Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-11-25

Search This Log:


sgcb: rv770 here with latest git, mesa is falling back to software rasterizer
sgcb: ah here we go, LIBGL_DEBUG=verbose glxinfo reports: libGL error: dlopen /usr/lib/dri/r600_dri.so failed (/usr/lib/dri/r600_dri.so: undefined symbol: radeon_bo_is_referenced_by_cs)
eosie_: sgcb: get the latest libdrm from git
sgcb: I have it... unless it's not installing correctly
eosie: it's an inline function in libdrm headers
eosie: after that, recompile mesa
cxo: better git to bed
Luzipher: sgcb: gentoo ? does your glxinfo report mesa 7.7 or 7.8 ?
sgcb: eosie: okay, I did a make clean distclean ...etc. on libdrm and mesa still no dice
sgcb: Luzipher: debian - 7.8
Luzipher: sgcb: hm, ok, it's different then from my problem a few days ago i guess. deleting my git directory helped back then.
eosie: sgcb: you installed libdrm incorrectly then
sgcb: eosie: I'm not sure what I did then, both libdrm and mesa were set with --prefix=/usr so mesa should've found the correct hearders?
hifi: sgcb: did you build libdrm with --enable-radeon-experimental
hifi: I'm not sure if I typed that correctly
hifi: it does not build the radeon bits without that flag
sgcb: ya i believe it's --enable-radeon-experimental-api
hifi: anyway, it did not install correctly
hifi: or your mesa did not link against it at least
sgcb: eosie(or anyone): could you link me to the header or otherwise provide me with the inline function name that should be linked in mesa, maybe a manual check would reveal something
hifi: your error message stated the function name
sgcb: err yes... ha ha
sgcb: alright I installed libdrm to /usr/local and mesa no longer falls back on swrast
sgcb: I find it odd though that mesa looks in /usr/local/include regardless of what prefix is set to...
nightmorph: oy. did someone mess up RadeonProgram on the x.org wiki?
nightmorph: what's all this stuff about 7.8-dev or 7.8-git -- shouldn't it say 7.7-dev like every other entry, or was 7.7 really released in the last couple of days?
nightmorph: it's all the changes by some guy named KenMays -- he added these weird inconsistent versions and a buncha games i've never heard of
hifi: also I think some general slow flag should be added there
hifi: where something works graphically as platinum doesn't mean it actually runs well
nightmorph: hifi: that should be in footnotes. as the text at the top says, the ratings are for rendering accuracy, not speed
hifi: sure, but a general footnote
nightmorph: so as long as there are no visual artifacts, and it all loads, then it should get a top rating. performance is something you'd put in a footnote :)
hifi: that also might have some different ratings
nightmorph: on the other hand, i am rather surprised at the rating UT2004 has received from this inconsistent wiki updater. on r700, no less
nightmorph: i can't imagine it'd have good performance, but there's no real detail
nightmorph: you're right, it'd be nice if there was a separate mark to give each app for performance
hifi: something like 3 general speed rating
hifi: ratings*
nightmorph: but then that begs the question of "well, running it under KMS? dri1? dri1?" etc.
nightmorph: er dri2
hifi: true
hifi: maybe we need a separate page for performance? :)
nightmorph: eh, that'd be a lot of copying stuff over and hoping people update it :)
nightmorph: but you're right, there's a serious lack of an "Anecdote central"-type place
hifi: or just rate everything on KMS only
hifi: until gallium is finished, KMS only, then KMS+Gallium or something like that
nightmorph: ooh, that'd be nice. problem is not everyone's using it -- some folks don't since they want the power management stuff, or they find KMS unstable
hifi: true, but thats better than nothing
nightmorph: which reminds me, anyone know what the status is of those new KMS power patches?
spreeuw: nightmorph: why?
spreeuw: I think I updated the ut entry
nightmorph: i'd like to see 'em pushed in
spreeuw: you have to disable s3tc
spreeuw: in driconfig
nightmorph: spreeuw: it's not enabled by default anyway, is it?
spreeuw: it runs fine for the rest
spreeuw: it is enabled by defualt
nightmorph: also, have you tried it on KMS+R700? if so, what are your git checkout revisions
spreeuw: giving you junk
hifi: on the other hand, a database with all this stuff would be more flexible
spreeuw: the s3tc detection is broken in mesa for r600
hifi: editing that massive table is pain
spreeuw: I complained about it before
spreeuw: aside from the fact that it isnt supported
spreeuw: the handling is broken too
nightmorph: hifi: yes. i hate wikis for this reason.
nightmorph: spreeuw: isn't s3tc really only for nvidia chips, anyway?
spreeuw: I have not tried kms
spreeuw: cant get it to run
spreeuw: no older radeon supports s3tc just fine
spreeuw: in mesa
spreeuw: r200 and r300 worked in the past
spreeuw: wit hte plugin lib
spreeuw: but its broken for higher
nightmorph: hmm, so it'll run on my old X1950 Pro, but not on my 4550? pity.
spreeuw: I only added 7.7-dev
spreeuw: a while ago
spreeuw: the new edits you mentioned are not mine
spreeuw: nightmorph: it runs on r600 too
spreeuw: just disable s3tc in the driver
spreeuw: you can do it with driconfig
hifi: so would current git master be 7.8-git as in the page currently?
hifi: I'll update an entry if so
spreeuw: 7.8-dev is the current git yes
spreeuw: sortof
spreeuw: they cant diversify it to checkout level
spreeuw: the table is a mess as is
spreeuw: and the gui editor is broken
hifi: yeah, just tried gui...
nightmorph: spreeuw: shouldn't it be 7.7-dev?
