glisse: agd5f, MrCooper: teaching EXA about this could make pageflip vblank working, thing is out of sync buffer, we might just always sync front & back buffer but i don't see anysane way to do this today in dri world with current drm
glisse: dri2 with no cliprect and private back buffer should be doable
glisse: for instance each time we queue a flip we could also queue a blit to update old front to be the same (using damage informations i guess)
agd5f: glisse: yeah. also we really need shatter to fix multihead
glisse: agd5f: yup that would be very helpfull
glisse: maybe we should talk with ajax to make sure he thinks to us
glisse: about vsync
spreeuw: hey, ohh
spreeuw: radeon hd has no 3d accell in 1.2.2?
spreeuw: changelog says dri support
jcristau: depends on the hardware
jcristau: r5xx should be ok
spstarr_work: glisse: how much work is switching the radeon DRI drivers to DRI2?
glisse: spstarr_work: most of the work is porting radeon to cs
glisse: dri2 bit should be easy
glisse: i hope to have time this week
glisse: but as usual i keep receiving things to do
AndrewR: downloading linux/kernel/git/airlied/drm-2.6.git (for testing modesetting/cs on rv280) ... link is slow.
agd5f: spstarr_work: airlied's radeon_gem_cs branch already has cs support, it's just a matter of hooking up the dri2 stuff
spstarr_work: sounds good so we're in shape for DRI2 RSN
glisse: agd5f: i was referring to porting r300-bufmgr branch to cs :)
agd5f: glisse: ah, yes, 3D side
airlied: glisse: I started doing it last inight :)
airlied: glisse: or at least started moving nh's code towards a CS relocation style.
airlied: with inline relocations.
airlied: if I have time I'll probably just make the current r300 cmdbuf relocateable for now.
airlied: and move to cs later.
spstarr_work: hello airlied
nitrotrigger: what's up with linux driver support for hd4870x2?
kdekorte: nitrotrigger, mode setting only, and probably only using 1 gpu
kdekorte: for more advanced features you'll need to use the ATI fglrx driver
nitrotrigger: why not?
kdekorte: nitrotrigger, ATI has not released the programming documentation for it yet
nitrotrigger: how does it affect my gaming on linux?
nitrotrigger: just that I'm considering buying a HD4870x2
kdekorte: fglrx, will give you 3d support -> games...
nitrotrigger: so it'll run about as fast as with the windows drivers?
kdekorte: free drivers that support 3d for that card are probably a ways out
kdekorte: I have a 3650 card with the fglrx and it works well
nitrotrigger: or should I still prefer GTX280 for it's linux support
kdekorte: that is an nvidia card right? Well you're in the same boat there, if you want 3d, you have to use the nvidia drivers.. no free drivers that support 3d
nitrotrigger: what are the free drivers good for?
kdekorte: the best card to have free 3d is the ATI 1xxx series (r500 chip)
nitrotrigger: is it just that it's open source?
kdekorte: free drivers are good for keeping your kernel free of unknown binaries and you don't have to worry about kernel upgrades breaking your driver
nitrotrigger: or are the fglrx drivers bad?
nitrotrigger: sounds quite good
kdekorte: personally I think the fglrx drivers are ok, but I know some people hate them... although I would prefer to use free drivers
kdekorte: on my laptop I use the free drivers (r500 series), but my desktop has an r600 so if I want 3d, I have to go with fglrx for now
nitrotrigger: I think I'll manage with the fglrx drivers until they release the documentation
nitrotrigger: they are going to release them right?
kdekorte: the lawyers have the docs now... but even when the docs are released it will take some time
nitrotrigger: about a year maybe?
kdekorte: hoping for less, but you never know
nitrotrigger: I'm going to army anyway
kdekorte: good luck with that
nitrotrigger: it'll take about a year, half if I'm lucky
nitrotrigger: I could buy the gpu then, but hardware is in general cheap nowadays
nitrotrigger: here in eu
nitrotrigger: because the dollar is cheap
nitrotrigger: all I wanted to see now is some comparison between the fglrx and windows drivers
kdekorte: http://www.phoronix.com/scan.php?page=home has some stuff, just search the site
nitrotrigger: "When there is the support though, it will be limited to one GPU. We wouldn't expect to see CrossFire before 2010, so if you intend to be using any open-source ATI driver at least partially, you will be better off using a graphics card with a single GPU."
