Gnutoo: yes I have
Gnutoo: I can also modify the config file or paste them
bridgman: there's a "Disable Low-Impact Fallbacks" option you should be able to choose
Gnutoo: ok thanks a lot
Gnutoo: still slow as hell but dosn't emit the message anymore
bridgman: which GPU do you have ?
Gnutoo: by the way I've an old xorg...mabe it could be the cause...
Gnutoo: I'll upgrade xorg
bridgman: ok, wanted to make sure you had vertex shader hardware; GE is very slow on IGP parts that have to do vertex processing on the CPU
bridgman: but yours should be OK
bridgman: can you try with flightgear ?
Gnutoo: with the config change I tried gnash
Gnutoo: (was fast with pure software rendering called agg and now it's slow with opengl rendering)
Gnutoo: flightgear is ok
Gnutoo: I'll try smc
Gnutoo: it works
Gnutoo: smc finally works
Gnutoo: maybe it's a gnash problem then
Gnutoo: thanks a lot!!! I'll investigate for gnash
Curtman: is amazed at the quality of video playback with this driver. You guys are the best.
Curtman: I wish I could have the day of my life that I spent fighting with fglrx back. :)
mcgreg: Curtman: nah .. dont say that. you propbbaly made interesting experiences :)
mcgreg: and *every* experiences is a worthy one. if it's good or bad
Neo_The_User: Curtman: ping > what exact driver?
Curtman: Neo_The_User, Does 6.12.2 sound right?
Neo_The_User: yes thats the stable radeon (non HD) driver
bridgman: not fglrx
Curtman: Yeah, I just tried the one that Gentoo gave me.
Curtman: Screw fglrx.. Nothing but trouble with that stupid thing.
Neo_The_User: I'm about to try the master branch in a little bit
Curtman: I don't even really care about 3D.. But my old nVidia card looked terrible compared to this 3200HD onboard video. I'm pretty surprised by that.
Neo_The_User: Curtman: you have a radeonhd card?
Neo_The_User: Why not use the stable radeonhd driver?
Curtman: It's onboard video on my new mobo..
Curtman: I haven't even tried it yet.. This seemed to look so nice.
Neo_The_User: If you ever want to upgrade to git: http://neo-technical.wikispaces.com/xf86-git (start from 2a.)
da1l6_: Neo_The_User: readonhd lacks tear-free video support, doesn't it?
Neo_The_User: I wouldn't know. I'm on r300
mjg59: There's no especially strong reason to prefer radeonhd over radeon for most hardware
mjg59: Any functionality disparity is likely to be treated as a bug by both teams
da1l6_: the xorg wiki page for radeon mentions the differences: http://www.x.org/wiki/radeon
Neo_The_User: da1l6_ so I guess it does support tear-free video ;)
bridgman: I think the main differences right now are watermark support on radeon, hdmi audio on radeonhd, and the power saving options on radeon might go a bit further
bridgman: not sure about tiling support
Neo_The_User: Color tiling?
bridgman: radeon has tvout on many GPU
bridgman: memory tiling; mapping memory pages onto square blocks rather than horizontal lines to make memory accesses more efficient
Curtman: radeonhd is supposed to have new support for sound on my hdmi jack.. But I have nothing else with hdmi to try it.
Neo_The_User: bridgman: How do I turn memory tiling on?
bridgman: write a bunch of code in the driver
agd5f: Neo_The_User: it's on by default on radeon fo r1xx-r5xx
Neo_The_User: oh ok so I have it
Neo_The_User: agd5f: I see you write tons of source code for xf86-video-ati and radeonhd. I've been wondering why you don't do so much with mesa (dri) to make more use of the driver.
bridgman: I guess that should show up in benchmarks on a 5xx between radeon and radeonhd
agd5f: Neo_The_User: lack of time
Neo_The_User: ah ok
bridgman: mesa is like another million lines of code to master ;)
spstarr: morning bridgman
bridgman: hi spstarr
icewaterman: hm, i am using 6.12.1-0ubuntu2 version of the radeon driver on amd64 and since i do, i.e. max payne and max payne 2 and other 3d games became awfully slow
icewaterman: before that i was using the version from intrepid (8.10) on i386
icewaterman: glxinfo says direct rendering is enabled
agd5f: icewaterman: 3D is handled in mesa
agd5f: so it was probably related to a 3D driver update
icewaterman: agd5f, mesa was of course also upgraded (now 7.4)
icewaterman: btw. i am using kernel 18.104.22.168
agd5f: icewaterman: what chip? there was a pretty big performance regression for IGP chips between 7.2 and 7.4
agd5f: it's fixed in radeon-rewrite branch of mesa
icewaterman: agd5f, its an rv530 i think
icewaterman: the onboard stuff is too slow for those games even if properly supported by the driver.
Neo_The_User: icewaterman: try upgrading the driver
icewaterman: Neo_The_User, unlikely. since compiling this stuff for both 32 and 64-bit is a real payne. been there done that,
Neo_The_User: What version of Ubuntu? Maybe try 9.04?
Neo_The_User: It has a later version of Xorg server (1.6)
icewaterman: Neo_The_User, this is 9.04
icewaterman: 8.10 worked fine, but i wanted to upgrade to amd64 version (which works fine most of the time). just wine apps seem to have a problem
Neo_The_User: Maybe in the repository there is package called atitool
mcgreg: hmm I read so much about tearing problem with xv/fglrx and generally even radeon drivers .. but I still have no idea and I dont see anything special. never did
mcgreg: am I too blind?
agd5f: mcgreg: I have trouble seeing it too
mcgreg: I'm glad I'm not the only one
yangman: depends on the media, monitor, and a whole lot of other things
agd5f: only way I can see it is to generally test with a video that just flips back and forth between black and white
yangman: I see it all the time
mcgreg: hmmm could you describe it somehow?
mcgreg: perhaps make a screenshot?
yangman: still happens with rv770 on radeonhd, actually. the vsync code for Xv still needs a bit of tweaking
mcgreg: I have a radeon4670 .. which is 730 I think
agd5f: mcgreg: imagine the screen scanning out the image when only say the top half has updated to the next frame
agd5f: so the top half of the video is frame 1 and the botton half is frame 0
mcgreg: hmm okay. but I could imagine you should see this quite "intensive" .. I should have recognised that.
agd5f: to avoid it you can either never render to the same buffer you car scanning out of, or stall rendering until the scan out has passed the area you are updating
mcgreg: perhaps I shouldnt care. IF I dont see it anyway, I dont need to bother ;)
Neo_The_User: Im going to try the experimental video output drivers for vlc with the open source drivers and see if I can get it to use more GPU over the CPU with the latest and greatest ati open source drivers from git master
agd5f: Neo_The_User: to use more gpu features you actually ahve to write driver code
MostAwesomeDude: Neo_The_User: You mean, with VDPAU?
MostAwesomeDude: Well, which video output were you thinking of?
Neo_The_User: I'm going to try all the experimental video output drivers. OpenGL stable and experimental, XCB experimental and stable, X11 exp. stable and see what uses the least amount of CPU usage in LXDE
bridgman: ooh, I can tell you ;)
MostAwesomeDude: Neo_The_User: Spoilers: Xv is fastest, followed by OpenGL.
