prunedtree: would this be an appropriate place to ask a technical question about ATI hardware ? I'm wondering how many registers there are per SIMD core in the hd4800 series
Terman: prunedtree: have you looked at the ati documentation for the r7xx chipset?
prunedtree: well, everything I could find
prunedtree: CAL doc, r6xx ISA
prunedtree: is there any other documentation than this: http://developer.amd.com/documentation/guides/Pages/default.aspx ?
Terman: I think that's it: R6xx_3D_Registers and r600isa
Terman: the first one is qite new
rhodan: i can't run vesa with my X1400. is this normal?
rhodan: it says (EE) VESA(0): Driver can't support depth 24
rhodan: i'm running newest git of xorg-server
adamk_: Why do you want to run vesa on it?
rhodan: because of the faster kde4
rhodan: i'm sure vesa was running fine once
oga: wait... vesa because it'll make kde4 faster?
prunedtree: the doc on 3d registers doesn't really help...
rhodan: jep, it does
adamk_: rhodan: I can't believe that KDE4 is faster with vesa... But if you have problems with vesa, you might want to ask on #xorgt.
adamk_: #xorg, rather.
chithead: kde4 is fast with EXA, xorg-server-1.5, xf86-video-ati-6.9.0 and later
prunedtree: It strange, I would have thought that they would document the size of the shared register file so you know how much register pressure you can afford
ufa: i was seeing the results in free3d.org
ufa: and it happens that my card (x600) is not archiving (not even near) the same results of the same cards
ufa: is there something I can do?
bridgman_: prunedtree; I think there are 16K registers per SIMD, ie enough for 2 sets of threads in flight if you use all 128 registers, and proportionally more threads if you use fewer registers
glisse: bridgman_: i am pretty sure i saw this documented somewhere
bridgman_: yeah, IIRC this was discussed during the R600 rollout in one of the interviews
bridgman_: 256KBytes per SIMD or something
bridgman_: the graphics guys tend not to care about this so much, but I think we put more info out in the stream sdk... haven't had time to read it all yet though
glisse: still very important for compiler :)
bridgman_: ah yes, good point
glisse: it's also pretty easy to find with a small test app
bridgman_: maybe for you ;)
bridgman_: yeah, I agree
bridgman_: ahh, here we go... anandtech article talks about registers :
kcodyjr: what license seems appropriate for that edid parser library? bsd? mit? lgpl? i don't much care, i just want it to fit in with wherever it might wind up.
nanonyme: kcodyjr: You can multi-license it if you like so you don't have to care.
chithead: most stuff in x.org is under MIT license
glisse: if it's to go into kernel it must be gpl at least
glisse: you can dual license to bsd too
kcodyjr: heh. so i have to multi-license to hit my first two target audiences. great. ;)
chithead: yes, kernel developers require that lgpl or gpl are at least one of the licenses
kcodyjr: if i want to multi license, i need to write a speech into each source file, right?
glisse: kcodyjr: looks at files in shared-core in drm repo
glisse: their should be example
jcristau: chithead: err. mit/bsd is just fine for the kernel
chithead: it is gpl compatible, but for some reason kernel developers want additional lgpl/gpl
kcodyjr: i was going to go something like: this software is primarily released under the LGPLv3, however, any large-scale open-source project is also granted permission to distribute according to any OSI-approved license which is organic to that project
kcodyjr: so i don't even need to care what the license list is ;)
glisse: kcodyjr: kernel people might not lack the absence of explicit license
kcodyjr: plus, it occurs to me, that a commercial entity could just say "ok. i pick bsd."
jcristau: kcodyjr: only "distribute"?
kcodyjr: jcristau, it's not my intent to restrict
jcristau: that's pretty restrictive then
kcodyjr: just to piss off anyone evil like, say, SCO
jcristau: but really, pick gpl or 2-clause bsd..
kcodyjr: as far as i'm concerned, any open source work is free to use it any way the hell they please
kcodyjr: gpl not lgpl?
kcodyjr: and are bsd and mit compatible - ie should i 3-clause it and explicitly add mit to the list
glisse: kcodyjr: really there are example in shared-core that should fit your wishes
kcodyjr: that looks like a straight-up bsd license to me
kcodyjr: unless not all files are the same
glisse: not all files are the same
kcodyjr: ok i'll keep looking
jcristau: a straight-up bsd license sounds a lot like what you seem to say you want..
