[newb]MIki: http://ubuntuforums.org/showthread.php?t=840858&highlight=7500 how do i do that last post?
lymeca: Does this driver support all R500-based cards such as my X1650?
[newb]MIki: what?
lymeca: Was that a hard question for some reason?
lymeca: I have a Radeon X1650 and I run Debian and I want to know if this card is supported by the radeon driver
[newb]MIki: nvmnd
MrCooper: lymeca: yes, the current version in sid supports it
lymeca: MrCooper: Is the sid version the latest version?
lymeca: 6.9.0?
MrCooper: yes
Zajec2: z3ro: agd5f told me that ddcprobe can be used only for LCD (LVDS for sure, don't know about DVI) - it means i can no use this for VGA-0
Zajec2: z3ro: that mean i still need valgrind-mmt
Jyzygzy-: there is somekind of problem with R300 and StepMania game
Jyzygzy-: there is this certain effect the game makes that raises the CPU usage trought the roof and it shows up as a huge lag spike
MostAwesomeDude: Jyzygzy-: Prolly a fallback of some sorts.
xsacha: hi
xsacha: the 6.9.0 driver has full compositing support for IGPs? what do i have to update? mesa? X?
osiris_: xsacha: yes. X and drm, and for indirect rendering mesa
xsacha: k thanks, the whole shebang.. what version of X?
Zajec2: Oh, btw... Can I run 6.9.0 using X Server 1.4.0.90? Or do I need something newer?
xsacha: thats what im wondering
Jyzygzy-: MostAwesomeDude is there anything i can do about that?
xsacha: thanx heaps for the IGP fix! (i think david?) pretty sad it was just 3 lines this whole time
MrCooper: Zajec2: yeah, that should work
Zajec2: MrCooper: ok, thanks
MostAwesomeDude: Jyzygzy-: I'll look into it tomorrow.
osiris_: xsacha: check the git log to see when was last exa related commit, and pick the version, but if you are going to get X from git just pick the git head
Jyzygzy-: MostAwesomeDude maybe i could provide you with some logs?
MostAwesomeDude: Jyzygzy-: Sure. Stuff like WARN ONCE or FALLBACK would be important.
Jyzygzy-: hmm so whats the logfile where these errors would be written?
xsacha: hey all of the sudden gitweb (freedesktop) is telling me 403 forbidden - no project
xsacha: in projects dir
osiris_: xsacha: yeap, try cgit
xsacha: oh a new one?
xsacha: i dont like this one.. cant sort by last commit?
xsacha: mmm but i see some IGP commits after 6.9.0 :(
xsacha: reverts something
xsacha: k thx guys
Jyzygzy-: MostAwesomeDude where do i look these warnings?
MostAwesomeDude: Jyzygzy-: The console that you ran stepmania from.
Jyzygzy-: oh ok
Jyzygzy-: ill go produce some logs
Jyzygzy-: hmm well when i start the game i get few fallbacks
Jyzygzy-: but when the actual lagging is happening there is no errors in the console
Jyzygzy-: File r300_render.c function r300Fallback line 472
Jyzygzy-: Software fallback:ctx->Point.SmoothFlag
Jyzygzy-: its like WARN ONCE so maybe its not warning me again when its happening on the actual gameplay situation?
adamk: Jyzygzy-, Use 'driconf' to "disable low-impact fallback"
Jyzygzy-: ok
Jyzygzy-: adamk i disabled thing called "SmoothLines" from the Stepmania Graphics settings
adamk: That would probably help too, I think.
Jyzygzy-: ok ill go try
Jyzygzy-: adamk well there is no errors anymore and the thing still lags
adamk: Well I've never tried StepMania, so I can't really say how well it should perform.
Jyzygzy-: with fglrx it works but those drivers are horribly unstable with R300
Jyzygzy-: imma try different versions :\
Jyzygzy-: would be really nice to get it work with radeon drivers cause they are way more stable
Jyzygzy-: adamk ohh yeaa! the latest CVS of Stepmania beta works! ^-^
Jyzygzy-: now im happy happy happy
adamk: Ahh, cool.