Neo_The_User: Yeah Xvideo also has an experimental output module
icewaterman: hmm I identified parts of my problem: Option "DynamicClocks" "on" is causing even tremulous to severly slow down
Neo_The_User: (I just cloned vlc from git a minute ago :)
agd5f: icewaterman: might work better with new clock gating code in git master
icewaterman: agd5f, as i said compiling for my system is a real pain, so i will not bother
bridgman: edgers package ?
agd5f: icewaterman: yeah, just fyi. your distro may provide testing packages if you were interested
Neo_The_User: be right back. Video testing
icewaterman: agd5f, might be the case for the alpha version of the next release, but especially for ubuntu these are extremely unstable
Neo_The_User: unstable xvideo driver = less than 5% max
Neo_The_User: when i run the video at 2x speed, it only takes up 20%
icewaterman: ok, tremulous seem to work fine, also counterstrike 1.5 is fast enough (so 32-bit wine works). i dunno though, why max payne doesnt
hifi: max payne uses some magic the open driver doesn't support?
icewaterman: hifi, it worked before, so i dont think so
icewaterman: on 32-bit ubuntu 8.10 it worked fine. now on 9.04 64-bit it is extremely slow
hifi: should buy it, great game
icewaterman: hifi, yes, i can recommend it. it is available for approx 1GBP from ebay.co.uk
icewaterman: plus shipping of course
hifi: I think I can get it for the same price locally
icewaterman: i got both the first and the second part for <10€
hifi: I don't like MP2
icewaterman: hifi, thing is, in germany it was impossible to get it at all.
hifi: it is?
icewaterman: hifi, first time i played it i also didnt like it, but when i played it the second time it was really nice.
hifi: no wonder why torrents are so popular
icewaterman: hifi, afaik Max Payne 1 was never sold in germany
hifi: restrictions or just no interest from the distributor?
icewaterman: and the german version of Max Payne 2 was very crippled from what i have heard.
icewaterman: hifi, probably because of some stupid law i guess
icewaterman: not that it was hard to order it from outside the country.
hifi: torrent ratings to up
hifi: *go up
yangman: laws regarding depiction of violence, probably
icewaterman: and it was also cheaper because the GBP was pretty much down the day i bought my copies there.
icewaterman: unfortunately i cannot try my other GPU as it is not supported by the driver yet
icewaterman: brb, i'll check if radeonhd works
icewaterman: ok, radeonhd suffers from the same problem, so it is likely to be mesa-related indeed
Gnutoo: hi, I still have a problem with opengl...I've 2 computer both ati: laptop->x700 desktop->r200
Gnutoo: and on the laptop gnash is slow(compiled with opengl)
Gnutoo: on the desktop I've weired refresh problems inside the gnash window
rah: I have a dual screen setup where one screen is shorter than the other, making a gap below it, like so:
rah: there is a problem in that the mouse can go off of the screen, in to the gap
rah: [one LVM filesystem-resize later]
rah: so my mouse goes off the screen into the gap
rah: should I file a bug?
yangman: rah: xserver limitation. all the drivers do it
rah: fglrx doesn't
rah: so I should file a bug against xserver rather than radeon?
yangman: fglrx doesn't use the randr cursor calls iirc
yangman: it manages multiple screens on its own
rah: so I should file a bug against xserver?
MostAwesomeDude: rah: There might already be one opened.
yangman: yeah, I'd expect there one to be open as well
rah: there may
yangman: it's been known for years
rah: still open :(
spstarr: yangman: so what do you do @ AMD?
yangman: I'm not involved with AMD at all
yangman: just an average user with a lot of time ;)
rah: I'm someone
bridgman: we're the same people as before, we don't know what the deal is with Gnash ;)
MostAwesomeDude: Gnash's OGL isn't perfect.
Gnutoo: ok thanks
chithead: the gnash agg backend does better than the others in my experience
Gnutoo: I thought nobody read my message...lol
Gnutoo: yes but I wanted to find a way to increase performances so I tried opengl...
MostAwesomeDude: Yeah, AGG is still much faster than Flash's renderer.
MostAwesomeDude: The OGL backend is faster than AGG but buggier and uglier.
Gnutoo: thanks a lot
icewaterman: MostAwesomeDude, gnash itself is far from perfect
MostAwesomeDude: icewaterman: Yeah, but it's gettin' there.
icewaterman: MostAwesomeDude, i hope so, but i seriously doubt it
icewaterman: would be cool though
Gnutoo: icewaterman, should be ready at the end of the summer...there is a huge but simple work to do...write a lot of classes...there is a fund raising for that and it will be done by students(the class writing)
icewaterman: Gnutoo, yeah, thats what google's summer of code is for :)
Gnutoo: I don't think it's google summer of code because not 1 but many students will be involved
icewaterman: Gnutoo, i am not aware of the requirements for participating in this so I dont know if google would fund it.
icewaterman: Gnutoo, what programming language is it written in? you said classes, so it seems to be object oriented
MarcOChapeau: evening everyone :)
Gnutoo: gnash itself is written in C++ but that's not what the student will code...they will code in action script
MarcOChapeau: quick question: does xf86-video-ati allow for audio over HDMI or is that with the radeonhd driver only ?
bridgman: just radeonhd right now
bridgman: only 6xx and above, of course
MostAwesomeDude: really needs to port that to radeon KMS
MarcOChapeau: bridgman: thanks. I thought so
MarcOChapeau: I'll wait then...
MarcOChapeau: MostAwesomeDude: audio support ? that would be cool :)
MostAwesomeDude: MarcOChapeau: It really should be in the kernel.
MarcOChapeau: MostAwesomeDude: it certainly looks weird to see it in the driver. is there any active code in the driver, or just a switch to flip ?
MostAwesomeDude: MarcOChapeau: Yeah, there's a bit of card setup involved.
MostAwesomeDude: But it's really not that involved, and it really should happen when DRM loads.
MarcOChapeau: MostAwesomeDude: but the encoding itself is transparent to the dev right ?
MostAwesomeDude: MarcOChapeau: Yeah, it just kind of appears magically to ALSA.
