EruditeHermit: airlied, the patch did not fix the problem; the outcome is the same LVDS is blank and VGA is like the photo I took.
EruditeHermit: I am going to run hte commands now
Nightwulf|work: hi all
EruditeHermit: airlied, if you are about, I have a few questions about what you asked me to do on the bug report
EruditeHermit: when am I supposed to reload radeon
EruditeHermit: or should vbetool load it automatically
airlied: EruditeHermit: so the attached patch does nothing?
airlied: and after you vbetool post you dn't get text mode back?
EruditeHermit: airlied, so you had me unload radeon
EruditeHermit: will vbetool post do anything if no graphics driver is loaded?
EruditeHermit: yes it didn't do anything if that is what you meant for me to do
EruditeHermit: radeontool regset CONFIG_CNTL 0x0xc0100
EruditeHermit: the register was already set to that
airlied: EruditeHermit: yup it only really works with no driver loaded
EruditeHermit: ok
EruditeHermit: so am I supposed to load radeon after that?
EruditeHermit: VGA has no signal after the vbetool post
EruditeHermit: and the LVDS is blank too
EruditeHermit: ./radeontool regset CONFIG_CNTL 0x0xc0100
EruditeHermit: OLD: CONFIG_CNTL (00e0) 0x000c0000 (786432)
EruditeHermit: NEW: CONFIG_CNTL (00e0) 0x000c0000 (786432)
EruditeHermit: is the output of the regset command
EruditeHermit: no textmode
airlied: EruditeHermit: oops try radeontool regmatch 0x3cc
EruditeHermit: airlied, do I have to redo the whole thing or can I just set the register and vbetool post again?
airlied: EruditeHermit: I need the value in 0x3cc
airlied: so I can tell what to set somehere else
EruditeHermit: 0x3cc 0x00000000 (0)
airlied: okay radeontool regwrite 0x3c2 0x2
airlied: then rerun the regmatch for 0x3cc and CONFIG_CNTL
EruditeHermit: do you mean regset instead of regwrite?
airlied: oops yes
EruditeHermit: OLD: 0x3c2 (03c2) 0x07d00000 (131072000)
EruditeHermit: NEW: 0x3c2 (03c2) 0x00000002 (2)
EruditeHermit: ./radeontool regmatch 0x3cc
EruditeHermit: 0x3cc 0x00000000 (0)
EruditeHermit: http://pastebin.com/aLGUavFm
airlied: now retry the vbetool post maybe
EruditeHermit: airlied, does it matter that I tried vbetool post before on this boot
EruditeHermit: there is no output signal to VGA and LVDS is still blank
EruditeHermit: but I didn't reboot in between vbetool post attempts
airlied: probably worth trying from scratch, though I'd like to fix that ptach
EruditeHermit: ok
EruditeHermit: do you want me to try it once you've put up a new patch?
EruditeHermit: if I do patch -p1 < some.patch, will it reverse a previously applied patch?
twnqx: no
twnqx: it might interactively ask you
EruditeHermit: git reset?
twnqx: patch -p1 -r or git reset, yeah
twnqx: err
twnqx: patch -p1 -R
airlied: EruditeHermit: http://people.freedesktop.org/~airlied/scratch/bios_post_test2.patch
airlied: give that one a god
airlied: go even
C2H5OH: hi
C2H5OH: is anyone else getting random white screen when turning modesetting on for an RV516 ?
C2H5OH: (I have kernel 2.6.33.2
EruditeHermit: airlied, btw your kernel is 2.6.34-rc2 based correct?
C2H5OH: (by random, I mean that sometimes happens and sometimes doesn't)
EruditeHermit: airlied, what steps do you normally use to build a kernel?
EruditeHermit: the distribution method I am using seems to suck for rebuilds
EruditeHermit: as in is there a tutorial webpage you'd recommend following
airlied: EruditeHermit: with Fedora, I just generate a cut down config file for my machine with menuconfig then make ; make modules_install install
EruditeHermit: ah but that requires spending 3hrs reading through menuconfig and figuring out what drivers are required for your system =p
airlied: EruditeHermit: generally lsmod
EruditeHermit: I've spent days/weeks in the past trying to figure out what modules were missing for my machine to boot
EruditeHermit: ah but lsmod only lists modules
EruditeHermit: some things have to be built in
EruditeHermit: thats the tricky part
airlied: yeah I start from the distro config and work my way back
airlied: at some point I juts give up
airlied: and accept it takes w hile to build
airlied: I normally nuke filesystems + networking protocols first
Ke: a bit easier nowadays when you just need ahci +filesystem in addition to basics
Ke: gotta love standards
airlied: EruditeHermit: but generally when I rebase I just accept the impact of recompling pain
EruditeHermit: the main reason I stopped compiling my own kernels was that it was a guessing game
EruditeHermit: changed so frequently
EruditeHermit: and distros do a better job
edwin: I've always been compiling my own kernels
edwin: just take .config from previous one
edwin: run make oldconfig
edwin: and you're set
edwin: ccache helps a lot
edwin: in speeding up the builds
EruditeHermit: edwin, do you know how to use ccache with Ubuntu package building tools?
edwin: with package building tools no, for normals compiles it sufficces to have /usr/lib/ccache as first component in PATH
EruditeHermit: airlied, sorry, I was talking to the ubuntu people about how to do this properly
EruditeHermit: I think I figured out a way to build faster and using their packaging scripts
EruditeHermit: might save time in the future
ssb: MrCooper, around?
MrCooper: yeah
ssb: MrCooper, could you please give me your .config and kernel commandline for a powerbook's 2.6.33 kernel? I'm having troubles getting KMS working here
MrCooper: ssb: http://people.freedesktop.org/~daenzer/config-2.6.33.2
MrCooper: radeon.dynclks=1 radeon.agpmode=1 video=LVDS-1:1440x960-8
ssb: MrCooper, thanks!
MrCooper: np
ssb: bah, still fail
glisse: ssb: how does it fail ?
ssb: glisse, right after yaboots white screen, it stops. no messages from kernel on screen, and no sign of any activity.
marvin24: agd5f: thank for creating a pm2 branch!
marvin24: agd5f: small question ...
marvin24: agd5f: the last commit disables clock switching if there is only one clock per state
marvin24: agd5f: this effectivly disables clock switching on all igps
marvin24: agd5f: which is a waste of energy - is it?
glisse: ssb: with same command line as MrCooper ?
marvin24: agd5f: in such a case, it could be helpfull to merge all power state into a single one
MrCooper: ssb: make sure the necessary firmware gets built into the kernel, otherwise it'll time out trying to load that because the root filesystem isn't mounted yet
MrCooper: it should still bring up a display after a couple of minutes though
edwin: hi
edwin: after testing the fedora live cd for the fedora test day, the radeon r600 driver is in much better shape than the intel, or the nouveau one!
edwin: so good job!
omero: hello all
omero: I got a dell d600 which has a Radeon RV250 (Mobility FireGL 9000)
omero: when I modprobe radeon (kvm enabled), the lcd panel will turn black
omero: I've also connected an external lcd on the vga connector... blackness there as well
omero: dmesg doesn't expose any problem
adamk: Did you mean KMS enabled insted of kvm enabled?
omero: ^^ yes...
