MostAwesomeDude: agd5f: Please add 4122905d to xf86-video-ati. This will add r5xx bicubic video filtering to textured Xv. I suggest fetching bicubic-xvideo from my repo, and then creating a new branch, maybe "git-push -v origin 4122905d:refs/heads/bicubic-xvideo" would be useful.
MostAwesomeDude: Everything after that commit is r3xx/r4xx, and may not work correctly (I haven't personally tested it.)
MostAwesomeDude: Thanks. :3
gentoofu: if only it's in the overlay (x11 layman), i would test it...
MostAwesomeDude: gentoofu: It's pretty easy to build by yourself. "git-clone git://anongit.freedesktop.org/git/~csimpson/xf86-video-ati"
MostAwesomeDude: "git-checkout -b origin/bicubic-xvideo"
MostAwesomeDude: Or is it people.freedesktop.org/git?
MostAwesomeDude: Hm, can't remember.
gentoofu: and sh uninstall.sh to uninstall or something like that?
MostAwesomeDude: Well, if you wanna switch to the overlay again, I'd just re-emerge from the overlay.
MostAwesomeDude: Like, in the directory, "./autogen.sh --prefix=/usr"
MostAwesomeDude: And then "make && make install"
MostAwesomeDude: And then to go back, something like "emerge -1av xf86-video-ati"
MostAwesomeDude: That way, you're never without a video driver.
gentoofu: hmm..
gentoofu: will try it sometime. not now :x
LeastAwesomeDude: Goddamn Comcast.
gentoofu: gross
LeastAwesomeDude: ?
gentoofu: Comcast... :S
gentoofu: I have Cox but it's not as gross xD
LeastAwesomeDude: Yeah, Comcast is a real pain.
diogo: hey, using the dri 7.0.4 with ATI 6.9.0 on paldo (so no patches pure products) and with blender I have 2 problems till now (my card is ATI Xpress 1150 RS482M)
diogo: let me do a screenshot of the problem
diogo: here is the grid problem
diogo: http://www.imagepaste.com/img/efae18010815e235859e891719640779.png
diogo: and here is if you try to get inside of an object every thing gets confused
diogo: http://www.imagepaste.com/img/42e25a91d2fe0973ca52bb92374c9e81.png
diogo: trying to upgrade to the mesa 7.1-rc4 now but does anyone knows a xorg conf option to fix problems like that
diogo: ?
diogo: so does anyone has an answer for the blender visual bugs?
diogo: hey can't compile the mesa 7.1 git version on my paldo linux the i830 doesn't let me
MostAwesomeDude: diogo: If you don't need it, then configure with --with-dri-drivers=r300,swrast
MostAwesomeDude: Or r200,swrast if you have an r200.
diogo: i have a rs482M...
diogo: its because the mesa 7.0.4 is not working for me and blender and also the 7.1rc3
MostAwesomeDude: Hm, that's right.
MostAwesomeDude: Anyway, that config will let you avoid the Intel stuff.
diogo: ok... is there any big difference between the 7.1rc3 to the git version
diogo: ?
MostAwesomeDude: Mm, not too much. Were you hitting a fallback?
diogo: I was getting white hashes on my screen... the 7.1rc3 fixed my problems of the mesa 7.0.3/4 but it got new problems with dri2 I think
diogo: what do you mean hittng a fallback?
MostAwesomeDude: It'll say "fallback" on the console.
diogo: you mean It would use the swrast instead of radeon?
diogo: yes it says fallback
diogo: on the 7.0.3 and on the 7.1rc3
diogo: got your question
MostAwesomeDude: Pastebin, please. I'll be back in a few.
diogo: ok
diogo: not using it... be back soon to
diogo: its not giving the fallack anymore
diogo: got the mesa 7.1 git version to install let me try it
diogo: back soon
diogo: MostAwesomeDude, I thought maybe I was using the wrong driver :q
diogo: :/
diogo: the best thing I can tell you right know is these images
diogo: http://www.imagepaste.com/img/42e25a91d2fe0973ca52bb92374c9e81.png
diogo: http://www.imagepaste.com/img/efae18010815e235859e891719640779.png
diogo: its what happens with the mesa 7.0.3/4
MostAwesomeDude: radeon works for *all* Radeons.
MostAwesomeDude: Hmm.
diogo: ok...
