unic0rn: ok... i guess i've nailed it. haven't tested with 2.6.32 yet, but i guess it'll be exactly the same. i have 2.6.33 running now
unic0rn: r200 + KMS + DRI2 working fine. with one small exception. 1024x768
unic0rn: 800x600 works
unic0rn: 1280x1024 works
unic0rn: 1024x768 - no picture.
unic0rn: vrefresh is correct
unic0rn: it just... isn't working.
unic0rn: looks like kms has some problem setting that mode, in both 2.6.33 and 2.6.32
unic0rn: ideas?
Tommeh: Crude, but have you tried a recent testing kernel?
unic0rn: i guess i'll try with different refresh rate, like 75. maybe it'll help... but 85hz vrefresh was the reason why i was using it. i have old crt monitor
unic0rn: well. i didn't.
unic0rn: but 2.6.33 in sidux is rather up to date i think..
Tommeh: Not when radeon dev is concerned :)
Tommeh: I use a build of 2.6.34-rc1 at the minute.
airlied: not sure what might have changed in there anywats
Tommeh: And I still have some strange issues with KMS + compositing, so it's worthwhile.
airlied: trying a lower refresh might help
Tommeh: airlied: did Linus accept your pull request for -rc3? (or is it still -rc2?)
unic0rn: well, i'll play with refresh rate for now.
airlied: not yet, though it had a problem i nit
unic0rn: 1280x1024 60hz kills my eyes.
Tommeh: airlied: not good :)
unic0rn: heh. guess what.
unic0rn: 75hz works
unic0rn: facepalms
unic0rn: as for compositing artifacts with KMS, the only thing i saw is that window textures in compiz seem to get heavily screwed after using xv video overlay. considering the fact it's actually textured xv, well. it should work.
unic0rn: but gl_nosw in mplayer works just fine
unic0rn: same effect with task-switching using cairo-dock btw. some of it's rendering problems seem to get fixed by DRI2 on radeon, but then it screwes up other things :)
Telek: Wow, nice job on the R600/R700 driver. I can run the darkplaces version of quake1 with a number of the lighting effects at ~60fps @ 1600x1200 on a HD4770. Only appears to have some artifacting regarding some lighting (gives a blue outline sometimes.)
Nightwulf|work: hi all
Tecan: wow those radeon drivers run quake well
Tecan: 1 2 and 3
Tecan: plays eduke32 (dukenukem3d) really good too
hifi: they do
Telek: Tecan: Yep.
adamk: Has anyone gotten the r600 driver to provide 3D acceleration under vmware? vmware hangs here when I start windows VM if I have 3D acceleration enabled in the guest configuration. The vmware log file shows "VMGL Panic: Caught signal 6 in glCompileShader"
glisse: adamk: you are using the vmwgfx for the guest on a host which has r600 driver ?
adamk: glisse: I haven't even gotten that far. vmware just hangs within a second or two of hitting "power on"
adamk: It never even posts. To be honest, I really don't care if I get 3D acceleration going on the guest (though it would be nice) but it seems like either a driver bug or vmware bug that's causing the ahng.
MrCooper: more likely the former, works on intel
adamk: That's kind of what I was thinking.
adamk: Guess I should file a bug report then.
MrCooper: it relies pretty heavily on GLSL, guess the r600 driver GLSL support isn't quite there yet
adamk: Yeah, it'd be nice if it failed nicely, though :-) Rather than forcing you to kill vmware-vmx. Which is why I wasn't sure if this really was a driver bug or vmware bug.
adamk: Or, I guess most likely, one of each.
glisse: w
AstralStorm: hmm, I'm doing something wrong now with opengl
AstralStorm: driver says this:
AstralStorm: [826240.952235] radeon 0000:01:00.0: z/stencil with no depth buffer
AstralStorm: [826240.952245] radeon 0000:01:00.0: r600_packet3_check:1055 invalid cmd stream 1165
AstralStorm: I did call glEnable(GL_DEPTH_TEST);
AstralStorm: do I need something else?
AstralStorm: oh, wait
glisse: AstralStorm: maybe you choosed a visual which hasn't a depth buffer
AstralStorm: probably yes ;p
AstralStorm: although a crash is bad
AstralStorm: it should drop me a GL_INVALID_OPERATION
glisse: yeah driver should ignore z stuff and report gl_invalid
AstralStorm: oh btw, is there a generic OpenGL help channel around?
AstralStorm: found it :>
MrCooper: adamk: as you can see vmware is trying to recover from a crash in a GL call
unic0rn: what's the current state of AccelDFS and page flipping for r200@kms? should i have those enabled at all?
adamk: MrCooper: Gotcha... So driver bug.