MostAwesomeDude: Just like a USB soundcard, completely hot-plugged.
icewaterman: Gnutoo, ah, i thought the gnash guys needed help with coding, instead they seem to need money. with the first i could have helped
MarcOChapeau: I see
Gnutoo: icewaterman, they need money to employ people for coding...so you could still help coding
Gnutoo: but you must wait for the summer because the foundations for the actions script 3 has not been finished yet
Gnutoo: if I understood well
icewaterman: Gnutoo, i dont know action script, i am more of a C/C++ guy
Gnutoo: (else there would be a killer feature thing: fix Xvideo performances...I would love to do it but It's over my skills)
agd5f: the setup in the graphics driver is for muxing the audio stream in the video stream. the sound device itself is handled by alsa
Gnutoo: basically there were an initial patch for xvideo that had huge performances...they made it integrate less ugly and the performance were lost
icewaterman: brb, i am afk for lunch
Gnutoo: icewaterman, ok see you later
rah: builds iceweasel again for the 5th time :/
bridgman: the GPU chip includes generic HD Audio controller hardware (with some enhancements) so standard audio drivers can find it and use it
MarcOChapeau: agd5f: I should probably take a look at the code. it looks interesting. I wonder how much that involves
MostAwesomeDude: agd5f: Am I right to think that it should be setup at DRM load time?
bridgman: whenever the rest of the display hardware is set up, which AFAIK should be DRM load time
agd5f: MostAwesomeDude: probably whenever hdmi is enabled
agd5f: so either at load or when randr enabled/disables hdmi
agd5f: should probably only do it when hdmi is requested as opposed to dvi
agd5f: since they sue the same encoders in most cases
MostAwesomeDude: Oh, is there a separation in xrandr?
agd5f: MostAwesomeDude: no, but the driver knows when it sets up the port
agd5f: you can tell from the edid
MarcOChapeau: I wonder if sending audio over dvi actually breaks the monitors or just displays crap
MostAwesomeDude: Neither IIRC.
agd5f: MarcOChapeau: neither
MostAwesomeDude: It just reduces DVI bandwidth.
agd5f: just one less thing to worry about, less bandwidth used, etc.
MarcOChapeau: hmm, so it's all a legend
MostAwesomeDude: Which is really the only reason that I'd agree on turning it off when HDMI's not present.
MostAwesomeDude: Otherwise, it's policy as to when it should be enabled, and userspace should worry about ti.
MarcOChapeau: sounds reasonable
MostAwesomeDude: *it, even.
agd5f: should probably just be an attribute on the connector in xrandr
agd5f: so you can force it on/off if you want
agd5f: defautl would be auto
MarcOChapeau: that would allow me to finally delete my xorg.conf :)
agd5f: IIRC, there's a whole spec of stuff you can mux into the hdmi datastream
agd5f: device discovery, progam info, closed captioning, etc.
DanaG: oh yeah, how likely is it we'll be able to, at some point, do hotplug detection on the HDMI port?
agd5f: DanaG: monitor hotplug definitely, not sure about remote device discover stuff
DanaG: what that'd do: not give ALSA the device until an HDMI cable is inserted.
agd5f: DanaG: also always has the sound device, sound and graphics are exposed as separate pci devices
bridgman: yum, salsa
ajax: agd5f: the extended hdmi stuff is called 'cec', there's a spec for it, it's a straightforward slow serial bus.
ajax: it's hopefully wired to some standard gpio on the gpu?
DanaG: bridgman: do you know R6xx and R7xx hardware itself well? I'm curious if one would be able to take an RV635 MXM board that gives just stereo HDMI, and replace it (assuming BIOS support and all that) with an R700 card and get 6-channel HDMI support.
agd5f: but you could have a script that looked for hotplug events from the drm and enabled hdmi audio when an hdmi monitor was detected
ajax: the sane thing is to just enable hdmi audio immediately when you detect an hdmi sink
ajax: i started porting the hdmi code from rhd at one point but then got distracted
bridgman: DanaG; obviously thermals/power etc would also be a factor
bridgman: I don't think we have Linux support for >2 channel audio
yangman: DanaG: PCM over HDMI only has enough bandwith for "standard stereo"
ajax: yangman: depends on the mode.
agd5f: ajax: IIRC, I think the hw handles the transation, you just hand it bytes.
ajax: 1080p, yes, that's true. 720p you could do 5.1 if you wanted.
yangman: ajax: right
DanaG: Too bad nobody makes a simple HDMI audio-only decoder.
DanaG: ... without a full-sized receiver, I mean.
yangman: DanaG: typically 5.1 and higher is done with multiple channels compressed down using EAX or DTS
ajax: agd5f: for dumping the audio onto the hdmi cable? yeah, that's what it looked like, you just don't want to do that if it's a dumb dvi sink, they get confused.
DanaG: On R700, you can do real 6-channel LPCM.
agd5f: ajax: no the audio is handled directly by the hw, I was talking about generic transations for things like cloased captioning
ajax: ah, gotcha
agd5f: transactions even
yangman: DanaG: got a source for that?
agd5f: DanaG: the r6xx/r7xx hdmi audio bits are pretty similar
kig: does the video card take shader machine code as input or straight-up glsl?
kig: oh, probably machine code, otherwise adding support for newer versions of glsl would require firmware updates or other craziness
bridgman: kig; the shader core on the GPU has its own instruction set, designed to support both GLSL and HLSL but lower level than either one
bridgman: more details at http://www.x.org/docs/AMD/ file r600isa.pdf
bridgman: using an RV770 as an example, each instruction performs 5 floating point operations on separate data, typically the RGBA components of a pixel or the XYZW components of a vertex
bridgman: 16 pixels run the same instruction at a time on each SIMD engine
bridgman: and there are 10 independent SIMD engines
bridgman: that's where the 800 ALU number comes from - 5 operations x 16 per SIMD x 10 SIMDs
bridgman: so 1600 FLOPs per clock if you're running Multiply/Accumulate (MAC) instructions
kig: bridgman: i see. i'm reading the R6xx_R7xx_3D.pdf and it's interesting
bridgman: sounds good; once you finish that, the r600isa.pdf manual gives you the instruction details
bridgman: guaranteed to put you to sleep every time
bridgman: one thing to keep in mind; each SIMD performs 5 ops on each of 16 pixels per clock, but the 3D doc talks about wavefronts of 64 pixels
bridgman: each wavefront takes 4 clocks to run, so 64 pixels/vertices in 4 clocks is 16 per clock
bridgman: R600/RV670 have 4 SIMDs, RV770 has 10 SIMDs etc..
mcgreg: and R730? :)
mcgreg: RV730XT to be exact
bridgman: 8 SIMDs but they're smaller; 8 pixels/vertices per clock rather than 16
bridgman: RV610/620/RS780 have 2 SIMDs, each 4 pixels/vertices per clock
bridgman: RV710 has 4 SIMDs, I think
bridgman: RV740 has 8 SIMDs, 16 pixels/vertices per SIMD per clock
bridgman: (all the above "xxx per clock" should read "xxx per SIMD per clock"
kig: the split between (ES->GS->null | VS)->PS is because VS writes to the pixels shader (and is different from GS) while the ES only writes to GS input memory or something?
bridgman: yeah, the naming shifted around a bit trying to find something that was easy to understand
bridgman: basic idea is that with no GS you can go directly from VS to PS and keep the intermediate results on chip
bridgman: once you plug in a geometry shader (which can insert a bunch of new vertices) you need to go off-chip to get enough buffer space
bridgman: so the last shader before PS just reads in from the off-chip rings and plugs the results into the on-chip buffer to feed PS
bridgman: something like that anyways
bridgman: the real problem is that the term "export" ends up being confusing when sometimes you're going off-chip and sometimes you aren't
kig: ahh, so it's like [vertex inputs in ram] -> ES on chip -> [ram] -> GS on chip -> [ram] -> pass-through VS -> PS
bridgman: yep, although some of the docs reverse ES and VS ;)
bridgman: I need to check where vertex processing is done when using GS; I think it's after GS
bridgman: I have a nagging thought that the r600isa doc may be wrong in that respect
bridgman: that's one of the reasons we didn't document GS yet ;)
bridgman: of course I may be mixing up GS and tesselator again
bridgman: particularly since the primary use of GS seems to be tesselation ;)
kig: my tangential interest to this is that i'm writing a glsl validator for the mozilla gl canvas, as pretty much every driver has a different idea of what is acceptable glsl
kig: "hey nifty, a graphics card programming manual"
kig: google's o3d team went with the intersection of Cg and HLSL, using Cg to compile them on the gl backend
kig: the problem with that is that cg's closed source. but eh, it's probably the best solution for today.