MostAwesomeDude: Looks like bugs, yeah. Hm.
diogo: these images are about the cube the main cube on the blender screen
MostAwesomeDude: My guess? Something poorly supported in the r300 DRI.
diogo: if I do a Option "DRI" "off" this stops happaning
MostAwesomeDude: Don't know what, though.
diogo: but the mesa 7.1 fixed this bugs though... but on the 7.1 I couldn't get to select anything at all
diogo: (talking about the Fedora which I used and the gentoo)
MostAwesomeDude: By any chance, did it give you the "render mode != GL_RENDER" fallback?
diogo: it was something like that
MostAwesomeDude: Yeah, that one? *Not* fixed.
diogo: let me just end upgrading to paldo unstable to see if it works here if not I may use the fedora live CD to see it
MostAwesomeDude: Mostly because it's a lot of code and not needed for most stuffs.
diogo: but talked to some people on fedora 9 (since they use the 7.1) and they said it happend on my card only and they didn't have it.. they didn't said what cards did they have but it seems that it happens only on my chipset
MostAwesomeDude: Hm.
MostAwesomeDude: That's *very* odd.
diogo: got the mesa 7.1rc3 to work in here just a minute to see if paldo is really the greates distro for my card
MostAwesomeDude: Distro shouldn't matter. Very few people (except Ubuntu) mess with Mesa/DRM.
diogo: well since paldo is pure linux (no patches at all) so it has been a while that it has made some difference here like for example the response of my printer
diogo: like also the detection of my touchpad by the xorg
diogo: so it has been a good experience till now so if the paldo mesa 7.1rc3 doesn't have this bug it means that the select bug is fedora one
diogo: let me reboot to see just finished the upgrade be back soon
diogo: so no it didn't fix
diogo: I'll imagebin the screenshot of what happens
diogo: http://www.imagepaste.com/img/255c6a01e685f1f5fd1fb391685db799.png
diogo: all the time I try to select an object with the right click it doesn't select it and do this hashes on the screen
diogo: and your bug you said is here
diogo: I'll paste bin it
diogo: http://pastebin.ca/1182572
diogo: here is the bug
MostAwesomeDude: Hey, there it is.
MostAwesomeDude: Probably wants GL_SELECT.
nha: diogo: that message itself is not a bug
diogo: I'm using the mesa 7.1rc4 which has a fix bug comparing to 7.1rc3 about the white hashes with selection box on blender the only bugs still here are this and wrong screen color
nha: we should really fix the message to make that cleaer
MostAwesomeDude: nha: I don't know much about RenderMode. How would we make GL_SELECT work?
nha: MostAwesomeDude: I have my doubts whether it's actually GL_SELECT mode that is the culprit
nha: basically, we should just get out of Mesa's way and let the software rasterizer do its thing
diogo: ok do you think that maybe the gdb can make some difference
nha: which is what we used to do
nha: diogo: no
diogo: I'll paste the wrong colors screenshot to show
nha: so either the fallback logic got broken by something, or it's more of an interaction between different problems
nha: e.g. software fallbacks access some of the buffers incorrectly
nha: diogo: this kind of problem really needs to be analyzed by someone with the necessary background understanding of OpenGL
MostAwesomeDude: nha: So, not me. :3
nha: :/
nha: I have an exam in a couple of days, so I'm not going to do it either
MostAwesomeDude: Just kidding. But I don't know where to start.
diogo: here is a shot of inside of the gray cube but it seems that one wall is blueish
diogo: http://www.imagepaste.com/img/55e6c97e76b749f4c06ba53cb3d027af.png
MostAwesomeDude: Actually, no. I know where to stare.
MostAwesomeDude: I should look at the code for fallbacks.
diogo: this other bug I just said caused something else too
nha: one of those days when I'm bored I will sit down with Blender and figure out those bugs and write test cases
nha: work on the r300-bufmgr branch goes first, though
diogo: as you may see this is a project of mine and it's done all in gray only the doors are brown...
diogo: http://www.imagepaste.com/img/db1fe803ad79bf769ef8e144787ce83c.png
diogo: but as you can see it is still blueish
diogo: kinda greenish too
nha: especially now that it seems Mesa master will be free for development again soon
nha: diogo: that's quite an interesting bug
nha: not really unexpected though - Blender uses quite different GL codepaths than games, for example
nha: so it's understandable that those paths are less well tested
MostAwesomeDude: True.
diogo: nha... please one of these days no... please be soon I need this laptop for a workstation and also can't trust fglrx guys (they suck... not true they are awesome on being slowness on blender :P)
nha: diogo: I understand. Unfortunately, I'm just doing this part time when I'm not studying, and so are most (all?) of the others working on the driver
diogo: yeah I know... the selection is not possible only if I use the mouse
diogo: if I use the outliner it does work which is weird..