MrCooper: unic0rn: AccelDFS is enabled by default with KMS, page flipping not supported yet
Ke: btw doesn't dri2 mean that one can implement opencl (without openGL bindings obv) totally independent from X
Ke: perhaps even without any display and also shutdown rop
Ke: and furthermore use assembler to compile symbolic code given by ati stream sdk
MrCooper: DRI2 mostly refers to an X11 protocol extension, so not really :)
Ke: well corresponding drm driver
suokko: unic0rn: AccelDFS works if you have kernel 2.6.33. Page flip is not implemented yet
MrCooper: KMS should allow that, though there are still a few loose ends
unic0rn: i guess mesa has something like accelerated framebuffer, but that's in testing stage basically. kinda limited glx api or something around that
unic0rn: suokko: yeah, i have 2.6.33 currently.
suokko: Then DFS should work
unic0rn: well. i have it enabled in xorg.conf all the time. just wondering if it's supposed to cause any problems with agp card.
suokko: If it doesn't it might be some agp bug which could be workedaround with lower agp speed
MrCooper: it should be more reliable with KMS though some AGP bridges may still screw up
unic0rn: well. i'm getting rendering artifacts, screwed up (overbright, lack of details) windows to be precise, with compiz, but only in 2 cases.
unic0rn: case one: video on xv
unic0rn: case two: glx usage (glxgears are not causing problems)
unic0rn: for example, cairo dock causes that when i'm switching apps with it
suokko: MrCooper: btw, Even sid is leaking if I enable compiz and open/close gl apps
unic0rn: launching some fullscreen opengl apps (via wine) also messes things up
Ke: MrCooper: how about practice, does that sound like crazy talk
Ke: aren't assemblers kind of trivial, if you have the proper compiler to produce the symbolic code
BioTube: Ke: things get double compiled - once to the symbolic code and once to to machince code
Ke: or is there any hope of oss opencl
glisse: Ke: compiler is not trivial
MrCooper: Ke: something like that will probably be possible at some point, faster if you help :)
Ke: glisse: but amd has a closed source compiler
glisse: yes they do
glisse: no we won't use it
BioTube: plus there's the fact that the protocol wasn't designed with OpenCL in mind
glisse: opencl won't use dri2
Ke: MrCooper: any entry level assignments?
MrCooper: err, I was referring to running GPGPU workloads independently from X
glisse: at least not if you dont use cl<->gl bridges
BioTube: glisse: I thought something new would be needed
glisse: BioTube: works is in drm side
Ke: yes, my only interest was gpgpu, not graphics
glisse: to allow drm client other than root
twnqx: glisse: you read that my "KMS crashes" were actually just X crashing, and the driver not restoring after loss of client? :X
MrCooper: Ke: airlied recently wrote up a nice summary, either on the dri-devel list or his blog
twnqx: my libpixman caused BUGs
schmankerl: hello, is there support for opengl 2.1 in radeon-driver? opengl 2.0 seems to work...
suokko: schmankerl: not yet
Thunderbird: version numbers are meaningless .. look for specific extensions your app needs
adamk: And you should really specify the Mesa driver you are using.
adamk: radeon typically refers to the Xorg DDX.
schmankerl: OpenGL version string: 2.0 Mesa 7.9-devel
suokko: schmankerl: yes. That is the newest code created so far
schmankerl: GL_ARB_map_buffer_range would be nice
suokko: Unigine?
schmankerl: among others
schmankerl: i read thats specified in 2.1
schmankerl: extension
suokko: schmankerl: I suspect unigine&others requiring it would require something more from driver too
schmankerl: ok
MrCooper: the i965 driver supports 2.1 but not everything the recent Unigine demos want
schmankerl: should radeon-driver support most wanted extensions or include every opengl extension in a milestone (2.0, 2,1, 3.0...)?
Ke: i965 is way awesome, I can even play GR with full functionality
GNU\colossus: GR?
Ke: Ghost recon (on wine)
Thunderbird: though we don't recommend intel hardware for Wine though (as a wine developer); for too much games (even older ones) there are big performance issues / bugs
Ke: well the only open source one that works
Thunderbird: it is likely the best open source driver at this point; the main issue for wine is that its shader / fbo performance is not good at all (most games even old d3d8 games really need this) and other areas are quite slow as well (GPU readbacks at ~50MB/s or so; various directx games need those regulary)
Ke: Thunderbird: btw raven shield refuses to start while graphics card is "unsupported", is it likely that it greps vendor string oslt?
Thunderbird: that can easily be the case
Ke: any way to fake it?
Ke: well offtopic
Thunderbird: directx reports the pci ids and gpu name and so on to apps
hifi: oh, Thunderbird hangs at #radeon too
Thunderbird: we have our own logic for setting that information but you can override it using some registry keys
Thunderbird: see, http://wiki.winehq.org/UsefulRegistryKeys but if you can code better look at wine/dlls/wined3d/directx.c and also figure out what the game really wants
Thunderbird: various games ship with a database in text files
Ke: well it's known that it has no database of supported cards
Ke: while it supports cards way newer than the game
Thunderbird: it might be detecting gpu capabilities
Ke: unless it has been patched
Thunderbird: capabilities are a bit like opengl extensions; most of our capabilities are based on those
Ke: yup
Thunderbird: it could be that your opengl drivers don't offer what this game wants
Thunderbird: which capability it doesn't like can be hard to find out
suokko: Thunderbird: Doesn't wine have trace for capability requests?