bridgman: so the app uses HLSL and then uses CG to generate GLSL ?
bridgman: no, that doesn't sound right... (reads up on Cg)
bridgman: sorry, vertex shader *does* run before GS
bridgman: I guess HLSL and Cg seem to be pretty close anyways
nassrat_: I am running jaunty, and I recently bought a hd2400
nassrat_: i did everything i found, other than upgrading the kernel
bridgman: what does "did everything you found" mean ?
bridgman: jaunty should work out of the box on 2400 with the open source drivers
bridgman: no 3d yet though
bridgman: ie it should install the open source radeon driver + drm automatically
nassrat_: i dont care for 3d. with the radeon driver
nassrat_: i get serious graphic corruption
nassrat_: i cant see the cursor
nassrat_: and the graphics are corrupt
bridgman: does the corruption show up in a screen snap
nassrat_: whats a screen snap
nassrat_: oh i wouldnt know
nassrat_: let me try
bridgman: that tells us if the problem is with drawing or display
bridgman: drawing errors show up in a screen snap, display problems do not
nassrat_: i will try, also with radeonhd (just compiled from git) and fglrx its completely broke
nassrat_: radeonhd i get lines and colors i see the cursor as two vertical red lines
nassrat_: fglrx i get nothing
nassrat_: most of the time i get no keyboard response
nassrat_: so i have to raise skinny elephants
bridgman: is the image ok during boot ?
bridgman: ie when the card is in text mode ?
nassrat_: yeah up untill the last second, and then the logo goes all colourful
nassrat_: text mode is perfect (unlike 8.04 which i botoed from live cd which is more broke)
bridgman: ok, maybe the watermark fixes that went into radeon recently will help
bridgman: try building xf86-video-ati (aka radeon) from git
nassrat_: do i need to build mesa as well
nassrat_: i built dri for radeonhd
nassrat_: should i just build the ati driver
bridgman: nope, mesa and drm are the same for radeon and radeonhd
bridgman: just build the radeon driver
bridgman: ati is actually a wrapper around radeon, r128 and mach64
bridgman: don't really need it any more
bridgman: but the repository is still called ati
nassrat_: but i didnt build mesa
nassrat_: thats ok right
bridgman: mesa won't run on your card yet anyways
bridgman: no problem
bridgman: the copy of mesa that ubuntu installs will do software rendering
kig: bridgman: o3d passes the shaders straight to directx on windows, on mac and linux it uses the cg lib to compile them to arbfp1/vp1 assembly shaders
nassrat_: oh btw, i am running kubuntu (kwin) if that makes a diff
nassrat_: bridgman: the corruption is in the screenshot as well
nassrat_: i will now compile the new radeon driver
bridgman: oh, that's interesting
bridgman: there's an option something like "EXANoDFS" you should try
bridgman: check the radeon man page
bridgman: kig; that's good to know; going to arb_fp/vp rather than glsl, good idea
nassrat_: im going to upload the image to confirm its actually corrup and not double display corruption
nassrat_: or is that nonesens
agd5f: nassrat_: if it's just small pixmaps it's probably DFS related. Does Option "EXANoDownloadFromScreen" in the device section of your config help?
nassrat_: i think the corruption went away :) but i still cant see my cursor
nassrat_: it went away
agd5f: nassrat_: is the cursor corrupted or just not visible?
nassrat_: not visible
nassrat_: i have to hover on something to find it
agd5f: nassrat_: try Option "EXANoDownloadFromScreen"
nassrat_: yeah that took the corruption away
nassrat_: but i still have no cursor
nassrat_: find another option please
nassrat_: if we keep going this way, i wont have to return this card
agd5f: nassrat_: could try software cursor: Option "SWCursor" "TRUE"
nassrat_: no way
nassrat_: u guys are awesome
nassrat_: so what did i just enable exactly
nassrat_: i should ve came here to start
nassrat_: so what do these two options mean
rnoland: for some reason on a few pci based cards it looks like the scatter gather pages are getting overlapped on the framebuffer...
rnoland: that is why i'm getting corruption on those cards...
bridgman: ok, my understanding is that EXANoDownloadFromScreen uses software copies from frame buffer to system memory rather than using hw acceleration
nassrat_: thats no good
bridgman: and sw cursor paints the cursor into the frame buffer rather than using the on-chip hardware
agd5f: nassrat_: and the second renders teh cursor with software
agd5f: rather than using the hw cursor
nassrat_: thats crappy
nassrat_: so should i return this thing
bridgman: both are workarounds but neither should have that big an effect on performance
agd5f: very strange the the cursor's not working for you
nassrat_: ok then
nassrat_: so i guess ill take the hit and keep it
nassrat_: what about multiple cards
nassrat_: does anyone think that works?
bridgman: the first one is definitely a known problem, not your card's fault
bridgman: I'll let agd5f answer that one
nassrat_: i have onboard openchrome agp
nassrat_: it would be nice if i can get that working but i heard with the current kernel on jaunty i cant
nassrat_: not sure why
agd5f: nassrat_: as for multiple cards... maybe.
agd5f: depends which is the primary
nassrat_: beginning the experimentation
nassrat_: which should be
nassrat_: i can make either from the bios
nassrat_: i actually got a cloned screen (back when things were bad though) with openchrom as primary
nassrat_: i think i have the corg.conf file
agd5f: nassrat_: probably want to make the unicrhom primary since we can post the radeon without vbe
nassrat_: define vbe
agd5f: nassrat_: vesa bios extentions
agd5f: nassrat_: most cards need to be initialized prior to use. the bios generally does this when the card is posted during boot up
nassrat_: cool, ill reboot and switch it
nassrat_: ps i got corruption when the os tried to fade the background when i wanted to reboot
nassrat_: you know when the shutdown, reboot suspend options show up
agd5f: if the card is not posted, on radeons at least, we can initialize the card via special init tables in the video rom
nassrat_: woooosh ...
nassrat_: you mean flash the card
agd5f: just means that the radeon can be used as a secondary card more easily than some others
nassrat_: oh nice
nassrat_: what about that video corruption on fade out
nassrat_: do you have a magic option
agd5f: nassrat_: Option "EXANoUploadToScreen"
agd5f: just a guess
agd5f: I haven't seen any corruption on fade out
nassrat_: still the same
nassrat_: just slower
nassrat_: can i take it out
agd5f: nassrat_: sure
nassrat_: ok the org conf i had previously isnt working
nassrat_: should i generate one using X -configure
nassrat_: i got Cannot read V_BIOS
nassrat_: but it created a new config though
nassrat_: ok when i try to start x i get no video on both cards?