MostAwesomeDude: nha: I'd like to think that this is like the greatest experience evar. Unfortunately, I sorta suck at it. :c
nha: MostAwesomeDude: But when you show the ladies the rotating cube, you can say that you're behind it all ;)
MostAwesomeDude: nha: Ladies?
nha: MostAwesomeDude: Seriously though, OpenGL is huge, try to find the small but satisfying things...
MostAwesomeDude: Yeah.
diogo: going to try the aiglx be back
MostAwesomeDude: I *am* proud that I'm responsible for turning one of the squares on the feature matrix from "TODO" to "DONE"
nha: :)
MostAwesomeDude: But all in all, I'm just a guy who writes code when he gets bore.
MostAwesomeDude: *bored, eve.
MostAwesomeDude: *even, even.
nha: *even
nha: ;)
MostAwesomeDude: Gotta clean out my keyboard.
RTFM_FTW: oh come on... OpenGL driver development is fun! :D
MostAwesomeDude: This is true. I probably wouldn't enjoy it as much, if it were my job. :3
_neuron_: sigh, wish the nvidia dev's felt that way, I've gotten two double free crashes from glibc in libgl today (got an ati card ready to go when pal tvout starts working on R500 :p )
_neuron_: but asking women over to watch movies is the only way I'm getting any action lately, so that has to work :p
MostAwesomeDude: You mean nvidia, or nouveau?
_neuron_: nvidia, nouveau isn't really in a functional state, well, last time I looked into it anyway
cxo: So i bit the bullet an ordered a 4870 instead of a 9800gx2 like i should have
cxo: is there a homepage for this driver?
cxo: cant resist shiny red things
MostAwesomeDude: Homepage?
cxo: yeah for the driver?
MostAwesomeDude: Not really. Well, not up-to-date.
cxo: or what's the git url?
MostAwesomeDude: puts "update radeon page" on todo list.
MostAwesomeDude: The git URL is git://anongit.freedesktop.org/git/xorg/driver/xf86-video-ati .
cxo: so does this driver work with wine stuff?
MostAwesomeDude: Yeah.
MostAwesomeDude: No 3D, though.
MostAwesomeDude: No acceleration for the HD 4870.
MostAwesomeDude: Yet.
cxo: what happened to all that atombios stuff, i thought that's supposed to abstract the hardware differences away
MostAwesomeDude: Oh, yeah, it does.
MostAwesomeDude: But it only covers modesetting.
MostAwesomeDude: Not acceleration.
cxo: oh
MostAwesomeDude: Acceleration can't be done without the 3D engine.
MostAwesomeDude: And for that we need docs.
MostAwesomeDude: And AMD should be giving us docs in about two weeks.
cxo: My understanding is that, the card has all the opengl functions on it, so its just a matter of sending the appropriate commands to it
RTFM_FTW: heh no
MostAwesomeDude: XD
MostAwesomeDude: Nope.
cxo: oh
MostAwesomeDude: The card has a pipeline.
RTFM_FTW: there isn't anything on the ASIC itself really tied to OpenGL
cxo: oh
MostAwesomeDude: You have to set up the pipeline, and then send it card-specific commands.
cxo: so you have to do all the rendering by hand depending on what functions the card gives you?
RTFM_FTW: mmm you guys already have some of the shader docs in the public realm for R6xx right?
MostAwesomeDude: RTFM_FTW: We have the ISA, so I could write a shader compiler, yeah.
RTFM_FTW: yep
MostAwesomeDude: cxo: Sorta. Read the r5xx docs to get an idea of how the stuff works.
RTFM_FTW: has someone started on that yet? :D
RTFM_FTW: heheh
MostAwesomeDude: RTFM_FTW: Some guy's got standalone C++ code for one of the GPGPU languages, but I don't know what happened to it.
MostAwesomeDude: It was some research scientist.