Thunderbird: sure we have traces but it is not useful for this
hifi: Ke: oh, I know that
Thunderbird: basically d3d provides apps with structures which contain the capabilities
hifi: Ke: it's because the propietary texture compression, what was it called
Thunderbird: structures filled with variables which contain bit masks
hifi: s3tc?
Thunderbird: dxtc/s3tc
hifi: raven shield refuses to start if the driver doesn't announce support for that
Thunderbird: enabling support using driconf might do the trick
Thunderbird: unless the game asks for encoding but not sure if much of them do
hifi: http://hifi.iki.fi/ravenshield-r7xx-wine.jpg
hifi: looks a bit like matrix ;)
GNU\colossus: wow, nice. matrix-style
hifi: I don't know if thats related to s3tc or some other issue with the radeon driver and wine
Ke: hifi: heh, there just was a story on phoronix on that
Thunderbird: a crappy story like most of the 'news' there
adamk: Alright, so since there are both radeon and wine developers here :-) Why do I have to LD_PRELOAD libGL.so.1 on FreeBSD for wine to get direct rendering?
Thunderbird: we dynamically load libGL for most purposes (only for opengl32 we directly link against it for GL 1.1 calls)
Thunderbird: I guess wine isn't able to locate your libGL.so.1
adamk: If I don't, I get r600_dri.so: Undefined symbol "_glapi_Dispatch" and it reverts to indirect rendering.
Ke: Thunderbird: yup, but at least they track news in a somewhat diligent manner
adamk: I don't know why it wouldn't fine libGL.so.1.. It's in /usr/local/lib and ldd shows it.
adamk: Well at least for other apps :-)
Thunderbird: a lot of the news isn't news and makes developers quite angry because it is in progress stuff
Thunderbird: and it tends to give people too much hope
Thunderbird: it makes a lot of people believe that lets say radeon drivers are nearly ready for serious opengl usage
Thunderbird: they tell too much fairy tales
schmankerl: i think opengl becomes more and more important
Thunderbird: if the windows opengl drivers from companies outside nvidia/amd (I'm looking at you Intel and others) were good, it would help
twnqx: do you claim that intel has relevance as soon as 3d is involved?
Thunderbird: they have a big market share; if lets say windows opengl drivers were good webgl -> opengl would work fine but now google and others are looking at webgl -> d3d wrappers
Thunderbird: and if people had good drivers, perhaps more apps would ship again with opengl
Thunderbird: it seems the windows drivers support more stuff now than I expected but the issue is that most users still have old drivers
Jonimus: Thunderbird: well and some XP users could still be on cards with no OpenGL support at all, but those cards would also stink in the D3D dept as well
hifi: hm, HL2 is not working, I get "WRITE DOMAIN RELOC FAILURE 0x1d 6 4" after the level loading screen
Ke: Thunderbird: still there is hope for opensource, if you only want to play games 10+ years old
Ke: ;o)
GNU\colossus: we're lucky that "recent" games have a tendency to SUCK, aren't we? ;)
Ke: indeed
hifi: valve still does great games, though
adamk: I played HL2 all the way through with the r300 driver at some point a while back.
hifi: this is frustrating, what the heck causes this
fiyawerx: Hi guys, maybe if this isn't the right place you can pointme where to go. I have an ATI 4890, Fedora 12, using radeon driver + mesa-dri-drivers-experimental. My problem is that I'm trying to get my HDMI output to act as primary when two displays are attached, instead of the DVI. (There are always two on this system), but it appears to make my smaller montior connected via dvi the primary.
fiyawerx: I've seen mention of "aticonfig --swap-monitor" for fglrx, but I'm not sure if there is something similar I can do for radeon
fiyawerx: or if that even would fix my problem
hifi: xrandr --disable the other output
Jonimus: fiyawerx: I believe xrandr assumes the rightmost monitor is the main.
suokko: fiyawerx: xrandr --primary
adamk: Yes, xrandr --output HDMI-0 --primary
adamk: Our whatever name you have for the HDMI port.
hifi: though, it doesn't matter if your software doesn't support xinerama
adamk: Is there software out there that doesn't support xinerama? :-)
fiyawerx: alright, I'll give that a try, thanks
hifi: like, everything?
hifi: starting from any game
hifi: I tried to use xinerama but all my software opened on the secondary window, always
hifi: even if I made my primary --primary
hifi: zaphod, instant success \o/
hifi: though, I need to use two cards (integrated intel and radeon) to work with zaphod
Jonimus: lol
fiyawerx: hm, how can I test my panels are still on the smaller screen.. but if I restart X then that command won't still be in effect, right?
adamk: You'll have to add that command to your sessin.
adamk: Or simply move the panels to the other monitor :-)
adamk: hifi:Once I used --primary, all my games started working fine.
schmankerl: .config/monitors.xml?
