nha: looks like 7.1/7.2 have been branched
MostAwesomeDude: TobiasTheCommie: Gitweb or cgit of mesa/mesa.
TobiasTheCommie: just wondering, anyone running the git driver on (open)solaris?
nha: that would probably be DDX/2D only, if it works at all
TobiasTheCommie: should be fun to try :)
TobiasTheCommie: my playing computer has win2k(solely for fifa08), ubuntu, ubuntu alpha, gentoo..
TobiasTheCommie: what's one more OS
nha: man I'm looking forward to when those exams are over
nha: then I'm going to beat r300-bufmgr into a shape where we can merge it into master (unless airlied or somebody beats me to it ;))
TobiasTheCommie: uh, btw, just to be really annoying
TobiasTheCommie: nha: found anything on my crash in the RING
nha: unfortunately, no
nha: it just looks really, really weird and I have no clue what's going on :/
TobiasTheCommie: is ok
TobiasTheCommie: i mean, it isn't like i'm crying myself to sleep every night because it hasn't been fixed yet
TobiasTheCommie: nope, not at lal
TobiasTheCommie: zfs alone is so cool that i feel like converting my computers to solaris or freebsd
_neuron_: not as impressive from a performance point of view as it is from a technical point of view from what I've heard
MostAwesomeDude: Yeah, the big deal is that it's easy to set up.
MostAwesomeDude: That's all.
MostAwesomeDude: LVM is not any slower, it just takes more commands to get initialized.
TobiasTheCommie: i don't care bout speed
redeeman: i believe the "big deal" is it's checksumming and distrust of the storage medium
redeeman: i don't trust my harddrives
_neuron_: yeah, the big deal is checksumming, but you can get that with other solutions aswell
redeeman: and i would like the filesystem to not do either
TobiasTheCommie: redeeman: indeed
redeeman: which solutions are currently available for linux?
_neuron_: reiser4 (no known issues with data corruption mind you, yes it was rubish, but it's not since the kernel developers picked it up)
_neuron_: with the cryptcompress code path
redeeman: i find reiser4 to be a great deal more unstable these days than before
TobiasTheCommie: one problem left though...
TobiasTheCommie: i do raids
_neuron_: really? I've had 0 problems with current 2.6.25
_neuron_: patch from kernel.org
TobiasTheCommie: so, zfs, or reiser4 with dmraid...
redeeman: why dmraid?
_neuron_: TobiasTheCommie, dmraid?
TobiasTheCommie: ehm, sorry, md raid
redeeman: yeah i use md raid aswell
TobiasTheCommie: i just like, teh distrust and integrated raid in zfs
_neuron_: only way you can guarantee data safety is tons and tons of backups anyway ;)
TobiasTheCommie: also, it can do timemachine, compression and douplicate copies of data, on specific files, directores, or entire tree structures
TobiasTheCommie: _neuron_: true, but even with that, i'm still paranoid
_neuron_: well, I sure as hell wouldn't trust all my data to one system anyway :p
_neuron_: aka one huge nasty zfs setup that can break at any given point :P
TobiasTheCommie: not any different in that regard to one huge nasty ext3 on md raid setup that can break at any given point
TobiasTheCommie: except the zfs is better
redeeman: i don't care for integrated raid in filesystem
redeeman: IMO that's a different layer
redeeman: even backups of data, is vulnerable to silent corruption of files from the main storage medium
_neuron_: well, zfs is technically better, but untested, ext3 is fairly stable (on a ups)
redeeman: ext3 is kindof mediocre IMO
redeeman: it scales extremely bad
_neuron_: doesn't like filesystems in general, hd vendors need to implement a battery, and some functions to ensure writes are completed to the disk, so journaling can be done differently :P
redeeman: but that's only a minor problem
redeeman: much worse is it when data is actually written different places than intended
redeeman: or if the disk silently corrupts
_neuron_: yeah silent corruption is scary, checksums are making it's way into several filesystems though
redeeman: but if i am to do a fileserver today, what choices do i have on linux?