RTFM_FTW: interesting
MostAwesomeDude: I have *no* idea how we're gonna do GLSL.
cxo: so these drivers must be pretty complex
MostAwesomeDude: In some parts.
cxo: so the HD3xxx family had 3d support now?
RTFM_FTW: yes a modern OpenGL driver is very complex
MostAwesomeDude: The most complex part, in my opinion, is the part that synchronizes all of the writing to the card.
MostAwesomeDude: cxo: No, no acceleration of any kind.
MostAwesomeDude: The r6xx ISA lets us write part of the compiler for shaders. There's still a lot more we need.
cxo: why are they so stingy with docs?
MostAwesomeDude: Actually, they're being *very* open.
_neuron_: they arnt, they just need to run it through legal first, make sure there arnt any patent issues that they licensed from other companies
loswillios: hi guys
_neuron_: intel and AMD are the only ones with open specifications for gpu hardware
cxo: ati_drv and radeon_drv, whats the difference
RTFM_FTW: MostAwesomeDude nah that is fairly trivial AFAIC
MostAwesomeDude: Patent issues, readability, etc.
_neuron_: ati loads radeon if you got a radeon card, and mach64 for example if you got hardware from the 1700 century
MostAwesomeDude: RTFM_FTW: Well, yes and no. Supposedly the Suse guys have the DRI working, except it doesn't work. :3
cxo: so if something is patented, it cannot be open source?
loswillios: I got what I deserved when I upgraded from f9 to rawhide today. X won't start anymore (just black screen) :p
RTFM_FTW: much of the complexity comes into play when you start having to manage resources (texture management for example) and you have to compile and optimize shader code
_neuron_: not if they licensed the patent from someone else no
_neuron_: it'd be illegal for them to open source it without owning it
RTFM_FTW: in fact nearly all of a driver falls into some level of "clever" resource management :)
cxo: is AMD an american company? i thought software patents was an american invention
_neuron_: it's an american company, but software patents isn't just in the us, and we are talking about hardware implementations :P
cxo: so gay
cxo: in my xorg i put ati_drv ?
MostAwesomeDude: cxo: Please don't say that. Not all of us are straight. :3
MostAwesomeDude: Anyway, I can guarantee you that AMD has more lawyers than any of us, and they know what they're doing.
MostAwesomeDude: So we just have to trust their judgement.
MostAwesomeDude: And so far they've provided us sufficient documentation.
cxo: whats the name of the driver i put in my xorg.conf?
_neuron_: "ati"
_neuron_: or "radeon", whichever is fine
cxo: that's confusing
_neuron_: ati just loads the driver needed, ati has a seperate driver for really old hardware, such as mach64
_neuron_: in which case ati would load mach64, so for them "ati" or "mach64" would work :p
MostAwesomeDude: cxo: "ati" tries to load all different ATI drivers. "radeon" is for Radeons.
cxo: i think the best ATi card i have run on linux was the Radeon 8500LE
cxo: all those cedega games they say "does not work on ati graphics cards", all worked
MostAwesomeDude: My X1600 and X1950 work fine.
_neuron_: just missing pal tvout support :p
cxo: MostAwesomeDude, can you play Call of Duty 4?
MostAwesomeDude: cxo: Not a fan.
MostAwesomeDude: So I haven't tested.
cxo: yeah, but you should try it out, a lot of people are having trouble with that card on that game
MostAwesomeDude: I'm a student, I don't have much cash, and I don't spend it on games.
MostAwesomeDude: Hm.
cxo: try the demo
_neuron_: having a pcie card + an old pci 9200 card should work just fine right?
MostAwesomeDude: What kind of trouble?
cxo: check the wine app db, most of them have posted logs
_neuron_: had some problems with that combination
_neuron_: if I wasn't so lazy I'd walk into the next room and try it, but, unfortunatly, I am
cxo: can you run 2 pci-e graphics card at the same time?
_neuron_: think you can, this is pcie+pci though
cxo: yeah i know
cxo: like i could pick up one of those x1950s now cos they are cheap and use it on linux
cxo: and have the 4870 work in windows
MostAwesomeDude: cxo: Not seeing any logs. Linky?
cxo: ok i will find it
cxo: MostAwesomeDude, this thread has been going on for a few months now http://bugs.winehq.org/show_bug.cgi?id=12929
cxo: woops, i think thats the fglrx driver
MostAwesomeDude: ...That's fglrx.
cxo: yeah i forgot
cxo: why doesnt he use the radeon driver
cxo: will he have the same problem hes got?