nassrat_: now im back to raising skinny elephants
nassrat_: whats the ideal xorg look like, for jaunty
agd5f: nassrat_: you might try the noint10 option with the radeon driver. I haven't tried multiple cards in a while
nassrat_: agd5f: just Option "noint10"
nassrat_: wierd thing is there are no errors in the log file
nassrat_: agd5f: with "Int10" "False"
nassrat_: the primary is now working
nassrat_: and the radeon (card2) is powering the monitor but no video
nassrat_: ok if i try to off the right of my screen
nassrat_: i get multiple cursor trals on my screen
nassrat_: i should also note
nassrat_: that it seems both my screen are somehow overlayed ont he same screen, or thats what it looked like when i was logging in
nassrat_: agd5f: any idea
nassrat_: oh the corruption went away
nassrat_: and now my cursor goes to the other side into the radeon connected monitor
nassrat_: on a black screen
nassrat_: i cant drag windows though ?
nassrat_: any idea, or pointers to where i can get more info
nassrat_: ok now i cant go anymore to the other screen, and i get corruption on the one screen again
nassrat_: its like its trying to draw the cursor ontop of my screen
nassrat_: agd5f: you still there
nassrat_: yeah im definitely getting both screens on one monitor, HELP
nassrat_: seems to have used up all agd5f
spstarr: nassrat_: bridgman is around :)
nassrat_: i gtg ill be back to bother them in a few mins
_Groo_: hi/2 all.. any radeon dev alive?
rah: $1000000 if you can figure out what's going on there
_Groo_: hi MostAwesomeDude.. btw are you alex?
_Groo_: MostAwesomeDude: well anyway, i sent the updated regs to my bug about the garbled screen
_Groo_: could you take a look please?
MarcOChapeau: I feel something is wrong since my last update... : http://rafb.net/p/NYxFnm98.html
MarcOChapeau: ideas ?
_Groo_: MarcOChapeau: are you using dri2/kms?
MarcOChapeau: _Groo_: if I am, I'm not aware of it. unless it has been merged into the branches I use
_Groo_: MarcOChapeau: what branches do you use?
MarcOChapeau: is checking what he has...
_Groo_: thinks that might help
MarcOChapeau: mesa is radeon-rewrite, libdrm master, xf86-video-ati master
_Groo_: MarcOChapeau: that might be the problem...
MarcOChapeau: _Groo_: what did I miss ?
bridgman: spstarr; bridgman was in Bowmanville grocery shopping and trying to find a new car battery
_Groo_: radeon-rewrite if for dri2, drm master doesnt have the radeon-rewrite bits yet, mesa does it, xf86 also doesnt have it yet..
bridgman: oh, and buying beer
spstarr: bridgman: car batteries ugh
bridgman: in theory radeon-rewrite will work on either legacy drm or kms/mm/dri2
MarcOChapeau: silly me
_Groo_: you need this 4 branches http://jglisse.livejournal.com/1822.html
bridgman: it's supposed to bridge the transition, ie the next year
_Groo_: bridgman: true :) but he his mixing wrong libs
bridgman: but if you want all the spiffy new stuff, then all new branches as Groo says
_Groo_: bridgman: so hes breaking drm
_Groo_: bridgman: and getting the errors in glxgears like he posted
MarcOChapeau: bridgman: actually it used to work 3 or 4 days ago with legacy
bridgman: carry on, I was out shopping ;)
_Groo_: bridgman: did you see my last update to my garbled screen bug?
_Groo_: oh, btw radeon devs..
MarcOChapeau: _Groo_: the simple alternative would be switching to mesa master ?
_Groo_: would be happy as a clam, if someone could fix xv too oO
_Groo_: MarcOChapeau: actually you need a new kernel, and the remaining branches
bridgman: Groo; no but will read through the backlog
_Groo_: or go to mesa master
bridgman: btw Alex is agd5f
_Groo_: bridgman: ah ok :)
_Groo_: and MostAwesomeDude?
bridgman: MostAwesomeDude is MostAwesome
_Groo_: airlied i already know.. and you i know from phoronix
_Groo_: bridgman: ahhh great... now i know :P ¬¬
MarcOChapeau: _Groo_: I wish I could just mess with my kernel, but I'm a -rt patch user... at least on this machine
_Groo_: MarcOChapeau: just patch 2.6.29 then...
_Groo_: MarcOChapeau: its worth the effort, dri2 in radeon is quitte amazing...
_Groo_: and very stable!
MarcOChapeau: _Groo_: ah, interesting
_Groo_: the only problem so far is xv and my stupid )(@**#&@(*#@(*# 200m garbled bug on suspend
MarcOChapeau: _Groo_: in fact I do have 2.6.29 (rt patched)
_Groo_: MarcOChapeau: then you are all set, just git pull drm-next , patch it with rt and be happy
MarcOChapeau: ok, let's give it a try
_Groo_: MarcOChapeau: building it is half the fun
_Groo_: brb, upgrading to kde 4.3 beta1
Neo_The_User: Has anyone tried libdrm 2.4.11 with radeon?
Neo_The_User: I was curious if it works with the xf86-video-ati driver from git master is all
MostAwesomeDude: I don't see why not.
Neo_The_User: Is Maciej Cencora in here?
MostAwesomeDude: Neo_The_User: His nick is osiris.
MostAwesomeDude: And yes, he's in here.
Neo_The_User: Oh ok. thanks
Neo_The_User: I love the new fixes
Neo_The_User: for radeon-rewrite. works good with ati drivers from git
nassrat_: So i got a multi-card setup semi working
nassrat_: for some reason both screens seem to appear on the same monitor
nassrat_: or is it the other way around
nassrat_: and I can move the cursor to the other physical monitor
nassrat_: yeah 2 screens on monitor0
nassrat_: and monitor 1 has no screens but i can move the mouse to it
nassrat_: anybody here got multiple cards working before?
nassrat_: hey, if I have randr 1.2
nassrat_: i cant use 2 graphic cards?
nassrat_: is this true
MrKeuner: hi all, I have a x300 card after jaunty update I cannot use fglrx anymore. I am using radeon module, and sometimes I get this strange behaviour that the screen colors slowly change from different colors and lasly going into full white. Everything freezes at that time. Can I fix this somehow?
MrKeuner: is there a keyword for this behaviour? I could not find a way to search the net
chithead: is it a laptop panel or external monitor? for the former I guess the display controller stops talking to the display
MrKeuner: chithead, laptop panel
MrKeuner: chithead, it happens most of the time when i run googleearth, but it also happened when i was switching users in Gnome
chithead: it is very strange that everything freezes, I would expect that only the display stops updating and the rest of the computer continues to run normally
chithead: back in the day, I had such problems with an x700 mobility. but it has gotten much better and I encounter this very rarely now
MrKeuner: chithead, I cannot know actually, may be the system was working at that time I did not try to ssh to my laptop when that happens. you guess it would work...
MrKeuner: do you think that is something I can fix as an end user?
chithead: maybe restarting the x server will help
chithead: other than that, wait for kernel modesetting to reach the distros. this should at least prevent this from happening with user switching
MrKeuner: Then I should be able to switch to the terminals as well?