omero: http://pastebin.ca/1861608
adamk: Sounds like you don't have fbcon support loaded (or compiled into the kernel).
omero: this are the relevant lines from dmesg
omero: what's the symbol... config_fbcon ?
adamk: CONFIG_FRAMEBUFFER_CONSOLE
adamk: That's the kernel config option.
omero: checking ..
omero: CONFIG_FRAMEBUFFER_CONSOLE=m
adamk: So was the module loaded before you loaded radeon?
omero: shouldn't get modprobed automagically if I modprobe radeon?
adamk: Not really sure.
omero: anyhow this is a path to investigate, adamk ... many thanks
omero: adamk: is fbcon always mandatory with radeon? this might explain another problem I have on anoter laptop (x1400 there)
adamk: It's mandatory when using KMS.
omero: ahh
omero: adamk: fbcon helped!
omero: I'll add a line in modprobe.d/modprobe.conf to have radeon preload fbcon
omero: this cannot be the problem on the other laptop though because that one's kernel has CONFIG_FRAMEBUFFER_CONSOLE=y
adamk: Well we don't know what the problem on the other laptop is :-)
omero: adamk: I had the console going black after a while... but I found out that 'modprobe radeon' before startx made the console not turning black after a while
omero: now however, the problem has become that afer a few hours the display shows some glitches
omero: like line of pixels being redrawn quicly and sometimes it will dim a bit then brighten up a bit
omero: the system becomes sluggish
omero: it doesn't happeng right away... but after some time the system hasnt been touched
omero: I'll be back in a while
adamk_: So when I use KMS on an x850 with these two LCD monitors, which ever monitor is connected to a specific port ends up with washed out colors.
adamk_: If I reboot with KMS off, no problems.
adamk_: If I use an x1900 with KMS, no problems.
mjt: is it intentional that _text_ mode vt with KMS is SO slooooooow?
mjt: when radeon.ko gets loaded every screen update slows by a factor of 100 or so...
mjt: so scrolling one screen-page for example takes about 10 sec
omero: adamk_: I have a x850 xt and noticed that colors are a bit different on one of two identincal (samsung) lcd but I thought that the difference in colors was to be blamed on the lcds and not the card/connector
Nightwulf|work: mjt: isn't the case here...it's just the opposite
omero: I don't know I'd use 'washed out' as a definition but indeed colors are a bit paler/brigter on one head (if I remember correctly the dvi one)
mjt: Nightwulf|work: text-mode scrolling is always fast, on any vga I've seen since early 1980x
mjt: Nightwulf|work: so it can't be "the opposite" ;)
Tommeh: Well, that was interesting.
mjt: Nightwulf|work: well, if you were using some framebuffer, which is even slower.... ;)
Nightwulf|work: mjt: it can...if i compare it with ums mode
mjt: hah
mjt: what IS ums??
mjt: there's no modesetting involved at all when you don't load radeon and x
Tommeh: nomodeset, or no KMS
Nightwulf|work: mjt: the "old" method to to modesetting, without kms
mjt: and while in text mode, scrolling is very fast
mjt: and always has been
chainsawbike: ums = userspace mode setting i think...
Tommeh: I just tested my radeon R600 card on the Fedora 13 beta disk - interestingly almost all the same problems I have with Ubuntu.
mjt: (the only "modesetting" in there is vga=foo kernel parameter)
Tommeh: Wish I'd had time to go through the tests, but I'm at work and I'm a day late anyway. :(
Nightwulf|work: mjt: youre partly wrong because you assume, that the consoles aren't affected by KMS...but they are
mjt: I _know_ kms is affecting consoles because it makes them VERY slow.
Nightwulf|work: mjt: why do you tell me then, it can't be that "the opposite" is the case? i can see it very, very clear here...when switching to UMS mode and try to use a higher resolution, it is terribly slow...with vga mode it's not visible so i can't say wether it's really slower
mjt: i compare, yet again, _TEXT_ mode with the graphics mode radeon.ko gives
mjt: TEXT-mode console is fast
mjt: but once I load radeon.ko which switches to some graphics mode (even if it's still 80x25 in text lines), the whole thing becomes very slow.
mjt: loading of radeon.ko takes a fraction of second
mjt: but when loaded, it slows down whole boot process from 28sec 1m 20s
Nightwulf|work: and that's exactly what I mean....I don't see such an effect here...80x25 or anything bigger is at least as fast as text mode but higher resolutions are by far faster with KMS than with UMS
adamk: omero: Hmmm.. I guess it could be the same issue. I wonder if any of the dev's have an x850.
Nightwulf|work: that simply should have told you, that it can't be intended, what was what you asked for
mjt: Nightwulf|work: aha, now it makes sense ;)
Nightwulf|work: mjt: at least ;-)
Nightwulf|work: uhm...last i mean...damn :P
mjt: so... any idea how to speed it up? :)
Nightwulf|work: mjt: which card do you have and which kernel version do you use?
mjt: it's built-in amd780g
mjt: all kernel versions up to 2.6.33
Nightwulf|work: hmm...dunno which type of chip is built into 780g but since it's not the newest chipset, it should be <5xxx
mjt: r600
xming: rv610 I think
Nightwulf|work: perhaps you didn't load the correct firmware from within the kernel?
Nightwulf|work: mjt: check here please: http://wiki.x.org/wiki/radeonBuildHowTo#TroubleshootingExtraFirmwareforR600.2BAC8-R700.2BAC8-Evergreen
Nightwulf|work: brb
mjt: what's 'correct' fw? Anything more correct than the one which comes with kernel and mentioned on x.org wiki?
mjt: wug
xming: mjt: rv670 here and kms in fbcon is slower thatn VGA text mode, but not as slow as you described
mjt: without firmware _everything_ is dog slow, and withoit the ucode thing it waits for extra 60s at boot
xming: w/o firmware I don't think it even works
mjt: rlc firmware comes with kernel
mjt: it works
mjt: but very very slow
xming: yes and you need an additional fw for interrupt
mjt: that's the ucode one
mjt: required since 2.6.33
Tommeh:
xming: it doesn't come with the kernel
Tommeh: Was gonna say :)
xming: me and pfp come with the kernel
xming: rlc isn't
Tommeh: I couldn't use 3D w/ KMS without the rlc firmware.
mjt: ok, at least mine does not complain about _missing_ fw. Maybe incorrect one, but not missing
mjt: i've pfp, cp, rlc and me files in /lib/firmware/radeon/
xming: mjt: you need these 3 -> http://pastebin.com/kU9WZ1Ay
mjt: exactly what i have
xming: then it's not a fw issue
mjt: it were slow even when i first tried kms (about 2.6.30 or so), when where were no fw required.
mjt: the same slow as now
mjt: or even 2.6.28, i don't remember...
Tommeh: mjt: at least running Ubuntu, I had to have a symlink from /lib/firmware/radeon/R600_rlc.bin to /lib/firmware/`uname -r`/radeon/R600_rlc.bin
Tommeh: It didn't actually find them otherwise.