Jyzygzy-: indeed ^-^
[newb]MIki: Is there anyone that wants to take over my pc to fix it? i got a 7500 mobility radeon card in ths laptop
surfer: damn it... i updated to the latest version of the repository Xorg libs/server and it's seg faulting >:|
surfer: should i rebuild everything with debugging turned on and do a BT?
mcgreg: how is it seg-faulting?
mcgreg: it crashes here too.not directly but when it started into login nscreen and I press whatever button on my keyboard it freezes
surfer: well, all i can "see" from my workstation is the screen going blank
surfer: but when i run Xorg by itself from an ssh session it does that
surfer: this
surfer: Unblank CRTC 0 success
surfer: Blank CRTC 1 success
surfer: Disable CRTC 1 success
surfer: Output 51 disable success
surfer: Blank CRTC 0 success
surfer: Disable CRTC 0 success
surfer: Blank CRTC 1 success
surfer: Disable CRTC 1 success
surfer: Segmentation fault
surfer: i don't hvae debugging turned in so gdb is not helpful
surfer: should i rebuild everything with -O -g?
glisse: osiris_: if you got fglrx stipple line dump i might give it a look
osiris_: glisse: ok, will send you an email
osiris_: glisse: sent
glisse: osiris_: you double check that it pickup proper program with stipple ?
glisse: looks strange that it doesn't use texkill
spstarr_work: everyone loves stipples
osiris_: glisse: double checked right now. only tex lookup instruction
osiris_: glisse: maybe my revenge test is bad: http://pastebin.com/d4f43e238
surfer: the latest x tools + xf86-video-ati = seg fault
glisse: osiris_: do you see stippled line ?
osiris_: glisse: no, the tests are too fast. but I wrote simple program that uses stipple lines (pretty similar to that in revenge test) and it works on fglrx
glisse: osiris_: i will do dump with other tools than revenge
osiris_: glisse: ok. one more thing, maybe its important: I made dumps on my rs690, so it doesn't have tcl unit. maybe stipple lines are done differently on that hardware.
glisse: i don't think so
nha: so which vertex attributes are sent by fglrx?
nha: I recall you said that it doesn't use texturing? Or just no kill instruction?
osiris_: nha: no kill instructions, only texture lookup
nha: have you looked at the texture specification? does it perhaps include a chroma key?
nha: or anything else that looks suspicious? ;)
osiris_: nha: no chroma key, but fglrx uses bits that aren't described in r300 docs in TX_FILTER1(0x4440) reg
nha: osiris_: what is that register set to?
osiris_: nha: a0300000
nha: those values are related to anisotropic filtering optimizations (i.e. quality vs. performance tradeoff)
nha: they're not in r300_reg.h yet, but they shouldn't matter
nha: actually, I'm not sure I've seen bit 21 before
glisse: nha: yup bit 21 is suspicious
glisse: everythings should stop at bit 14
osiris_: texture size is 64x16
osiris_: everything seems normal
nha: it would be interesting to capture the texture data
nha: unfortunately, I don't know enough about revenge to do that :/
osiris_: z3ro: ping
nha: osiris_: come to think of it, did you check for blend modes and other stuff that they might use to remove the pixels?
nha: osiris_: technically, the OpenGL spec says that the stippled-away fragments are killed, but perhaps they're trying to use blending instead when they can because it's faster, or something
glisse: nha: other solution they do stipple in software and accelerate it only on professional card
glisse: nvidia is doing such kind of things
glisse: even if hw got the capability they don't use it on gamer card
nha: so you mean the dumps we're seeing might in fact be bogus... could be.
MostAwesomeDude: Morning, guys. Um, bit 21 on filter_1 is the aniso threshold, I thought.
glisse: the dump we see are just other rendering that stipple line but i don'tthink so, gb_enable is programmed to stuff things
nha: I though that was 19:17
osiris_: glisse: if they do stipple lines in software, what are GA_LINE_STIPPLE_(CONFIG|VALUE) regs for?
nha: anyway, off to buy food
MostAwesomeDude: It's one of those things that we don't set because there's no real reason to set it. Maybe the fglrx people know something we don't about texture optimization.
nha: MostAwesomeDude: I believe it's simply one of those quality vs. performance tradeoffs
MostAwesomeDude: nha: Precisely.
MostAwesomeDude: nha: Ooh, have you checked out my fogcoord code at all?
nha: i.e. setting the threshold higher decreases the number of pixels where multiple samples are taken, thus increasing performance while reducing the image quality a little
nha: MostAwesomeDude: where is that?
osiris_: nha: which regs should I check for blend modes?
glisse: osiris_: what i mean is that they use this register only on firegl card
glisse: even though normal card does have the register
glisse: but i don't think this is the case here
MostAwesomeDude: nha: http://gitweb.freedesktop.org/?p=users/csimpson/mesa.git;a=summary
MostAwesomeDude: Pull either radeon-fog-coords or master.
