dmb__: agd5f, nope, radeonhd has the same issue
dmb__: i would figure other people would have the same issue with the same card
dmb__: agd5f, do you know others with the same video card (GPU) where it works?
agd5f: dmb__: I know that GPU works, not sure about your card specifically
dmb__: anything else i should do?
agd5f: dmb__: can you try a different monitor?
dmb__: hmm, don't really have any around
agd5f: other option is to load vesa with the dvi attached and dump regs using avivotool and compare with reg dump from radeon
agd5f: avivotool regs all
dmb__: agd5f, i'll do that tomorrow and attach it to the bug report
dmb__: goes to be dnow
agd5f: dmb__: sounds good
agd5f: airlied: http://www.botchco.com/alex/xorg/0001-drm-radeon-kms-atom-add-support-for-spread-spectrum.patch
agd5f: adds support for spread spetrum. I don't have an atom laptop at the moment
agd5f: so only tested on desktop cards
airlied: I'll try it on mine
dileX_: hi
dileX_: 2.6.32-rc5 released
fly_: hm
fly_: hey osiris are you here?
osiris: just for a few minutes
fly_: Oh, hmm well there was some sort of problem with the driver (you wanted the crash dump)
osiris: yep
osiris: did it go any further with my patch?
fly_: http://pastebin.com/d69cbfd06
fly_: it gets into game for 2 seconds shows the world (which it didnt before) then dies
osiris: fly_: ok, this is a completely different bug. ping airlied, he's the master behind the lower level stuff like bo handling
fly_: airlied, ping
fly_: ::)
fly_: so this probably easier to fix
fly_: hmm another error in dx7 : drmRadeonCmdBuffer: -22. Kernel failed to parse or rejected command stream. See dmesg for more info.
ConnorBehan: is ~220fps in glxgears to be expected when using a 256MB R500 card with KMS?
ConnorBehan: most people I've asked have at least 1000
fly_: er i get 3000
ConnorBehan: and the funny thing is, on my laptop which has an 8MB r128 card, I get 600!
ConnorBehan: what options are you using?
ConnorBehan: I'm usng AccelMethod EXA, SWCursor, EnablePageFlip and RenderAccel
fly_: we well i do have a few options but its really more about the kernel and the driver version
fly_: are you using KMS?
ConnorBehan: ya
fly_: kernel version?
AStorm: still, that's too slow
AStorm: sounds like your Mesa is too old
ConnorBehan: 2.6.31
fly_: yeah mesa version
AStorm: glxinfo should do
ConnorBehan: OpenGL version string: 1.5 Mesa 7.7-devel
AStorm: hmmh
fly_: with a 7.7/// thats odd
ConnorBehan: does it matter that this is a PCI card?
AStorm: but still, older kernels are slow w/ KMS on 3D
AStorm: ^
ConnorBehan: not PCI-E
AStorm: shouldn't.
fly_: pci does matter.....
fly_: kind of
AStorm: not much.
AStorm: what matters is that you have an old kernel
AStorm: w/o drm-next.
fly_: yea make a backup and try the new rc (thats what i did
AStorm: or just drm-next itself
ConnorBehan: oh... I thought any kernel recognizing the modeset param had drm-next
AStorm: wrong
ConnorBehan: what exactly is drm-next?
fly_: the implementation changes
AStorm: drm radeon development kernel branch
ConnorBehan: has it been merged to kernel.org?
AStorm: whatever is going to be pushed to next mainline
AStorm: it has.
AStorm: (and it is being pulled into mainline)
ConnorBehan: sorry I lied
AStorm: (every once and then)
ConnorBehan: I'm using 2.6.31.2
AStorm: doesn't matter.
fly_: its in the 2.6.32-rc4
AStorm: it still doesn't have most of drm-next work on KMS
ConnorBehan: do I have to enable anything in 2.6.32-rc4 besides CONFIG_DRM_RADEON_KMS?
AStorm: not even that
ConnorBehan: oh awesome
AStorm: this switch toggles the default radeon.modeset value
AStorm: :>
ConnorBehan: ah... the function I think it should've had in .31 :)
ConnorBehan: so that should improve 2D and 3D right?
AStorm: 3D definitely
AStorm: 2D... maybe some
AStorm: assuming it works, there have been reports of failures with r5xx
ConnorBehan: 720p plays worse than on my laptop's 8MB card
AStorm: ConnorBehan: -vo gl and be done with it ;p
AStorm: latest driver even supports the fragment shaders
AStorm: for extra nice scaling
ConnorBehan: trying that now... I love this channel
fly_: theres a few problems with the r300 driver still as ive come to find out in the last few says
AStorm: (so, using vo=gl:yuv=4:lscale=4:cscale=4::slice-height=0:filter-strength=0.4:ycbcr:rectangle=1
fly_: lol
AStorm: for example)
fly_: days
AStorm: not sure about ycbcr option
AStorm: whether it's faster or slower with r6xx
AStorm: (esp. lscale and cscale are useful)
AStorm: (these control scaling shaders)
ConnorBehan: -vo gl actually makes it slower right now
AStorm: because your 3D is slow
AStorm: thanks to your old kernel drm
ConnorBehan: ok, I'll compile it already
ConnorBehan: thanks alot
AStorm: 2D with KMS is *still* slower than w/o though on r6xx
AStorm: visibly, but not as bad as earlier
faLUCE: adamk: yesterday, after I went out, you wrote: [18:06]
ConnorBehan: this is the radeon x1550 PCI-E chipset on a PCI to PCI-E bridge thingy
AStorm: doesn't really matter
AStorm: not sure though if your card will manage the powerful scaling shaders that do unsharp mask
AStorm: the bicubic scaler it should handle well
AStorm: (lscale=1:cscale=1)
fly_: [drm:r100_cs_track_check] thats interesting im not using a r100
AStorm: fly_: are you sure? what's the card?
fly_: x1950xtx thats r580
AStorm: maybe they share that code
AStorm: I don't know, I lost track of the devel some time back
AStorm: have to relearn the code
fly_: it appears as though i only get my kernel parsion error within dx7 emulation when there are other players in the map (counterstrike source in wine)
fly_: parsing*
AStorm: hmm
AStorm: maybe wine does something... special
fly_: i dont believe so, i didnt get the message before applying osiris's patch
AStorm: doesn't mesa have a debug dump of opengl functions called?
AStorm: hmmh
fly_: i dont know where
fly_: really new to the graphix on linux scene
fly_: if you need to know how to make an awesome game server i can help lol ;)
hifi: fly_: you play 1.6?
fly_: i can but i moved to source long ago
hifi: can you get steady 100fps in de_aztec
hifi: any resolution, any settings
fly_: havnt tried on linux
hifi: thats what I'm asking
hifi: with the patches you applied, if they matter
fly_: dont know.. downloading it now
fly_: the patch i applied is a param fix
hifi: I'm stuck with windows ATM because I get 30 fps in de_aztec in the water
hifi: I have a X1950 Pro, RV570
fly_: we'll see soon
hifi: 30-40 fps when you look down towards the water from bomb site A
fly_: hifi yo
fly_: had to do some stuff its done downloading tho
hifi: yeah, np
fly_: appears to be about 45-71 at 1024x768
fly_: in water
fly_: im sure with a very few tweaks i could get that to 71 flat
fly_: gonna try with 800x600 now
fly_: yeah 800x600 im actually getting less fps
fly_: but then again i am playing windowed not fullscreen
vehemens: i see there's a fix for r600 compiz xterm text corruption, but it doesn't fix the other corruption.
taiu: vehemens: what's the other?
vehemens: gnome menus and xchat are the two i've seen so far
vehemens: was wondering if i need some of the other 7.6 patches, or if the problem is still in work.
taiu: vehemens: did you get the 7_6 branch with the fix
vehemens: incorporated the 7.6 patch into a master snap shot from around the 5th, hence the question.
taiu: ok,
taiu: gnome menus went ok for me, I currently looking into kde - which has nice blockiness all over
taiu: trying to understand R600DownloadFromScreen, i suspect there's something fishy there
taiu: you could also try if EXANoUploadToScreen or EXANoDownloadFromScreen makes things better
Omar87: Hi
Omar87: Is such an error known to you? http://pastebin.com/d6415a399
Omar87: http://pastebin.com/d6415a399
tb`n: Omar87: which error? Theres loads of text there.
Omar87: tb`n, this is the output from the dmesg command on my machine.
tb`n: yes, but where's the error I ask?
vehemens: page allocation?
mzz: lemme dig up his other pastebin
Omar87: tb`n, I use Ubuntu Karmic, and I'm having the problem that the system crashes and logs out on me. I was told at #ubuntu-bugs to show you the output of dmesg.
Omar87: mzz, hi.
mzz: his X is printing "drmRadeonCmdBuffer: -12. Kernel failed to parse or rejected command stream. See dmesg for more info." according to gdm
mzz: I referred him here. It's a ubuntu karmic system, I don't know the exact version numbers of components involved.
mzz: err, it *was* karmic, right?
adamk: thinks it's pretty clear what error he's referring to in dmesg, but has no idea what's causing it.
mzz: his X is apparently killed by whatever this is. Xorg.0.log(.old) is just cut off.
adamk: Googling "page allocation error" comes up with some results, suggesting this is due to memory fragmentation.
adamk: http://www.cyberciti.biz/faq/linux-page-allocation-failure-erro/
tb`n: Omar87: is it karmic with patches up til today?