Tommeh: The clues were in /var/log/dmesg, IIRC
mjt: i'm certain it loads the fw
mjt: because, well, i implemented the udev replacement that does the actual load ;)
Tommeh: Fair enough.
xming: mjt: is it slow when you are using UMS?
mjt: and again, w/o fw it runs entirely differently in all modes
mjt: with ums my text consoles stayed in text mode
mjt: which is fast
xming: and fbcon in ums?
mjt: i never used any fbcon or that crap
xming: well kms is using fbcon
mjt: yeah
mjt: but without kms it's not activated
xming: so you are using that "crap"
xming: :p
mjt: never useD
mjt: and don't use actually
mjt: in my dictionary it's a side-effect, a bug in radeon, -- the graphics mode of the text console that it forces on me.
mjt: so i'm trying to compensate for that somehow
xming: I have no idea why it's so slow on your machine, fbcon has always been slower, kms + fbcon seems alright to me (bit slower than vga text mode)
xming: well rc670 is a bit faster than rv610
mjt: aha. I just booted a default debian kernel on another machine, -- fbcon with ums is even slower... ;)
xming: I always had the impression that xterm is faster than fbcon
mjt: yeah, definitely
xming: mjt: http://phoronix.com/forums/showthread.php?t=19391
mjt: heh. the sympthoms are, well, the same... ;)
xming: so kms + fcon is not h/w accelerated :(
xming: you need a faster CPU :D
mjt: that's a... progress.
mjt: i used to work in text mode over 2400 modem
mjt: that was fast enough for me, >10 years ago
mjt: no i need faster... _processor_ for that!
mjt: s/no/now/
mjt: so the summary is: Not Possible.
mjt: 'hwell ):
Nightwulf|work: re
mjt: maybe i'll try to switch back to ums... if it's still supported
Nightwulf|work: it is
stringfellow: agd5f / anyone who knows r600: just curious, but why does r600 only report GLSL 1.10 / GL_ARB_shading_language_100?
adamk: Probably because that's all that's supported in the driver :-)
mjt: ok, another question. If radeon is modular, can i specify video mode and font when loading the driver? Now, when it loads, it switches to very high resolution and the default font looks really tiny. Later I load the proper font. But this way the screen gets cleared up twice.
chithead: do you really mean font? or rather dpi? you can specify this in xorg.conf
mjt: it's all about CONSOLE still...
mjt: no xorg.conf setting affect console
adamk: You can specify a resolution with video= on the kernel command line.
adamk: Not sure about font. You could check the framebuffer console docuemntation.
mjt: video= takes effect before radeon.ko gets loaded
mjt: is it really video= not vga= ?
stringfellow: I don't think GLSL 1.20 introduces a lot of extra hardware requirements, if any
stringfellow: mostly laguage grammar changes like fixing arrays
glisse: stringfellow: the hw is not the issue
glisse: the driver is the issue
stringfellow: doesn't it get already parsed shaders from Mesa?
glisse: yes
glisse: GLSL -> IR -> driver -> hw
Tommeh: glisse: are my results from the radeon Fedora 13 beta still required?
Tommeh: I think I've found someone that's had the same problems as me anyway.
Tommeh: (Already in the wiki)
glisse: i haven;t look at test day yet
glisse: i am fighting rs600
mjt: who's winning? :)
glisse: obviously rs600
glisse: all it wants to do is corrupt memory and then corrupt hd
Tommeh: Heh
Tommeh: The weirdest problem I have is my 2nd monitor (w/KMS) is fine without compositing, but when compositing is on it flickers when you make any changes to resolutions on or graphics settings _and_ then it's contrast is jacked up really high compared to the other screen.
Tommeh: If it wasn't for that, KMS would be usable.
Tommeh: I'm a sucker for compositing on a daily basis :$
ssb: MrCooper, ahh.. thanks! it worked with built-in firmware. Not useable though -- system hangs after I start glxgears or x11perf.
glisse: ssb: you need agp change
omero: it seems that the problem I'm seeing with the display flickering depends on some kind of powersaving/dpms ...
ssb: glisse, which one?
glisse: they were on dri-devel but they need change
glisse: in the mean time radeon.agpmode=-1
glisse: should work
ssb: thanks! lemme check...
MrCooper: glisse: assuming you're referring to binding the scratch page instead of unbinding, the lack of that doesn't explain an immediate hang
MrCooper: glisse: also I still don't understand how your patch would actually achieve that :)
glisse: MrCooper: why it would not explain hang ?
glisse: MrCooper: my patch force the use of scratch page for all bridges
MrCooper: glisse: because it would only hang after a few minutes for me, and the radeon driver change which triggered that has been reverted
glisse: it just rely on the fact that the infrastrucutre is already there
ssb: same hang with radeon.agpmode=-1
MrCooper: glisse: need_scratch_page on its own doesn't mean the scratch page will actually be used for anything
glisse: it means that on unbind it will be used instead of 0
MrCooper: I can't see any code to that effect
glisse: i changed bridge with custom unbind function
glisse: writel(bridge->scratch_page,
ssb: hang doesn't seem to be an immediate: I start glxgears, gears appear but doesn't rotate and the system completely hangs about one second later
glisse: i pretty much every where agp_remove_memory callback
MrCooper: ssb: sounds like https://bugs.freedesktop.org/show_bug.cgi?id=27525 , but I have no idea what's going on - I'm writing this on my usual X session with compiz, has been running for 8 days
ssb: same with x11perf -shmput500
MrCooper: glisse: didn't see any such change for uninorth
ssb: wonder how to debug this stuff
glisse: MrCooper: ok i did my patch on top of the other uni patch so it got lost
glisse: forget to merge both patch
glisse: thus uninorth was the only one left out
glisse: ...
glisse: anyway i think best is to just enable scratch page on bridge basis i can't test all the bridge and AGP code is completely nutt
MrCooper: glisse: also, PGE_EMPTY won't work as is for uninorth, and I still don't see any initial binding of the scratch page to all entries
glisse: last time i check pge_empty would have worked with uninorth too
glisse: i remember thinking on that ...
MrCooper: ssb: which version of Mesa are you using?
MrCooper: glisse: for U3 maybe
ssb: 7.7.1, current debian sid
MrCooper: ssb: does the machine respond to ping afterwards? Can you log in remotely?
ssb: MrCooper, nope
ssb: haven't tried with plain ethernet though
MrCooper: shouldn't really matter
MrCooper: as I mentioned in the bug report above, the Debian binaries work here as well
dileX: hmm, flash10 movies in ffx-3.6.3 run x-times faster as it should (r300g st/dri)
Tommeh: x86_64?
dileX: no, i386
Tommeh: Probably why flash works then :)
evil_core: Will I get perf diff. 2.6.34-rc4 vs d-r-t for r500?
dileX: will try mesa 7.8 GIT branch
evil_core: dileX: is it master?
evil_core: I dont understand, 7.8 is released verions, am I rigt?
evil_core: right*
dileX: the flash/ffx problems are w/ mesa master
dileX: 7.8 GIT branch has fixes for mesa-7.8.x
evil_core: what is ffx?
evil_core: ffx is firefox?
dileX: yes
agd5f: marvin24: I'm still tinkering with the interface, the final result will do something useful
agd5f: omero, adamk_: the washed out color stuff is fixed in drm-next
evil_core: I dont use flash anyway
evil_core: but should I use 2.6.34-rc4 or d-r-t for r500?