nightmorph: since 7.7 isn't released AFAIK
nightmorph: so 7.7-dev really means "current git"
spreeuw: no current git is named 7.8-dev
spreeuw: in the gl renderer string
hifi: just when I saved
hifi: well, 7.7-dev is better than 7.5
spreeuw: 7.7 or 7.8 doesnt really matter
nightmorph: 7.7-dev
hifi: it doesn't really matter as the problems are still there :p
nightmorph: as all the other entries use it. i'm looking at mesa git right now, and there's no 7.8 branch, so whoever put it in there was on crack and sloppy
spreeuw: just pick the highest version you see
nightmorph: 7.7-dev, please
nightmorph: for consistency's sake, as that's what it's been the whole time
nightmorph: spreeuw: also, i need to reexamine my renderer info, then
nightmorph: OpenGL version string: 1.5 Mesa 7.7-devel
spreeuw: you must mention the highest version tested in the table
spreeuw: ut2004 runs fine btw
spreeuw: it should atleast have silver or gold
spreeuw: you must disable s3tc
nightmorph: i'll reinstall it on my desktop, then. i've been trying it on my intel X3100 laptop, but performance is abysmal even at lowest settings
spreeuw: in the driver
hifi: I really think we should have a database
spreeuw: for the 3th time
nightmorph: yes, we heard you
hifi: with all the quirks, hacks and comments
spreeuw: yep a db would be the best
nightmorph: hifi: agreed, but...i doubt fdo will implement one
spreeuw: but diversifying further messes the significance up too
spreeuw: its a tradeoff
hifi: it does not need to be hosted at fdo
spreeuw: someone could maybe php something up
hifi: though, ideally you could "dump" the database as wiki text to fdo :)
spreeuw: with smarty or so
spreeuw: or with some ugly web 2.0 engine
nightmorph: ugh, more javascript fail
nightmorph: no?
hifi: also the database could host screenshots of apps running on radeon
nightmorph: there's a linux hardware query site out there...can't remember what it's called offhand, but they seemed to ahve the right idea
hifi: that way you could actually see and read how well it performs
hifi: I smell a weekend project
nightmorph: hifi: that'd be good, but the current pace of the git work would make screenshots out-of-date on even day to day checkouts :)
nightmorph: at least here, i've seen urbanterror go from platinum to gold, on some maps, as there are a few texture regressions, but they work fine in the next day's checkout
hifi: sure, but you could TAG it with a commit revision ;))
nightmorph: true
hifi: I like how wine's appdb works
hifi: the overall structure of a applications page is good
nightmorph: yeah, that is convenient, though the comments at the bottom, while useful, are a little unwieldy
hifi: true, but comments can be removed and only a freetext field left where you can write comments how to get more speed or similiar
hifi: like HOWTO and NOTE at appdb
nightmorph: hmm, i like it
airlied: agd5f: yeah I thought I had most of those cases, it cuold also be DFS/UTS stuff, will track it more tomorrow
AndrewR: hey all ...
AndrewR: not sure if this page still have anything interesting (for developers) ... http://homepage.mac.com/arekkusu/bugs/invariance/index.html (discusses 2d antialiasing with OpenGL. Found it while looking for some tutorials about simples bilinear filtering)
fpoibaf: airlied: wrt bug 22851 how can I do that
fpoibaf: should I attach gdb to the running app
airlied: fpoibaf: over ssh, that should work best
fpoibaf: but how can I run an OpenGL app under ssh?
airlied: export DISPLAY=:0
airlied: and it should run the same
airlied: the problem with most games is they have wrappers
fpoibaf: I'll try
airlied: so its easier to start the game and attach gdb afterwards
fpoibaf: airlied: it runs but I don't see nothing, tried also glxgears
fpoibaf: airlied: I tried it from a Windows machine with putty and from an Ubuntu livecd
airlied: fpoibaf: does glxinfo work?
airlied: make sure the screensaver hasnt kicked in
fpoibaf: airlied: glxinfo works fine
airlied: reports hw rendering?
fpoibaf: airlied: screensaver was not run
airlied: wierd then gears should work fine
fpoibaf: airlied: direct rendering yes
fpoibaf: airlied: sorry, my bad, it works fine, I'll give you the backtrace
fpoibaf: airlied: I simply get this: http://pastebin.ca/1686626
fpoibaf: and the game is stuck now
airlied: can you backtrace?
airlied: is gdb working still
fpoibaf: airlied: http://pastebin.ca/1686627
adamk: cxo: That runs fine, but then ./configure still gives that same error when I run it.
TobiasTheCommie: just wondering, if i want to enable ClockGating on my card, do i have to use an xorg.conf? or is there another way to enable it?
airlied: fpoibaf: I've attached a patch to mesato test
fpoibaf: I'll try it
fpoibaf: airlied: tested it on 7.7 branch: it works! thanks!
fpoibaf: can you push it also to 7.6 branch?
Zajec: TobiasTheCommie: KMS or not?
fpoibaf: airlied: I confirm it works also on the 7_6 branch
airlied: cool I'll figure out th mesa comit sequence
hifi: I did some database design for the "RadeonDB" project, would someone be interested to give comments?
fpoibaf: airlied: if you can, could you then look at this bug: https://bugs.freedesktop.org/show_bug.cgi?id=25179 ? I remeber you fixed a similar bug time ago.
TobiasTheCommie: Zajec: kms
TobiasTheCommie: Zajec: but i falsely thought the xorg.conf was being phased out, it appears i was wrong, so i just created one and added the option to the xorg.conf
airlied: fpoibaf: pushed to 7.6, 7.7 branches, will end up on master when someone pulls it
fpoibaf: airlied: ok, thanks
Zajec: TobiasTheCommie: i think ClockGating is enabled by default on newer hardware (r6xx-r7xx)
Zajec: TobiasTheCommie: for r5xx i think you have to enable this
Zajec: TobiasTheCommie: check parameters on radeon module
Zajec: (modinfo radeon)
Zajec: then modproble radeon param=value(before you have it loaded)
airlied: not sure about the other one right now, probably need hw to reproduce it
TobiasTheCommie: Zajec: i have a rv530, but the radeon module is compiled into the kernel.. and modinfo radeon gives me naught :(
TobiasTheCommie: Zajec: i'll google
TobiasTheCommie: or try
Zajec: TobiasTheCommie: uhhhh, did you compile radeon into kernel?
Zajec: TobiasTheCommie: or some... khem stupid? distribution?
TobiasTheCommie: Zajec: i did
Zajec: ah, so just your mistake :)
Zajec: don't know anything about passing params to in-compiled modules
TobiasTheCommie: you can do that from grub
jcristau: radeon.option=value on the command line
TobiasTheCommie: already pass kms stuff through grub to it
jcristau: /sys/module/radeon/parameters/ has the current values
TobiasTheCommie: ah, thanks
TobiasTheCommie: can i get a list of available values? i mean, is the option going to be called ClockGating for the kernel module as well? or is it lowercase... that is actually really annoying...