Omar87: tb`n, yes.
vehemens: maybe he need's patches up to tomorrow :>
Omar87: tb`n, oh, wait. I think there's a new load of updates that I wasn't aware of.
mzz: I haven't really looked into how recent the radeon stuff in karmic is
Omar87: Yeah, I'm updating the system now.
tb`n: Omar87: looka at kernel-karmic-kms there https://wiki.ubuntu.com/KernelTeam/ReleaseStatus/Karmic#Milestoned%20features
tb`n: Omar87: "KMS for ATI Radeon is partially available but not of sufficient quality in 2.6.31 to enable for release."
Omar87: tb`n, so what is the solution now/
Omar87: ?
Omar87: It just happened again now!
Omar87: Please, either tell me what to do to fix it or tell that there's no fix for it at the moment.
Omar87: or tell *me*..
Omar87: tb`n, it just crashed on me now.
Omar87: tb`n, and that was a second time.
Omar87: tb`n, pleas help.
Omar87: please*
Omar87: tb`n, and a third time.
fly_: ok man i hope airlied can give me some kind of fix for this issue :/ cs 1.6 is SO ugly lol
Omar87: tb`n, it appears to happen every time I launch Firefox now.
Omar87: will be back in an hour or so.
tb`n: Omar: the kernel on ppa seem to be old, and if you want to use KMS with your Radeon you'll just have to compile it yourself, it seems.
tb`n: oh
tb`n: I'm slow. :-P
fly_: ooo 2.6.32-rc5
fly_: download
fly_: lol
fly_: is kernel happy
fly_: loo loo loo compiling
mmark: hi !
mmark: i have a problem with xv on radon 4770
mmark: i have compiled latest 6.12.4 driver, but xvinfo shows noting
mmark: any sygestions - how to make it working ?
soreau: mmark: Do you have HW direct rendering working?
soreau: glxinfo|grep render <-- should output two lines at least
mmark: no, i hAVE NOT hw DIRECT RENDERING
mmark: is it conected to xv / textured xv ?
soreau: I dont think its connected per say but, which driver are you using?
mmark: i am not interesed in HW rendering, games etc - i need xv working with mplayer
mmark: i am using ati xf86-video-ati 6.12.4
airlied: mmark: you need kernel support
mmark: but my X serwer is much older, 1.4.2
mmark: do i need kernel support for xv ?
adamk: Yes.
mmark: i dont need hardware accel
adamk: For r500 and higher GPUs, you need kernel DRM support.
adamk: Xv is done via the 3D engine on those cards.
airlied: xv is hardware accel
airlied: v2.6.30 is first kernel for r600 to have Xv support
mmark: 4770 is r740
airlied: same difference,
adamk: Right, but you still need 2.6.30 for DRM support.
mmark: anyway, i need to check kernel support
mmark: thx sor tips
mmark: thx for tips
mmark: ok, i will compile DRM and will check
mmark: thanks for help
microm: my KDE icons and desktop text and items are all a bit screwed unless I use EXANoDownloadFromScreen and EXANoUploadToScreen. Is this a solvalble problem with the radeon 4650 ?
tb`n: microm: are you running KMS with your 4650? What is your experience apart from this problem?
microm: apart from this problem I am very happy with it. What is KMS (I need to google it)?
tb`n: Kernel Mode Setting, the new stuff in the kernel.
tb`n: microm: since you ask, you probably don't have it :)
tb`n: microm: I don't think any distributions use it by default yet, except perhaps the upcoming Ubuntu (Karmic) which only supports KMS on Intel graphics so far.
microm: I'm at gentoo 2.6.31 and xf86-video-ati-6.12.4 but I have not seen anything labeled KMS, maybe I missed it
adamk_: microm, It requires a newer kernel for your card.
adamk_: It's part of the DRM.
microm: I don't need a solution right now, I just want to know if the problem is known and if it will go away some time in the future, or if I should return the video card to the store and buy something else.
gimzo: archlinux uses kms by default (I just did an update a few days ago and noticed kms
microm: ie will I always have to use EXANoDownloadFromScreen and EXANoUploadToScreen ?
adamk_: Does using those options actually cause a significant performance drop?
adamk_: Is that an AGP card, btw?
microm: adamk_: yes it causes a performance drop, and yes it is an AGP card (but I set BusType to PCIE in xorg.conf as agd5f told me)
adamk_: I had to disable various EXA options back on an older AGP system that I no longer use.
microm: I still have quite a few AGP system 3 in use and a 4th one as spare motherboard - so I am very attached to AGP for good or for bad
Pallokala: I have also some corruptoins with AGP-card and git-sources
Pallokala: usually with forward/bak-buttons in FF
Pallokala: after running it for a long time
mzz: Pallokala: certain glyphs in font rendering getting corrupted too?
mzz: I might've been seeing the same (along with annoying lockups after a while)
mzz: I haven't tried turning agp back on recently though
mzz: I'll probably try that again once I'm on a newer drm and/or entire kernel.
mzz: (I'm on the oh-so-ancient 2.6.31.latest :)
Pallokala: well not always, but g.ex. occasional lines of text are corrupted in thunderbird
Pallokala: I haven't experienced any lockups yet
Pallokala: doing bustype "agp" causes really wild corruptions for me
microm: one the kde pannel, the virtual desktops have some corruption, the taskbar is almost unreadable, the iconify,minimize and close buttons have corruption and the desktop items. But I have not had locks up since I increased the aperture to the max allowable by my bios, dri works and text is fast.
mzz: Pallokala: it's fine here until I've used it for a while, at which point it either locks up or starts corrupting certain characters and then locks up not too long afterwards
mzz: I haven't seen either since I turned off agp (agpmode -1)
fly_: I appear to no longer get that kernel error with 2.6.32-rc5
fly_: anyone else have huge cpu usages with OpenGL?
adamk_: That's kind of a broad question. Different applications could hit different fallbacks, and some operations might heavily hit the CPU.
fly_: glxgears gives me ~50% cpu usage
fly_: that cant be normal
mzz: no?
Pallokala: well I use Gnome, but I see the gradual corruption too
mzz: fly_: I'm not sure glxgears in non-vsynced mode would be a bit of a cpu hog
mzz: err
mzz: it might be, I mean.
Pallokala: glxgears is not a benchmark... and it consumet a lot of cpu
fly_: hmm how do i turn vsync off (thats another thing ive been meaning to ask)
mzz: with vsync on I wouldn't expect it to use that much cpu, at least it doesn't here and it's not like this is an awesomely fast system
Pallokala: Hey, with earlier radeon-models one could build a special vga->scart-cable (with resistors and everything) and get nice RGB-display on old-fashioned CRT-TV. Is that still possible with rv630 and OSS drivers?
mzz: I forgot what the trick for that is
hoo-hah: soreau: thanks man
Pallokala: afaik vsync is a nice thing to have
honk:
honk: are you sure you're not seeing the average of 2 cores?
fly_: canmmm xorg i see
soreau: hoo-hah: Glad you figured it out. Don't know why it would start without dbus, but I will make a note of it
soreau: wouldn't*
fly_: what? why should it be 100%
mzz: glxgears with vsync on is only a few % of cpu load here, with fps pegged to my monitor's refresh rate pretty much
mzz: if fps is considerably higher than refresh rate I'd expect it to load one core fully, I think
fly_: glxgears reads 3200 fps for me
mzz: then yes.
jaykay: hi. is there a problem with the git-repository?
fly_: are we not understanding the meaning of hardware acceleration ... this is why the video card has a gpu
mzz: but that's a pretty daft benchmark, as people keep saying.
jaykay: i always get "fatal: The remote end hung up unexpectedly" when checking out
fly_: the video cards gpu is supposed to be 100% not my cpu
honk: no
honk: glxgears is not a benchmark
mzz: amount of work done cpu-side isn't going to be zero
honk: (neither for your gpu nor for your cpu)
fly_: well i dont doubt it creates "some" cpu load but 50% is a bit much
mzz: and with vsync off it's pushing frames as fast as it can, which'll end up with it being limited on either end.
honk: fly_: like I said: 50% is too low :}
fly_: how can that be low? thats HUGE
honk: no it's not
mzz: I'm not sure which end I'd expect to limit for this case, but it's a silly case anyway.
mzz: I really don't understand what people have against vsync.
honk: while (true) doStuff();
honk: that'll generally give you 100% load on something :]
mzz: yep
honk: and drawing a 2d surface is not something that'll cause a lot of load on your gpu ;p
mzz: I'd expect you to end up with the cpu waiting on the gpu or vice versa. I can't really predict which it'd be. Apparently it's the gpu waiting on the cpu.
fly_: well how can we expect to have a game preform well when its using up 100% of the cpu
fly_: thats a recipe for bad gaming experience
honk: glxgears is not a game
honk: or a benchmark
mzz: do you realise what 3200 fps *means*?
fly_: no not glxgears i bet i could load up tuxracer and itll give me high cpu
fly_: lol
mzz: if it's not vsynced: yes, I'd expect to see 100% cpu load.
honk: anything that doesnt have a built in limit will hit 100%
honk: that's kinda obvious isnt it?
mzz: that's what not being vsynced *means*. It'll keep pushing out frames as fast as it can, pretty much by definition using 100% cpu.
honk: if your system does something as fast as it can, it'll cause 100% load
Pallokala: but if it is single thread program on dualcore?
honk: Pallokala: then he'll see the 50% he's reporting ;p
mzz: and yes, I'm assuming your 50% is actually 100% of one core.