kcodyjr: pretty much; i just want to be sure it's compatible with both the kernel and X
jcristau: it is.
kcodyjr: cool. glisse, 10/10, i keep seeing variants of the same straight up bsd license - which is fine. i'll just grab one of them, add an explicit endorsement for MIT/BSD/GPL/LGPL redistribution, and add it to each file
kcodyjr: wow i'm having a bad day if i forgot to close a regex ;)
kcodyjr: ok, now that i've returned the spanking git gave me for not reading the docs before creating a public repository... http://master.vegaresearch.com/git/libedid contains an early, but "does something ;)", version of the parser
kcodyjr: along with both a short and full dump from my own main monitor, so it can be played with without having to dig out your own edid data.
Guest25878: the kernel devs don't like the GPLv3, is there actually any only GPLv3 code? Most seems GPLv2 (or later)...?
kcodyjr: i have no idea. i could possibly go change that to 2+, but that clause applies only to a commercial project using it anyway
kcodyjr: any open source project may include it under that project's own license
kcodyjr: i -think- i wrote it right that way anyway
nanonyme: Guest25878: Not surprising. I think Linus has confronted GNU idealism pretty openly.
kcodyjr: yeah i've read that too
nanonyme: And I personally think he's right.
nanonyme: They're trying to make software, not save the world. :)
kcodyjr: that's why i granted permission to relicense under anything osi approves
Guest25878: yes. just go sure that they accept gplv3 only programms before setting the license.
kcodyjr: the idea was simply to force corporate lawyers to fart ;)
kcodyjr: it isn't v3 only, that's the beauty of it
kcodyjr: it's only v3 only if you're not doing something related to another open source project with it
Guest25878: oh, nice :)
kcodyjr: the terms are that it automatically relicenses itself to whatever the larger including project uses, so if the xorg guys pick it up, for them it's MIT
jcristau: kcodyjr: once it's mit anyone else can use it as mit. if not, it wasn't free in the first place.
kcodyjr: jcristau, i know. i also know corporate lawyers, it's just my way of twisting their tail.
jcristau: that's just silly
kcodyjr: yes it is. but you weren't there when the lawyers told us we couldn't update our ubuntu desktops because they're afraid of the gpl.
kcodyjr: that was also silly ;)
kcodyjr: also, i'm unsure how exactly incorporation/usage is going to go. this maximizes flexibility for now.
kcodyjr: ... the header includes the provision that the header may be modified to reflect whatever license it's being incorporated into.
kcodyjr: did i mention very early version of it? ;)
kcodyjr: anyone taking a look at the interface, rather than the license? ;)
spstarr_work: MadMerlin: anything I can try with the r3xx gallium driver?
spstarr_work: nanonyme: i have an r1xx (Radeon 7500 AiW) on my other box since the TV tuner works ;-)
spstarr_work: orkid: lol "
nanonyme: spstarr_work: Are you answering to comments from the backlog in order from oldest to newest?
spstarr_work: er MAD
spstarr_work: nanonyme: yep
spstarr_work: finallly caught up
spstarr_work: clears buffer
MostAwesomeDude: spstarr_work: Yeah?
spstarr_work: MadMerlin: anything I can try with the r3xx gallium driver?
spstarr_work: 'MAD:' translates to him anything I can try with the r3xx gallium driver?
spstarr_work: MostAwesomeDude: anything I can try with the r3xx gallium driver?
nanonyme: MostAwesomeDude: Is today pretty awesome?
MostAwesomeDude: spstarr_work: Enjoy your lockups.
MostAwesomeDude: nanonyme: Fairly awesome.
spstarr_work: MostAwesomeDude: hmm, i'll pass :)
StR|Sangreal: please, i need advice
StR|Sangreal: i have centrino duo architecture, kubuntu intrepid AMD64 and ATI Radeon Mobility X1450
StR|Sangreal: could sb tell me which driver is most suitable for me?
MostAwesomeDude: Well, radeon would work fine.
Enverex: radeon will work best
StR|Sangreal: could you please explain simply how should i install and load it?
chithead: it should be loaded by default already
StR|Sangreal: well in that case sth is wrong
StR|Sangreal: i mean as i install kubuntu, i get much flickering, some lags in performance, and i am not able to play video
StR|Sangreal: (and i didnt try to run graphically demanding applications...)