TobiasTheCommie: or i could look in the kernel source
airlied: what gpu? its dynclks I think, but really > r600 is probably already on by default
TobiasTheCommie: i just found dynclks in the source
TobiasTheCommie: int radeon_dynclks = -1;
TobiasTheCommie: it's rv530, but i'll add a radeon.dynclks=1 to grub and reboot
TobiasTheCommie: lets see if this breaks everything
TobiasTheCommie: muahahahahhaha
TobiasTheCommie: hm, i get nothing in dmesg about it, though it is set in sys/..... so that looks fine
TobiasTheCommie: though i would love to enable clockgating in xorg, and that still fails... oh well, i'll live
airlied: in kms xorg isn't different
TobiasTheCommie: no no, i'm just being really really stupid
TobiasTheCommie: it hasn't been enabled for r500+ in the driver version i'm using.
TobiasTheCommie: i think
TobiasTheCommie: at least the news from phoronix is less than 7 days old.. bah, ignore me
Zajec: TobiasTheCommie: yeah, disabled by default (do not know why)
Zajec: TobiasTheCommie: and you may need it o r5xx
TobiasTheCommie: well, i put it in my xorg, and the log says
TobiasTheCommie: (WW) RADEON(0): Option "ClockGating" is not used
airlied: if kms x has no effect
TobiasTheCommie: none?
TobiasTheCommie: hm
TobiasTheCommie: hm, just wondering, is there any known gl regression with the kms driver on r500?
TobiasTheCommie: projectm is really slow, but i don't feel confident blaming the driver yet... since i've changed a load of crap since the previous install i had
glisse: kms is lot slower on the 3d side
airlied: glisse: I commented on phoronix article
airlied: didn't enable r3xx/r5xx colortiling for f12
TobiasTheCommie: just disabled kms as a test(and all power savings as well), still slow. but as i said, i don't feel confident blaming the driver, so i'm not doing that.
airlied: which was silly since I think they worked
TobiasTheCommie: btw, i must say that i'm really impressed and happy with the radeon driver, have been using the opensource driver constantly for about a year now... and kms has just made it an even bigger joy to use.
glisse: airlied: also page flipping is likely a great contender for regression on games
Jonathan_L: Yay, it works now! I had to edit /etc/default/grub to include radeon.kms=1 in the LINUX_KERNEL_CMDLINE_DEFAULT variable, and now it works!
Ronis_BR: hi airlied
Ronis_BR: I read MIT license
Ronis_BR: and... I prefer GPL :)
MostAwesomeDude: 'k.
Jonathan_L: Finally no flickering in OpenGL apps, woho! Bye now, got nothing else to say
Jonathan_L: MostAwesomeDude: srzly? I doubt you're as awesome as me. :P
Jonathan_L: gwahahah.
Jonathan_L: Gonna go now, bye
stikonas: MostAwesomeDude: your glxinfo.py madsnippet greps for EXT_texture_non_power_of_two. Shouldn't it grep for ARB_texture_non_power_of_two?
stikonas: that is for OpenGL 2.0 support
MostAwesomeDude: stikonas: You're probably right.
MostAwesomeDude: It seems to match up with what Mesa does, though.
lordheavy: i've configure my MB with only sideport (rs880) instead of sideport+uma and it seem the lockups are gone
mokoloko: Anyone playing quakelive here?
mokoloko: Game starts and looks fine except for being unable to see menus ---> can't join
soreau: Which version of mesa?
mokoloko: Therefor I don't think it's anything major stopping it for working correctly
mokoloko: latest for F12 from koji
soreau: card?
mokoloko: radeon 4850
mokoloko: btw kde4 with effects works perfectly now :) also quake 2k4 and nexuiz are mighty playable
mokoloko: So in quakelive I can join game but only thing I can do there is to spectate
mokoloko: 2d stuff missing
mokoloko: *UT2k4 of course :)
tball: Hello
tball: Anyone know how to change the mesa opengl string?
tball: As far as I know drm is not far from supporting opengl 2.1 on r600+?
tball: So I wonder if Heroes of Newerth (needs opengl 2.1) would run if I changed the opengl sting to 2.1
agd5f: tball: drm is fine. you need mesa from git
tball: agd5f, You mean that current drm supports opengl 2.1?
tball: ahh ok now I udnerstand
tball: mesa is what defines opengl version :P
tball: agd5f, Well do you know if mesa supports opengl 2.1 on r600+?
agd5f: tball: some part of it
lordheavy: opengl 1.5 currently here
glisse: i guess here it's glsl support which is needed
agd5f: tball: to try it you need to uncomment R600_ENABLE_GLSL_TEST in r600_context.c
tball: Could I in theory change the reported opengl version in mesa src code?
agd5f: and rebuild
tball: agd5f, Ok thx.
tball: Do you think there is a change of getting Heroes of Newerth to run?
agd5f: tball: depends what it uses
tball: agd5f, but it tests if I got opengl 2.1 support. So I might need to change the reportet opengl version of mesa
agd5f: tball: just uncomment that line
tball: ahh :)
tball: Understood
hifi: btw. why radeon naming is a bit weird with older chips
agd5f: hifi: how so?
hifi: according to wikipedia I mean
hifi: http://en.wikipedia.org/wiki/Comparison_of_ATI_Graphics_Processing_Units
hifi: I mean, in "Radeon R100" there are boards with code name "RV200" etc
hifi: is the article wrong or how RV200 is related to R100 and so on
hifi: same goes with R200 with RS300 chips in the list
biotube: tball: the code for GLSL has frozen my computer with the glslnoise test, so run it at your own peril
agd5f: those are just internal model numbers. rv200 was a variant of the r100
agd5f: rv100, r100, rv200 are all "r100" variants
agd5f: same with rs100 rs200
hifi: so would it be wrong to categorize rv200 in "R200 series"?
agd5f: hifi: yes
hifi: that makes things a little more complicated...
agd5f: hifi: the page is correct
tball: biotube, Well
hifi: I'm doing a "RadeonDB" project which I mentioned earlier, a database of performance reports on different setups
adamk_: Is there an xorg.conf option to disable coherent_mode? Or is it only possible via xrandr?
tball: biotube, What version of opengl does glxinfo report on your machine?
hifi: agd5f: would you care to give some input?
agd5f: hifi: like what? the names are what they are
hifi: may I pm you?
biotube: tball: with the GLSL code enabled, 2.0
agd5f: hifi: IGP chips almost always don't follow the standard convention. e.g., rs480 is r300 based. rs690 is r4xx based, etc.