_neuron_: solved this years ago with a little find md5sum diff script :p
redeeman: lol but that creates MASSIVE io
_neuron_: yeah, run it on off hours ;)
redeeman: it would be TEH HORRIBLE to do on ~5tb data
_neuron_: reiser4 with cryptcompress, which isn't a good choice given it's current state, it hopefully will be a good option though
_neuron_: it is, I ran it on 2.5tb of encrypted data
_neuron_: it will finish overnight though .P
redeeman: but with dmcrypt setup and such, you only get 100MB/s, so that's probably cpu limited
_neuron_: I actually put the encryption layer under the raid
redeeman: as do i
_neuron_: so I had 5 kcryptd's with 4 cpu's
redeeman: but i wonder if i will keep doing that
_neuron_: it feels a bit hackish doesn't it :p
redeeman: well for raid6 it will create some definite overhead
redeeman: ofcourse the overall throughput will be better with a quadcore, much better
_neuron_: only running raid5 here atm, all that's needed really
_neuron_: well, all I need that is
TobiasTheCommie: raid5 as well here
redeeman: i currently have raid5 aswell, but im sure gonna get raid6 next time
TobiasTheCommie: but, i don't really like teh whole "silen corruption"
redeeman: im getting paranoid in my old(lol) days :D
TobiasTheCommie: i'll go zfs2 next i think
redeeman: i just don't really trust any of the solaris distributions
redeeman: mostly because i don't know them
TobiasTheCommie: redeeman: now that's a really horrible argument :)
_neuron_: heh same, not so much about trust as it is about the pain of working with stuff you dont know though :P
redeeman: well i find it to be a reasonable argument
redeeman: if stuff goes wrong, i cannot fix it
TobiasTheCommie: hence why i'm starting to try solaris and zfs on non production systems
redeeman: or well.. i cannot count on it atleast
TobiasTheCommie: so i can test it out
redeeman: yeah i might just do that aswell :)
TobiasTheCommie: see if it is worth the hassle
TobiasTheCommie: figure it out
redeeman: all i really need is sftp access
redeeman: so as long as i can get solaris up with zfs properly, and some form of good sshd, im set ;)
redeeman: loves sshfs
TobiasTheCommie: sshfs? why not just nfs? not on a trusted line?
redeeman: well.. i used to use nfs
_neuron_: why not encrypt when you can though :P
redeeman: but i also used to have crashes because of it
TobiasTheCommie: crashes because of it?
redeeman: since i changed to sshfs the number of network storage related crashes have dropped to zero
redeeman: yeah, if i like.. loose network for even 0.1s
redeeman: it crashes my client
redeeman: i've concluded that linux's nfs client stinks
TobiasTheCommie: _neuron_: don't need the overhead on internal ethernet network
TobiasTheCommie: on wireless, yes, sshfs is a good idea
redeeman: just patch your ssh with the none cipher
redeeman: then authentication can be done with encryption, and then you can drop to no encryption
redeeman: not that i care though
redeeman: i care more about not having my client crash
TobiasTheCommie: and agreed
redeeman: really, the BIGGEST annoyance i have with sshfs, is that it only tells me i forgot to modprobe fuse AFTER i enter my ssh pass :)
TobiasTheCommie: hm, i want crisps
redeeman: (and also, the ONLY annoyance)
TobiasTheCommie: redeeman: i just have an encrypted partition on my laptop, with a keyfile that allows me to ssh without password(only from one specific ip, so no security hole if computer is stolen)
_neuron_: one can always use a keychain aswell
TobiasTheCommie: is reading up on reiser4
TobiasTheCommie: i have no idea how secure a keychain actuall is
airlied: nha: I finall checked the f/b/z patches into bufmgr
TobiasTheCommie: i have no idea how secure a keychain actually is
airlied: nha: I hope my 7.2. branching was what Brian wanted.
airlied: if so we should go bufmgr into master asap..
airlied: I started playing with a GEM+cs backend in the DDX
_neuron_: biggest annoyance right now is that reiser4.fsck will run even when the filesystem is marked clean, and will warn about a load of filesize != actual filesize (on cryptcompress), because the filesize obviusly isn't the same as actual size, but filesystem safety is guaranteed by checksums.