Pallokala: is the maximum load 100% or 200%?
Pallokala: Solaris and Linux did this differently, right?
honk: Pallokala: depends on the tool you use ;)
fly_: yes its one core
fly_: but only 40%
fly_: now
fly_: for some odd reason
fly_: lol
honk: are you sure you're not running HT or something? ;p
honk: or a really ancient card? ;p
fly_: x1950xtx is old but not ancient
honk: *nod*
Pallokala: r5xx, right?
honk: whatever..
fly_: yup
mzz: still, this isn't an interesting thing to measure.
honk: glxgears is not a benchmark. end of discussion ;p
mzz: more interesting is if some app you actually care about can push out frames at a rate that's smooth to the human eye (closer to 100 than to, say, 10)
fly_: my wine games push 50+
mzz: once you start pushing out more than 100 frames per second it stops being interesting no matter what app it is, imho.
fly_: capped at 85 (vsync)
honk: fly_: okay, tell us again what you're complaining about
fly_: the cpu usage
fly_: will
fly_: effect
fly_: gameplay
honk: your games run smooth, dont they?
fly_: not entirely
honk:
honk: you just said they did
fly_: ok now stop a minute and think about this
honk: rotfl
fly_: games need cpu..
honk: and the games cause the load on your cpu
fly_: if theres no more cpu to give
mzz: games frequently find a use for all the cpu power you have available.
fly_: then the game doesnt run well
fly_: its a simple concept
mzz: again, this isn't all that interesting a measurement.
honk: they are the ones that decide how much cpu will be used for drawing and how much for AI 'n stuff
honk: fly_: seriously, you're not making any sense
honk: glxgears uses all cpu because it's written that way
fly_: no YOU arent listening or at least not understanding how game performance works
honk: I'm well capable of writing a game myself, kthx
fly_: and im not talking about glxgears
honk: glxgears on nvidia is using 100% cpu, too
fly_: ^ see above comment
mzz: if the games are targetting a framerate above 50 and currently getting about 50 any improvement driver-side will show up as more frames per second, not lower cpu usage (it'll still be 100%)
honk: actually NOT using 100% cpu is the real problem here
honk: that means either the driver is doing weird stuff or your gpu is too slow ;p
mzz: if cpu usage drops below 100% that means the game ran out of stuff to do (it's drawing all the frames it wants to draw, and doesn't have anything to do on other fronts either)
honk: (then again, this is just glxgears we're talking about and a gpu doesnt need to be fast in drawing that to be fast in games. again: glxgears is not a benchmark)
fly_: ok honk listen to yourself.... you are saying its normal for a game to use 100% cpu due to graphics usage when IT needs to preform cpu side calculations
mzz: since "all the frames they want to draw" frequently equals vsync or isn't limited at all, seeing 100% cpu usage is still the rule, not the exception.
fly_: wheres the cpu for calculations huh
fly_: ?
fly_: thats not normal
honk: fly_: a game's main loop: while(true) { doAI(); drawFrame(); }
honk: that'll yield 100% cpu no matter what
honk: and it'll do any necessary calculations anyway
mzz: honk: only way it wouldn't would be if (a) it blocked on the gpu somehow and (b) there aren't any other threads that want to run
rah: fly_: I don't think he said it's normal for a game to use 100% cpu due to graphics usage; he said it's normal for a game to use 100% cpu
mzz: in practice I haven't seen this happen
fly_: rah, i could understand then but ive never seen any OS use 100% cpu for a game before linux
honk: fly_: properly coded games use as much cpu for the logik as they need - and ALL THE REST for gfx
mzz: I have
honk: that's 100%. no matter what
fly_: so what good is my gpu then
honk: it's good for getting more than 0,0001fps
rah: fly_: how often have you measured CPU usage on linux and other OSes?
honk: fly_: btw, why the heck are you still using a CRT? *g*
fly_: i measure cpu usage on any OS i use
fly_: honk, little think about if it aint broke dont fix it
rah: fly_: while playing games?
fly_: rah all the time
rah: fly_: why?
honk: fly_: operations are offloaded from the cpu to the gpu. but that doesnt mean that the cpu will be completely idle either. it still has to tell the gpu what to do (and lots more)
fly_: because it is one of many things that tells me how i can improve the way a game plays
honk: fly_: if you want idle cpu, limit the game's fps
honk: that's all there is to it
mzz: fly_: your hardware specs must significantly exceed mine, or you must run stuff at "lower settings" (for lack of a better description) than I do
mzz: both are actually likely, I must admit.
fly_: its only pushing 50fps at times.. how should that create 100% cpu
honk: wtf?
honk: it's only pushing 50fps because it's at 100% cpu
mzz: fly_: it's hitting 100% cpu *because* it's only hitting 50fps at times
mzz: heh
mzz: fly_: if the driver was better and/or the game less demanding it'd be able to consistently push out whatever its target framerate is (vsync?) and you'd see cpu usage drop
fly_: target framerate is 85
mzz: fly_: as long as it's not reaching its target framerate cpu will be pegged at 100% (and/or vice versa depending on how you look at this)
honk: fly_: vsync is off - or glxgears wouldnt hit more than 85fps
honk: ;p
fly_: if its off its off by default
honk: of course it's off by default
mzz: honk: not here actually (my glxgears is vsync-limited, and I don't recall me doing that explicitly)
mzz: doublechecks
rah: fly_: what is the target frame rate for the games you're running under windows and the games you're running under linux?
rah: s/what is the target frame rate/what are the target frame rates/
mzz: fly_: yeah, what are both fps and cpu usage like under windows for the same game?
fly_: i easily push 100 fps in counterstrike source in windows (given theres no emulation going on) i actually have it capped at that or itd go to 150 or so but even at 150fps the cpu usage for the game will only be about 30%
mzz: ok, so tell it to target 1000 or so
honk: fly_: what you should learn from this is that 100% cpu usage for games is absolutely normal
mzz: and/or remove the cap, depending on how this game works
fly_: it just wont
rah: fly_: and what are the target frame rates for the games you've measured cpu usage for under linux?
fly_: target framerate is whatever i set it to
fly_: currently 85
mzz: alternatively: try again under linux and tell it to limit frame rate to something below what's actually achievable
Pallokala: and yet all the fps which exceeds the refreshrate of the display is useless
fly_: the refreshrate is 86
fly_: 85
rah: fly_: are you talking about the monitor framerate, or the game's frame rate limit?
mzz: they may match, obviously
rah: they *may* match
honk: rah: he's just assuming that vsync is working
fly_: Game fram limit! : 85 monitor refresh rate : 85
mzz: I'm not :)
fly_: lol
mzz: fly_: ok, so limit the game to 10, and remeasure cpu usage
rah: fly_: what games?
honk: css ;p
fly_: it icounter strike source
rah: there's a linux client for counter strike?
fly_: nope
mzz: fly_: so apparently this thing runs better under windows, which means there's still progress to be made on the driver and/or wine front
honk: both probably ;)
rah: you're using wine?
honk: rah: yes he is
rah: lol
fly_: ugh itll do the same if i run google earth
mzz: fly_: once you exceed 100 or so fps the numbers stop interesting me, and once you hit the game's target framerate cpu usage drops. So far I'm not surprised.
rah: fly_: try capping the framerate of a game that runs natively on both windows and linux, and measure the cpu usage on both OSes
mzz: fly_: I'd expect cpu usage to also drop under linux as long as you limit the game to a number of fps that's below what's actually achievable, which is apparently (unfortunately) lower than what you get under windows.
mzz: rah: there's a bunch of overhead imposed by wine, sure, but is that really an order of magnitude?