nitrotrigger: does that mean
nitrotrigger: they will limit the hd4870x2 open source drivers to use only one of it's cores?
yangman: nitrotrigger: no. there's no info available on enabling CrossFire properly, so the best an open driver can do is initialize and control a single GPU core
nitrotrigger: is hd4870x2 a single gpu?
chithead: no, x2 cards are two gpus on one pcb
chithead: you can still use the second core for computation tasks
nitrotrigger: so will the things be different from 2010?
nitrotrigger: I mean, it has a some sort of crossfire combination inside the 4870x2?
nitrotrigger: so that when they get crossfire working on open source, they can make a driver to use both of hd4870x2's gpus?
yangman: nitrotrigger: 4870x2 is 2 sets of 4870 components with a on-board CrossFire bridge
nitrotrigger: so yes to my last question?
yangman: nitrotrigger: if/when CrossFire programming info gets released, yes
nitrotrigger: hmm one think I have in mind
nitrotrigger: you know OCZ makes DRAM's "for crossfire"?
yangman: RAM has nothing to do with it
nitrotrigger: so those are all bullshit?
nitrotrigger: the fancy crossfire-logo on 'em
yangman: I personally don't see how CrossFire certification would matter for anything except motherboards and power supplies
nitrotrigger: the price isn't very bad
nitrotrigger: for those crossfire modules
nitrotrigger: just a bit more expensive than the regular ones
yangman: the certification is just a marketting gimmick, AFAIC
nitrotrigger: looks like nVidia has one of it's own
nitrotrigger: SLI-certification against. Crossfire
nitrotrigger: I must say they look good :D
nitrotrigger: but that's probably not the first thing I want from DRAM modules :D
nitrotrigger: anyone still awake?
nitrotrigger: can hd4870x2 use all of it's 2GB GDDR5?
yangman: nitrotrigger: you'd assume so, since it's there
yangman: nitrotrigger: like I said, two sets of components. each GPU gets a GB
chithead: nitrotrigger: each of the gpu cores can only use its 1gb though
nitrotrigger: so there aren't any "completely useless memory"
nitrotrigger: so it's not different from crossfiring two hd4870's?
nitrotrigger: no, it can't be
yangman: it's different
yangman: the on-board CrossFire bridge makes the card appear as a single entity to the OS
ovoskeuiks__: There is some bridge chip that makes it one card isn't there??
yangman: FWIW, a lot of hardware reviewers are recommending 2 4850 over 4870x2
nitrotrigger: it's more expensive
nitrotrigger: takes much more room
nitrotrigger: eats more electricity
nitrotrigger: and the mobo needs to support it
nitrotrigger: but yet
yangman: afaik, all consumer mobo with more than 1 PCIex16 slots support CrossFire
nitrotrigger: doesn't hd4870 have 512MB memory?
yangman: power consumption doesn't vary that much. you're driving about the same amount of hardware
yangman: there's new 1GB versions of the 4870
chithead: there also exist 4870 variants with 1gb
yangman: the main factor is price
nitrotrigger: I'll check that from my local dealer
nitrotrigger: hd4870x2 is about 450-500 euros
nitrotrigger: one is about 250-300 euros
chithead: advantage of the 4870x2 and upcoming 4850x2 is that they do not need a crossfire capable mobo
nitrotrigger: the mobo would be 126 euros
nitrotrigger: but what would I get?
nitrotrigger: 01:43 < yangman> FWIW, a lot of hardware reviewers are recommending 2 4850 over 4870x2
nitrotrigger: hmm I was thinking of 2 4870
ovoskeuiks__: Apparently the interconnect on the 4870x2 is faster than the crossfire bridge so the 4870x2 'should' be faster than 2x 4870
nitrotrigger: so it's cheaper and faster
nitrotrigger: to buy hd4870x2 over two hd4870's
yangman: nitrotrigger: actually, I may have been thinking of advantage over a single 4870
chithead: the internal bridge exists, but current versions of ati's driver do not activate it
yangman: nitrotrigger: in any case, do what fits your budget :)
nitrotrigger: chithead: even the windows drivers?