hifi: funny, maybe my window manager at the time just sucked (openbox)
hifi: need to retest with xfce
adamk: Or maybe it depends on the game or the library. For example, some games will use SDL, so if SDL isn't built with xinerama support, for some reason, that could cause problems.
adamk: I think I ran into someone on this channel who had WITHOUT_XINERAMA set in gentoo because they were using xrandr.
adamk: They couldn't figure out why everything was opening in the middle.
adamk: Of course I'm sure you're not doing that :-)
yoshi314: hi there, anybody around?
adamk: Maybe.
yoshi314: if kms is reporting that EDID invalid, but still works - should i be worried ?
yoshi314: *is invalid
Jonimus: yoshi314: it could be your monitor has a corrupt EDID table
yoshi314: http://pastebin.com/d6tYKwhF - third edid attempt gives different result
adamk: Generally if everything is working, I'd say be happy :-)
yoshi314: which is even more strange
yoshi314: well kms is stuck for ~2seconds and then sets a reasonable resolution
yoshi314: but this problem is worrying me a bit
yoshi314: i'll have to look around for some edid parser info to figure out what exactly is wrong there
hifi: no, even with xfce --primary with xrandr has no effect
hifi: all software I open, like a terminal window, opens in the secondray display
evil_core: http://carme.pld-linux.org/~evil/varia/dri_report/upgrade-winecrash.log
adamk: Well in most window managers, don't windows apps typically open in whatever screen the mouse is in?
evil_core: after drivers upgrade most of wine games stopped working
adamk: Anyone other than me still getting that cursor corruption on r600 hardware? THe pointer turns into a vertical line of tiny triangles?
adamk: It's only solved with a reboot.
adamk: It used to be much more common, and then I remember hearing that it was fixed.. And, indeed, I didn't see it for a long time.
hifi: hm, using dual head with a 1080p display seems a bit sluggish
hifi: drawing firefox when changing desktops takes a second
thansen: agd5f: have another moment to look at my displayport woes? airlied helped me a bit last night..
thansen: Jonimus: updated my dd-wrt and it's all working now :)
roysjosh: yoshi314, http://wiki.mandriva.com/en/Tools/monitor-edid (fedora12 package name: monitor-edid)
roysjosh: yoshi314, that doesn't look like good EDID at all, in your dmesg
yoshi314: roysjosh: i found something called read-edit packaged for gentoo, i'll try that first
agd5f: thansen: can you try the patch on the bug?
thansen: agd5f: this should be applied to the git drm kernel?
agd5f: thansen: yes
agd5f: and of course try with kms
thansen: right
thansen: agd5f: is it ok to be running the ddx with the other patch still applied?
agd5f: thansen: yeah
zhasha: holy crap guys, there's an auto parts store in this very chatroom!
zhasha: http://www.thansen.dk/
thansen: whoa! those are some sweet scooters
[Enrico]: mhm the PM algorithm seems to be improved (i just compiled drm-radeon-testing), it is a lot less choppy now.... i have to try with a video
zhasha: thansen: never buy from thansen, their stuff falls apart when it leaves the store
thansen: looks like even before you leave the store it falls apart
[Enrico]: cool it is smooth now! (composite it is not 100% smooth but better then before) good work, keep improving it. thanks :)
[Enrico]: bbl
thansen: agd5f: ok, done building with just the attached patch..brb
thansen: agd5f: I think we're making progress.. here's dmesg http://paste.pocoo.org/show/195286/
thansen: so I still don't get a picture, but the screen doesn't go into 'powersave' mode either though (which it was before)
agd5f: thansen: phase 2 of link training still fails the same way :/
thansen: agd5f: :( on the bright side, at least the monitor thinks *something* is getting sent to it
agd5f: thansen: can you try changing the mode with that patch applied? using xrandr?
thansen: agd5f: by mode I'm guessing you're talking about the resolution?
agd5f: yup
thansen: agd5f: so, by default it was setup to mirror (using the gnome-display-properties)
thansen: I tried to flip it using that and didn't get a picture
thansen: non-mirrored higher and lower resolutions that the LVDS
agd5f: thansen: don't worry about mirror vs dualhead, just try another mode like 1024x768
agd5f: xrandr --output DP-0 --mode 1024x768
agd5f: or whatever the output is called in your case
thansen: agd5f: yeah, I don't get anything :(
suokko: "WRITE DOMAIN RELOC FAILURE 0xe 2 4" :/
thansen: just fyi, I did an install of ubuntu 10.04 last night and forced the binary drivers to install and got the screen working at its full res (for the first time since owning it) so I know the hardware combo works
GNU\colossus: should I be getting a radeon x300 or a x1550? or doesn't it really matter?
agd5f: GNU\colossus: x1550 is more capable
GNU\colossus: agd5f: in any significant way?
spreeuw: get the fastest with a passive card
spreeuw: cooling
spreeuw: 4670 or so
agd5f: GNU\colossus: r5xx vs r3xx. more advanced shaders, larger max texture size, etc.