evil_core: I want get better performance/compatiblity
marvin24: agd5f: your approach makes certainly sense for most cases (linux can switch power state as it also does with the cpu)
marvin24: it just fails in some
agd5f: marvin24: I know. I was mostly just playing with things in those patches
mjt: fun. the emulated vga in kvm guest (the text mode) is significantly faster than "text" console.
marvin24: agd5f: what's good is, that it actual make some sense out of different power states ;-)
dileX: the same w/ 7.8. dmesg reveals troubles w/ iwl3945 wlan device. so could be a network problem.
evil_core: adding pageflip for UMS would need changes in kernel also?
agd5f: evil_core: yes
evil_core: thats bad
evil_core: agd5f: so if I would try polyaiung with it, then I need to play with drm code, later with ddx and later also with mesa?
agd5f: evil_core: yup
agd5f: well, mostly just drm IIRC. the ioctls are already in place in the ddx and mesa
agd5f: although it's not been tested in so long, I'm not sure how well it works anymore
evil_core: hmm...but do you think its possible to adapt r400 code for r500 within a few days, without programming skills?
evil_core: hmm..they are for r500?
agd5f: evil_core: should be pretty easy.
agd5f: you just need to update the crtc base addresses
evil_core: whats crtc?
agd5f: evil_core: the hw block that reads out the image from the framebuffer and sends it to the monitor
evil_core: and I can get those addresses from some docs?
agd5f: you can point it at different parts of the framebuffer. pageflipping is just changing the crtc base pointer
agd5f: evil_core: yes. you can just copy it out of the ddx as well
evil_core: so all is ready for r500 in ddx and mesa? yay
agd5f: evil_core: well, maybe. You might need to fix up some stuff as it's not currently used, so I'm not sure what's missing
evil_core: some weeks ago some peoploes scaed me that if nobody did it up to now, it must be realtively not so easy or too much to do
agd5f: but in theory it should work
evil_core: "in theory theres no difference between theory and practise, but in practise there is ;)"
evil_core: anyway thanks, I will try to look over that code
agd5f: evil_core: at the mimimum, you'll need to edit the ddx to enable pageflipping on r5xx chips
oldmerovingian: Is it possible to use the open source ati driver on 2.6.27?
Ke: yes
Ke: I bet most distros ship it by default
Wizzup: oldmerovingian: Possible, yes. Recommended, no. Better get a newer kernel
Ke: (I think it would be recommended to go with what the distro ships, unless you need some specific features)
oldmerovingian: Wizzup: Well I am having problems with newer kernels so I was thinking of going back to a lts kernel
Wizzup: What problems?
evil_core: oldmerovingian: believie me its a bad idea, especially for raeoen driver
oldmerovingian: Well I am having random kernel panics
oldmerovingian: I am running out of ideas
Wizzup: What distro?
oldmerovingian: Arch
Wizzup: Hmm ok, and how recent kernel?
oldmerovingian: 2.6.32 and 2.6.33
Wizzup: Did you manage to see what component panic?
oldmerovingian: But the 2.6.32 kernel with radeon works on Ubuntu fine
oldmerovingian: All the logs are empty
oldmerovingian: It seems to happen so fast that nothing gets recorded
Radioactiveman: Hello, is it possible to use powermanagement AND kms with ati rv770 and latest git packages?
Radioactiveman: I ask because I have read here: http://bbs.archlinux.org/viewtopic.php?id=79509
Wizzup: oldmerovingian: Well, there can be several problems. Are you using up to date xf86-video-ati, mesa/libdrm?
agd5f: Radioactiveman: the pm stuff is still a wip, but you can enable it with dynpm=1
Radioactiveman: with KMS enabled or not?
agd5f: Radioactiveman: with kms
Wizzup: with kms
oldmerovingian: Wizzup: Yea, I know. Thats what makes it so frustrating. Yes, all my drivers are up to date
evil_core: only with KMS
evil_core: Automatic merge failed; fix conflicts and then commit the result.
Wizzup: oldmerovingian: I don't really know a lot about arch. Do you compile your own kernel?
Radioactiveman: ok, because one guy writes in the forum you have to disable kms to use DynamicPM and ClockGating
soreau: oldmerovingian: What makes you think its the graphics driver causing the panics?
agd5f: Radioactiveman: correct. those are ums options, kms handles it directly
oldmerovingian: Wizzup: No, its a pretty vanilla kernel with a few patches.
oldmerovingian: soreau: I dont think it is since it works fine in Ubuntu
soreau: oldmerovingian: A panic could be caused by any kernel module.. when does the panic usually happen?
Radioactiveman: so I could install all git packages and use dynpm=1 in grub and I'll get working pm with KMS? Without any changes in xorg.conf ...
evil_core: how to reset git d-r-t?
oldmerovingian: soreau: It seems to be random. Mostly when I am using a web browser, but that is most of the time
Radioactiveman: and do I need another git tree for this?
suokko: evil_core: git reset --hard origin/drm-radeon-testing ?
soreau: oldmerovingian: Are you using kms or not?
oldmerovingian: soreau: No, I added radeon.modeset=0 to the kernel line
evil_core: 3a44d81fb0e85b18484cae6a20f0661710b6c1db
soreau: evil_core: git reset --hard
evil_core: is it last commit then?
suokko: evil_core: That will reset your local branch to match exactly what the remote branch looked like when you last did git fetch
soreau: evil_core: or git reset --hard
soreau: oldmerovingian: Does it happen when using kms?
oldmerovingian: soreau: I cant use kms when I use compiz, so I am not sure there. I have openbox now, so perhaps I could enable it and see
evil_core: ok, thanks, it looks like nothinf changed in the last 5 days
soreau: oldmerovingian: Why cant you use compiz?
oldmerovingian: soreau: I get a white screen when I log in and KMS is enabled
adamk: Then your drivers are installed properly.
soreau: oldmerovingian: White screen with compiz is almost always caused by drivers using software rasterizer. My guess is you probably need to update the userspace components but can you pastebin your X log from the kms session?
soreau: adamk: +not? :)
adamk: D'oh.
adamk: Yeah... Stupid fingers.
oldmerovingian: soreau: I dont use GNOME or compiz anymore. They arent even installed
oldmerovingian: soreau: I can try to get it with openbox
soreau: Ok so let me get this straight..
evil_core: how to copy locally drm-2.6 to take as less space as it can?
soreau:
evil_core: to not got all dev history, etc
evil_core: I want to play on copy
oldmerovingian: soreau: Yes, I removed them because I was testing to see if it had anything to to with the panics
agd5f: evil_core: git checkout -b my_test_branch origin/drm-next
agd5f: replace drm-next with whatever remote branch you want to track
soreau: Well, as I said, panics could be caused by any kernel module. I still stand by my guess that you need to update your user space components though
oldmerovingian: Can I compare modules from one distro to another?
soreau: I guess you could try the vesa driver to see if the panics dont happen, which would be an indication that the panics are caused by radeon somehow
agd5f: oldmerovingian: also make sure arch doesn't do something like build all your apps with -O3
oldmerovingian: Nah, I dont think it is the video card. It works fine in Ubuntu. No crashes.
oldmerovingian: Maybe I will compare the modules on each distro to see which are different
evil_core: agd5f, ok, thanls
omero: is anyone else having a dimming/brightening display after a while?