rah: fly_: if there's a difference, that's noteworthy
rah: mzz: wine's overhead may not be the only thing effecting cpu usage
mzz: sure, but iirc his "easily push 100/150" game under windows and his "stuck at ~50 with the cpu pegged" under linux are the same game (using wine)
rah: if the same things without using wine, then I would take notice
rah: otherwise, there's an unknown in there
rah: (and an unknown which, in my experience, can have a wildly varying impact on game performance)
mzz: so far I'm not all that surprised, although it'd be nice if the performance gap in linux vs windows was smaller, and I don't know how much of that is wine and how much is the driver
fly_: agreed wine creates overhead
rah: fly_: wine's overhead may not be the only thing effecting cpu usage
mzz: especially since iirc source is very much a d3d thing, so you're hitting wine's d3d -> opengl layer
mzz: so this isn't entirely an apples vs apples benchmark :)
fly_: capped fps at 30 no difference in cpu
mzz: (since with that layer in there what's being asked of the card may not be the same either)
mzz: ok, that starts being a bit odd, but try capping at 10, and making sure it's consistently meeting that
fly_: mm i dont believe the games engine will allow me to cap it at 10 but ill try
mzz: or at least cap it as low as possible and make sure it's stuck at the capped value, not dropping below it
rah: fly_: nexuiz will run natively under windows and linux
rah: fly_: http://www.nexuiz.com/
fly_: that would require me booting windows which i am not into lol
fly_: im trying to replace windows
fly_: i bookmarked it i might look into it after doing some tests here
honk: just complain properly next time: "my game is not running fast enough" - that'd lead to "well, d'uh!" and the discussion could have ended there ;P
mzz: so far the oddest thing you've told me is that your glxgears isn't using vsync by default while mine is
fly_: lol
mzz: other than that: what honk said, although I'd s/d'uh/sucks/
fly_: just nope the games engine has a minimum frame cap of 30fps
fly_: im gonna try some kernel adjustments
mzz: well, if you use a less demanding game or one that the drivers like better I'd expect cpu usage to drop (as long as the game supports vsync or some other kind of limiter)
Omar87: Hi
Omar87: I'm back
Omar87: Can you help me please?
fly_: using a different game is not an acceptable solution to a problem
mzz: Omar87: tb`n told you roughly that you need a newer kernel if you want to use kms right after you left
mzz: Omar87: that is: either disable kms or use a kernel (or at least drm modules) newer than 2.6.31
mzz: at least iiuc.
Omar87: mzz, where do I find that new kernel? www.kernel.org?
mzz: I'm just trying to channel tb`n with a little help from my /lastlog :)
mzz: can't help you there. Your best bet currently might be to just drop kms and wait for 2.6.32, imho.
Omar87: mzz, and how do I go about that?
mzz: I'm actually a bit puzzled with kms being on by default. Is it?
Omar87: mzz, I'm not sure.
mzz: actually nvm. I'm gone, sorry.
Omar87: mzz, how do I disable it?
Omar87: mzz, I'm searching for it now, but I didn't find anything so far.
fly_: omar
Omar87: fly_, yes?
fly_: the the newest kernel can be found here http://www.kernel.org/pub/linux/kernel/v2.6/testing/
fly_: KMS setting can be found in the config --> Drivers --> Staging --> Youll see the radeon one there its somewhere near the bottom iir
Omar87: config?
fly_: have you compiled your own kernel before?
Omar87: fly_, nope. But i'll be happy to learn.
Pallokala: which distro you are using?
fly_: heh not from me :) theres wikis out the wazoo online
Omar87: Karmic.
Pallokala: ubuntu 9.10 beta, right?
Omar87: Pallokala, ahh.. of course.
Pallokala: you can get binarypackages for it from somewhere
Pallokala: so you don't need to compile the kernel yourself
fly_: (though you should learn at some point)
Pallokala: if you end up doing it, ubuntu forums probably have multiple good guides how to do it right
fly_: and BACK UP BACK UP BACK UP
Pallokala: s/should/might have to/
fly_: make beeping noises if you gotta
Pallokala: well... usually kernel upgrades dont break anything
Omar87: fly_, back up. That's what I hate most about such things.
hnsr: I guess from http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.32-rc5/
fly_: no but improperly compiling/referencing them wrong in grub can make you think you broke it
fly_: lol
fly_: i had a friend reinstall 5 times because he didnt know he could repair
Omar87: fly_, but it doesn't actually break it, right?
fly_: ive never managed to "break" a system compiling a kernel you can certainly make it not boot though
Omar87: ok, thanks a lot for the help.
Pallokala: search the net for proper guides
fly_: which is why backups are handy i have a /boot tree in a folder within boot if something goes wrong (horribly) i can just cp it over
Pallokala: why did he remove old kernels?
fly_: who where
Pallokala: your friend
fly_: friend?
rah: [13:24]
fly_: OH no he didnt delete them just changed their grub entries or what have you wrong
rah: reinstalling 5 times?
rah: you'd think he'd try something different after the first time
fly_: finally the 5th time i came over and booted into a rescue shell mounted his drive and went poof and he went OHHHHHHh
rah: kind of, you know, learn what he was doing..
fly_: hey man he saw me using linux and decided to try for himself lol
fly_: i TOLD him to read first
fly_: silly windows users you know
rah: sadly, I do
fly_: i had an arguement with another buddy the other day he asked me why im using linux when windows is so much better, i told him its not and hes like then why do you have so much trouble getting stuff to work , i told him its because linux is all about manually setting things up and he maintained that windows was a better OS because billions of people use it and that means its better
fly_: facepalm
muep_: imo linux isn't all about doing manually something that could be done automatically
Pallokala: if million flys love a stack of poo...
fly_: eh.. gentoo would disagree muep_
fly_: lol
muep_: gentoo != Linux
mzz: "linux" is a bit pointlessly broad
fly_: well im just saying gentoo is ALL About manual
mzz: yep
fly_: its the stickshift of the linux wold
fly_: world*
mzz: and stuff like ubuntu very much isn't, and is occasionally easier to get things working on than windows
muep_: linux is just a quite versatile kernel, which you can get well pre-configured from multiple vendors
stikonas: or compile it from scratch using Linux From Scratch :)
fly_: yeah lol ive been meaning to try a LFS
muep_: or without LFS, just reading the components' readmes
muep_: ...
fly_: brb reboot
soreau: wonders if 7 is any better at including drivers than xp
muep_: why not?
soreau: My pride with linux is most all drivers are included with the kernel
fly_: yeah all that driver support in a teensy 2-3mb kernel
biotube: soreau: if my experience with it is true, then no
fly_: even less if you are REALLY into trimming the bloat
biotube: soreau: no support for rt2500 out-of-the-beta-box
octe: fly_: all the drivers in a 2-3mb kernel?
octe: most of the drivers are seperate modules
octe: $ du -sh /lib/modules/2.6.31-14-generic/
octe: 111M /lib/modules/2.6.31-14-generic/
fly_: not ALL just the ones you need for a system
fly_: octe thats if you dont compile most built in
soreau: biotube: Some chipset companies son't care to work with linux devs to produce drivers. Others embrace linux banking that if it justwerks it will sell better and get less negative reviews
fly_: my modules folder is 500kb
octe: fly_: true i guess, if you actually take the time needed to only compile in what you need
biotube: soreau: i was talking about windows 7
octe: i haven't done that in several years :)
fly_: yeah it takes some doing but make defconfig helps a lot
octe: i kust don't see the gain for that kind of time investment
fly_: its like kernel lipo
octe: not to mention i won't get automatically updated kernel images
soreau: biotube: Well you did use some pronouns
octe: from my distro
soreau: biotube: By 'it' you must have meant 7
octe: fly_: have you compiled out module support completly?
soreau: has never used this 7 OS
fly_: i have before not currently tho
fly_: my servers dont load modules and run with 1.8mb kernels
biotube: soreau: I don't know of any other '7'
biotube: my bad
octe: is your 2-3mb kernel everything you need or do you have an initrd too?
soreau: biotube: By contrast, does rt2500 (whatever that is) work OOTB with linux?
biotube: soreau: yes
fly_: octe for a server most definitely
fly_: my current kernel is about 2.8
soreau: In that case, my point is exemplified :)
octe: for a server i can see the use in stripping it and removing module support completly
fly_: oh yeah my game servers run amazingly despite the horrible coding valve does
fly_: rediculous number of usleep calls
soreau: thinks it might be interesting to see how many drivers included by default for lin compared to win
fly_: i dont know.... win has a LOT of drivers
soreau: namely network devices. "Ok, I have windoze installed but I need the net working to get my network drivers..."
fly_: lol i love that
fly_: -me beat head on wall
hnsr: what's wrong with a high number of usleep calls? out of curiosity
biotube: fly_: windows *lists* a lot of drivers, but a lot of them are probably the same under the hood(or should be)
soreau: obviously you will need 'The CD' or another machine already with network working and a way to put the drivers over to the poor winbox (but I digress)
fly_: hnsr, it creates a lot of cpu load and causes server side fps to be unstable
hnsr: because I use that in my game loop when fps is being capped, if enough time is left so I don't end up eating 100% cpu for nothing
fly_: this is only a factor of course for 1000 server side fps games
fly_: ie cs cs:s
octe: i find it funny that there are NO 64-bit windows drivers for my pci sata-card
octe: works out of the box in linux
hnsr: basically when I run at 60 fps, and have little to render, having the cpu sleep a little lowers cpu load from 100% to just a few %
fly_: hnsr oh yeah for sure
soreau: octe: None whatsoever, eh?
octe: soreau: yup.
fly_: hnsr the valve games have engine caps of 1000fps though
hnsr: fly_, ah, yeah at that point i stop calling sleep altogether
fly_: they dont though lol
octe: soreau: not that i found at least..
fly_: actually removing usleep altogether will result in VERY HIGH fps but unplayable
soreau: On another plane, I get the general feeling that 64bit sw sucks for the most part even though I've yet to test 64bit firsthand. If I did, I'd probably run 32bit OS on it
fly_: yo-yos lagouts heavy load times
octe: 64-bit seems to work fine these days
fly_: unless you want to play 32bit games then it can be a pain in the butt
octe: no problems for me with linux or windows 64bit
octe: i haven't had any issues with 32-bit games really..