_neuron_: a lot of people just think of the one driver
cxo: so the x1950 is fully supported with the radeon driver?
MostAwesomeDude: Yes.
cxo: whats the equivalent of that on the nvidiia world, in terms of performance?
MostAwesomeDude: Dunno.
MostAwesomeDude: I don't really think it's a fair comparison.
_neuron_: in my experience, nvidia is faster, but a hell of a lot more broken
_neuron_: 3d is faster, 2d is damn useless
cxo: i mean to a particular model
_neuron_: my 9200 PCI card is faster in 2d than my 8800GTS if that puts things into perspective
cxo: i mean in 3d of course
_neuron_: for 3d my 8800GTS is generally fast, but for example I've been testing it a bit today, and libGL has crashed twice due to a glibc detected double free
cxo: i wonder if you could LD_PRELOAD free() and monitor all the address that are freed :)
cxo: Do you need a special board for Cross fire? or will any board work?
_neuron_: probably could, but working around bugs in a closed source driver is crap :/
_neuron_: think you need special motherboards for that still
cxo: if it lets you kill the boss on the last level, i'd do it
cxo: so how does this work, App->libGl->driver->hardware ?
_neuron_: yeah
MostAwesomeDude: For non-nVidia drivers, app -> Mesa (libGL) -> DRI -> DRM (kernel) -> HW.
cxo: and ati/nvidia supply their own libGL?
MostAwesomeDude: nvidia's official drivers subsume the Mesa-DRI stages and just use a massive blob.
MostAwesomeDude: fglrx uses a patched Mesa for SW fallbacks, and their own DRI.
cxo: which is the cause of all this misery?
MostAwesomeDude: Both binary drivers also have a kernel blob.
MostAwesomeDude: Anything glibc catches is still in userspace.
cxo: so at the rate things are going, when do you think the newer HD48xx cards will get the kind of support your x1950 has?
MostAwesomeDude: A month or three.
MostAwesomeDude: Hard to say for sure. Depends on how clean the DRI Suse's been working on is, and also depends on docs.
cxo: AMD has got novel people on their payroll?
MostAwesomeDude: No, they have radeonhd devs under NDA. :3
MostAwesomeDude: agd5f is an AMD employee, though.
MostAwesomeDude: So is bridgman, but he's on vacation.
cxo: do you know why the fglrx driver package is twice the size of nvidia's?
MostAwesomeDude: cxo: Double the bugs?
cxo: a friend of mine a few years back, went for an interview at ATi (before the merger), to work on Driver code, he said that the interviewer was so arrogant, that he just got up and left
cxo: he's working at nvidia now
cxo: each new catalyst never mentions a bugfix in their changelog
cxo: do they actually fix bugs?
MostAwesomeDude: We may never know.
cxo: like what is it, compiles, boots, yay lets ship it
_neuron_: did you have an ati card before the merger? :P
_neuron_: AMD has done SO much to improve the quality of the driver :P
cxo: i did, and i said it was the best ATi card on linux for me, 8500LE
cxo: but yeah amd has done a lot
_neuron_: had a 9700 card back when it was new, and it was absolute rubish, I switched back to nvidia within a month
cxo: 9500, 8/16 pipeline hack was fun
cxo: heh who writes the changelog
cxo: "The entire X desktop including the mouse cursor is choppy during playback of video files or suzi test app "
cxo: the guy cant even spell SUSE?
cxo: or is suzi something funky cool
cxo: "Connecting an HDMI display device may result in the audio failing to be heard" <-- "failing to be heard"? that's a problem with your ears
cxo: wait i should be trolling in #ati
MostAwesomeDude: Yes.
_neuron_: hehe
_neuron_: gah, gonna go yank out the nvidia card in this box and try 2 ati cards (old pci one for tvout)
_neuron_: was gonna wait for r500 tvout pal support, but the latest nvidia driver has some text corruption bug which is driving me insane :P
cxo: 177.68?
cxo: oh he left
_neuron_: for R500 do you need mesa 7.1 or 7.0.3 ok?
MostAwesomeDude: _neuron_: Mesa 7.1rc3 or newer IIRC.