omero: I'm sure it's related to some @#!& powersaving because this happens only after I have left the laptop unattended for a (long) while
edwin: agd5f: 2 radeon blogposts on same day? :D
agd5f: edwin: it started out as one, but ended up making more sense as two
Radioactiveman: how many people work actively on the driver?
agd5f: Radioactiveman: the drm? maybe 3-5
Radioactiveman: the xf86-video-ati driver
agd5f: Radioactiveman: probably the same people
Radioactiveman: ok, thanks for answer. cu
edwin: and the r600 one?
edwin: (mesa)
agd5f: edwin: maybe a few more
evil_core: atombios is used also in UMS?
jcristau: yes
evil_core: agd5f: r100d.h and radeon_reg.h > r500_reg.h?
glisse: evil_core: rxxxd.h are asic specific header
glisse: radeon_reg has grown from various asic up to a point where it's sometimes hard to know on which asic each reg exist
ponyofdeath: hi, guys I am trying to configure my ati card with and intel onboard i have 2 monitors on the ati and 1 on the intel here is my xorg.conf, lspci, and xorg log http://pastebin.com/z0PAisR6
ponyofdeath: wondering what i need to chage to get all 3 monitors working and one with and 90 degree rotation?
ponyofdeath: seems that fglrx does not like the intel driver in xorg and unloads it
mjt: ponyofdeath: i mentioned this channel in context of radeon driver, not fglrx -- see /topic here
ponyofdeath: mjt: ok I am trying to find out if its possible to do 90 degree rotation with radeon driver
mjt: yeah, that's the right question here ;)
ponyofdeath: if yes then I would like some help to switch my config over to the radeon driver i dont think xrandr supports multiple vid card right
soreau: ponyofdeath: That's the wrong x log for starters
evil_core: glisse: I want to backport pageflip for r500 from r400
soreau: ponyofdeath: That log is with fglrx
ponyofdeath: soreau: yeah I havent switched over yet reading up on xinerama right now wondering if i can do an 90 degree rotation on one of the monitors first?
glisse: evil_core: you mean for kms ?
glisse: you would have to look at intel flip stuff
glisse: to reuse the drm facilities
evil_core: I want backport it in UMS
evil_core: I am not programmer agd5f told me it would easy to backport from r400 to r500
evil_core: I even must google for asic to read whats that ;)
Tommeh: Way to jump in at the deep end :)
glisse: for ums yes back port is easier
glisse: shouldn't be too hard, thought i don't remember the ums flip code
glisse: note that we consider ums dying
evil_core: KMS is unusable for know - so slow, etc
glisse: i don't kms is that slow here i am not that far from ums
glisse: isn't
evil_core: KMS is nightrmare on mine r500, not because of low fps, but its laggy(frames looks like accumulated in q3a/et), and many things doesnt work in wine(Deathkarz/Ignition) in KMS
glisse: anyway bottom line is i think in 1 year from now we will likely be deaf to ums
evil_core: and XV is laggy usually
glisse: evil_core: you have everything from git ? did you opened bug ?
evil_core: no, it doesnt makes sense for me, no body will understand me anyway
evil_core: I suck at english
evil_core: and KMS XV problem is widely known
evil_core: they were even patches, but didnt worked on mine r500
evil_core: and for strange frames flow in KMS, I got answer thats rprobably quake3 bug, etc
evil_core: its unacceptable for me, because q3 worked smoothly on TNT2 and even older cards(really wide variety)
glisse: well one day or the other you will have to jump on the kms wagon ...
agd5f: evil_core: you should still report the problems. I'm not aware of any issues with Xv on kms
evil_core: "* Destroy all the FBs associated with @filp."
evil_core: its filp, or should be "flip"?
evil_core: ouch, I found whats that ;)
evil_core: agd5f: do you know in which files I should hunt for code to copy?
evil_core: hmm...there were also hyperz support for r300/r400?
agd5f: evil_core: radeon_cp_dispatch_flip in radeon_state.c
agd5f: evil_core: for r5xx, you'd want D1GRPH_PRIMARY/SECONDARY_SURFACE_ADDRESS/D2GRPH_PRIMARY/SECONDARY_SURFACE_ADDRESS
agd5f: rather than RADEON_CRTC(2)_OFFSET
evil_core: agd5f: it means I got choose D1... or SECOND.. and simply replace it?
agd5f: evil_core: replace RADEON_CRTC_OFFSET with writes to AVIVO_D1GRPH_PRIMARY_SURFACE_ADDRESS and AVIVO_D1GRPH_SECONDARY_SURFACE_ADDRESS
agd5f: and replace RADEON_CRTC2_OFFSET with writes to AVIVO_D2GRPH_PRIMARY_SURFACE_ADDRESS and AVIVO_D2GRPH_SECONDARY_SURFACE_ADDRESS
agd5f: so you'll need to write 4 regs instead of 2
evil_core: agd5f: so I should copy that function to other file, rename them, or make some IFDEF for r500?
agd5f: evil_core: something like this: http://people.freedesktop.org/~agd5f/r5xx_ums_pageflip.diff
agd5f: untested
agd5f: you'll need to define the reg offsets, etc
evil_core: ok, thanx, and thats all is needed?
agd5f: evil_core: that and making sure the ddx enabled it and calls the right ioctls
evil_core: anyway, how to clean only radeon dir in kernel?
evil_core: make drivers/gpu/drm/radeon clean <- this cleaned everything
suokko: evil_core: make SUBDIRS=drivers/gpu/drm/radeon clean
suokko: SUBDIRS is always used for limiting kernel compilation to specific directory
evil_core: suokko: ok, thanks
mmp: glisse: hello; sorry for bugging; did you manage to get that RS600 motherboard?
sgcb: hey, does anybody know how to fix a bad EDID checksum error... http://pastebin.com/Nea1apzE
fcami: sgcb: check the cable (plugged correctly?), change the cable, or create xorg.conf
fcami: probably in that order, but YMMV.
agd5f: sgcb: is this a recent issue? might want to bisect and see what broke it
agd5f: might also be that the drm is now more verbose about checksum errors
glisse: mmp: yeah i am stilling trying to make it works
mmp: glisse: ahh; can I help anyhow?
sgcb: it just started happening a few days ago.
agd5f: sgcb: try bisecting
sgcb: agd5f: okay I'll see what I can come up with, thing is I have 2 of these monitors, the newer one works but the old doesn't
sgcb: would they have unique edids?
bdmurray: I'm having an issue with the radeon driver on Ubuntu Lucid part of my dmesg can be found at http://pastebin.osuosl.org/32450
agd5f: sgcb: maybe.
glisse: mmp: don't think so unless you start playing yourself with the code
agd5f: bdmurray: GPU has hung
mmp: glisse: well... I'm not very optimistic, as I have a master's thesis to finish :-(
bdmurray: agd5f: okay, is there some debugging I could do?
agd5f: bdmurray: do you have a reproduceable test case?
bdmurray: agd5f: yes, I believe so. I'd need to try one more time to be certain. I belive the same thing has happened 3 times and this is the 1st time I've been able to ssh to the system.
evil_core: what means that something is bitch to do/make? (hard or easy)
Wizzup: hard, or annoying
evil_core: airlied told that about HiZ, and was wondering what he wanted to say up to now ;)
evil_core: anyway, theres no pageflip for KMS in any radeon driver, am I right?