fly_: eh its all in having the right libs/drivers
octe: i suppose
octe: and another fun thing with properietary systems like windows is my x1900xt card, obsolete by amd/ati
octe: and the last released drivers from them that support it do not support windows 7
octe: i installed them anyway, but they seem glitchy sometimes
octe: linux has a pretty high works-out-of-the-box factor for my hardware config right now ;)
octe: compared to windows
soreau: so fglrx AND windoze ati drivers dropped support for r3-5xx at the same time?
soreau: or just for 7 or what
muep_: soreau: for all platforms, the same time
muep_: but unlike many Linux users, XP/Vista users can keep using the old drivers for quite long afterwards
glisse: muep_: you can too if you don't update your kernel or xorg
fly_: yup
fly_: no not for 7
fly_: because ATI felt their pockets werent geting fat enough
muep_: glisse: I know I can, but I don't think many would want to do that
muep_: but yeah, it's a possible workaround
fly_: is seriously mad about ATI putting his video card into the legacy class...
rah: fly_: you could buy another manufacturer's card to register your protest
muep_: is happy with specs being available for both of his legacy ATI GPUs
rah: fly_: but then they'd just do exactly the same
fly_: yep
fly_: well nvidia doesnt seem to..
rah: besides, they've dropped support from their proprietary driver
rah: about which the best thing that can be said is: meh
fly_: yeah i wasnt impressed with the fglrx driver
rah: luckily, we have hardware documentation
rah: and hence, good free drivers
fly_: i find it strange my fan doesnt power on..
fly_: doesnt that mean its not working hard enough to get hot?
hnsr: what really annoys me is intel windows drivers, I can't even freaking install them, says my system vendor isn't validated or something
rah: I'm guessing that would come under "power management" which is To Be Done
hnsr: so all I get is opengl 1.5 support form whatever drivers came with my laptop
rah: mmm... windows
soreau: muep_: Linux users can keep using fglrx on r3-5xx if they keep old versions of X and kernel. Hardy, Debian *cough*
fly_: yea screw that
muep_: soreau: I find that quite limiting
fly_: i liked my answer better
muep_: and I prefer to use the free drivers anyway
fly_: lol
fly_: NEW YORK CITY! -bang-
fly_: actually im finding my card runs better with the free drivers
fly_: the fglrx have always been a sad attempt to make linux users not complain about their hardware not working
soreau: muep_: Of course, I was just making a point to your previous comment (insinuating linux users are forced to quit using fglrx if they prefer not to)
ferret_: I was basically forced to quit using it
fly_: i read a way to force fglrx to work with higher versions
ferret_: It was either that or never upgrade things again
fly_: i dont know how well it works that way tho
ferret_: there are dodgy patches
fly_: i caught a huge spider
fly_: its at lwast |---this---|
fly_: big'
soreau: never heard of the proprietary driver referred to as a spider, but whatever
fly_: lol
fly_: anyone got an opinion on kernel hz settings in regards to graphics acceleration?
chithead: the exact value is only important for video, should be a multiple of the frame rate if possible
honk: graphics acceleration == desktop usage == high hz :}
fly_: yeah but what about dyn ticks
fly_: i changed my kernel hz from 1000 to 250 and am seeing no difference
honk: dyn ticks are a good idea either way
honk: energy saving ftw :]
fly_: unless you happen to need high hz for a server
fly_: in which case in order to stabilize server side frames dyn ticks is preferrably off
honk: says who?
fly_: says me after testing it
honk: yah right
fly_: lmao go ahead and download SRCDS make a server set fps to 0 and watch the fps at www.fpsmeter.org
honk: and I say pigs are flying .]
fly_: ill wait
fly_: :)
honk: I'm not gonna run any tests myself
honk: and I'm not just gonna take your word for it
fly_: well then your point is moot
fly_: :) i HAVE tested this
honk: my point is not moot :}
honk: I just dont believe your values
honk: (judging from what else you say in here, they are simply not trustworthy)
fly_: so nice of you to base an assumption on a completely unrelated topic
fly_: me = server guy, you = graphics guy
fly_: two different worlds
fly_: only i get paid to do servers ;)
fly_: Omar87 werent you the one doing kernel stuff>?
fly_: oh quit
fly_: not join
fly_: hah
fly_: maybe he finally finished :x
honk: I'm not a graphics guy
agd5f: Pallokala, others with UTS/DFS issues, make sure you have the latest drm-next bits from airlied's tree
fly_: how often do mesa drivers get updated these days?
agd5f: fly_: almost everyday
chithead: see http://cgit.freedesktop.org/mesa/mesa/ for the commit log
fly_: hmmm i forsee many gits in my future
fly_: and multiple times a day from the looks of it lol
fly_: finds it refreshing to see an open source project so active
Pallokala: agd5f: UTS/DFS? you mean display corruption? or the AGP issues?
Pallokala: also, do you know the answer to my question about vga->scart-output of rv630 (HD2600Pro)
Pallokala: I tried to look for this, but didn't find anything definite
agd5f: Pallokala: display corruption with kms
agd5f: Pallokala: what about scart? if it accepts an rgb signal, it's just a matter of choosing the right modeline
Pallokala: I dont use kms
agd5f: Pallokala: ok
Pallokala: ok, so new radeons can still output this (intel cant)
Pallokala: I will make an r5xx-compatible cable and try it
agd5f: well then with agp and non-kms you will likely have to disable UTS/DFS
fly_: display corruption like what?
agd5f: small pixmaps like fonts icons
Pallokala: well I'll update to 2.6.32-rc5 and turn kms on
fly_: hmm yeah thats what i did
agd5f: Pallokala: not sure that has the fix yet
Pallokala: well I shall find out
agd5f: your best bet is airlied's drm-next
fly_: i dont know if it fixes your problem but it helped my card a heck of a lot
Pallokala: ok, I will pull it over the vanilla-sources
fly_: hmm so im trying the newest libdrm and mesa(yes im compiling it again) from git.. does it make any sense to update xf85-video-ati
fly_: 86*
adamk_: fly_, Depends on how new the xf86-video-ati is that you are currently using.
fly_: that one i dont know how to check :)
adamk_: /var/log/Xorg.0.log would list it.
fly_: and would i possibly need that for any compatibility reason as 32bit ( or is it noarch )
fly_: im 64bit
adamk_: No.
adamk_: YOu only need the 64 bit version of the Xorg DDX.
adamk_: You would needs 32 bit versions of libdrm and Mesa, of course.
fly_: mmk just checking (im going to be updating my 64bit as well
fly_: lots of compiling gonna happen. .. zzzZZzz
fly_: uhh what the configure errors
fly_: ./configure: line 12434: syntax error near unexpected token `XINERAMA,'
fly_: Futurama?
adamk_: You likely need an updated protos package.
fly_: hmmm good call lemme look into that
agd5f: fly_: need xorg macros
agd5f: package is usually called something like xorg-utils
fly_: indeedhmm
adamk_: Ahhh.. macros, not protos.
fly_: i just yum-builddep'd it
fly_: ^^
fly_: save me some trouble
rhodan: If I want to try Gallium3d, which packages do I need? Any special branches?
MostAwesomeDude: Nope, it's all on Mesa master.
rhodan: MostAwesomeDude: so I just emerge mesa with USE="gallium" and change the driver in the xorg.conf to - what?
rhodan: I recall something like "xorg-tracker" or the like
zhasha: rhodan: you don't need to use the xorg st to use the 3d driver
MostAwesomeDude: rhodan: "modesetting", but if you want to experiment, I really recommend testing the DRI driver instead.
zhasha: I'd recommend you don't. It's another possible point of failure
rhodan: So what do I do now? It's not a productive environment, I don't mind things breaking. Im a console monkey anyway.
rhodan: And what's the DRI driver?
MostAwesomeDude: I have no idea how to do this from the ebuild.
rhodan: MostAwesomeDude: i set USE=gallium and re-emerge mesa. Then I have a mesa with gallium3d-support.
MostAwesomeDude: ./configure --enable-debug --with-driver=dri --with-dri-drivers=swrast --disable-gallium-intel --enable-gallium-radeon --with-state-trackers=dri,egl --disable-glw --disable-glut --without-demos
rhodan: MostAwesomeDude: yeah, that's what it does.