_neuron_: airlied, your the reason I got 2 ati cards in this computer right now :P, was reading up on your kernel modesetting stuff from an nvidia card, with text corruption (which was driving me absolutely nuts), and decided screw this crap :P
airlied: _neuron_: hopefully we don't get any problems on -ati :)
airlied: zzzz &
nha: airlied: yep, I saw those patches, nice; I'll get my eliminate redundant texture storage patch into shape next week after my exams
TobiasTheCommie: EXTERMINIEREN! HALT! SONST WERDEN WIR SIE EXTERMINIEREN! ... EXTERMINIEREN!!!
nha: ... what?
redeeman: TobiasTheCommie: that was brilliant :D
TobiasTheCommie: that it was
nha: ah... now I get it
nha: I've never gotten around to watching Doctor Who
TobiasTheCommie: nha: the imperfection is yours
z3ro: airlied: ping
z3ro: I think maybe you or ajax forgot to add radeon/radeon_buffer.h on the latest commit on the r300-bufmgr branch. just checking with you as you're the committer and ajax isn't around. :P
AndrewR: MostAwesomeDude, i will try to test multipass bicubic scaler from virtualdub tomorrow on friend's machine (full radeon 9200/2.4 Ghz p4 HT) .... still no free time for test your r3xx bicubic code, sorry.
TobiasTheCommie: i'm really seriously thinking of getting meself a new GPU, the current rv530 is extremely slow(even in windows with normal windows drivers)
TobiasTheCommie: i just dont' really know what to get
TobiasTheCommie: any recommendations, and any cards out now where the UVD will have the specs released?
redeeman: TobiasTheCommie: i don't believe there is ANY documentation for UVD on any cards..
MostAwesomeDude: AndrewR: That's fine, take your time. :3
MostAwesomeDude: TobiasTheCommie: There is no UVD/AVIVO documentation, period.
MostAwesomeDude: (Well, okay, not true, the AVIVO modesetting stuff is out there. I mean, the video decoding.)
TobiasTheCommie: MostAwesomeDude: i think i heard that it would be released for newer versions, where the DRM wasn't integrated in the UVD.. i just hoped a) that it was true. and b) such a card had been released by now.
TobiasTheCommie: i know it isn't out for the r600 or less.. i didn't know if they intended to release it for r700, or 800, or if i have to wait for r1000... and i don't even know if it is true
TobiasTheCommie: i guess i could send an email and ask them... but thought i would ask here first..
MostAwesomeDude: Well, bridgman could answer definitively, but he's not here, so I'll just repeat what he said before.
MostAwesomeDude: UVD docs can and will eventually be released without divulging the DRM (AES, HD, etc.)
MostAwesomeDude: But r6xx docs are top priority.
TobiasTheCommie: good, what i had heard..
TobiasTheCommie: wait, ehm
TobiasTheCommie: but can the UVD docs(sans DRM) be released for the r500-700 series? i thought it was too tightly integrated
TobiasTheCommie: or is that, again, something i should ask bridgman?
TobiasTheCommie: i mean, i don't necesarily need a card with UVD support now, i just want one where i know it will come at some point
MostAwesomeDude: Well, the UVD is different pre-r6xx.
MostAwesomeDude: I think they were going to try to only release iDCT and MC blocks.
MostAwesomeDude: For XvMC.
TobiasTheCommie: well, r700 is new as well(pretty sure r600 is uvd+ and r700 is uvd2, may be wrong though)
TobiasTheCommie: one to many wells
TobiasTheCommie: oh well, i'll ask brigdman :)
MostAwesomeDude: He'll be back in like a week, I think?
TobiasTheCommie: i can wait that long :)
TobiasTheCommie: honestly, my current rv530 card can't even play fifa08 properly unless i'm at the very lowest settings
TobiasTheCommie: and even then, it sometimes lags a bit
MostAwesomeDude: I haven't had any problems with any of the games I've thrown at it, for the most part.
TobiasTheCommie: well, i have an x1600 from asus, and their drivers also crashes a bit... very unstable
TobiasTheCommie: for windows that is
TobiasTheCommie: card is fine in linux
TobiasTheCommie: i only paid like 50bucks for it.