chithead: even the windows drivers
nitrotrigger: you know if they are going to be activated?
chithead: ask in #ati that is where the proprietary ati driver is supported
nitrotrigger: yangman: I still can't see why would they recommend two hd4870's over one hd4870x2?
chithead: cooling issues, support for more monitors, ability to use the cards independently, ...
yangman: nitrotrigger: they wouldn't. I meant 2 4850s over a single 4870
nitrotrigger: well I'll buy the x2
nitrotrigger: what vendor do you prefer?
nitrotrigger: I was thinking of Asus
yangman: *shrug* I haven't bought a video card for years
yangman: just avoid vendors with consistent known problems
nitrotrigger: I don't know any :D
nitrotrigger: any forums I could check for that?
nitrotrigger: but now I'll go to sleep
nitrotrigger: thank you for the help
chithead: nitrotrigger: ##hardware
nitrotrigger: oh thanks
nitrotrigger: I'll check that in the morning
spstarr_desk: tries new kernel
xnguard: Kernel upgrade time for the workstation.
spstarr_desk: let's see of kms + agp works
spstarr_desk: airlied: wedged :(
spstarr_desk: it's a no-go.. awaiting your next patches
spstarr: hullo MAD
airlied: spstarr_desk: damn, where'd it wedge?
spstarr_coding: airlied: same thing with kdm, logged in then it locked
spstarr_coding: I can try with gdm also
airlied: try gdm + gnome session if possible.
spstarr_coding: one moment
airlied: just to see if I fixed at least that..
spstarr_coding: trying now...
spstarr_desk: airlied: nope
spstarr_desk: wedges at gdm prompt
spstarr_desk: even before that, I tried with 'rhgb' and it got stuck in plymouth
spstarr_desk: if i didnt enable plymouth it got to at least gdm
airlied: spstarr_desk: wierd.. what kernel -30?
spstarr_desk: latest one .. it is
spstarr_desk: there's a -30 now? :)
spstarr_desk: you have additional changes?
airlied: I think -30 contains the r300 fixes :)
spstarr_desk: doh, they aren't in the static repo.
spstarr_desk: oh my
spstarr_desk: airlied: GNOME started, but i had compiz enabled
spstarr_desk: X crashed i got black and white stripes but its letting me back into gdm
spstarr_desk: maybe that race...
spstarr_desk: trying kdm now..
spstarr_desk: hmm, switching VT then running init 3 gives me a garbled keyboard
spstarr_desk: ssh's in
spstarr_desk: yeah cannot vt switch with keys
spstarr_desk: though chvt 7 with ssh (as root) worked
spstarr_desk: airlied: something broke with keyboard and kms mode on?
spstarr_desk: oh, it broke only when X crashed restarting it worked
airlied: spstarr_desk: X left something in a bad place maybe on crashing.
airlied: so things work if you don't hit the compiz?
airlied: my r300 even seems to suspend/resume.
spstarr_desk: ok compiz or kwin crash with X + kms
spstarr_desk: going to disable kwin composite.. sec
spstarr_desk: crashed with no composite... there is something be written to log
spstarr_desk: [ 137.359485] [drm] bios LVDS_GEN_CNTL: 0x30ff24
spstarr_desk: [ 142.207106] [drm:drm_mode_getfb] *ERROR* invalid framebuffer id
spstarr_desk: [ 187.926719] [drm:drm_mode_getfb] *ERROR* invalid framebuffer id
spstarr_desk: [ 71.445182] mtrr: base(0xe1144000) is not aligned on a size(0x5a4000) boundary
spstarr_desk: [ 129.629286] [drm:r300_emit_carefully_checked_packet0] *ERROR* Offset failed range check (reg=4540 sz=1)
airlied: spstarr_desk: still something doing 3D there.