MostAwesomeDude: rhodan: I doubt the ebuild renames radeon_dri.so to r300_dri.so.
rhodan: Just much more comfortable
zhasha: rhodan: r300-gallium (src/gallium/drivers/r300) is the pipe driver, build it with the DRI state tracker
MostAwesomeDude: And you'll need to do that in order to get the driver loaded.
zhasha: and the radeon winsys
rhodan: winsys?
zhasha: rhodan: gallium drivers consist of 3 parts, winsys, pipe driver and state tracker
zhasha: for radeon cards, use the radeon winsys, softpipe and r300 pipe, and a state tracker of your choice (for 3D use DRI)
rhodan: zhasha: Ok, let's assume Gentoo did that for me. I just write Driver "modeset" into my xorg.conf?
adamk_: You either need to enable modesetting by default in the kernel config, or pass radeon.modeset=1 on the kernel boot line to enable KMS.
rhodan: adamk_: I have that since _months_ :)
rhodan: So 'Driver "modeset"' is good?
rhodan: Bah, I'll just try it, brb.
rhodan: nope, didn't work, there's no such driver ;)
dileX: MostAwesomeDude: you got "modesetting" run successfully?
zhasha: rhodan: why don't you listen? DO NOT USE THE 'modesetting' XORG DRIVER
chithead: no need for shouting
fly_: What use the modesetting driver? (sorry habit)
zhasha: rename radeon_dri.so to r300_dri.so, and watch in awe
rhodan: ain't been seing no not
zhasha: chithead: MAD and I both told him not to use that driver, he didn't listen
rhodan: I have both radeon_dri.so and r300_dri.so in /usr/lib64/dri/.
zhasha: remove r300_dri.so
rhodan: ok.
zhasha: rename radeon_dri.so
rhodan: I'll just rename the r300_dri.so, no need for removing, right? I'll name it peter.so
rhodan: brb, re-X-ing.
rhodan: OpenGL renderer string: Gallium 0.3 on RV515
zhasha: there you fo
zhasha: go*
rhodan: Gears are a bit glitchy, though. Does 2D run over gallium now, too?
fly_: how far are the gallium drivers from being done? i havnt even read what they are doing differently
rhodan: Oops, kwin effects crashed my X.
zhasha: rhodan: it's very buggy
rhodan: Can somebody explain to me what I just did? ;)
rhodan: Does everything I see run over Gallium3d now or just the 3d part?
fly_: just gotta ask.... why would they name it after a "poor" metal
fly_: i mean gallium has a really low melting point
rhodan: fly_: They themselves are Tungsten.
fly_: what who is
rhodan: Which has a really high melting point
rhodan: fly_: Tungsten Graphics Inc.
fly_: oh but tungston is a way higher quality metal
fly_: just wondering why the new driver would be called gallium
fly_: lol
rhodan: fly_: Because Tellurium3d sounds wrong.
fly_: hah nice
fly_: so i want to change the libdrm for 64bit but if i yum it its going to kill like 58 dependencies
fly_: should i just try to overwrite?
fly_: i suppose i could just write down what depends on it that i actually want and reinstall them after i do this
fly_: -yawn- the git for mesa is slooow (not the download the compression)
ajax: yeah, the thing about having eleven years of history in your git repo...
fly_: heh
fly_: hey at least its progressing.
fly_: unlike a lot of the open source community
fly_: plays circus music while compiling mesa for a laugh
MostAwesomeDude: rhodan: gears is glitchy because I just committed a thingy that apparently exposes a few glitches in vertex setup.
MostAwesomeDude: fly_: I asked about that choice of words. Apparently gallium was the only metal they could find that's actually used in computers and isn't already used by some other company.
fly_: wow really
MostAwesomeDude: Yep.
biotube: copper? iron? tin? aluminum?
soreau: MostAwesomeDude: Would things be different if I use r300g on a per-app basis by setting LIBGL_DRIVERS_DIR as opposed to using it as prefix/lib/dri/ for X altogether?
MostAwesomeDude: soreau: Very much so. I don't have it installed and I don't recommend people install it.
MostAwesomeDude: Which is why I'm very confused as to Gentoo having it in their USE flags.
soreau: MostAwesomeDude: The only reason I ask is because dileX said he had OA working with it but wouldn't provide any details I asked about - card, which repo he was using, how it was being used etc
soreau: For me, OA r300g just makes a black screen on my rv350
jemark: does someone has a ATI Mobility 9700 card here? how does this one perform in the latest linux distros?
soreau: jemark: It should perform decently with anything that doesn't require >ogl1.5
osiris_: jemark: it should perform quite well on all games that don't use glCopyTex[Sub]Image or FBOs
MostAwesomeDude: soreau: AFAIK OA doesn't work yet.
soreau: MostAwesomeDude: That's what I naturally assumed. Maybe I misunderstood what dileX was saying
mjt: hmm. Apparently font loading is broken with kms on 64bit kernels and 32bit userland.
jemark: osiris_, thanks. it's a bit slow in Debian testing with the opensource driver..
mjt: (in console)
soreau: jemark: Debian packages = old
mjt: when switching to 32bit kernel fonts are all ok, but with 64bits it displays question marks on red background instead of (international) characters.
fly_: MostAwesomeDude, I think thats a sign to mvoe away from naming things after metals... we should move on to Krypton3D now
fly_: gasses FTW
MostAwesomeDude: fly_: I think that names are not very important.
soreau: jemark: If you are at least using X 1.5 or later, I'd make sure to at least use exa as opposed to xaa (see your X log)
jcristau: soreau: is mesa 7.6 really that old?
fly_: jcristau in the scope of development it is
MostAwesomeDude: No, it's not.
MostAwesomeDude: All the stable drivers included in Mesa 7.6 are nearly unchanged in current git master.
MostAwesomeDude: Only the unstable and experimental drivers have seen serious development.
MostAwesomeDude: Check the log; most of the commits have been to llvmpipe, i915g, st/xorg, st/egl.
fly_: ive been thinking about TRYING the gallium but yall would have to give me a place to send errors because im sure ill find em
rnoland: really wishes that someone would merge 7.6 to master....
osiris_: rnoland: I'm on it
rnoland: osiris_: cool, thanks.
rnoland: i suppose that i could do it, but...
MostAwesomeDude: fly_: I started accepting r300g bugs a few weeks ago. Feel free to file them.
fly_: whats the comparison between the gallium and the standard r300?
fly_: i still dont know what exactly is different
MostAwesomeDude: r300 is a stable GL 1.5 driver. r300g is an experimental Gallium driver.
fly_: lol thats not what i meant
jcristau: fly_: git diff src/mesa/drivers/dri/r300 src/gallium/drivers/r300 will tell you exactly what's different
fly_: lmao
osiris_: rnoland: sorry, I won't do it. there's a conflict in shader parser stuff, and that's way beyond my understanding
MostAwesomeDude: Well, from my point of view, Gallium's got a very generic, easy-to-program-for API, and doesn't force me to convolute my code to fit the GL API.
MostAwesomeDude: From your point of view, r300 is stable, and r300g is experimental.
biotube: fly_: gallium supports GLSL
rnoland: osiris_: ok... i sent brian a note at like 5 this morning...
fly_: r300 doesnt?
osiris_: nope
fly_: interesting.. maybe thats why the GLSL setting in wine doesnt work
fly_: @)2
biotube: the infrastructure doesn't support it
MostAwesomeDude: The infrastructure *does* support it.
MostAwesomeDude: Mesa totally supports GLSL.
jcristau: swrast and i965 do glsl :)
biotube: give me a minute to remove my foot from my mouth
MostAwesomeDude: The shader compiler is what's lacking.
fly_: so that would probably hinder my playing dx 9 emulated games maybe?
osiris_: fly_: no, dx9 can be emulated with vp/fp
fly_: i dont know ive been getting a particular crash osiris told me to talk to airlied about it because he deals with bo handling. anyways im still dx 8+ less
fly_: you osiris
fly_: lol
osiris_: and wine supports it since few months IIRC
fly_: he hasnt ponged me back yet
fly_: osiris_ eh.. kinda supports it.. its a work in progress from what ive heard
fly_: id be happy with 8 lol thats all i need to not get kicked from competition servers
osiris_: fly_: what game are you talking about?
fly_: Counter-Strike Source
fly_: the zblock plugin (which every team uses on match servers) kicks players with DX level < 8
fly_: some anti- hacking mumbo jumbo for paranoid people
fly_: i played Counter-Strike 1.6 today without much incident. though i forgot how ugly that game is
fly_: 40+ fps
fly_: and btw i must say compared to the last time i tried to game on linux you guys are kickin arse
osiris_: we're doing our best :)
fly_: i fell complete freedom from windows is merely a few patches away :) anything i can do bughunting wise to speed up the process you name it
osiris_: fly_: file a bug report about the bo problem for easier tracking
fly_: umm yeah i just dont know where to file it
dileX: soreau: maybe, you overread ist yesterday, I gave you the infos (mesa master and rv515 w/ dri/st). OA runs... very slow. but runs. its broken somehow the last days.
osiris_: fly_: bugs.freedesktop.org
UnNamed: hi, any idea of why mesa dri 7.6 is so slow with irrlicht based game? i had the fallback env var and i see "r300DrawPrims: Rebasing primitives. 0xbfd25720 nr_prims 1 min_index 36 max_index 216" (lots, but always small numbers)
fly_: sweet ill file it after this cigarette
osiris_: UnNamed: can you profile it (using the sysprof would be the easiest way)?
osiris_: UnNamed: it probably hitting some software rendering fallbacks
UnNamed: osiris_: only other message about fallback is about smooth line one (no idea what uses that in the game, and i have enabled the env var to check what fallbacks blender was using)
UnNamed: osiris_: and the "funny" thing is that 7.5 worked fine
osiris_: UnNamed: make sure you have enabled "Disable low-impact fallbacks" in driconf
soreau: dileX: Can you say how it is broken?