Neo_The_User: KMS into a distro? That'll be awhile
chithead: if ctrl+alt+f1 does not work, ssh into the computer and run "chvt 1"
chithead: fedora 11 has kernel modesetting for amd, intel, nvidia
MrKeuner: that's a radeon problem or has my video card got broken somehow?
chithead: I think it is a driver issue. you may want to file a bug
Neo_The_User: speaking of fedora 11, howcome I get dri2 support and more GLX visuals and stuff than I do in any other distro?
Neo_The_User: on radeon r300 card
bridgman: Neo; install & run, should come up automatically
bridgman: F11 has all the KMS/MM bits except for newttm
Neo_The_User: oh thats why
Neo_The_User: i just need to build KMS
Neo_The_User: haven't tried that in awhile. been waiting for it to stabilize
nassrat_: bridgman: have you setup multiple vcards before?
bridgman: nope, only have one monitor ;(
davi: The drivers at http://people.debian.org/~bgoglin/rebuilds/
davi: are older than the sid ones
davi: The sid ones work better.
Neo_The_User: davi: should I use that?
Neo_The_User: or was that not directed to me
bridgman: that's not unusual; once drivers reach a certain point people stopp making "special" builds and just go with what's already in the distro
bridgman: generally newer = better
bridgman: Neo; don't think that was for you
Neo_The_User: oh ok
davi: Not specially directed to you Neo_The_User. It was just a general comment.
bridgman: F11 should just install and run, complete with KMS and TTM and radeon-rewrite etc.. (and DRI2)
bridgman: you probably just need to rebuild drm to get newttm but not 100% sure
Neo_The_User: ...I'll switch back to fedora 11 when its stable and offcially out
davi: bridgman, It would rocks if somebody backport the sid radeon driver to lenny, becouse the lenny one does not support my hardware
davi: My hardware is very new
bridgman: if it's just the radeon driver you need you can probably just install a package
bridgman: drm is usually the trickier part
davi: I will try, but last time it broke if I install it without the rest of the xorg drivers, kbd, etc.
Enrico|ITA|: davi: it is normal, there can be binary incompatibility between debian releases
davi: I understand. Maybe what I should do is compile by hand the driver and replace just that part
davi: Does the radeon driver include the OpenGL support. What must I install to get OpenGL support, being in hardware or in software?
Enrico|ITA|: davi: mesa provide opengl
Neo_The_User: davi: it provides it
Enrico|ITA|: davi: the radeon driver contain the ddx driver and the 2d accel part, 3d are contained in mesa
Neo_The_User: you just need mesa
davi: I suppose it it libgl1-mesa-glx
Neo_The_User: I use radeon-rewrite. works like a charm
Neo_The_User: yeah thats it
Neo_The_User: i just love doing things outside the package manager. my personally prefernce
Enrico|ITA|: well the package manager is usefull, i prefer rebuild the rpm/deb source
Enrico|ITA|: but i'm on gentoo now, so the problem don't exsist :D
Neo_The_User: I can't get Xorg and lxde working on gentoo so I gave up
davi: What command to check the OpenGL feature?
hax0r1337: Anybody knows a good way to debug progs/demos that come with mesa? I'm using gallium3d and there are many progs that lock up my GPU.
Enrico|ITA|: mhm i've fogotten the trick, it was a env var if i remeber well
Enrico|ITA|: somethin like OPENGL_DEBUG or MESA_DEBUG i can't remeber.... may be LIBGL_DEBUG
hax0r1337: cool, also MESA_DEBUG=1
Enrico|ITA|: i'm smart then :D
Neo_The_User: it works too
hax0r1337: what about debugger, gdb or valgrind?
Neo_The_User: i perfer gdb
Enrico|ITA|: gdb ftw
hax0r1337: gdb seems to freeze along with the apps/progs
Neo_The_User: hax0r1337: ssh?
nassrat_: should i disable or enable composite with the r600 (hd 2400) on the radeon driver
Neo_The_User: I have it on for fun
Neo_The_User: I have r300 though
Neo_The_User: seems fine
Enrico|ITA|: nassrat_: i don't think radeon support composite on r600 yet
Enrico|ITA|: nassrat_: couse the glx component is still missing in mesa
_Groo_: glisse: glisse are you alive?
Enrico|ITA|: nassrat_: but it should come soon
nassrat_: cant wait
Enrico|ITA|: nassrat_: these smart guys are doing a great work on it :D
nassrat_: i was about to return this thing, but i think the open source will come around
nassrat_: thats good to hear
Enrico|ITA|: nassrat_: eheheh i'm sorry but amd has released the doc with one year late..... so don't bother this great developers, the fault is of amd
nassrat_: yeah, atleast it works unlike fglrx
nassrat_: do you think i should turn off videooverlay and textured video
nassrat_: and tirn on opengloverlay
Enrico|ITA|: nassrat_: why turn off textured video?
Enrico|ITA|: nassrat_: but indeed you can try who works better
nassrat_: i saw somebody doing it online
Enrico|ITA|: nassrat_: and btw, fglrx for me work quite well, i prefer the FOSS radeon driver indeed, but i'm using an r600 too, and i like kde4 composite :D
nassrat_: wait what
Enrico|ITA|: nassrat_: and i play nexuiz, so i need 3d
nassrat_: is it a pci card
nassrat_: not express
nassrat_: but normal pci
Enrico|ITA|: nono pcie
nassrat_: yeah, mine doesnt work
nassrat_: and im using kde as well
nassrat_: i have not yet experienced the 3d
Enrico|ITA|: nassrat_: well fglrx should be tuned a lot to work well
Enrico|ITA|: radeon just works most of the times :D
nassrat_: i think i need to get myself an nvidia
Enrico|ITA|: nassrat_: i will never buy a nvidia card
nassrat_: why not
Enrico|ITA|: nassrat_: bad legacy policy, no FOSS official support
Enrico|ITA|: nassrat_: nouveau work indeed, but not as good as radeon can
nassrat_: yeah, thats what i thought
Enrico|ITA|: nassrat_: but the point is radeon has an official ati doc now, nouveao don't
Enrico|ITA|: nassrat_: and the nvidia blob driver isn't so far better than fglrx
Enrico|ITA|: nassrat_: any way radeon dev should continue to reverse engineer some part of the driver, like tvout and powersave..... THANKS RADEON TEAM YOU ARE GEAT GUYS!
nassrat_: i cant seem to play any vidoe :(
Enrico|ITA|: nassrat_: couse amd gives only 2d and 3d + register guide and some code, nothing more
Enrico|ITA|: nassrat_: the lastest radeon has EXA and Xvideo support (with lastest drm) but indeed you should disable opengl overlay (no 3d yet remenebr?)
nassrat_: its not working for me i think
nassrat_: is there any option i need besides AccelMethod EXA
Enrico|ITA|: nassrat_: indeed, no 3d for r600/700 atm
bridgman: Enrico|ITA|; it's actually AMD folks doing the reverse engineering ;)
Enrico|ITA|: nassrat_: well enable xvideo in section extension
bridgman: most of it anyways
Enrico|ITA|: bridgman: amd rever engineer they own driver?