StR|Sangreal: so it means i get no acceleration after distro install
StR|Sangreal: and i am offered to install fglrx, but it is not compactible with neither the device nor xorg 1.5
StR|Sangreal: so could you pls tell me steps to follow to get the 3D acceleration?
MostAwesomeDude: The Kubuntu people will probably be more helpful...
StR|Sangreal: once i was already mislead and the result was a terminall collision between the video driver, compiz and kdm...
StR|Sangreal: and so that kdm is not working now at all
Enverex: StR|Sangreal, Load the r300 DRI module, install the xf86-video-radeon/ati driver and that should be pretty much it
adamk_: StR|Sangreal: Intrepid installs and uses the radeon driver by default.
adamk_: StR|Sangreal: If you are having problems with it, you should use a service like http://pastebin.com to show us your /var/log/Xorg.0.log file.
StR|Sangreal: my default xorg.conf is very very simple
adamk_: I'm sure it is, but that's not what I asked about :-)
adamk_: In any case, I'm going to read in bed now :-)
StR|Sangreal: well my system is broken now... i will reinstall it and then call for help again
mastamind: hi. i am using xf86-video-ati version 6.10.0. tv out works quite well but is still only black and white. is that a common problem?
mastamind: i am setting the tv standard in xorg.conf to pal. which should be the correct standard for austria.
chithead: mastamind: I think only ntsc is supported properly with xf86-video-ati at this time
mastamind: pal support just misses color :-)
chithead: my guess is that your tv reacts to ntsc signals by displaying black and white
mastamind: so you think setting tv_standard to pal does not really change something? it is still ntsc?
chithead: mastamind: is there a visible difference between the "pal" and "ntsc" setting?
agd5f: mastamind: probably the wrong cable
agd5f: mastamind: ntsc and pal should work except for igp where only ntsc works
mastamind: chithead: yes. there is. the ntsc setting makes the picture flicker.
agd5f: some cards need special cables to connect the right pins
mastamind: agd5f: i will check the cable with another pc.
mastamind: agd5f: you are right. the cable is a black-and-white-cable... :-(
huffman: How can I tweak my monitor settings like brightness/contrast?
agd5f: huffman: laptop panel or plug in monitor
huffman: plug in
huffman: I want to do it in software
huffman: Plus I don't have contrast settings on my monitor when using DVI
mastamind: agd5f: why do people sell black and white cables?
agd5f: mastamind: some cards are pinned out funny so you need special cables
mastamind: the problem is, that my tv does not understand svideo on scart.
agd5f: huffman: does you monitor actually have sw contrast controls?
agd5f: generally those are only available on CRTs
huffman: If I use vga, yes. But can't you usually modify these settings in software?
huffman: I guess I was mainly looking for any ati utility to modify settings.
King_InuYasha: the new release of fglrx is still broken on Ubuntu 8.10
EruditeHermit: works for me
King_InuYasha: my computer has Radeon 9600 Mobility M10
King_InuYasha: and because as of now, the radeon driver does not yet support OpenGL 2.1, I would have to use fglrx
King_InuYasha: which doesn't work ever since I started using Ubuntu intrepid
King_InuYasha: it always says "no screens detected"
EruditeHermit: yes I had that problem too
EruditeHermit: the leaked 9-2 beta driver worked with that though
EruditeHermit: did you try that?
EruditeHermit: 9-1 works with my 4850
EruditeHermit: I prefer the open radeon driver though
huffman: Why is the open source driver better than the fglrx one, in terms of functionality
huffman: particularly for a 9800pro user
EruditeHermit: it isn't necessarily
kcodyjr: but it's more likely to fail gracefully ;)
huffman: Does the fglrx one have better opengl support?
huffman: Which versions do each support>
kcodyjr: fglrx supports more features especially as you get newer, but it also bombs like yesterday's mexican food
EruditeHermit: huffman: radeon supports OGL 1.3 up to r500 generations
huffman: EruditeHermit: Hmm, what about fglrx?