soreau: Good to know it's broken at least and I'm not doing something wrong (assuming you get black screen too)
UnNamed: osiris_: uh, damn, big diff, from one frame every 30 secs to 20fps (still worse than 7.5)... i really wonder why the smoothline thing
UnNamed: osiris_: i really wonder why no so big problem in blender where stippled lines, etc are common and so problematic for a game
UnNamed: osiris_: so big and radical has been the 7.5->7.6 change for r300?
osiris_: UnNamed: well, the whole driver was rewritten under common memory manager and some major GL extensions has been implemented
fly_: osiris_, I assume Id be filing this under mesa yes?
osiris_: fly_: yes, product: mesa, component: Drivers/DRI/r300
fly_: kk ill have it up in a jiffy
UnNamed: adds per app config then
UnNamed: osiris_: thanks for the idea (never would had guessed the game was really doing something with lines)
osiris_: UnNamed: np
dileX: MostAwesomeDude zhasha soreau: commit-id
osiris_: UnNamed: it doesn't have to be the stipple lines. it could be smooth lines, stipple polygons or smooth polygons
soreau: dileX: What is this all about?
dileX: [19:09:52]
dileX: soreau: ^^
dileX: r300g: openarena: vbo/vbo_exec_draw.c:223: vbo_exec_bind_arrays: Assertion `exec->vtx.bufferobj->Pointer' failed.
UnNamed: osiris_: "WARNING! Falling back to software for smooth lines" :]
UnNamed: osiris_: stipple lines is with blender
soreau: dileX: What I was asking is what happens for you there. Does it cause black screen or just fail to start altogether?
dileX: soreau: black screen. escape to VT-1, kill xorg-server.
UnNamed: osiris_: still no idea what a game does with lines, all are triangles
fly_: hmmm so before i can file this i need to recreate it but im in the middle of compiling git and getting that libGL.so.1.2 error again
soreau: dileX: Ah ok.. that's pretty much the same that happens here. Thanks for clarifying
fly_: hates that pesky libGL.so.1.2
osiris_: UnNamed: maybe it's using wireframe mode for some outlining
UnNamed: osiris_: i doubt it, one coder from the game also wonder what it could be
fly_: How did we get it to stop looking for libGL.so.1.2 again
fly_: n/m
fly_: brb
zhasha: dileX: something tells me mipmaps are to blame
dileX: hmm, I downgraded my kernel to 32-rc3 and an elder mesa-master commit. seems to be f-up, too. (I hope I was not on swrast and gave wrong info. this sometimes happen, when I rebuild one of the involved sw-components, like ddx, xorg-server, etc. I noticed and told MostAwesomeDude already).
spstarr: pulls drm-next
spstarr: nothing too interesting in drm-next
spstarr: looks at mesa git
rhodan: My glxgears does this blue-flashy thing when on gallium3d. Is this normal?
spstarr: alls quiet
rhodan: somebody ban simplexe
rhodan: No ops in here?
rhodan: btw. how do i use the xorg state tracker?
spstarr: that DDX isn't ready yet?
rhodan: Do you get red window borders with Chromium, too?
octe: "drmRadeonCmdBuffer: -13. Kernel failed to parse or rejected command stream. See dmesg for more info" anyone know what that means
octe: doesn't seem to be any relevant info in dmesg..
octe: http://pastebin.ca/1624078
octe: dmesg stuff..
octe: could the mtrr-stuff be related?
rillian: WhiteRabbit56: thanks. I found the same blog post. Next step is to build the current ati driver and see if any of the power options help
WhiteRabbit56: rillian:right now in devel, what i want to work most is power man.... only reason that i still run fglrx is because it gives me about an hour longer battery life on my laptop
WhiteRabbit56: so as soon as we get good power management i can say goodbye to fglrx
rillian: have you tried agd5f's new options?
rillian: (my issue is the fan in a desktop, but yeah, same thing)
mjt: have a fanless desktop running radeon driver for over a year now...
rillian: in theory I bought the card to play with opencl, which I guess requires the fglrx driver for the forseeable future anyway, but in the meantime I prefer a free driver. the 2d support works fine)
WhiteRabbit56: not yet... I've been working all day and just found those last nite... and i have to go to work again in about an hour so I'm not gonna mess with just yet
WhiteRabbit56: mjt: rillians fan is running 100% using the radeon driver... right rillian?
rillian: I don't know if it's 100% but it's certainly the loudest thing in the machine
WhiteRabbit56: rillian what card do you have?
rillian: HD 4850 from asus
WhiteRabbit56: hmm... have you tried to install fglrx and see if there is any difference?
rillian: mjt: and yes, in retrospect probably have gotten a fanless card. however, my previous nvidia was nicely quiet unless an opengl app was running.
rillian: WhiteRabbit56: I haven't no
mjt: my "card" is built-in amd780g graphics ;)
rillian: there you go
mjt: i tried to find a discrete card which's fanless AND with good cooling. Failed.
rillian: installing a bigger heat sink also also an option, but it's hard to tune that without temp sensor feedback
WhiteRabbit56: rillian:You should try that... if it supports your current kernel...
rillian: hence the current quest
rillian: WhiteRabbit56: yes, and I could tune a replacement heatsink that way. Just want to try the free driver first
rillian: otherwise I might as well switch back to the nvidia card
WhiteRabbit56: true
mjt: by "good cooling" i mean, for example, this -- http://www.nix.ru/include/view-photo.html?good_id=48730&pid=2258
mjt: note how the heatsink will be cooled by normal aifrlow
mjt: it's quite effective
BeteNoire: grate
mjt: hello BlackBete
BeteNoire: oh, high :)
mmark: hi !
mmark: well. i have just managed to enable xv on mu radeon 4770
mmark: with 2d acceleration and xv
mmark: this is more then 80% my needs :)
mmark: gret work, folks - especialyy for such a new card like rad 4770
mmark: hope you will manage to make full hardware rendering work on 4770 soon :)
mmark: anyway - GREAT WORK - THX for all radeon driver programers ...
rhodan: Is there any power management with KMS and DRI2?
rhodan: DynamicClocks seems to be ignored.
kdekorte: rhodan, nope
rhodan: :(
rhodan: My battery life sucks.
kdekorte: mmark, 3d is possible on the card... read the wiki in the subject
kdekorte: rhodan, I know it is being discussed and there is work started, but nothing real yet
rhodan: grabs himself an alcohilised coffee.
rhodan: *holised
tlp: lol
tlp: sure you didn't already grab one? :p
rhodan: tlp: I'm pretty sure that was the first time in my life i typed that word. Give me a break ;)
Ghworg: The unison gtk gui causes X to suck up 90% of my CPU when transferring files, wonder what weird stuff it is doing to cause that
Masca: I have just upgraded the driver and now my X is really messed up (I get a cloned desktop in the upper side of my LVDS and it have funny colours). Do you know if it's Xorg, KMS, or the drvier?
rhodan: Masca: cosmic radiation
Masca: =(
rhodan: np: New Order - Temptation
kdekorte: Masca, what version of xserver do you have, should be near the top of /var/log/Xorg.0.log
Masca: 1.7.0.901
Masca: kdekorte: ^
kdekorte: haven;t see that version... was expecting 1.6.0.9xx or something like that
kdekorte: as the odd colors may have been fixed with 1.7 of the server with the recent ati driver
Masca: it's on testing in archlinux.
Masca: Well.I guess actually recent ati driver broke the colors.
rhodan: 1.7.99.1 here :=
kdekorte: there was a conversation about this in that recent at broke the 1.6 colors but they were supposed to be fixed with 1.7 server. So you may want to open a bug on it
Masca: kdekorte: The problem here is that I'm not really sure where is the problem => dont' know to wich project open the bug.
kdekorte: start with the driver
kdekorte: I know I don't have color issues and I'm in KMS, DDX (driver from pretty recent git) and xerver 1.7
Masca: I'm trying to downgrade packages.
EruditeHermit: tormod, hi
tormod: hi
EruditeHermit: with your packages KMS does not work on kernel 2.6.32-rc4 or newer
EruditeHermit: actually KMS works, but it defaults to software rasterizer for 3D
mjt: your packages?
EruditeHermit: xorg-edgers
mjt: i'm running 2.6.32-rc4 here right now
mjt: works quite fast
EruditeHermit: mjt, ubuntu with xorg-edgers?
mjt: i don't know what's xorg-edgers and don't use ubuntu
tormod: what chipsets are you on?
EruditeHermit: I am using rv350 Mobility M10
EruditeHermit: M10/11
mjt: r600-based thing here
EruditeHermit: tormod, did you test it with 2.6.32 at all?
EruditeHermit: the kernel is a ubuntu kernel ppa kernel btw
tormod: the mainline kernel ppa I guess?
tormod: no, I haven't tested it
EruditeHermit: yeah mainline kernel ppa
tormod: so it works with 2.6.31 but not .32?