bridgman: what would you call distilling a few lines of information out of millions of lines of code ?
bridgman: or tens of thousands of pages of doc
Enrico|ITA|: well that's true
bridgman: it's not really reverse engineering but it's close, and at the end agd5f ends up having to write code and then tinker until it works anyways
bridgman: didn't agd5f write the power mgmt code and most of the tvout ?
bridgman: airlied found the parser bug that was holding up tvout though
bridgman: just curious, other than 2d and 3d guides and register specs and working sample code, what else is there than twe couldn't have given ?
bridgman: a pony ?
nassrat_: video still isnt playing?
nassrat_: is it just OPtion "XVideo" "Enable"
bridgman: nassrat_ what do you get when you run xvinfo ?
nassrat_: adapter 0 Radeon Textured Video
nassrat_: and a bunch of stuff
Enrico|ITA|: bridgman: my girlfriend likes pony :D
Enrico|ITA|: so i will ask it on amd site :D
bridgman: Enrico; I'll see what we can do
bridgman: file a bug
Enrico|ITA|: bridgman: thanks :D
Enrico|ITA|: sure :D
bridgman: nassrat_; what happens when you try to play video ?
nassrat_: i dont see pictures moving
nassrat_: but if i go away and come back i get a new frame
Enrico|ITA|: nassrat_: are you sure you have the lastest drm?
bridgman: what player and output are you using ?
Enrico|ITA|: nassrat_: and indeed also the lastest libdrm
bridgman: youtube doesn't go through video acceleration AFAIK
nassrat_: yup i am using that
bridgman: use a real video and a real player
Enrico|ITA|: nassrat_: flash stuff? ehi don't make things harder try an ogv or ogg first
bridgman: mplayer or something
bridgman: mplayer -vo xv should use Xv output
Enrico|ITA|: nassrat_: since you are using kde, you can use smplayer
bridgman: mplayer -vo X11 -z or something like that bypasses Xv and scales in software if you want a comparison
_Groo_: bridgman: is xv fixed in latest radeon-wrewrite?
bridgman: xv is in the ddx (X driver, ie radeon or radeonhd), not mesa
bridgman: radeon-rewrite is actually a rewrite of the radeon subtree of mesa
bridgman: so no connection
Enrico|ITA|: in short is fixed :D
bridgman: um, yeah ;)
nassrat_: ok seems to work fine
Enrico|ITA|: or better, is working :D
nassrat_: actually playing a wmv using kmplayer straight up
bridgman: oh good
Enrico|ITA|: wmv........ crap :P
nassrat_: first thing i could find
nassrat_: so how do i fix utube
Enrico|ITA|: nassrat_: that's a flash plugin problem i think
nassrat_: i think it was working fine on openchrome
_Groo_: really? ist fixed????? git pulling NOW! :D
_Groo_: hum.. its fixed in what cranch?
bridgman: hold on Groo, what are you pulling ?
_Groo_: im using glisses
_Groo_: im using kms/dri2/ddx/mesa from glisses
bridgman: I don't think Xv was ever broken, was it ?
_Groo_: which has xv broken..
nassrat_: bridgman: yeah i can watch video on the onboard card
_Groo_: bridgman: i can assure you it is..
Enrico|ITA|: bridgman: i think that too
bridgman: I don't think that's the Xv code though, more likely a memory manager problem
_Groo_: bridgman: gl2 works fine, but xv cant "scale".. in a window it shows only partially, and at full screen it shows in partially also
bridgman: AFAIK nassrat is not using KMS/MM/DRI2/blah/blah
bridgman: yeah, I've seen that
_Groo_: oh yeah, it the "default" driver, xv works like a charm, so does textured
bridgman: was going to pull down the latest F11 over the weekend to see if it was still there
_Groo_: bridgman: if they are using the same branches, its still broken :P
bridgman: (confused) but they're not using the same branches
bridgman: only branch in common is radeon-rewrite branch of mesa and that has nothing to do with xv
bridgman: DDX is different and DRM is different
bridgman: sorry, you're talking about F11 right ?
bridgman: if so, they probably yes, just wanted to check on another system
bridgman: the video problems seemed to be system-specific from earlier reports
_Groo_: im talking about f11 yes
nassrat_: i am still getting a touch of video color corruption at the end of the kibintu usplash
_Groo_: bridgman: glisse told me is the way the driver is interpreting the input.. aparently textured doesnt work the same way xv is used to, lol
_Groo_: nassrat_: dri2/kms still has a lot of rgba color problems
_Groo_: i mean, bridgman
Neo_The_User: I learn so much idling
nassrat_: should i not use UXA
Enrico|ITA|: btw is hdmi audio output working in radeon?
_Groo_: nassrat_: uxa in radeon???
Enrico|ITA|: nassrat_: as far as i know uxa is for intel only
_Groo_: nassrat_: isnt it just for intel for now?
Enrico|ITA|: nassrat_: you should use EXA
Neo_The_User: EXA FTW!!!
Neo_The_User: agd5f made it better
_Groo_: bridgman: can you look at xv? or my bug? ill buy you cookies!
Neo_The_User: thats why hes so awesome
_Groo_: bridgman: and kittens, so you can have fresh meat!
Enrico|ITA|: well it is time to sleep here. good night all :D
_Groo_: the nice thing about radeon devs (you can write this down kids), is that they are always up for little fresh kittens!
Neo_The_User: good night
_Groo_: just say the forbidden word in here, and you'll see the throwing kittens around like ping pong balls
bridgman: don't need kittens, there's a big ol' racoon prowling around outside
bridgman: at least I know what he eats
_Groo_: bridgman: :D
_Groo_: thinks no one will ask what the forbidden word is
bridgman: I should really just put it out in a bowl so he doesn't have to pry the lids off the garbage cans and tear through the bags
bridgman: poor little guy could break a nail
_Groo_: bridgman: just eat him already :D
bridgman: need to check if my Small Game hunting permit lets me shoot racoons
_Groo_: will give everyone a hint about the forbidden word.. it's green.
_Groo_: bridgman: raccoon city.. so many memories, so many stories
_Groo_: nassrat_: auehauehhuae not even close... radeon devs dont care about the environment, they develop for ati aeuhauehauehuahe
bridgman: you have to like his attitude though... last time I chased him away from the garbage he scurried up one of the posts in my carport and sat on a crossbeam watching me
bridgman: and slowly eating a chicken leg
bridgman: yum yum this is good
_Groo_: bridgman: lol... they are funny :D
bridgman: I think racoons & bears are too smart to shoot
bridgman: which is unfortunate from a garbage and general carport tidiness point of view
_Groo_: bridgman: i couldnt shoot a living thing if it wasnt to save myself... i much prefer shooting criminals
bridgman: vegan ?
soreau: bridgman: You can't get some sort of locking mechanism for the garbage canister?