agd5f: hifi: sure
agd5f: rs740 is r4xx based
tball: biotube, Ok,, Well Heroes of Newerth might need opengl 2.1
tball: Will opengl 2.1 come with stable mesa 7.8?
biotube: that's the goal for the software renderer
biotube: but likely for r600 as well
biotube: if Mr. Li fixes whatever's wrong with the GLSL code
hnsr: i think HoN needs 2.0
hnsr: but I might be mistaken
hnsr: OpenGL 2.0 / GLSL 1.20 it seems
tball: well
tball: it might actually run then :)
tball: biotube, Is it unstable in general or were the crash caused by the test you ran?
hnsr: I think it also ships the GLSL shaders as plain text so maybe you can apply workarounds for issues :p
biotube: tball: the fslight test works, but glslnoise locks up the GPU(except for the cursor)
biotube: haven't tried anything else
glisse: agd5f, MrCooper you saw the things with macbook & delay issue ?
MrCooper: yeah but I'm not sure how the change can cause that
agd5f: glisse: yeah. may need to blank the crtc when updating the linebuffers is the only thing I can think of
glisse: what puzzling is how the change lead to such delay
glisse: we don't even call atomfunc iirc
tball: biotube, with or without kms?
glisse: thus it's not some kind of delay inside an atomfunc
tball: Have anyone tried the new poermanagement patch for kms?
biotube: tball: with KMS
tball: If I install kernel 2.6.32, radeon-git, mesa-git I should be able to have opengl 2.0 + glsl
tball: is that correctly understood?
tball: maybe libdrm-git
biotube: yes, if you compiled with -DR600_ENABLE_GLSL_TEST in your CFLAGS
tball: ok :)
adamk: Or you uncomment R600_ENABLE_GLSL_TEST in r600_contexct.c
adamk: r600_context.c, rather
lordheavy: OpenGL renderer string: Mesa DRI R600 (RS880 9710) 20090101 TCL DRI2
lordheavy: OpenGL version string: 2.0 Mesa 7.8-devel
lordheavy: nice :-)
lordheavy: adding -DR600_ENABLE_GLSL_TEST in CFLAGS
tlp: 2.0 eh
wirry: Oo
wirry: boots his gaming-rig
rnoland: agd5f: ok, resolved the panic... stupid locking....
biotube: don't forget your debugging rig ;)
wirry: (gaming here means gaming with it, not on it);)
tball: wirry, :D
tball: I really feel this driver is moving faster than fglrx
crespo: is there a command to manage the fan/temperature of the GPU? (fedora 12, HD4850)
biotube: no, but there's a dynamic clocking patch for KMS
agd5f: crespo: not at the moment
crespo: thanks
wirry: dammit, still only 1.4
tball: biotube, how well does that patch work
tball: ?
biotube: tball: dnamic clocks? haven't tried it but the email says it causes minor corruption for a short bit on reclock
tball: Well if thats all, it might be worth a try. I'm on a laptop
ni1s_eee: does radeon DRM provide framebuffer funtionality yet?
evil_core: ni1s_eee: WTF? KMS?
evil_core: radeon with KMS yeas
wirry: hmm, i added -DR600.. to my cflags and reemerged mesa: http://pastebin.com/m4062188f - but what wents wrong?
evil_core: but dunno if its connected to DRM anyway
ni1s_eee: you know, like inteldrmfb
biotube: wirry: it looks right to me
wirry: ähm...i want opengl 2.0 - everything else works (but i have to remove the swrast-file manuelly as its choosen before r600_dri.so)
agd5f: ni1s_eee: yes, radeon kms provides a kernel fb interface
kdekorte: I just booted with the latest DDX from git and I get a working plymouth screen, but X shows up with two blackscreens
wirry: hey gues, stop playing hon, help me instead (i want to play too)
kdekorte: latest patch is 94bcbf that causes this
evil_core: ni1s_eee: see 'modinfo radeon' and enable KMS
agd5f: kdekorte: what patch?
kdekorte: the last one that converts to using dri2
kdekorte: 05551295c5e0946745163f17e5c1d3d41b94bcbf
kdekorte: Can't even switch to console when that patch is in
agd5f: kdekorte: need updated libdrm
kdekorte: as of today?
kdekorte: I have the one from a couple of days ago... if updated drm is needed, note in the log would be nice
agd5f: kdekorte: yeah from yesterday I think
kdekorte: agd5f, I see the compilation error now about the missing symbol...
kdekorte: agd5f, I heard a rumor that r6xx interrupt support was coming soon... any truth to that?
agd5f: kdekorte: got sign off, just waiting on one last thing
kdekorte: agd5f, cool... looking forward to seeing the difference (if any) it makes
wirry: dammit...why doesnt opengl2.0 doesnt work here
kdekorte: My app (fosfor a gstreamer+clutter) video player is working quite well on the current r600 stuff..
Zajec: do you know how wine changes resolution for games?
kdekorte: Zajec, poorly
Zajec: :D
Zajec: i can not play Tony Hawk Pro Skater using wine and not-KMS *and other resolution than default*
Zajec: only resolution 640x480 which is also a resolution of menu works
kdekorte: Zajec, I know that even when I run games in a window, it still messes up my dual screen setup when the game quits
Zajec: and with KMS I can even see menu 640x480
Zajec: *can not*, sory
kdekorte: wirry, did you edit the file or just set the C flag?
kdekorte: I had better luck editing the file than using the C flag
wirry: i tried booth
wirry: same result
wirry: hmmm...emerged motif and one line changed: GLw: yes (was no)
spstarr: is running Mesa 7.8 (trunk) and testing mipmap
Wolfi: hi! how should I compile MESA to get opengl 2 instead of OpenGL version string: 1.5 Mesa 7.8-devel ?
Wolfi: is that possible ?