GNU\colossus: the card will be used for a compositing desktop and casual video viewing only
[Enrico]: spreeuw: afaik 4670 has the fan so it is not passive
GNU\colossus: agd5f: thanks, that settles it :)
spreeuw: it doesnt
spreeuw: not mine anyway
spreeuw: club3d pcie
spreeuw: 1GB ram
GNU\colossus: spreeuw: I'm picking up a video card on ebay for my grandfather's machine, that's why I won't be getting a high end model
spreeuw: only 256 of which is used with floss driver
GNU\colossus: I have a passive HD3650, it's sufficient for what I'm doing
spreeuw: 4670 isnt high end
GNU\colossus: depends on your point of view ;)
spreeuw: if its just for some desktopping a radeon will do fine
spreeuw: 9250
spreeuw: or if you can get one 2nd hand a 8500
GNU\colossus: it's gotta be a PCI-E board though
spreeuw: I used to have an x700 works ok
spreeuw: r300 driver
spreeuw: so with s3tc/dxtn support
spreeuw: not sure how ati behaves with all that ubuntu 3d desktop jizz
GNU\colossus: it works very fine with my (Gentoo) 3D desktop jizz ;)
thansen: agd5f: will it do me any good to apply the patch from the link you supplied?
agd5f: thansen: you can try it, but I doubt it will change anything
thansen: applied that patch..brb
thansen: agd5f: your suspicions are correct :(
thansen: agd5f: any other ideas or avenues for me to persue?
agd5f: not at the moment
thansen: agd5f: how/when is the best way for me to follow up with being annoying :)
Telek: Can anyone here tell me if they're heard of the possibility of a Radeon HD5450 PCI Card?
glisse: Telek: arg what good it would be ?
chithead: club3d sells radeon 4350 in pci and pcie x1 variants, but it took them quite some time after launch
Telek: glisse: Well for me, I've still got some PCI based systems that I could get low end 3d acceleration on. And more importantly having just read up on the OGL4.0 API specs, it has the potential to allow almost ALL scene rendering to take place on the card, so it should be pretty fast even with a low end system as long as all the 'big' data gets cached to the card?
glisse: nowadays cpu still mess a lot with the rendering data
glisse: vertex,index,....
Thunderbird: I doubt there will be PCI cards and if they will, they will be expensive
glisse: and this need bandwidth
Thunderbird: really upgrade your system
Telek: I wasn't asking for opinions, I was asking for availability.
glisse: from theoritical pov this is doable
glisse: then it's up to manufacturer to decide if they think there is a market or not for that
otaylor: Telek: if you come into a channel and ask off-topic questions, you have to expect off-topic responses as well...
glisse: i surely hope they won't do AGP version
Thunderbird: there were AGP 3xxx cards and perhaps 4xxx as well
glisse: yeah 4xxx agp exist
glisse: and it hurts my feeling
spreeuw: hehe
glisse: hd5xxx agp would break my heart
Telek: Lack of AGP 1x/2x support is what breaks my heart! :D
otaylor: glisse: question - could the problems you found with your gpu reset code have been causing system hard locks?
otaylor: glisse: my r3xx system has been much more crashy over the last few months
glisse: otaylor: what you mean ? that the new reset code lead to hard lockup ?
glisse: otaylor: agp ?
otaylor: glisse: not sure exactly what I mean ... but it sounded like you found some problems there recently, so wondering if it could have been * before ~2-3 months ago, working gpu reset code saving my system * after 2-3 months ago, system tries to gpu reset, hangs solid
otaylor: glisse: agp, yes
glisse: i think we will need the bind scratch page on unbind for agp as i did for pci gart, i was hoping MrCooper to submit such patch :)
glisse: for some reason we unbind agp page while gpu is still messing with them
Telek: otaylor: Check your power connector? I had an AiW with the floppy style connector and it turned out the solder joints for the pins had started breaking loose from having to be disconnected to work on the system.
glisse: if all gpu lockup were hw related it would make my life easier :)
otaylor: glisse: I once saw it print a couple lines at the end a kernel oops a ssh session (but nothing useful) - other than it's just a solid hard lock every 4th or 5th time I restart gnome-shell
glisse: otaylor: laptop ?
otaylor: Telek: laptop. It's very clearly related to what I'm doing at the time on the system, so unlikely to be hardware.