TobiasTheCommie: i'll go 300 or so for the next one
TobiasTheCommie: MostAwesomeDude: http://www.pcpro.co.uk/reviews/94171/asus-eax1600pro.html <- my card
TobiasTheCommie: may actually be just 256mb
MrCooper: TobiasTheCommie: I'm afraid you're greatly underestimating the time it takes from GPU design to the final product
TobiasTheCommie: MrCooper: i'm afraid of a lot of things :)
TobiasTheCommie: but yeah, i probably do, and it doesn't surprise me all that much
spreeuw: Error: couldn't get an RGB, Double-buffered visual
spreeuw: getting this one again
spreeuw: dont tell me I have to upgrade xserver again
spreeuw: it's with mesa git
MostAwesomeDude: spreeuw: Double-check your linking; make sure you're using the right libGL/DRI combo.
odz: spreeuw you need git xorg-server for 3D to work
spreeuw: oh bummer because it's build is broken
MostAwesomeDude: It is?
MostAwesomeDude: You need libXtrans from git.
odz: you need to build xorg-server with git libX* and *proto
MostAwesomeDude: At this point, it might be almost easier to build on LFS or Gentoo than on Debian. I feel kinda bad for distros 'cause everything's gonna bump at once.
spreeuw: I'm on lfs
spreeuw: its the only sane distro, but yeah it's a bit much work sometimes
odz: its pretty easy with gentoo
agd5f: TobiasTheCommie: pre-r6xx and real r600 had a separate idct engine. MC was always doen on the 3D engine. r6x and above (and a few older chips rv550, rs6xx) have the uvd block, newer r6xx chips have uvd2
agd5f: we haven't had a cahnce to review UVD yet so I can't say how much if any we can release
agd5f: Ori_B: as for the clone stuff, the xrandr tool tends to do the wrong thing. I think you have to turn off the output first then turn it on with the chose crtc
agd5f: xrandr --output DVI-0 --off
agd5f: xrandr --output DVI-0 --crtc 1 --mode 1024x768
Ori_B: agd5f: hm? *tries*
Ori_B: agd5f: yay! it worked.
Ori_B: agd5f: but.. why did it stop working?
TobiasTheCommie: agd5f: frell, thanks
zhick: has anyone else problems with wine and latest mesa? for me nothing 3dish is working (no dri, only software-rendering).
TobiasTheCommie: zhick: yes, though i thought it was a regression in wine since the game in question had problems before(but now no 3d works, before everything sans the game 3d worked)
zhick: hm, too bad. i was hoping it was some kind of obvious misconfiguration on my side :>
TobiasTheCommie: well, considering my problem was with a game that didn't work before, don't take it my statement as being the truth
zhick: what does "libGL warning: 3D driver returned no fbconfigs." mean?
robokopp: thank you radeon developers, my r480 runs great :)
agd5f: Ori_B: it "stopped working" because we used to not allow cloning. we forced one output per crtc
Ori_B: agd5f: ahh. ok. any idea what it'd take to fix?>
agd5f: Ori_B: the driver's not broken. the xrandr utility is just wonky
Ori_B: well, the behavior is broken =)
Ori_B: so should I make xrandr disable the CRTCs before reassigning them?
agd5f: Ori_B: you shouldn't have to. disabling the output works around wonkiness in xrandr
Ori_B: er, sorry. that's what I meant.
Ori_B: make xrandr do that itself
agd5f: what should happen is xrandr should check for when you request dualhead and make sure it assigns separate crtcs for each output
Ori_B: ok. I may have to look into that sometime later then
robokopp: hm, is it possible to switch dri between the xservers if i have 2 running at the same time?
agd5f: robokopp: nope
robokopp: ok, thank you
dmb: airlied, that was not a nice email linus sent you :(
EruditeHermit: TobiasTheCommie: http://www.gossamer-threads.com/lists/linux/kernel/965267
airlied: dmb: I was expecting it though..
EruditeHermit: will it just wait till 2.6.28?
EruditeHermit: well that'll probably be out before xorg 7.5 anyway