Wolfi: is this the right channel btw? xD
spstarr: is testing
stikonas: Wolfi: yes, you have to uncomment GLSL_TEST in r600_context.c, but glsl still locks up on many demos
spstarr: OpenGL vendor string: Advanced Micro Devices, Inc.
spstarr: OpenGL renderer string: Mesa DRI R600 (RV635 9591) 20090101 TCL
spstarr: OpenGL version string: 1.5 Mesa 7.8-devel
spstarr: sluggish
stikonas: OpenGL renderer string: Mesa DRI R600 (RV730 9480) 20090101 TCL DRI2
stikonas: OpenGL version string: 2.0 Mesa 7.8-devel
stikonas: OpenGL shading language version string: 1.10
spstarr: how did you enable GLSL?
stikonas: uncomment #define GLSL_TEST in r600_context.c
Wolfi: oh yeah how did you do ? xD
spstarr: ahh its not ready though
stikonas: only some of demos works
Wolfi: kay will try thx
spstarr: turns of VBOs still
spstarr: still broken
hifi: does any web designers hang around here who are also radeon enthusiasts?
spreeuw: hifi: now you make me curious of the follow up question
spreeuw: for
hifi: I'm doing it
hifi: you know what :)
hifi: agd5f helped me out with the database design, it'll be really cool
spreeuw: oh that! nice
spreeuw: design doesnt matter, make it look unsexy
spreeuw: to avoid stereotype gamers
hnsr: pff im a gamer and I like unsexy things!
spreeuw: me too
hnsr: i like my websites unstyled :D
spreeuw: you must be the target audience then
hifi: god, I play counter-strike
hifi: did I win something?
spreeuw: hey so you avoid all the cpu/fps/res debacles by making it a 1-4 subjective perfromance rating
evil_core: hifi: cs always worked under wine
hifi: not under radeon
hifi: (graphically yes, speedwise no)
evil_core: IT works for me
evil_core: q.6nonsteam
evil_core: 1.6*
hifi: on which board
evil_core: on R500
hifi: 100fps all the time?
evil_core: maybe not
evil_core: dunno, I love VSync
evil_core: its not perfectly smooth
evil_core: but its 1600x1200
hifi: heck, I'd love even vsync
Wolfi: OpenGL version string: 2.0 Mesa 7.8-devel
Wolfi: yay! thx
hifi: but 30fps with 800x600 window makes me cry
evil_core: hifi: what card do you have?
hifi: RV570, X1950 Pro
hifi: all git baby
evil_core: kits R500?
evil_core: and what res?
hifi: 800x600, as I said
hifi: resolution doesn't matter
hifi: it's as slow/fast on any resolution
evil_core: you told about 30fps 800x600 ;)
hifi: and it is
evil_core: I dont believe thats not dependant on resolution
hifi: it's some kind of optimization issue
evil_core: DO you use64bit OS?
hifi: osiris looked at some debug traces and found out HL1 engine uses the slowest possible drawing method that the open source driver can do
hifi: I don't know if he had the time to do anything about it, probably not
evil_core: hifi: because by default you gotaccelerated indirect rendering in 32 bit env probably, and you must install 32bit Mesa R300 DRI driver
evil_core: I can check if it works in 800x600
hifi: I do know my way around the driver, it's not a configuration issue
hifi: it might just be that your card is so much faster that the slow code on mesa half isn't slowing you down so much
evil_core: IMHO 30fps is not low
evil_core: I will check it now, and wil ltell you
spstarr: FPS with today's builds is slow
tball: Who is making the powermanagement part in kms?
glisse: tball: who ever is actually writting the code :)
hifi: evil_core: doesn't matter, 30 is too low for me, I want stable and smooth 100fps
hifi: game engine is dependant on fps, the physics are dependant on fps, it's a completely different game at 30fps and 100fps
Wolfi: is ati helping you guys sometimes ? I mean in other ways than just giving out a few pages of their spec ?
hifi: aren't some of the devs working at AMD
glisse: Wolfi: amd has 3 guys working on open source
tball: glisse, I thought there were one specific person on it :P
glisse: tball: it's not like there is hundred of coders and that we need to assign each one a task
Wolfi: but why are they still releasing a proprietary driver ??
Wolfi: it's such a pain in the ass, they could simply give you all their code and boom, ultimate open source driver!
glisse: Wolfi: i am sure their is the answer burried on phoronix forum
glisse: but to sumup they can't release source of fglrx
Wolfi: if they help mesa, it means that they aren't against open source at all, so what the point in doing two drivers instead of one open source ?
glisse: and they want to have fglrx to support workstation market
Neo_The_User: glisse: do you work for AMD?
spreeuw: hmm interesting I did modprobe radeon modeset=1 and then my screen went corrupt
spreeuw: console hung kernel too
glisse: Neo_The_User: no i work for Red Hat
evil_core: hifi: how to display FPS? ;)
Wolfi: lol is my question a sort of taboo ?
hifi: evil_core: net_graph 3
tball: Wolfi, Well fglrx is mainly for workstations
glisse: Wolfi: no it's not, just that we keep answering it again and again :) i guess we should write it down on the wiki
agd5f: Wolfi: the closed source driver is shared across OSes.
tball: tball: I am aware of that ;) But I just read about this guys, who made the powermanagement patch. But I don't remember his name
Neo_The_User: does anybody know if it would be possible to implement physx technology in the open source radeon drivers?
Wolfi: yeah all right sorry then xD
Neo_The_User: or at least in fglrx?
agd5f: Neo_The_User: sure
agd5f: one could do in open source
Neo_The_User: i hope nvidia releases the physx stuff :)
Wolfi: is physx a common technology or simply a gadget today ?
Neo_The_User: would be fantastic for need for speed shift
Neo_The_User: its a common technology used in all the games i have
Neo_The_User: need for speed shift, etc.
Wolfi: will look for the answer on wikipedia don't bother answering xD
yangman: Wolfi: the API for PhysX is proprietary
tball: Neo_The_User, well I guess it doesn't matter if wine doesn't support physx
Neo_The_User: ah true
Neo_The_User: it would be nice if the open source driver would work in windows
Neo_The_User: i dont think the windows kernel supports the DDX though
Neo_The_User: how do you spell noeavu again? google isnt correcting my spelling since im spelling it so wrong.
k]t: Linux NET 2.6.32-rc2 #2 SMP Wed Nov 25 07:50:30 PST 2009 x86_64 GNU/Linux
spreeuw: is dri2proto-2.1.tar.bz2 new enough?