otaylor: glisse: yes, laptop
unimatrix: I assume everybody saw this already? http://www.phoronix.com/scan.php?page=article&item=ubuntu_fglrx_open&num=1
glisse: otaylor: i am trying to finish cs analysis replay tools
glisse: once i am done with them i will get back to analyze gpu lockup
MichaelLong: good to hear :>
glisse: i have been hand analyzing cs stream for too long, it's inefficient work
MichaelLong: has still a locking rv250 after resume
glisse: otaylor: i might also do agp patch tomorrow i got a couple idea i want to test
glisse: unimatrix: yeah we are aware perf pov we suck
glisse: MichaelLong: likely an agp issue
MichaelLong: hmm
glisse: some bridge doesn't like suspend/resume
unimatrix: glisse: but what's the catch? i mean surely it shouldn't be tough to beat that bloated catalyst
glisse: unimatrix: it's a lot tougher than you think
MichaelLong: glisse, hm but why is it working with radeon.modeset=0
glisse: unimatrix: i don't know the size of the closed driver team but i think it btw 10 or 100 time bigger than # people working on opensource
glisse: also we are working on a larger range of hw
glisse: AMD software teams focus on 2-3 generations of hw at most
unimatrix: that explains a lot i guess
MichaelLong: yeah, i can't use any fglrx at all for quite some time
otaylor: unimatrix: and also, what catalyst is bloated with is, uh, stuff to make it go fast
unimatrix: otaylor: and bugs, they also beat you in bugcount
glisse: i don't think they have more bug than us
otaylor: unimatrix: fglrx definitely suffers on Linux for being a cross-platform code base
otaylor: unimatrix: I don't think it's a great choice if what you care about is general desktop operation (though I haven't used it recently), because it's just not a testing focus, and it has a ton of code in it for other things
glisse: last estimation i made for radeon hw was 300 real bugs
otaylor: unimatrix: but I would expect it to be a lot more accurate at implementing OpenGL specifications in general, because it presumably gets lots of conformance testing and its implementations of advanced OpenGL features are a lot more mature
otaylor: glisse: 'already filed' real bugs?
otaylor: glisse: who knows how many bugs exist potentially, but 300 would be a tiny number for any significant code base
glisse: otaylor: yeah btw fdo, redhat, kernel
unimatrix: so what's the trend like? are more people starting to get interest in radeon?
glisse: thought i think there is likely many duplicate in them
glisse: and if we ignore the mesa bugs than there is lot less
evil_core: O3 or Os is quicker for new gcc and Mesa/wine?
rehabdoll: Os just makes the binaries smaller, O3 faster
Wizzup: evil_core: I'd suggest -O2 though
rehabdoll: yup
Wizzup: -O3 is AFAIK unsure optimisations
Wizzup: They can cause (extreme) slowdowns and even different behaviour
rehabdoll: http://funroll-loops.info/
Wizzup: heh
chithead: -O3 is teh fast. except when it's not. -Os often hurts due to alignment
evil_core: O3 is default in PLD distro
evil_core: i windered if switching to Os for mesa and could help
chithead: http://www.linux-mag.com/id/7574/1/ is a comparison of -O3 -O2 and -Os, overall -O2 wins
evil_core: I know that O99 breaks wine and mesa
evil_core: ok, thanx, I will read it
evil_core: http://www.linux-mag.com/id/7574/2/
evil_core: its a joke
Thunderbird: there are reasons why such flags are not used by default anyway; optimizations are risky
evil_core: nvidia binary compilation, and games also mostly closed source
Thunderbird: for wine compile flags are not that important
Thunderbird: most time is spent in application code and in case of gaming in drivers
evil_core: so what about Mesa?
Thunderbird: if it hits software fallbacks then perhaps it might matter (but I don't know the mesa code)
evil_core: but I wonder which Os, O2 or O3 ;)
evil_core: I know that some would be faster :P
Wizzup: -O2
evil_core: Wizzup: thanks
evil_core: O99 shouldnt break Mesa with O2?
Wizzup: wat
Wizzup: evil_core: Please read http://www.gentoo.org/doc/en/gcc-optimization.xml
MostAwesomeDude: Mesa already sets a bunch of flags on its own.
MostAwesomeDude: IIRC the one that matters is -ffast-math, which Mesa uses by default.
evil_core: Wizzup: theres nothing about O99
Wizzup: evil_core: read the "What about -O levels higher than 3?" part
evil_core: its not really true
evil_core: O99 must be used with O2, Os or O3
evil_core: and it breaks wine
evil_core: it cannot load dlls then
Wizzup: wait....what? Are you talking about -O99 as in Optimisations level 99
cockroach: hey, is there any chance of getting a multi-screen setup to work with a 5770 atm or should i just wait?
evil_core: Wizzup: it must be used with O2, O3 or Os
evil_core: check it if you dont believe
Wizzup: -O99 does absolutely nothing more than -O3.
evil_core: it doest
evil_core: otherwise wine would work
evil_core: it doesnt after adding that one flag
evil_core: and it works with O3
Wizzup: :coolface.jpg: -> afk
evil_core: http://archives.postgresql.org/pgsql-patches/2003-10/msg00186.php
spreeuw: Nexuiz has been forked under the name Xonotic
FIReun: oldnews d-:
cockroach: is there any chance of getting a multi-screen setup to work with a 5770 atm or should i just wait? xrandr does report two outputs (DVI-0 and DVI-1) but no matter which one i modify, it's always the same screen...
cockroach: (also, no output on the other one)
evil_core: libGL: radeonCreateScreen: drmMap failed for GART texture area
evil_core: wtf?