spstarr_desk: lemme Enabled=false
spstarr_desk: I will restart again
spstarr: i am in KDE + kms!
spstarr: vt switches
spstarr: and back
spstarr: airlied: 2D looks good trying GNOME now..
spstarr: and now im in GNOME with no compiz
spstarr: cannot VT switch with GNOME
spstarr: X is not letting me
spstarr: (is this by design with GNOME?)
airlied: I think something busted in evdev or xorg maybe.
spstarr: in either case, KMS + GNOME/KDE work without 3D
spstarr: trying glxgears...
spstarr: that works
spstarr: so composite + GL
spstarr: oh interesting
spstarr: chvt # doesnt work in when in GNOME
spstarr: it aborts
spstarr: oh bootup
spstarr: 247.024999] [drm:radeon_mem_init_heap] *ERROR* heap already initialized?
spstarr: [ 247.031046] mtrr: base(0xe2250000) is not aligned on a size(0x5a4000) boundary
spstarr: back to kDE..
spstarr_desk: oh that wasn't good
spstarr_desk: it gpu wedged
spstarr_desk: i can ssh into the box though
spstarr_desk: (WW) xf86CloseConsole: KDSETMODE failed: Input/output error
spstarr_desk: (WW) xf86CloseConsole: VT_GETMODE failed: Input/output error
terracon: -30 not good on my R350
spstarr: fsck switching VTs is so fast
spstarr: im in KDE now
spstarr: terracon: doesn't work ?
spstarr: terracon: turn off desktop effects / kwin composite mode
spstarr: doesn't work at the moment
terracon: spstarr: desktop effects are off
spstarr: it locks?
terracon: I just tried running enemy territory, Lockup
spstarr: so you were in KMS + your desktop?
airlied: terracon: so you can log in..
airlied: thats working compared to yesterday.
spstarr: glxgears works trying vt switch with glxgears running
spstarr: full screen glxgears + vt switching, worked
spstarr: i can now remain with kms on, just no composite in either desktop
terracon: let me try booting one more time with kms enabled and see if login comes up. sec
spstarr: airlied: in text mode kms, is it supposed to be sluggish scrolling text?
spstarr: (this could be considered the same when using vgafb mode)
spstarr: ro root=/dev/VolGroup00/LogVol00 slub_debug=- printk.time=1 rhgb quiet :)
airlied: spstarr: yes until I write accel for it.
terracon: ok with kms enabled. kdm doesn't appear still
spstarr: airlied: no problem, so now i've caught up to your kms for r4xx
spstarr: terracon: it locks for you?
spstarr: terracon: no kwin composite enabled right
spstarr: odd.. which chip do you have
terracon: I'm pretty sure I can still ssh into the machine it's just kdm doesn't come up
spstarr: or non-mobile
airlied: terracon: got gdm? :)
terracon: Xorg says R350
spstarr: i should put in the Radeon 9800 Pro AGP card today? need me to try that one airlied?
airlied: spstarr: can't hurt..
spstarr: sure can do....
terracon: I'm not sure if I have enough gdm bits installed. I can try that
spstarr: before I do, im going to do a shutdown.. wanna see if the system locks up on rebooting
spstarr_desk: locked up
spstarr_desk: thats not good
spstarr_desk: airlied: were you able to reboot your r3xx laptop?
airlied: spstarr_desk: I think AGP reboot might have issues.
spstarr_desk: ok, i will put in the card in this box.. brb from laptop..
terracon: airlied: ok gdm works
spstarr: installing card...
terracon: but running glxgears results in a lockup
airlied: terracon: did you ever have to use agpmode setting in the xorg..conf before?
terracon: airlied: I've used agpmode in the past. Right now there's no xorg.conf
airlied: terracon: what agpmode did you have to use? try booting with radeon.agpmode=
airlied: for the one you used to use.
terracon: trying that now, back in a flash
spstarr: the box is getting new kernel..
spstarr: airlied: my agpmode is still 4x
spstarr: ok, the box is ready for testing...