_neuron_: modprobe radeon's drm module while being in X probably wasn't a great idea was it :p
MostAwesomeDude: Meh, just don't use -f. :3
_neuron_: I'll unmask that, thanks
_neuron_: hehe
MostAwesomeDude: _neuron_: For Radeons, running -9999 is safe.
_neuron_: fun, loading radeon while in X while having the fglrx module loaded = not a good idea :p
_neuron_: started a modprobe loop that ate up all my memory ;)
_neuron_: hm, if I got two display devices plugged in, and one of them not in xorg.conf, can I still use that tvout with xrandr?
spstarr: hmm
spstarr: airlied: modesetting support is in for r300?
spstarr: Linux segfault.sh0n.net 2.6.27-0.244.rc2.git1.fc10.i686 #1 SMP Fri Aug 8 13:26:20 EDT 2008 i686 i686 i386 GNU/Linux
spstarr: i get a garbled screen on boot, then it goes blank.. and then well.. i can't log in :)
spstarr: er
spstarr: 273
airlied: spstarr: I have a newer version coming soon, but I think 244 should've worked..
airlied: spstarr: what outputs?
spstarr: VGA-0 disconnected (normal left inverted right x axis y axis)
spstarr: DVI-0 disconnected (normal left inverted right x axis y axis)
spstarr: LVDS connected 1400x1050+0+0 (normal left inverted right x axis y axis) 0mm x 0mm
spstarr:
airlied: hmm LVDS should've worked.. you can stick text on the command line to make it boot old way.
spstarr: oh 'text'
spstarr: so by default it will use plymouth even with rhgb not listed?
airlied: spstarr: it'll use modesetting not plymouth
spstarr: ah
airlied: actually it always runs plymouth
airlied: rhgb stops it running the plymouth graphics bits
spstarr: im just gonna wait and restrict kernel/nash/mkinitrd and friends for a little while, until you say its gold to use
airlied: spstarr: how you going to know I fixed it then?'
spstarr: hmm
spstarr: well, true :)
cxo: you know what will be funny,
cxo: if wine implements the windows 7 api before microsoft does :)
spstarr: .net ;p
damentz: hahaha
spstarr: airlied: this is the mobile rv350
agd5f: I haven't actually tested LVDS on legacy chips yet
spstarr: ah
agd5f: need to build a more recent kernel on my laptop
MostAwesomeDude: agd5f: Get my earlier message?
agd5f: MostAwesomeDude: yeah. No time to merge tonight. I'll take a look tomorrow
MostAwesomeDude: Alright. No rush. :3
spstarr: decides to have a powernap
Ori_B: Ugh. it looks like xrandr 1.2 is broken when laying out the monitors
Ori_B: radeon 3870, 64 bit system, radeon 6.9.0
Ori_B: the mouse moves like the 2 monitors were there, but the display shows up in clone mode
Ori_B: command I'm using to set it up is:
Ori_B: xrandr --output DVI-1 --right-of DVI-0
Ori_B: anyone have an idea of how to fix it?
spstarr: XrandR is lying to me on my old GoldStar CRT
spstarr: it says i can only go 65Hz when 75Hz is just fine
spstarr: i can override this in xorg.conf though
spstarr: but still.. I thought XRandR could defeat such bogus EDIDs
spstarr: even worse, the monitor can do 1024x768 fine too at 60Hz
spstarr: but XrandR doesn't even list it
Ori_B: hm. it looks like it broke sometime betweeen 6.7 and 6.9
Ori_B: argh.
cxo: are u faster than the 4870x2?
Ori_B: since it doesn't have legs, I can assume that I'm faster than it.
cxo: http://forum.donanimhaber.com/m_25688835/tm.htm
Ori_B: however, it's far better at driving monitors, although not quite good enough.
cxo: basically a 100x 4870s are as fast as the brain
cxo: which mean by like 2015 integrated graphics would be more powerful than the brain
Ori_B: uh. basically a 8086 from the early 1980s is faster than the brain.