_Groo_: bridgman: i meant for sport.. no vegan, i like my meat
bridgman: so you pay someone to shoot 'em ;)
bridgman: soreau; I have bungee cords on the lids now; so far so good
bridgman: might just mean I didn't have any really tasty garbage though
_Groo_: bridgman: shooting to eat is one thing.. shooting for the fun of it, well , theres no point in it
soreau: Ah :)
bridgman: wouldn't shoot 'em for fun
bridgman: although I do throw the odd tin can for fun
bridgman: like to see their little eyes get wide
_Groo_: bridgman: everyone eats everyone, nature is like that.. but i dont see wild life killing for the fun of it... except maybe sharks and lawyers
bridgman: and I guess I might shoot *near* them for fun
bridgman: sharks don't kill for sport, they just get carried away
bridgman: might be the same for lawyers, not sure
_Groo_: bridgman: well if i were a shark i sure would looove to rip some stupid surfers apart
bridgman: you'd feel bad the next day
bridgman: if you remembered, which I guess you probably wouldnt
_Groo_: bridgman: yeah.. surf boards are hard to digest :P
_Groo_: bridgman: acording to NatGeo, sharks have very good memory...
bridgman: your bug was that big mess of debugger output in the pastebin ?
bridgman: interesting... didn't know that
_Groo_: bridgman: ???
bridgman: brains are about the size of a walnut IIRC
bridgman: aw, don't make me scroll back
bridgman: maybe it was someone else; pastebinned a bunch of stuff at MostAwesomeDude
bridgman: looked like debugger output
_Groo_: bridgman: its all about sinapses, not size.. women have smaller brains then man, and they are house trained!
bridgman: said something about updating register info in your bug, was that you ?
bridgman: but bigger than walnuts
_Groo_: bridgman: my bug is https://bugs.freedesktop.org/show_bug.cgi?id=16537
_Groo_: bridgman: must be me yes
bridgman: is that suspend with KMS or user modesetting ?
_Groo_: bridgman: user modesetting you mean x with dri1?
_Groo_: well the behaviour is the same with kms and with old model.. but with old model i could use vbetool to fix it, so i kinda forgot about it till now :P
nassrat_: has anybody on KDE4 ever lost his background and panels
nassrat_: trying to get multiple displays working
bridgman: looks like agd5f is already working on it ?
_Groo_: nassrat_: which kde4?
_Groo_: bridgman: on the bug?
nassrat_: if i turn on the radeon card
nassrat_: i get the background and everything
nassrat_: if i turn on the onboard card
nassrat_: i lose the desktop
bridgman: ah, ok, it was rah who did the pastebin right before you commented about the registers
nassrat_: i still have kwin
nassrat_: it seems something polluted my .kde dir
_Groo_: bridgman: since radeontool now allow to set registers, i could test some changes if you point me what to test
_Groo_: nassrat_: probably plasma screwed up, just remove/move plasma configs from .kde
DanaG: feeling random? www.instructables.com/id/Pie-in-a-Jar/
Neo_The_User: DanaG thank you
nassrat_: _Groo_: seems to be it thx
DanaG: Somebody in another channel I hang out in linked to that a few days ago.
_Groo_: nassrat_: you welcome
bridgman: didn't you just give the register dump to agd5f ? give him a chance to look at them ;)
_Groo_: bridgman: what else can i do? :D
_Groo_: im ansious by nature, sorry
bridgman: I don't know; go outside, see a movie, call your girlfriend
bridgman: it's saturday night (well, here anyways ;)
bridgman: you don't need to be sittin' around suspending and resuming ;)
bridgman: anyways, need to use the phone, bbl
_Groo_: lol, im married.. i have no life :F
Neo_The_User: 500 bucks says radeon-rewrite broke X on my machine. brb
spstarr: looks at phoronix's test of DRI2 (benchmark)
DanaG: heh, when a display gets up to around 147dpi, it starts to "feel" more "paperish" than lower-dpi displays.
DanaG: Well, at least on my matte display, it does.
spstarr: is taking for granted KMS and AGP non-corruption
spstarr: its like it never happened
DanaG: has R600. No 3d4me
Neo_The_User: omg! omg! it works
DanaG: yeah, I did just smush a few words (including a number-ified one) together.
spstarr: Neo_The_User: which chip?
Neo_The_User: 9800 PRO (R300)
spstarr: Neo_The_User: damin right it should
_Groo_: is going to bed, seeya all tomorrow
spstarr: Neo_The_User: AGP? Fedora?
Neo_The_User: good night
spstarr: Neo_The_User: KMS enabled?
Neo_The_User: no. debian sid. no
Neo_The_User: AGP 8X
Neo_The_User: no KMS
Neo_The_User: linux-2.6-tip kernel
Neo_The_User: i dont like fedora too much. its a heavy distro
_Groo_: Neo_The_User: loser;) real man use kms/dri2 ;)
Neo_The_User: well i like stable stuff
Neo_The_User: KMS only works in fedora for me
Neo_The_User: and its too heavy for me
Neo_The_User: i'd use arch but allanbrokeit broke gcc
spstarr: Phoronix is doing a good service, with the benchmarks it gives us a direction
Neo_The_User: he breaks everything he touches
_Groo_: Neo_The_User: lol
Neo_The_User: how do i turn on dri2?
Neo_The_User: That only works for me on fedora 11
_Groo_: you need to build glisse branch of kernel/ddx/drm/mesa
Neo_The_User: glisse's mesa branch???
Neo_The_User: zhasha told me radeon-rewrite in #dri-devel
Neo_The_User: oops. I pinged him by accident
Neo_The_User: I'll wait until it goes upstream
_Groo_: Neo_The_User: well... it will take a while
Neo_The_User: Do you think KMS will ever go into linus's kernel git tree?
Neo_The_User: I know. I like stable stuff
_Groo_: my guess.. about 3 more months at least
_Groo_: Neo_The_User: stable is no fun :P
Neo_The_User: well i use everything else unstable
Neo_The_User: distro (unstable) mesa (unstable) kernel (unstable) xf86-video-ati (unstable)
_Groo_: Neo_The_User: unstable is no fun :P
Neo_The_User: _Groo_ until I tested KMS, super unstable software was not my favorite thing to have
_Groo_: experimental, git pull every 10 min!!
_Groo_: carnage, destruction, mayhem
Neo_The_User: one day... your computer is going to eat itself
Neo_The_User: ugh! glxgears isn't working
_Groo_: well, i alays wanted to test a git pull of cgit.blackhole.com/fissiontest
_Groo_: but im waiting for switzerland to disappear first :D
Neo_The_User: drmRadeonCmdBuffer: -22
_Groo_: Neo_The_User: wrong ddx
_Groo_: are you using mesa master?
Neo_The_User: ah i'll use xf86-video-ati from git master
Neo_The_User: no. radeon-rewrite
_Groo_: Neo_The_User: thats why.. its broken that way, go back to master
Neo_The_User: eh who uses glxgears anyway!
Neo_The_User: I'm going to try openarena in high def
hax0r1337: Neo_The_User: OpenArena? This is what I'm playing atm: http://img154.imageshack.us/img154/1642/screenshot0000.jpg
Neo_The_User: i doubt my GPU could handle that
Neo_The_User: im upgrading soon though
spstarr: maybe we should just have one BIG fat fan covering the whole card? :)
spstarr_desk: X crashed with firefox again
spstarr_desk: that GPU lockup is not fixed on the r3xx