glisse: evil_core: yeah no pageflip for kms
adamk_: Hey guys... I'm helping a slackware user to get KMS and DRI working on his GPU. Dmesg is complaining that RV635_pfp.bin is missing, but I'm not familiar with that one.
adamk_: And, interestingly, I don't have it on my system either (but I'm using an RV700 GPU, so I assume it's not necessary).
glisse: it's a firmware
adamk_: Right, but I'm not seeing it here: http://people.freedesktop.org/~agd5f/radeon_ucode/
adamk_: Is this one included in the kernel tree?
glisse: weird it's in fedora firmware package
adamk_: It's not in the 2.6.31 kernel source tree.
agd5f: adamk_: make sure your initrd has the ucode files
glisse: 2.6.31 is kind of old
adamk_: Yeah, he's actually using 2.6.33.2... 2.6.31 is the last one I build on the only linux box I can access at the moment.
adamk_: So it should be in the source tree for 2.6.33.2?
jcristau: firmware/radeon/RV635_pfp.bin.ihex is there
adamk_: Hmmm... I guess either it didn't get installed properly or he's using an initrd and never added it... Time to find out which. Thanks for the tips.
sgcb: agd5f: I've tried a couple older kernels that I still have on disk (34-rc[234] and even 33). All have the same problem now apparently?
evil_core: bzImage in vmlinuz, right?
Ke: make install ;o)
evil_core: doesnt work
evil_core: # make install
evil_core: sh /mnt/download/soft/drivers/gfx_card/radeon/drm-2.6_pageflip/arch/x86/boot/install.sh 2.6.34-rc3airlied-59071-g3a44d81-dirty arch/x86/boot/bzImage \
evil_core: System.map "/boot"
evil_core: Cannot find LILO.
Ke: heh, I remember this problem
evil_core: Ke: and what helped?
Ke: can't remember
evil_core: so shut up, and dont make me more angry ;)
Ke: well anyways, you can copy the bzImage
Ke: http://www.gentoo.org/doc/en/handbook/handbook-x86.xml?part=1&chap=7 ← source gentoo hendbook
evil_core: its what I wanted to do anyway ;)
evil_core: s/i386/bzImage/
evil_core: can I restore some file from ext3 after deletion?
BioTube: evil_core: not neatly
evil_core: I deleted initrd file, and have better things to do, than write initrd scripts, etc
BioTube: you'd have to grep the filesystem for it and pray it's all together
evil_core: its not so important
evil_core: but its not fun to assembly initrd :/
bdmurray: agd5f: so I've recreated this a couple of times once with firefox and once with googleearth is there some other way that might be more reliable?
evil_core: agd5f: you made a nice wallhack when ddx used with nonflipped kernel :D
androide: hello, i have a X550 ATI Card, but i dont know the right version of the driver, which i'd use?
androide: xf86-video-ati 6.12.192-1
androide: when i Start the X it goes blue, and the blue layer keeps even on the TTYs
androide: any ideas?
androide: with vesa driver works, but not fluently
jdccdevel: I'm having a problem with my dualhead setup. Can anyone help?
jdccdevel: I have a X800, and I want to run it in zaphod mode
evil_core: agd5f: it looks like single buffering now, or even worse
evil_core: I need to cnagne something in mesa, or is it simple broken?
evil_core: pr maybe somebody else could help me with PageFlip on r500?
evil_core: androide: what kernel?
evil_core: androide: try switching between KMS and UMS
jdccdevel: Does anyone have experience with zaphod heads?
evil_core: [ 1641.138] (**) RADEON(0): Option "EnablePageFlip" "False"
evil_core: [ 1641.145] (**) RADEON(0): Page Flipping enabled on r5xx and newer chips.
evil_core: I've changed bot that var I shoould? ;)
jdccdevel: Question, is it still correct that DRI is not supported when using multiple screens (i.e. not mergedfb?)
adamk: With zaphod? Last I heard it would not be possible without KMS. And, even with KMS, the drivers needed work.
adamk: Not sure if that work ever happened.
jdccdevel: Adamk: KMS == Kernel Mode Setting, right?
adamk: Correct.
UnNamed: MostAwesomeDude: hi, i think i got something from oprofile, but no idea if valid as i had to "workaround" the missing lib (no complains on launch, tho)
jdccdevel: adamk: What sort of work do they need? I'm very interested in getting it working, and may be able to help
MostAwesomeDude: UnNamed: Okay. Could you pastebin opreport -l path/to/your/r300_dri.so ?
jdccdevel: adamk: At the very least, I could act as a tester.
evil_core: MostAwesomeDude: maybe you can help me with pageflip?
UnNamed: MostAwesomeDude: what do i do with te huge tree of files that appeared in /var/lib/oprofile/sample/...
MostAwesomeDude: evil_core: Pageflipping requires a bit of kernel code. I don't think anybody ever wrote it out for r500.
MostAwesomeDude: UnNamed: oprofile --dump, then opreport will work.
evil_core: MostAwesomeDude: I talkd to agd5f and he did it, it works, but looks like single-buffered
UnNamed: MostAwesomeDude: hmmm... that was just curiosity, it seems the other cmd printed a list of calls with some %
evil_core: for UMS, not KMS
MostAwesomeDude: evil_core: "Looks like?" How so?
MostAwesomeDude: UnNamed: Yep, it'll do that. Pastebin, please. :3
evil_core: I got interleaved frames
UnNamed: MostAwesomeDude: what is ":3"?
evil_core: in quake3 it looks like a wallhack
evil_core: it looks its drawing in the main buffer
UnNamed: http://pastebin.com/dF0izvMW
agd5f: evil_core: you hve to update the sarea like is done RADEONInitCrtcBase and RADEONInitCrtc2Base
MostAwesomeDude: UnNamed: It's just a smiley. Could you pastebin that opreport output?
evil_core: agd5f: in Mesa?
agd5f: no in the ddx.
UnNamed: MostAwesomeDude: yeah, what smiley? and you should read :P (<- simple smiley)
MostAwesomeDude: Sorry, lag. Reading.
evil_core: agd5f: and you think it will be enough?
MostAwesomeDude: So what app is this, and how are you making it go slow?
agd5f: evil_core: I don't know without testing
agd5f: could be
evil_core: agd5f: I dnot understand which functions/files I need to update
UnNamed: MostAwesomeDude: blender, "solid" mode (wire, shaded, textured are a lot faster), with monkey, subdivided so it had ... 32000 faces iirc
UnNamed: MostAwesomeDude: trying to rotate the view takes seconds
evil_core: I ot them in 3 files
MostAwesomeDude: UnNamed: Shader fallbacks.
UnNamed: MostAwesomeDude: so solid mode uses a different shading than others, and then goes slow?
MostAwesomeDude: UnNamed: RADEON_DEBUG=fall should print out some useful stuff.