Neo_The_User: speeuw: for what?
hnsr: nouveau
Neo_The_User: thanks!
k]t: cxo: im lost now... =\
hnsr: im really bad at spelling but somehow i manage to remember that
spreeuw: building KMS Neo_The_User
Neo_The_User: should be
Neo_The_User: checks
spreeuw: seems to be the last release
spreeuw: I dont get build errors
spreeuw: but somehow it wont work for me yet
Neo_The_User: yes
spreeuw: must be looking over something obvious
cxo: k]t, huh? I wasn't following, i just stepped in now
k]t: cxo: im the guy from yesterday... 3d ~ 3d rendering with radeonhd driver ~ ... you recommended ...
k]t: exec -o uname -a
k]t: Linux NET 2.6.32-rc2 #2 SMP Wed Nov 25 07:50:30 PST 2009 x86_64 GNU/Linux
k]t: and now....
k]t: gasp
spreeuw: (note theres an rc8 now)
cxo: k]t, dmesg | grep -i drm
k]t: exec -o dmesg | grep -i drm
k]t: [ 0.417078] [drm] Initialized drm 1.1.0 20060810
k]t: hates irssi
cxo: dont paste it here
k]t: sorry
cxo: use pastebin or something
k]t: say thats it... drm initialized....
cxo: (btw, you obviously need to build libdrm, mesa and xf86-video-ati as well ) you dont need radeonhd
evil_core: hifi:dont got 100FPS, how big your sources are.Changes in kernel(2.6.32) are needed?
spreeuw: the wiki says another build order
k]t: cxo: actually i have no clue how this turned into a conversaion, i was just saying gasp...
spreeuw: xserver, libdrm (with special flags), xf86-video-ati, mesa
spreeuw: and the wiki is silent about the needed kernel options
cxo: if you're on a new distro, you might not need to build xserver
k]t: i _was_ gasping because i have no nearly no modules, and boot sequence needs reconfiguration...
cxo: did you make a new initrd?
spreeuw: it would be helpful if someone idiotproofed the kms wiki a bit more
spreeuw: checks at the relevant stages
cxo: you have to assume some sort of kernel building experience, you cant teach everything
k]t: yep... thats what took soooooooo long... i realized i had no mkinitrd... so i found *yaird*... that seemed troublesome then eventually i realized that a million+ people _have_ the script =\
cxo: which distro are you on?
spreeuw: cxo: doh, but how about mentioning the required stuff for radeon
k]t: debian squeeze/testing
cxo: sounds new enough
spreeuw: hard kkernel cfg options
spreeuw: no descriptions in words
cxo: mkinitramfs i think its called on debian
cxo: so do, mkinitramfs -o /boot/initrd-2.6.32-rc2 2.6.32-rc2
k]t: ok, theres something wrong with git://git.freedesktop.org/git/mesa/mesa.... empty repo?
cxo: or something like that
k]t: checking
k]t: yep, mkinitramfs....
cxo: you obviously, did make modules_install as well?
Neo_The_User: k]t its anongit not git
cxo: git clone git://anongit.freedesktop.org/git/mesa/mesa
Neo_The_User: exactly
k]t: i did... make defconfig , make, make modules ; make modules_install ; make install ; initrd -o blah blah ; reboot
cxo: no thats not going to work
k]t: oh yea, i had to adjust lilo too =P
Neo_The_User: lilo? ewww!
k]t: hahaha
cxo: start by using the kconfig from your distro
k]t: kconfig?
k]t: as in kde?
cxo: cp /boot/config-(whatever came with ubuntu) .config
k]t: say... ahh!
cxo: then, make menuconfig
cxo: and configure the kernel for your machine a little
k]t: what about new features...
spreeuw: CONFIG_DRM=m
spreeuw: CONFIG_DRM_KMS_HELPER=m
spreeuw: CONFIG_DRM_TTM=m
spreeuw: CONFIG_DRM_RADEON=m
k]t: oh... sorry, you answered that ^^
spreeuw: is that enough?
cxo: firmware
Neo_The_User: yup
Neo_The_User: firmware blobs in kernel binary
cxo: this helps too, CONFIG_FIRMWARE_IN_KERNEL=y
bjaglin: i was asking here yesterday what could cause "[dri] RADEONDRIGetVersion failed because of a version mismatch" when using KMS, and it turned out to be some kind of initialization order in Karmic, since reloading the module after boot as explained there is working. Has anyone encountered the same problem on another distro?
k]t: well, could i just run make menuconfig from top dir again to modify the config?
spreeuw: bjaglin: interesting, I had this too
k]t: .config*
cxo: k]t, better start with your distro's kconfig first
cxo: defconfig these days on linux is almost useless
k]t: kernel.. not kde... got it...
k]t: hahaha
spreeuw: whats wrong if modeset=1 garbles your screen?
spreeuw: in console
Neo_The_User: spreeuw: you mean the text?
spreeuw: and in xorg
k]t: ok...
spreeuw: no it squishes my screen a nit
cxo: then kms is wonky
spreeuw: and puts blocks of color in
spreeuw: memory areas or so
spreeuw: cxo: is there anything thta would influence this?
spreeuw: kernel cfg wise
cxo: i think someone else had the same problem, i forgot what the solution was
spreeuw: or is it a hardware combination?
bjaglin: spreeuw: on karmic as well?
spreeuw: no on agnostic
spreeuw: unaffiliated
cxo: spreeuw, might as well update everything and try it again, build drm_radeon_kms statically
spreeuw: you mean built in?
cxo: yes
spreeuw: and then use the kernelflags to enforce kms?
cxo: no, just let it boot, the default is enable
spreeuw: ok will try that
cxo: oh now i remember!
cxo: the modules on your box are being loaded out of order
spreeuw: first drm?
cxo: yeah
cxo: just build that config statically and that should fix that
spreeuw: it seems to have worked
spreeuw: I just modprobed form within xorg
cxo: ok
spreeuw: screen went into some memory corruption mode
spreeuw: but after a fs frefresh it seems to work
spreeuw: hmm crash
spreeuw: do i need any of the FB_ options?
spreeuw: FB_RADEON etc
spreeuw: FIRMWARE_EDID?
spreeuw: TILEBLITTING?
cxo: radeonfb is deprecated
cxo: its been replaced with kms+radeon
cxo: brb
spreeuw: ok
k]t: libdrm is with drm from git right?
k]t: included with...
cxo: git clone git://anongit.freedesktop.org/git/mesa/drm
k]t: yep, i got it.
cxo: ./configure --prefix=/usr --enable-radeon-experimental-api --disable-intel
k]t: just waiting for kernel to rebuild, then reboot...
k]t: api or ati?