EruditeHermit: yep
EruditeHermit: well I have big problems with kms on 2.6.31 but its not packaging related
tormod: you think this is packaging related? :)
EruditeHermit: well
EruditeHermit: remember last time it wasn't working with either
EruditeHermit: did you patch some check out
EruditeHermit: something in sysfs was missing
EruditeHermit: or somesuch problem
EruditeHermit: tormod, do you remember this?
tormod: someone (you?) said something about /sys ... configD32 or something?
tormod: I never heard about it anywhere else and could not find it in any commit
EruditeHermit: oh
EruditeHermit: well
tormod: maybe something they added to Fedora to make it work :)
EruditeHermit: If you can grab the latest mainline kernel and test
EruditeHermit: that would be amazing
tormod: sorry not now, but I can try on RV410 later
EruditeHermit: or I could reboot into it and give you my log
EruditeHermit: tormod, in 2.6.31 kms do you notice corruption when opening png files?
tormod: to be honest I haven't tried kms for a while, but did not see corruption with it back then
EruditeHermit: ah ok
EruditeHermit: let me reboot and try both kernels again
EruditeHermit: and hope that it just works
EruditeHermit: =p
EruditeHermit: brb
EruditeHermit: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed because of a version mismatch.
EruditeHermit: [dri] radeon kernel module version is 2.0.0 but version 1.17.0 or newer is needed.
soreau: EruditeHermit: Doesnt it say what to do on the wiki in that case?
soreau: Also, have you tried if the png corruption happens with a more recent kernel or drm-next?
EruditeHermit: soreau, so its the same setup with kernel 2.6.32-rc5 and I get the above error
EruditeHermit: the libraries were compiled with --enable-radeon-experimental-api as stated in the wiki
EruditeHermit: we know this to be true because KMS works with 2.6.31
EruditeHermit: on the same setup
mjt: how do you know about kms?
mjt: you have to enable it on 2.6.31
EruditeHermit: mjt, what do you mean?
mjt: with kernel module parameter
EruditeHermit: yes
EruditeHermit: I did
EruditeHermit: I boot with radeon.modeset=1
EruditeHermit: and to verify
EruditeHermit: glxinfo states DRI2
EruditeHermit: which is only available if you have KMS
mjt: when i enable it here on 2.6.31 and load older drivers i get the same error about versions
EruditeHermit: my drivers are all within the last few days of git head
mjt: by "older' i mean without kms
EruditeHermit: libdrm-radeon1 2.4.15~git-0ubuntu0tormod
mjt: 2.4.14 here
EruditeHermit: soreau, I can't test with a newer kernel because i don't get DRI with a newer kernel
mjt: OpenGL renderer string: Mesa DRI R600 (RS780 9610) 20090101 x86/MMX+/3DNow!+/SSE2 TCL DRI2
EruditeHermit: soreau, no corruption with the software rasterizer for PNGs on 2.6.32-rc5
EruditeHermit: but I dont think thats a fair test
soreau: Then fix your dri ;)
EruditeHermit: well thats why I posted the error above
rhodan: What does --enable-radeon-experimental-api do? Which wiki?
soreau: rhodan: See the topic
mjt: see /topic
rhodan: I see.
rhodan: But I have KMS.
rhodan: Does that mean it's already done for me?
rhodan: Ah, it's set in the ebuild.
rhodan: Gentoo does everything for me \o/
rhodan: (II) RADEON(0): KMS Color Tiling: disabled ← what does that mean?
EruditeHermit: soreau, so any other suggestions>?
soreau: EruditeHermit: Double check everything, including your kernel config
rhodan: Is there any way to speed up 2D with KMS and DRI2?
rhodan: Still seems to slow.
soreau: I havent tested 32_rc5 yet but rc1-4 has worked fine here so far
soreau: as well as drm-next
mjt: 2d is indeed slower with kms than without ;)
EruditeHermit: soreau, its all compiled by Ubuntu
EruditeHermit: and their source packages say it should work
rhodan: I'd like to have a Windows®-like 2D experience. Without glitches and hangs :(
soreau: Then go bug whoever compiled the packages
Lenin_Cat: when is 3D support in R6xx gonig to be good enough for gaming usage?
EruditeHermit: soreau, I did, and he said it should work
soreau: EruditeHermit: Perhaps you have package manager issues/some old files still lingering
rhodan: Lenin_Cat: On 21. May 2010.
Lenin_Cat: you have that set?
rhodan: Lenin_Cat: I say it is.
rhodan: If it isn't, you won't mind. If it is, you'll be admiring me for my prophetic talent.
Lenin_Cat: rhodan, say what is
rhodan: Lenin_Cat: the date. It's correct.
Lenin_Cat: right, brb rebooting.
rhodan: http://xkcd.net/628/
rhodan: If he gets usable 3D that day, I'll be his hero ;)
EruditeHermit: soreau, thats not the case
EruditeHermit: it works with 2.6.31
soreau: EruditeHermit: Well I dont know what else to tell you other than build everything yourself
tormod: EruditeHermit, mostly canned responses here I see :)
EruditeHermit: tormod, huh?
tormod: EruditeHermit, but for what I know, maybe something changes if it's built against 2.6.32 kernel headers
EruditeHermit: ah
EruditeHermit: I wonder if you can build it to work for multiple kernel versions
mjt: works here on 2.6.32-rc4 and 2.6.30
tormod: EruditeHermit, you can just download the source packages from the PPA, and build them
tormod: did you pastebin your logs?
EruditeHermit: tormod, I pasted the part that gives the error
EruditeHermit: let me pastebin it
EruditeHermit: one sec
tormod: and dmesg as well
EruditeHermit: tormod, http://pastebin.com/m25aa6169
EruditeHermit: tormod, http://pastebin.com/m33d9532f
tormod: EruditeHermit, (II) [KMS] drm report modesetting isn't supported.
tormod: EruditeHermit, but I see it KMS is enabled in dmesg...
biotube: EruditeHermit: try loading fbcon before starting X
tormod: so the server then looks for an non-KMS drm but finds the KMS drm and complains about the version IIUC
EruditeHermit: ok
EruditeHermit: so let me log out
EruditeHermit: and stop X
EruditeHermit: and then load fbcon and log back in
mjt: wonders how many times EruditeHermit need to see the same thing told by different people...
mjt: damn
EruditeHermit: biotube, it worke
EruditeHermit: d
EruditeHermit: tormod, so it seems fbcon is not loaded early enough
biotube: EruditeHermit: I edited my dm's init script to load it
tormod: EruditeHermit, ok, put it in /etc/initramfs-tools/modules
tormod: dm?
biotube: display manager
EruditeHermit: tormod, do I need to rerun mkinitramfs
EruditeHermit: or something?
tormod: or maybe /etc/modules is early enough
tormod: if you put it in the first, yes
tormod: the second, no
tormod: you run update-initramfs -u
EruditeHermit: ok
EruditeHermit: let me reboot
EruditeHermit: and try
spstarr: ugh
spstarr: [20962.373782] [drm] Resetting GPU
spstarr: [21452.386277] [drm:radeon_cs_parser_init] *ERROR* cs IB too big: 16390
spstarr: [21452.386281] [drm:r600_cs_legacy] *ERROR* Failed to initialize parser !
spstarr: this is still bad
spstarr: badness
hangfire: Is there someone here using xf86 radeon drivers on a rv770 (or similar?). What is the 2d performance like? It's horribly slow for me - any advice / can I see your logs/config?
hangfire: (I have drm etc. loaded + working)
EruditeHermit: biotube, hmm fbcon in initramfs doesn't do it
biotube: EruditeHermit: I never said it would - I said i doctor my init scripts
biotube: s/doctor/doctored/
EruditeHermit: what did you put where?
biotube: modprobe radeon modeset=1 && modprobe fbcon && sleep 3
spstarr: man that bug is pissing me off
spstarr: [21452.386277] [drm:radeon_cs_parser_init] *ERROR* cs IB too big: 16390
spstarr: logs a bug now
spstarr: subsequently, closes another
spstarr: SHIT
spstarr: thats hot
spstarr: wtf
spstarr: the GPU is fucking H O T
spstarr: wtf
spstarr: geez
spstarr: bad
spstarr: the GPU will not overheat on its own?
spstarr: or the card will downclock if its too hot?
MostAwesomeDude: Well, joy. Looks like Newegg's been dragging their asses.
MostAwesomeDude: Most of my new box won't get here until Tuesday.
MostAwesomeDude: At least I've already got the r600.
Dr_Jakob: MostAwesomeDude: is that r600g I hear getting closer?
MostAwesomeDude: Dr_Jakob: Just a bit. I've gotten feedback from everybody except Keith, Jose, and Brian, on the blitter changes.
MostAwesomeDude: So I think I'm going to write a new patch that lets surface_copy/surface_fill be NULL, and then commit to master and wait for bitching and moaning in the morning. :3
MostAwesomeDude: And then I might start r600_screen.
MostAwesomeDude: I'd really like to track down a few r300g things first before I get too distracted...
Dr_Jakob: Ok
MostAwesomeDude: Moving to the hashed shaders kind of broke gears a bit.
spstarr: tries VBOs
spstarr: ERROR: mapBuffer: glMapBuffer returned NULL (no vertex data)
spstarr: hmm
spstarr: maybe not :)
spstarr: VBOs is broken still
spstarr: damin GPU is hot
spstarr: tell me its not supposed to get this hot