UnNamed: MostAwesomeDude: i have been using that for ages, all i got was WARNING! Falling back to software for line stipple + smooth lines + render mode != GL_RENDER
UnNamed: MostAwesomeDude: when using the app more, i remember that it also prints smooth poly and some other stipple things
UnNamed: MostAwesomeDude: if i use driconf to disable the stipple/smooth ones (!= GL_RENDER still appears), it still goes slow too
evil_core: anyway, I am starting understanding it :)
evil_core: so I need to write RADEONInitAvivoBase?
agd5f: evil_core: something like: http://people.freedesktop.org/~agd5f/r5xx_ddx_pageflip.diff
MostAwesomeDude: UnNamed: If *any* of those fallbacks happen, then we have to do software.
MostAwesomeDude: Including GL_SELECT, which Blender is a fan of.
MostAwesomeDude: Does it still lag as much on r300g?
UnNamed: MostAwesomeDude: why would it do select for rotating the view?
MostAwesomeDude: UnNamed: I'm honestly not sure.
UnNamed: MostAwesomeDude: it can make sense to do software to select items (i can live with slow select)
UnNamed: tries something and see if select happens
UnNamed: MostAwesomeDude: i tried just adding the monkey etc and no select op at all (no warning printed)... rotating was slow as always
UnNamed: MostAwesomeDude: for r300g i will have to enable kms etc, so it will have to wait at least until tomorrow
evil_core: agd5f: nothing changed, but it was only part of the fix?
evil_core: ouch, I understand, its missing function I asked ;)
gsedej: hi! I installed ubuntu 10.04 and special gallium3D PPA, so I have mesa 7.9-gallium
gsedej: glxinfo still says OpenGL 1.4
gsedej: how to get OpenGL 2.1?
MostAwesomeDude: Which chipset?
MostAwesomeDude: Actually, could you just pastebin the output of glxinfo?
gsedej: right rebooting
gsedej: I have R350 .9600 mobile
gsedej: do I need to do something at boot?
MostAwesomeDude: Hm. Yeah, I recall people saying that r300 didn't advertise right...
MostAwesomeDude: No, no special boot things.
gsedej: like radeon.gallium=1?
MostAwesomeDude: No, nothing like that. You do need to be in KMS though.
gsedej: kms is default in 10.04 right?
MostAwesomeDude: Dunno.
evil_core: radeon.modeset=1
evil_core: you see if console got native DFP res
gsedej: sec...
soreau: just check 'dmesg|grep modeset'
gsedej: yep, it is :)
gsedej: and it's FAAAST :D
UnNamed: what is fast?
gsedej: switch to console
gsedej: I think I have 2.1 sicne console log is to short to get all extensions
gsedej: just sec...
soreau: gsedej: What does 'glxinfo|grep nGL' say?
UnNamed: wonders how many times one has to switch consoles...
gsedej: OpenGL version string: 2.1 Mesa 7.9-devel :)
MostAwesomeDude: I really would like a pastebin of glxinfo on an r300.
gsedej: thanks guys! :)
UnNamed: MostAwesomeDude: r300 card? or r300 driver? or r300g driver?
soreau: gsedej: Can you pastebin the output of 'LIBGL_DEBUG=verbose glxinfo' ?
gsedej: ok... so how to "test"? phoronix test suite?
gsedej: glxgears is same speed (I know it's not benchmark)
UnNamed: gsedej: go use real apps (and not just games) ;]
gsedej: webGL works? OpenGL ES 2.0?
gsedej: which real apps?
soreau: google earth?
UnNamed: gsedej: that is one app :]
soreau: of course a crappy connection can make the performance bad
MostAwesomeDude: UnNamed: r300 card on r300g driver.
gsedej: http://pastebin.org/152883
gsedej: glxinfo output
gsedej: it's RV350
evil_core: tuxracer, icebreaker, xbill, mc, sc, vim ;)
MostAwesomeDude: Hah, so many configs and visuals. :3
soreau: Yea, I have that same card
gsedej: soreau, google earth should be faster? (much faster?)
soreau: gsedej: No, it just should render correctly
gsedej: MostAwesomeDude, what's that for? (visuals?)
UnNamed: i have a rv360 but all insists on calling it rv350
gsedej: soreau, how correctly? It did before on classic mesa
soreau: There's also apps like xbmc and blender (real world apps that aren't games)
soreau: gsedej: gallium is a new 3D driver
UnNamed: and wings3d, and meshlab and a bunch of chemistry/physics apps
gsedej: xbmc was slow with KMS and much flickery without KMS (but fast)
soreau: It's not as good as classic from my tests here on rv350
soreau: but, it's come a long way from what it used to be
gsedej: err, what basicly offers GL 2.1 compared to 1.4? Shaders?
gsedej: btw... GoogleEarth is slower... is this OK?
evil_core: yup
evil_core: KMS is slower than UMS generally
evil_core: and only xmoto works best with Gallium
gsedej: what about games in Phoronix test?
evil_core: agd5f: your patch were complete fix or only part?
gsedej: that even bets old FGLX driver?
evil_core: dunno, mine r500 isnt supported by fglrx
gsedej: http://www.phoronix.com/scan.php?page=article&item=ubuntu_lts_gallium3d&num=2
gsedej: I am going to bed, see you tommorow!
evil_core: I too
evil_core: gnight all
MostAwesomeDude: Hm.
MostAwesomeDude: Strange. rv350 appears to not correctly report the NPOT cap...
MostAwesomeDude: soreau: Do you get 2.1 advertised on your rv350 w/r300g?
soreau: MostAwesomeDude: Yes, it reports 2.1 again now
sgcb: is drm 1.1.0 20060810 supposed to be the latest kernel drm version?
MostAwesomeDude: soreau: Okay, that's what I expected.
MostAwesomeDude: So the PPA version is just out-of-date.
soreau: That's what I was thinking too
soreau: hm
soreau: MostAwesomeDude: What is it misreporting in the glxinfo pastebin?
MostAwesomeDude: soreau: ARB_npot. We used to have a conditional on it, but we took it out.
soreau: So it reports npot thought not implemented?
soreau: though*
MostAwesomeDude: Well, it's a lot more implemented now than it was before.
soreau: heh
agd5f: evil_core: you still need to enable pageflipping for r5xx
sgcb: agd5f: I've tried past kernels 2.6.34-rc[234] and 2.6.33. All now have the same problem. (they didn't a few days ago) I've also hooked the monitor up to an ubuntu/nvidia desktop and it worked there so I know the monitor is not borked.
sgcb: is there something else I can do to debug the problem
agd5f: sgcb: did it used to work at some point in the past?
sgcb: yes, 3-4 days ago i believe
agd5f: sgcb: sounds like maybe the drm edid checker just got more verbose? Try bisecting between whatever commit you were running a few days ago with what you are running now
sgcb: agd5f: I've tried my past kernels already (however not git-bisecting, just the ones I still have on my system) 2.6.34-rc[23] were in fact working, but now I'm seeing the same checksum error
sgcb: with those kernels
Jonimus: I've heard mesa sometimes has issues with OpenJDK, could that be why my app is failing while trying to setup a double buffer?
Jonimus: yeah, it appears to attempt to setup page flipping and fails to do so :/
soreau: MostAwesomeDude: This happens with m64+ using z64 vid plugin with gallium http://pastebin.org/153042
soreau: arachnoid and rice work pretty well now though
MostAwesomeDude: looks up that opcode
MostAwesomeDude: Oh, nevermind, I know that one. Predicated kill.