Ori_B: it's not speed. it's what you do with it.
cxo: no it isnt
Ori_B: and it takes about 0.1 seconds for an impulse to propagate from one side of the brain to the other.
cxo: what we do in a micro second huge hives of computers take days to do
cxo: classic examples are things like image or sound recognition
cxo: fuzzy logic
Ori_B: and what an 8086 takes milliseconds to do, we take days to do. (here, let me give you 1000000 numbers between 0 and 65535. add them.)
cxo: i was never taught to do that
cxo: so i cant,
cxo: but if i was taught to do that,
cxo: i could
Ori_B: you were never taught to add numbers? I'm sorry.
cxo: only small numbers
MostAwesomeDude: "Fuzzy logic" was one of the worst buzzwords of the nineties.
MostAwesomeDude: Anyway, image recognition isn't about speed; it's about quality.
cxo: and what i can do in a second, an 8086 will take eternity to do, i'm a jazz buff, give me like 5s of any western jazz tune between the the 1920s to 1970s and i;ll tell you all about it
Ori_B: cxo: sports cars are faster than dump trucks.
cxo: depends
Ori_B: so of course they're going to be much better at moving lots of construction materials, right?
cxo: mercedes make some killa trucks
Ori_B: we'll just load several tons of earth onto them and construction will become so much easier!
cxo: well depends on the sports car, american sports cars are basically dump trucks
Ori_B: the brain is far slower than any processor. it's also structured completely differently.
cxo: wait.. i think that was an oxymoron "american sports car"
MostAwesomeDude: Classic example: My Ford Escort gets 32 MPG. My neighbor's Hummer gets 5 MPG.
cxo: maybe your brain is
cxo: Cossworth made an RS Escort, beautiful car
MostAwesomeDude: Anyway, the thing about the brain is that it doesn't operate on quantized or discrete data.
cxo: fuzzz
MostAwesomeDude: It operates on *continuous* data.
Ori_B: the issue with AI isn't speed. it's how to process data.
Ori_B: to make a computer emulate the brain, you have to come up with completely rigid instructions on how to be flexible.
MostAwesomeDude: And there will *never* be a non-biological continuous data processor.
MostAwesomeDude: (Unless somebody comes up with something truly elegant...)
cxo: mad, do you have a license for that nick? it's too long to be legal
Ori_B: anyways. off topic.
cxo: way
MostAwesomeDude: cxo: Believe it or not, it's actually shorter than my full name.
Ori_B: MostAwesomeDude: have you ever looked into the XRandR code?
MostAwesomeDude: More importantly, you can always Tab it out.
MostAwesomeDude: Ori_B: Nope. What would I look for, if I were to go there? Got something in mind? :3
Ori_B: MostAwesomeDude: "make my dual monitors work again", maybe?
Ori_B: =)
Ori_B: MostAwesomeDude: I just updated my desktop (details above, I can pastebin my log too) and dualhead stopped working
cxo: is there a tool like aticonfig for the radeon driver to issue pplib commands?
Ori_B: it seems that X thinks it works, but the viewports are always on the secondary monitor.
Ori_B: cxo: xrandr, pretty much?
MostAwesomeDude: Ori_B: Have you confirmed that it's a break in radeon and not in X?
Ori_B: MostAwesomeDude: nope, but since it seems to be an issue with setting up the monitor viewport I assumed it was the radeon driver.
Ori_B: MostAwesomeDude: how would I debug?
MostAwesomeDude: Ori_B: Well, if you're confident of a previous working version...are you in git?
airlied: Ori_B: you might try adding --crtc 0 and --crtc 1 in some places.
MostAwesomeDude: First I'd check that the old version works.
Ori_B: MostAwesomeDude: no, I'm on Debian packages... and the older ones seem to have gone away =/
MostAwesomeDude: Hm.
airlied: i think the cloning support in xrandr sucks ..
Ori_B: I can build from git, but that's a pain.
MostAwesomeDude: Yeah, I know.
MostAwesomeDude: 's X1950 is in a Debian workstation
Ori_B: *tries*
MostAwesomeDude: Ori_B: airlied knows his stuff, try that first.
Ori_B: MostAwesomeDude: I know he does =)
MostAwesomeDude: airlied: So I heard a whisper that you've started on r3xx KMS?
airlied: MostAwesomeDude: well agd5f did..
airlied: MostAwesomeDude: its all in the modesetting-gem branches.