spstarr: trying GDM + GNOME with kms on the Radeon 9800....
spstarr: no desktop effects on
terracon: airlied: I tried agmode 8 and than stepped down by one. I tried enemy territory . Sometimes a little bit of it would render at the top of the screen than it would lockup
spstarr: X is not starting
spstarr: well plymouth is stuck...
spstarr: not sure what its doing..
spstarr: lemme kick it
spstarr: Fatal server error:
spstarr: xf86OpenConsole: VT_WAITACTIVE failed: Interrupted system call
spstarr: adjusts xorg.conf
airlied: yeah I think the vt switching migt be bustyed.
spstarr: that box had some ricer options on, turned them off (for r2xx they worked ok)
spstarr: rv2xx even
spstarr: i want to eliminate all variables so its basic
spstarr: X wont start
spstarr: trying to force it
airlied: is it kdm or gdm?
airlied: I do wonder if the making X work in VT1 is messing up kdm.
spstarr: i switched it to GNOME for testing
airlied: are you running lateset rawhide bits?
airlied: initscripts was messed up before.
spstarr: as of last morning
airlied: hmm should be fine.
spstarr: rawhide says: glibc xorg server update
spstarr: xorg-x11-server-Xorg x86_64 1.5.2-8.fc10 rawhide-x86_64 1.5 M
spstarr: some exa change
spstarr: gonna do a reboot on the box in a moment.. once updates are in
airlied: yeah only an optimisation
spstarr: i can try adjusting agpmode if you like
spstarr: [ 4.595260] pci 0000:01:00.0: putting AGP V3 device into 8x mode
spstarr: dropping down to 4x... on reboot
spstarr: airlied: i am noticing some corruption in text glyphs
spstarr: letters are becoming garbled
airlied: spstarr: yup known problem I think
airlied: ah compiz fails on my r300 as well.
spstarr_home: airlied: rv350 locked up.. reenabled EXANoComposite mode
spstarr_home: r300 - set to AGP 4x.. still locked up
spstarr_home: no X starting
spstarr_home: i run X on commandline and its just stuck
spstarr_home: Fatal server error:
spstarr_home: xf86OpenConsole: VT_WAITACTIVE failed: Interrupted system call
spstarr_home: as root it just won't ctl+c
spstarr_home: just after(--) using VT number 7 .. nothing
spstarr_home: cannot chvt with ssh to force console switch
spstarr_home: ok will put back my 7500 for now, we have VT issues
spstarr_home: ok, spstarr_desk is back on radon 7500
AndrewR: http://pastebin.ca/1232390 (it works, at least any xv-enabled player give me crashed X ;) this is agp rv280)
AndrewR: ..and moving big window (big for 1024x768) around with enabled composite (in e16 WM) crashed X too ... with "[drm:radeon_gem_ib_get] *ERROR* Major case fail to allocate IB from freelist ffff
AndrewR: [drm:radeon_cs_ioctl] *ERROR* ib_get failed"
spstarr: that's a new one :)
spstarr_home: out of kms:/
spstarr_home: lockups random, even with ExaNoComposite on
spstarr_home: it's interesting to discover which 'path' i use is more stable to use
spstarr_home: composite + exa == right now most stable with ExaNoComposite set to false (off by default)
spstarr_home: with no VT switching, im ok
spstarr_coding: airlied: progress comes in steps, each day is better then the next :-)
AndrewR: http://pastebin.ca/1232410 (last 3000 lines from syslog, with drm debug=1)
spstarr_coding: airlied: it might make sense to disable kms for r3xx for now if F10 comes soon?
spstarr_coding: that way by F11 it should be golden
airlied: spstarr_coding: it depends on what else I can figure out..
airlied: I think rs48x will definitely get no kms.
spstarr_coding: i will test whatever you thow at me
airlied: otaylor: I think I traced the fault to your avoid flushing for new buffers..
airlied: I think the in_vram stuff might not be correct :)
airlied: I'll give it a look now .
spstarr: airlied: it finally wedged with composite mode on
spstarr: now im in a very limited X session so i can do some critical task
spstarr: (fail safe)