spreeuw: api
spreeuw: ./configure --help
cxo: you have to be a little more proactive k]t
spreeuw: cxo: does KMS have any kernel flags of itself?
spreeuw: config variables I mean
spreeuw: it seems to be magically always on
spreeuw: always built I mean
cxo: probably, i'm not sure, check the source for params
k]t: ok, sry, about that.
spreeuw: nowhere is it explicitly mentioned in make menuconfig
k]t: well i have an autogen.sh script anyway...
spreeuw: the word KMS appears nowhere
spreeuw: except for some helper lib
spreeuw: argh helper mod
spreeuw: building everything in now
spreeuw: FB master group option can not be unselected
k]t: so video-ati instead of radeonhd?
spreeuw: video-ati is the usual one yes
k]t: and in xorg.conf in device section i assume i will still use: Driver "radeonhd"
spreeuw: no radeon
spreeuw: or ati
k]t: ok.
spreeuw: leave all other options on autoprobe
k]t: nick k[t
spreeuw: for the videocard driver
k[t: i probably won't be able to... i want to setup 3 screenz
k[t: screens*
spreeuw: lol awesome console now
spreeuw: it initializes properly now I built it all in the kernel
spreeuw: only xorg ddx gives me stripes now, will try a few other versions
agd5f: spreeuw: your ddx needs to be built with kms support
k[t: pwd
spreeuw: agd5f: it was reported at configuretime
spreeuw: I think my problem was a too old drm
spreeuw: libdrm I mean
spreeuw: switching between xorg and console is real smooth now with kms
tsamolotoff: Hello guys!
tsamolotoff: What's on KWin effects being broken?
tsamolotoff: http://bugs.freedesktop.org/show_bug.cgi?id=24131 - Is it fixed or not?
evil_core: hi, mine IM client has been killed, and I didnt saw the answers
spreeuw: use your logs
spreeuw: (II) RADEON(0): Direct rendering enabled
spreeuw: (EE) GLX: could not load software renderer
spreeuw: why doenst it pick r600
spreeuw: I moved swrast out of the way
spreeuw: nearly there ;p
spreeuw: HUH?! now its working somehow
spreeuw: seems aiglx has to be enabled
cxo: make sure mesa linked against drm properly
spreeuw: else it flals back to software
spreeuw: cxo: yeah must be somekind of buildorder or ld issues
spreeuw: it sortof works now
spreeuw: glxgears behaves erratic lol
spreeuw: cogs go back and forth
cxo: ahaha
cxo: wishes his did that
cxo: I get about 18,000 FPS without compiz and about 11,000 FPS with compiz enabled
cxo: (2.6ghz X3, HD4870)
evil_core: spreeuw: nothing in logs, but state showed that I got 2 direct messages(from lost terminal), but dunno which channel, and if it was irc ;)
spreeuw: sweet rebuilt mesa too now
spreeuw: and it works
evil_core: cxo: because its direct vs indirect
evil_core: compositing sucks
evil_core: especially memory ;)
spreeuw: yeah its sortof unnecessary too
spreeuw: I'd prefer some edge in games instead of a toy desktop
evil_core: WindowMaker rulez, the only problem is that it segfault while alt_tabing from wine fullscreen games
evil_core: but its not radeon problem, there were patches for that causing another problems :P
Julia_: so
evil_core: wepaes flickers in XAA, and some corruptions still occurs
evil_core: hifi: can you share your xorg.conf?
Julia_: I get low fps using radeon drivers on my laptop only when there is almost no CPU usage. At the moment I am running virtualbox and installing some software and I get 40+ fps
evil_core: lol
Julia_: Is it because of some laptop power saving feature?
evil_core: hmm..can be
Julia_: How can I make that I get normal fps all the time?
evil_core: but only on battery?
spreeuw: lookup freq scaling
Julia_: Please help me!
evil_core: is it Thinkpad/Asus or some Acer?
spreeuw: echo performance | tee /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
Julia_: HP
Julia_: with radeon x1200
spreeuw: it will cut batterylife
Julia_: and amd dualcore processor
evil_core: dunno if HP bioses are broken, but this is stupid that GPU power depends on CPU usage
spreeuw: and for the radeon modules theres a frequency scaling option too
Julia_: It is not because of BIOS. On WIndows everything was fine
spreeuw: evil_core: maybe mesa fallbacks?
evil_core: wtf? Athlon in Laptop is not smart decision, they are overheating
evil_core: spreeuw: so much impact?
kdekorte: is anongit down at freedesktop?
airlied_: fd.o just went offline
Julia_: How can I configure it?
evil_core: Julia_: on AC/battery makes difference?
Julia_: no
evil_core: spreeuw: maybe tell her to first dump values from proc, and then change them later
evil_core: Julia_: and what soft are you talking about? I hope its not glxgears or something like that
Julia_: no
Julia_: desktop with compiz
Julia_: evil_core,
evil_core: disable compiz first
evil_core: it can be caused by vblank
evil_core: I wouldnt mind if its constant 40fps, you will never see more than 24
Julia_: evil_core, I need compiz
evil_core: If it doent cause game lags, then why bother?
Julia_: What game lags?!
Julia_: Radeon drivers give me ~18 fps for desktop
evil_core: whats the problem, only "fps"?
Julia_: simple desktop with compiz
Julia_: I could only dream about games
Julia_: 18 fps for desktop is so lame
Julia_: When I minimize windows or scroll through webpages I see them lag
Julia_: that is the problem
kdekorte: Julia_, have you tried it with metacity or kwm instead of compiz?
Julia_: It is not an option.
Julia_: I need compiz
kdekorte: Need? seriously?
Julia_: and I get good fps with higher cpu usage
evil_core: Julia_: do you knoe how to set up xorg.conf? I can give you mine if you would be able to select options from it, but I definitely dont have time today to play with your.
evil_core: so maybe its pure software
Julia_: Only way?