Ori_B: airlied: hm. "xrandr --output DVI-0 --crtc 0 --right-of DVI-1 --crtc 1"?
airlied: should do r100 up to r500 now..
cxo: Ori_B, like, aticonfig --pplib-cmd "get temperature 0"
airlied: Ori_B: xrandr probably too many crtcs there.
airlied: you might need to set them separate first..
airlied: check with xrandr -v
MostAwesomeDude: airlied: Okay, *now* I have to get this working. Too cool-sounding to pass up. Should I just pound on the kernel till it works, or does anholt's tree work too?
airlied: if they are ending up cloned?
Ori_B: hm. that's.. interesting.
airlied: MostAwesomeDude: either anholts tree or my drm-rawhide tree.
cxo: mad, so is the x1950 the best ati card to get on linux?
airlied: http://git.kernel.org/?p=linux/kernel/git/airlied/drm-2.6.git;a=shortlog;h=drm-rawhide
Ori_B: airlied: so with the command I pasted above, it gives me the left monitor on, the right monitor off.
Ori_B: if I swap it, it gives me both monitors cloning the right monitor
airlied: Ori_B: wierd.. check with xrandr -v for the crtcs being the same.
Ori_B: and '-v' gives me the xrandr version; --verbose?
airlied: ah --verbose
Ori_B: Clones: DVI-1
Ori_B: CRTCs: 0 1
airlied: really needs to add don't clone support to xrandr
airlied: it sucks.
Ori_B: hm. ignore that.
Ori_B: ok. yeah, it seems on insisting on using CRTC 0
MostAwesomeDude: cxo: IMO, yes. For now.
airlied: &
Ori_B: airlied: ok. so:
Ori_B: randr --output DVI-0 --crtc 1 --right-of DVI-1 --crtc 0
Ori_B: ^-- gives me 0,0 for CRTCs
Ori_B: randr --output DVI-0 --crtc 0 --right-of DVI-1 --crtc 1
Ori_B: ^-- gives me 0,none for crtcs
Ori_B: *xrandr
airlied: Ori_B: the problem is the second --crtc1 doesn't sapply to the right-of
Ori_B: ok.
airlied: I think you need to set DVI-1 onto crtc 1 first
Ori_B: ok, it's doing funny stuff.
Ori_B: xrandr --output DVI-1 --crtc 1
Ori_B: that seems to switch my DVI-1 to crtc 0
Ori_B: and disables DVI-0
cxo: MostAwesomeDude, can i have both my 4870 and an x1950 in my pc at the same time and just use the x1950 when in linux?
MostAwesomeDude: cxo: Yeah, if you specify the PCI ID.
MostAwesomeDude: *PCI bus, even.
Ori_B: airlied: so, I'll pull down git in a few minutes and try that.
Ori_B: airlied: would I need a specific version of the server, or would 1.4.2 do?
Ori_B: or.. *tries what Debian Experimental has*
Ori_B: (...it can't hurt that much, can it?)
Ori_B: *cough*
Ori_B: well, here goes...
cxo: gives Ori_B a hit of his weed
Ori_B: well. that went badly.
Ori_B: input devices don't work now
cxo: call Dell Customer support
cxo: tell them you upgraded your x-server and now your input devices dont work, and your pissed
Ori_B: heh.
Ori_B: well, I recorded what I need to downgrade to get things half-working again
cxo: back porting x stuff?
cxo: gross
Ori_B: eh? no just getting rid of the experimental stuff
Ori_B: ie, all the git snapshots that apt-get -t experimental xserver-xorg-radeon pulled in
cxo: what to do on a saturday night and you dont feel like going out!?
cxo: i think i'm gonna work on a kconfig for my box
Ori_B: sighs and decides to try fglrx for the night
Ori_B: ...or does that have an amd64 port?
cxo: eh?
cxo: its just a blob, you can run it on anything
Ori_B: uh. it's a blob. so you can only run it on what the blob was compiled for.
odz: when you compile a new version of xorg-server you need to recompile all input drivers you are using
cxo: between x86, i don think that's a biggy
Ori_B: well, that didn't work =(
Ori_B: grumbles.
cxo: fglrx is -supposed- to work with anything up to xorg-7.3
Ori_B: well, all it does is give me a black screen.
cxo: 'In the same keynote, Yerli also revealed that they plan to release the enxt version of CryEngine by 2012 and that he expects it to look "three to five time better" than CryEngine 2 that powered Crysis. '
cxo: 5times better than crysis, would be like 1080p pr0n
Ori_B: tcoppi: you stalker =P
tcoppi: lol