Obscene_CNN: whats different between Nexuiz and Xonotic?
agd5f: cockroach: should work ok with kms
cockroach: agd5f: okay, i'm building 2.6.33 as we speak (i assume i need .33 for kms on that card, right?)
agd5f: cockroach: needs 2.6.34rc releases
cockroach: agd5f: oh. thanks.
evil_core: I got laggy q3a with testure compression disabled on UMS
evil_core: it tears and lags
Obscene_CNN: wonders why Nexuiz is far less sensitive to code changes than Torcs
evil_core: libGL: radeonCreateScreen: drmMap failed for GART texture area
evil_core: I got it with many games using glide wrapper
evil_core: what can cause that?
Obscene_CNN: the glide wrapper perhaps ;)
evil_core: but it works with Dethkarz
evil_core: and it worked on nvidia
evil_core: and all that games started working today in non-glide mode
cockroach: agd5f: indeed, seems to work just fine - thanks!
Obscene_CNN: has figured out that Nexuiz Spends far more time waiting on the GPU and that is the reason it is less sensitive than Torcs to code changes.
evil_core: I got black stripes when using se noai
evil_core: avf=expand=:-100::2se ai
evil_core: ain mplayer config
BioTube: yet another reason to hate autotools: configure thinks anything accepting -c is GCC
FIReun: they were a good idea at the time, but are now beyond their prime
BioTube: FIReun: I'd say they're actively rotting
BioTube: unable to recognize another compiler in this day and age is inexcusable
BioTube: s/unable/inability/
cockroach: is there some way to enable power management on a 57xx card (to get the fan to shut up)?
BioTube: cockroach: nope
BioTube: cockroach: DRI comes first
cockroach: BioTube: okay, sounds like a plan
Jonimus: cockroach: and even then we still need a driver for the built in thermal chip since the newer cards no longer use an external chip
cockroach: Jonimus: so basically i have to get better headphones, then :)
Jonimus: cockroach: atm yes
svenstaro: cockroach: or use mechanical means :)
cockroach: svenstaro: clip off the fan and place the box in the freezer? :)
BioTube: cockroach: or buy a big heatsink
svenstaro: probably just use a metal rod with a rubber block in the end to act as a fan limiter :)
cockroach: heh
BioTube: just keep your ear open for the sound of blades flying
cockroach: i guess i'll just be spending more times gaming in windows than working under linux - it's more fun anyway :)
cockroach: s/times/time/
svenstaro: you should spend more time making gaming on linux more fun :)
FIReun: I'd rather play GOW3 on a ps3
BioTube: somebody needs to write a Driver Programming 101
cockroach: svenstaro: now that you mention it, i haven't played openttd in a while...
FIReun: not it
FIReun: cockroach: 1.0 series has its own graphics data!
svenstaro: cockroach: it has come a long way, it now uses all its own media
Jonimus: BioTube: agreed
BioTube: I keep trying to wrap my head around what's going on and failing horribly
cockroach: FIReun: i know, i tried a mac build at some point (back when they still released those), quite nice
FIReun: cockroach: but it still needs work on the industry "management" and level populator
Jonimus: BioTube: there is actually a book about it
BioTube: Jonimus: where?
Jonimus: let me find it
FIReun: cockroach: and I think they should drop the TTD compat and go to hexagons! (:
cockroach: :)
cockroach: (also, last time i played was the first time the AIs outperformed me - must be getting old...)
Jonimus: http://www.xml.com/ldd/chapter/book/
FIReun: cockroach: they still mostly do, unless you confine them to using only boats in a land dominated map
BioTube: Jonimus: thanks!
FIReun: I tried allowing them only cars, but they smoked my trains and boats and planes
cockroach: about the thermal chip, is there any documentation around? i might feel like looking into it (I "wrote" a driver for lm-sensors ages ago, was kinda fun)
cockroach: FIReun: lol
Jonimus: BioTube: no problem, my friend showed it to me and I've been planing to read it for a while so yeah
cockroach: (sweet! even suspend-to-ram works, except that one screen is corrupted after wakeup. but nevertheless, looks promising)
FIReun: cockroach: I was like "woo, I'm up to 2M!" I look over and the AIs are at 12 and 40M$
cockroach: :D
FIReun: and hundreds of cars
FIReun: I think the AI cheats d-:
cockroach: obviously
Jonimus: BioTube: here is the new version but I'm not sure if there is a free online copy of it or not http://oreilly.com/catalog/9780596005900#description
cockroach: i think they should be equipped with a "things have to look awesome and be free of redundancy" obsession like i am, that should level the field...
Jonimus: BioTube: here found it http://oreilly.com/catalog/linuxdrive3/book/index.csp
Jonimus: ofc it doesn't have stuff about the in kernel video drivers and such since that is a recent addition but it is at least better than nothing
BioTube: Jonimus: thanks
Jonimus: np BioTube, now if only I knew C I could read up and help out as well :P
BioTube: the only formal schooling I have is two years of high school java
Jonimus: BioTube: yeah I'm about 2/3rds through my Java class, though I also know a bit of python and z80 ASM
BioTube: I learned my C off of tutorials(probably why I have so much trouble with it)
svenstaro: formal schooling is not at all helpful for freaky open source driver development :)
FIReun: oof, java is no way to learn [:
BioTube: FIReun: I think it's got more to do with the fact that it makes GUI programming so much faster to do(and most schools use Windows)
svenstaro: BioTube: QT makes GUI programming so much faster to do
BioTube: not to mention the fact that academia still seems to be drolling over OOP, so of course a language that forces you to use a hammer instead of a drill is perfect
Jonimus: most of my assignments have had more OOP related declarations and fluff than actual code :/
MostAwesomeDude: Welcome to Java.