Julia_: I don't like messing with corg.comf
evil_core: maye paste glxinfo, xdriinfo and log file
Julia_: I like when X configures it automatically
Julia_: but fine
kdekorte: if you're gonna paste use pastebin
spreeuw: nice nexuiz runs
Julia_: umm maybe other day. thx
kdekorte: Julia_, I don't use an xorg.conf either
MostAwesomeDude: Julia_: Pastebin 'LIBGL_DEBUG=verbose glxinfo'
evil_core: Julia_: maybe in simplicity its indirect
Julia_: it is late I have to get up early tomorrow
spreeuw: this week I get the 4670 wonder how fast it is in nexuiz compared to the 3450
spreeuw: need to compensate driver speed with hardware a bit
evil_core: spreeuw: R500?
spreeuw: rv730
spreeuw: or 720
spreeuw: biggest I could find wih passive cooling atm
evil_core: so its osmething definitely newer, hoped mine card can be quick too :/
spreeuw: evil_core: i'd get the fastest chip supported by an open driver if I were you
spreeuw: if you dont mind noisy coolers you can probably get faster than mine
spreeuw: they arent that expensive anymore
spreeuw: 75 euro with shipping
DanaG: yay, I got my firegl to call itself a firegl again... but it doesn't make any difference for Linux currently.
DanaG: ANd perhaps it might not ever make such a difference between Radeon and FireGL of same GPU.
spreeuw: thats r100 right?
spreeuw: or do you mean closed driver?
evil_core: spreeuw: its thinkpad, so I cant
DanaG: I mean, between what would be "M76" and "M76GL" in Windows... though my GPU is M86 (GL).
spstarr: Lenovo?
spstarr: W500? :)
evil_core: no, noew thinkpad sucks
spstarr: has a FireGL RV635
evil_core: its T60p
spstarr: oh
evil_core: last FlexView(IPS), 4:3 and UXGA
spreeuw: evil_core: if it uses system ram its going to be dogslow
evil_core: but its a bit upgraded, T7200, 4GB DDR3 and 400GB SpinPoint M7
DanaG: wishes he had a magic antigrav thing... then he could've gotten the 8730w instead of 8530w. Oh, but they don't let you get DreamColor LCD with ATI GPU.
spreeuw: tried a few radeon igps
evil_core: 500GB*
DanaG: Is there any REAL reason not to use RV635 with a 24-bit internal LCD?
evil_core: I mostlu run ZSNES, DosBox and Fizzwizzle
DanaG: Or does HP just want more money from people wanting the good LCD, by demanding they upgrade the GPU?
evil_core: so I would happy even with Intel
spstarr: im back to the Intel GPU mode again
spstarr: radeon too slow for 3D until they optimize the paths
spreeuw: DanaG: video quality should be good enough by both
spreeuw: image quality that is
DanaG: hmm, then I wonder why HP DEMANDS nv if you select DreamColor LCD. Artificial limitation?
DanaG: (Dreamcolor is full color depth lcd, plus rgb LED backlight.)
spreeuw: maybe they tuned them to eachother
spreeuw: but a monitor shouldnt care much what its fed digitally
spreeuw: it can only be good or not good
spreeuw: the tft itself is tuned of it self
spreeuw: by internal chips
DanaG: hmm, I hope some day (down the line, when I can afford it), I'd like to get such a laptop with that sort of display, and then swap to ATI myself.
DanaG: http://hpfansite.com/hp-elitebook/nvidia-quadro-fx3700m-update-do-not/
DanaG: hmm, they may be doing something funky there.
airlied: does displayport happy dance
Droste: when is the youtube-video to be expected? :-D
amarks: is ddccontrol meant to work with kms
amarks: Probing for available monitorsradeon_open: mmap failed: Invalid argument
flyback: is restoring one of these for his dedicated data recovery box http://www.xmdell.com.cn/index.php?gOo=goodspic.dwt&goodsid=683 probably the last major project before his mind shutsdown for good
cxo: laughs as the gayness
cxo: s/as/at
cxo: is restoring this http://tinyurl.com/yawqgpe the first of many major projects before I become an old man
flyback: why is it gay?
flyback: it's to pull bits off drives that can't be read normally
flyback: I wanted a rock solid chipset known for solid ide and pci
flyback: 440bx was it
flyback: speed means jack shit
flyback: when you get say 2bits/hr etd
flyback: etc
flyback: very nice indeed
flyback: this data is as priceless as that old car
flyback: it's lost contact info for friends online
cxo: associates gayness with nerdyness
flyback: 10 yrs I been waiting till I either had all the money for recovery or could do it myself
flyback: now I can
flyback: I have the tools and knownhow to do anything to a drive except platter swaps
flyback: actually I can do platter swaps but only single platter drives
cxo: why does the number of platters matter?
cxo: isnt it all one spindle with control shims in between?
flyback: yes
flyback: but if you release the screws
flyback: and the platters slide 1 atom out of alignment
flyback: not even the nsa will gets hit off
flyback: you need a special tool to clamp all the platters so you can release them from the spindle
cxo: oh so they need to be "balanced"
flyback: well locked in position
flyback: else they will slide
flyback: because the signals are written in colums
flyback: down thru the platters
flyback: what is that a jag?
flyback: bwm?
flyback: sorry my brain has more bitrot than my old hd's
flyback: almost 36 I doubt I will see 40
flyback: almost guarantee it
cxo: Only the good die young
flyback: yeah so true
flyback: I always played by the rules
flyback: and I got fucked anyways
flyback: oh well
airlied: wonders where the topic is
MostAwesomeDude: airlied: ^^^
flyback: start with this one
flyback: DO NOT move onto the other videos till you watch this one, all the others build
flyback: http://video.google.com/videoplay?docid=-396122032591841175&ei=swAOS7WuMaG8qgKC4s3zBg&q=scott+moulton&hl=en
MostAwesomeDude: Well, it's at the top of *my* screen, anyway.
flyback: just in case you are curious cxo
soreau: wonders why airlied has permitted this for as long as he already has
airlied: soreau: was in the kitchen
soreau: airlied: I meant the days pasty
flyback: doesn't give a fuck, he's a dead man anywyas
cxo: feels sorry for the dead that flyback
Droste: when building mesa i get a warning: http://pastebin.com/m2787e870 am i missing something? because the build runs fine and the result is working (for now...)
cxo: If it almost broke, don't fix it
cxo: Looks like an artifact of your distro's toolchain
Droste: is it bad? :-D i think im just going to ignore it :-)
Droste: I was concerned about the "depend" in makedepend. but it seems that it works without ioccom.h :-)
cxo: Droste, it has nothing to do with mesa. Its just how your toolchain works
cxo: I think its trying to get all dependencies resolved, even if it doesnt need them all
cxo: but its just warning you, so dont worry about it
Droste: good
k]t: might be going from x86-64 to i486
JohnDoe_71Rus: morning