Jonimus: hmm yeah it seems I can't do double buffering in Java with the current setup :(
Jonimus: does that sound right for a r7xx card?
soreau: MostAwesomeDude: and fwiw, z64 used to work with gallium, pretty slowly though
soreau: I'm installing cairo-dock to test it but last time I tested it wasn't too happy with gallium
soreau: had like duplicated texture problems
MostAwesomeDude: cairo-dock may or may not obey the GLX rules. The jury was out on that last time.
soreau: I talked with fabounet about it but didn't get too far..
soreau: I'll show a screenshot if it's still doing it
MostAwesomeDude: soreau: If you don't mind, could you pastebin the output of that z64 thing with RADEON_DEBUG=fp set?
soreau: Sure
soreau: MostAwesomeDude: http://pastebin.org/153079
soreau: idly wonders what 'predicated kill' is
MostAwesomeDude: You can kill a fragprog with KIL. It checks to see if any component is negative, and if so, it bails.
soreau: ugh. the g made the music stop and the screen go blank when trying to start cairo-dock
MostAwesomeDude: Ugh indeed. dmesg show anything?
soreau: Let me do it again and try ssh this time
soreau: it didn't happen this time
soreau: oh I set it wrong..
soreau: well LIBGL_DRIVER_DIR isn't working because even though it makes glxinfo report gallium, cairo-dock reports classic mesa
soreau: maybe it's whatever libGL it was built against?
sgcb: LIBGL_DRIVERS_DIR ?
soreau: yep
MostAwesomeDude: soreau: Would you kindly see if http://paste.pocoo.org/show/202429/ works?
soreau: looks
soreau: hmm
soreau: now rsp segfaults and arachnoid gets an assertion failed
MostAwesomeDude: Huh.
MostAwesomeDude: Oh!
MostAwesomeDude: You have to make clean in src/mesa/drivers/dri/r300/compiler.
MostAwesomeDude: And then re-make.
soreau: ok
soreau: now all vid plugins fail with mupen64plus: radeon_opcodes.h:228: rc_get_opcode_info: Assertion `rc_opcodes[opcode].Opcode == opcode' failed.
soreau: I had to apply the patch by hand because git apply didn't work
soreau: hope I didn't screw it up :P
MostAwesomeDude: No, I did. Try http://paste.pocoo.org/show/202434/
MostAwesomeDude: One new chunk.
soreau: MostAwesomeDude: Now z64 segfaults, arachnoid gives "mupen64plus: main/texenvprogram.c:194: translate_source: Assertion `0' failed." glide gives this http://pastebin.org/153150 and black/blankscreen and rice doesn't give any error but blackscreen too
MostAwesomeDude: soreau: What.
soreau: idk, it blew up
soreau: I did make clean in that one directory
MostAwesomeDude: One second, making clean in my entire Mesa and seeing if anything bad still happens here.
soreau: but maybe need to do full make clean
soreau: runs his ultimate rebuild script complete with distclean and config options :)
MostAwesomeDude: Builds and runs here. I don't have the app you have, though.
soreau: MostAwesomeDude: It ran for half a second with z64 but it blackscreened and now I'm ssh'd in
soreau: MostAwesomeDude: Here http://pastebin.org/153159
soreau: can I kill it yet?
soreau: I assume the problem would be reproducable..
MostAwesomeDude: Yeah, go ahead and kill it and back out that change.
MostAwesomeDude: I'll have to investigate that late.r
soreau: Well I haven't even tested the other plugins
soreau: this just happens to be z64 which is somewhat of a stretch compared to the others
MostAwesomeDude: Oh. Well, you can if you like, but I've a feeling that it's going to continue to misbehave.
MostAwesomeDude: wishes there were an option for dumping GLSL
soreau: This is the output while it was black screened http://pastebin.org/153166
MostAwesomeDude: You ran it full-screen?
soreau: yea, I have it set to open fs..
airlied: MostAwesomeDude: not some mesa deubg option?
soreau: MostAwesomeDude: I get flooded with these using glide while the game is running very smoothly http://pastebin.org/153168
soreau: I'm surprised because last couple times I tried with gallium, arachnoid ran but now it's actually very playable
soreau: and with your patch, ric runs slower for some reason
soreau: but it's cool to see the blur and fog effects doing nicely in the emulator too
MostAwesomeDude: airlied: Not that I know of.
MostAwesomeDude: soreau: http://paste.pocoo.org/show/202444/ should be much more successful.
MostAwesomeDude: Note to self: Unify r300_fragprog and r500_fragprog TEX transforms.
soreau: MostAwesomeDude: Ok I will try it now
soreau: MostAwesomeDude: Still does the same thing with z64 and rice chokes on it
soreau: but rice works fine with classic mesa
MostAwesomeDude: soreau: Figures. Could you pastebin with RADEON_DEBUG=fp so I can see how we're failing on shaders?
soreau: MostAwesomeDude: For the horrid z64 plugin?
soreau: or for rice going slow
MostAwesomeDude: Whichever you care about more, probably the former since that's what the patch is targeting.
soreau: right
soreau: MostAwesomeDude: http://pastebin.org/153184
MostAwesomeDude: soreau: Hm, looks like I'm just not doin' it right. I'll have to try to track down this app later and do it locally. Sorry.
soreau: MostAwesomeDude: It's no big deal, I just happen to be testing tonight. Thanks for looking
soreau: I am still trying to get cairo-dock running
soreau: and I have to start compiz with GALLIUM_ABORT_ON_ASSERT=false because I get a bunch of "r300_blit.c:137:r300_surface_copy: Assertion `dst->texture->format == src->texture->format' failed."
MostAwesomeDude: Ooh, don't hide those. Those are important.
soreau: Yea, it's spamming the terminal and causing some visual errors
soreau: yay, 0.9 works again with classic mesa
soreau: still trying to figure out why I get red tinted shadows with it though
MostAwesomeDude: Oh, copy from ARGB to RGBx, probably.
soreau: ah, I'll relay that to smspillaz..
airlied: need to fix tfp test in piglit to get compiz going properly I expect
EruditeHermit: airlied, from your patch
EruditeHermit: http://pastebin.com/zNG668LB
EruditeHermit: I get this error on compilation
airlied: add an #include "r100d.h" to radeon_combios.c for now
EruditeHermit: ok
jdccdevel: I have a question about zaphod heads on a R430, Is it possible to set up DRI on one or both heads?
agd5f: jdccdevel: with kms it should work
jdccdevel: agd5f: which kernel version do I need? I'm currently running (on gentoo) linux-2.6.32-gentoo-r7
agd5f: jdccdevel: any kernel with kms
jdccdevel: mesa 7.7.1
jdccdevel: xf86-video-ati-6.13.0
jdccdevel: libdrm-2.4.18
jdccdevel: sorry, that's xf86-drivers-ati-6.12.6
airlied: you need 6.13 for kms
Jonimus: does anyone else have issues with xf86-video-ati and openjdk?
jdccdevel: Ok. 6.13 with KMS.
DanaG: hmm, how does R600 KMS deal with multiple video cards?
airlied: DanaG: same as non-kms
jdccdevel: is there a good reference anywhere on how to set up KMS and make sure it is working properly?
Jonimus: I'm makeing a java app that uses double buffering and on linux it crawls but my friend on windows he says it flies