MostAwesomeDude: It is a rather masturbatory language.
FIReun: BioTube: its probably easier to teach than anything else
BioTube: my teacher mentioned using two classes for a simple board game
BioTube: I used an array and one class
FIReun: they should have stuck teaching 1st years Basic
FIReun: on old c64s
BioTube: FIReun: over here the AP requires the course be in Java
svenstaro: BioTube: Actually, it does make sense in more complex games. Two classes for a board game sounds reasonable.
FIReun: everything in school is java now
svenstaro: A real board game will have AI, graphics, probably sound, input, network
BioTube: svenstaro: not for sudoku
Jonimus: FIReun: heh I'd have rather done Basic than Java for most of my assignments so far
svenstaro: sudoku is hardly a board game
FIReun: and I still think of anyone who writes java for a living as a java monkey
FIReun: or a script kiddie
BioTube: I never did figure out how to generate a valid board
Jonimus: BioTube: heh I've done it in TI-Basic
BioTube: I just cheated and used a few pre-generated ones
svenstaro: BioTube: if in doubt, brute-force
svenstaro: BioTube: if you want to be awesome, use a genetic algorithm :)
BioTube: svenstaro: tried that, got invalid board
Jonimus: there are rules you can use to randomise a board based on a known correct one
FIReun: Jonimus: right, there ya go.. 1st year programming "how to write code for your TI"
Jonimus: FIReun: heh thats how I got through math class in High School without dying of boredom
BioTube: Jonimus: I was more interested in figuring out how to highlight certain numbers
Jonimus: heck I even worked with a friend on a sudoku game
Jonimus: BioTube: ahh
BioTube: never did figure out how to color terminal output in Java
Jonimus: I'm sure Java makes it a pita to do
FIReun: Jonimus: back in grade school one student accused me of cheating in science class because I figured out how to save a list of periodic elements in my hp calc
Jonimus: nice FIReun
FIReun: Jonimus: but if I was smart enough to do that.. hmm, did I really need to resort to cheating?
FIReun: Jonimus: unfortunately, all you need is a usb cable now days to load your TI up with whatever someone else wrote.
Jonimus: FIReun: my view on it is, if your smart enough to make a program to do it then its not cheating.
Jonimus: yeah I hang out with some other calc programers and we refuse to help those who come in looking for a program they could easily make themselves just to use on a test
FIReun: Jonimus: the first year we got IBM-8086's (love the amber monochrome monitors) and had a "computer lab" I wasnt allowed to take the class because seniors were given preference... so the teacher offered me the "assistant" position...
Jonimus: lol nice
svenstaro: You people are so old
FIReun: yup
FIReun: gettn older every day
FIReun: senility is actually kindof fun d-:
svenstaro: I'd like to die on my 50th birthday which is about the age brain decay picks up
Jonimus: heh I'm only a freshman in College, I ain't old ;P
svenstaro: How long is college?
FIReun: Jonimus: in this economy that doesnt mean anything, I'm a sophomore
FIReun: svenstaro: as long as it takes
Jonimus: FIReun: good point
svenstaro: Is college equal to university?
FIReun: if the economy hasnt "recovered" by the time I finish my BS, I'm on to a MS
FIReun: svenstaro: for most intents and purposes
FIReun: svenstaro: college aint no Yale tho
BioTube: FIReun: if it hasn't recoverd by then, we're very probably screwed
BioTube: Japan still hasn't recoverd
FIReun: BioTube: depends on your definition of recovered
FIReun: BioTube: I'd say we're already screwed, but I've been told I'm a pessimist
FIReun: I say I just know more than they do
FIReun: of course, knowing more just means you know how little you know (:
Jonimus: nice FIReun
ryoohki: where is the firmware for fedora 12?
ryoohki: also, what do i edit if there is no /etc/X11/xorg.conf?
Jonimus: ryoohki: nothing.
Jonimus: as for the firmware see the topic
ryoohki: Jonimus: thanks but then how do specify that i want hw gl and not mesa? i'm using fedora 12 with x server 1.7.6
ryoohki: Jonimus: the topic read "please make sure firmware is installed on free distros..." but does not say where these firmwares are.
ryoohki: i'll look on the wiki
airlied: in fedora the firmware is already installed
airlied: or you need xorg-x11-drv-ati-firmware
ryoohki: airlied: thanks
DanaG: fb: conflicting fb hw usage radeondrmfb vs EFI VGA - removing generic driver
DanaG: nice
DanaG: "nice" as in "good" -- it handed off properly.