EruditeHermit: huffman: fglrx supports OGL 2.1 and 3.0 on newer boards
huffman: Oh okay
huffman: Thanks for the info :)
EruditeHermit: radeon for me has better video playback
EruditeHermit: especially with compiz
EruditeHermit: and suspend is reliable
EruditeHermit: and 2D drawing seems faster
EruditeHermit: all for less than r500
EruditeHermit: r600-700 is still in development
EruditeHermit: for radeon
Enverex: radeon/hd generally "just work" but lack the 3D or advanced 3D that fglrx has (laymans summary)
kcodyjr: yeah pretty much
kcodyjr: advances directly to the porcelain altar
huffman: I only have a 9800 in my desktop and a 9600 in my laptop
huffman: Not upgrading soon :P
kcodyjr: the 9600 on my older frontend is solid as a rock on the radeon driver
EruditeHermit: unless you play games, stick with radeon
huffman: yeah, that's what I had been using
huffman: My laptop is out of order at the moment though :P
EruditeHermit: and even if you want to play games with 9600 and 9800, you shouldn't really be =)
Enverex: Exactly, the open drivers are great for 2D, they work without any configuration and work well, but aren't much use for 3D, especially on newer boards (which generally don't have any 3D support at all)
huffman: Nvidia has a lot more linux support, yes?
Enverex: nVidia has a better closed driver than ATi
kcodyjr: nvidia's equivalent of fglrx ... well... is stable
huffman: Ah, gotcha
huffman: what about open source?
Enverex: It doesn't have one
kcodyjr: consequently their open source teams have a harder time of it
Enverex: (well it does but it's so unusable that it's not worth mentioning)
kcodyjr: nouveau has to reverse engineer, they're really only usable on one chip family
huffman: So, open ati > open nvidia, closed nvidia > closed ati
kcodyjr: and marginally on that one (nv40)
kcodyjr: and ati image quality > nvidia image quality, imo
Enverex: btw ATi's closed driver doesn't work well with Wine, has major shader issues
huffman: Didn't realize that
Enverex: Anyway, night all
huffman: Thanks for the help
kcodyjr: i've been buying ati for years because it just looks way better
Enverex: kcodyjr, That was only true back in the Rage Pro days, they've all been identical for years
EruditeHermit: I just bought an HD4850 because of their open source documentation release
kcodyjr: and it never fails to bite me in the ass for the next 6 months after i buy it
kcodyjr: i disagree, having seen both recently
kcodyjr: but i've got a printer's eye, if i'm looking, i can pick up something being 1 pixel off
kcodyjr: it's a f'kin curse actualily
chithead: ever since getting burned by kyro, I only bought graphics which were properly supported by open source drivers. that meant intel and older amd. soon I may buy a current ati card
EruditeHermit: works for me
kcodyjr: the X* series are well enough supported by now
kcodyjr: and anything with a number 9xxx 7xxx... were there 8's?
chithead: 8500-9250 are r200
kcodyjr: i've got a 9600. i never notice it. ;)
huffman: I've never owned an Nvidia, so I don't really know the difference.
kcodyjr: had an older one that tended to get choppy on the edges
kcodyjr: for some reason, when the airplane was rolling over, the wings would sawtooth
kcodyjr: never happened on the ati in the other room, this was like 2003
kcodyjr: now, the nvidia's mostly bug me because they just will not accelerate on hardened gentoo ;) but their output does still seem to chop and studder to me... maybe they just don't like the open source drivers at all, even unaccelerated
chithead: kcodyjr: did you try nouveau?
kcodyjr: yes, which actually did seem to improve things, but had horrid stability problems
kcodyjr: caused 3s display hangups every time i minimized a window
kcodyjr: honestly, exactly the sorts of bugs i'd expect in a reverse engineered driver at its stage of development
chithead: I am stuck with a nvidia 7300 at work, nouveau started to work quite well for me some months ago
kcodyjr: other than the sluggishness, yes, it was, 3d and all
kcodyjr: airlied, is drmModeReplaceFB supposed to be fast enough to call during the retrace? or should i just be setting up my scanout at twice the height i need and flipping with a register write?
judoka: hello, any kind souls who can help explain some performance and dri stuff?
judoka: I have a mobility radeon 9600 (rv350), I use the open source driver
judoka: ubuntu Jaunty
judoka: is hyper-Z enabled by default? there is no setting in driconf
judoka: I see 50% cpu usage on a 1.6Ghz Pentium M w/ compiz
judoka: rotating the cube w/ transparency
judoka: glxgears gives ~1756fps
judoka: should I be seeing high Xorg CPU usage w/ open source driver and mobility radeon 9600 w/ compiz?