MostAwesomeDude: It's getting there, but there's still a lot of stuff to finish figuring out.
MostAwesomeDude: I'm splitting secondary color so RGB is the color and A is the fog.
MostAwesomeDude: has put the channel to sleep with talk of fog coords
osiris_: hehe
MostAwesomeDude: osiris_: No joke. It's like the one thing that nobody wants to talk about.
MostAwesomeDude: There have been patches floating around for this since 2005.
MostAwesomeDude: (None of the patches really worked, but that's not the point.)
osiris_: MostAwesomeDude: maybe glisse, and nha never have touched that area, so they can't comment.
osiris_: or maybe it's just too boring :P
MostAwesomeDude: osiris_: The latter. :3
glisse: MostAwesomeDude: what about the poor alpha channel ? :)
glisse: you killed him ?
osiris_: nha: actually blend modes, are probably the good shot. fglrx sets RB3D_CNTL to discard src pixel if alpha = 0
MostAwesomeDude: glisse: If I understand correctly, secondary color's alpha is not used in blending.
MostAwesomeDude: If I misunderstood, then I'll have to stuff textures.
glisse: MostAwesomeDude: i would need to reread spec but i think secondary alpha can be used
glisse: would have
MostAwesomeDude: glisse: I'm going off of Aapo's posts from ~2006.
MostAwesomeDude: But, y'know, whatever. There's actually a fog-to-tex stuffer in the GA/GB I could use.
z3ro: osiris_: pong
MostAwesomeDude: The real problem is getting the fog out of VAP correctly.
MostAwesomeDude: Or, no, wait. Now I'm being silly. I have an idea.
osiris_: z3ro: does revenge allows texture dumping? if not, would it be hard to implement that feature?
z3ro: osiris_: it's not completely implemented yet, but most of the framework is there...
z3ro: I've got a patch sitting around somewhere...
z3ro: http://pastebin.ca/raw/1059177
z3ro: note thats completely untested.
z3ro: and there is some other crap in the patch, too.
osiris_: z3ro: don't know if you haven't sent it to you yet, if not: here is gl_stipple_line test for revenge: http://pastebin.com/d4f43e238
z3ro: osiris_: ok. I'll merge them into my local tree.
z3ro: then I really should make a new gpg and ssh key...
z3ro: (I jsut woke up :P)
osiris_: z3ro: will tx dump work if fglrx uses gart texturing?
z3ro: osiris_: it should. it's using the same memory functions as everything else. so if reading the RB and IB's work, then textures should too.
osiris_: z3ro: ok, thanks
z3ro: no prob
z3ro: just keep in mind it is untesed, so it might not work if I screwed something up. :)
surfer: any idea why the latest ati driver causes X to segfault?
surfer: radeon hd works OK.
surfer: it's hard to tell exactly what's going on... my screen just goes into power save
surfer: and won't come back without a reboot
surfer: hmm.. maybe i'll try removing the option "dpms" from my config
osiris_: ok, I got to go. I will play tomorrow with blend modes. nha, glisse if you have any further comments on stipples, write it anyway. I'll read backlog
osiris_: bye
surfer: i think there's something in my config causing the radeon driver to disable my "working" display
surfer: Output TMDS2 setup success
surfer: Output 51 enable success
surfer: Enable CRTC 0 success
surfer: Unblank CRTC 0 success
surfer: Blank CRTC 1 success
surfer: Disable CRTC 1 success
surfer: Output 51 disable success
surfer: Blank CRTC 0 success
nha: MostAwesomeDude: cool, will look at that stuff eventually
surfer: Disable CRTC 0 success
surfer: Blank CRTC 1 success
surfer: Disable CRTC 1 success
surfer: but it's shutting off my LCD screen (on DVI-0)
nha: MostAwesomeDude: though figuring out wth is wrong with shadowtex is higher up on my todo list
nha: MostAwesomeDude: maybe I'll skim your code tonight
nha: mind you, I know next to nothing about the fog specs
MostAwesomeDude: nha: Neither do I.
MostAwesomeDude: :3
MostAwesomeDude: Right now I'm trying to squish arb_point_params.
nha: MostAwesomeDude: when working on stuff like this, I always find it useful to work with definite test cases, sometimes based on stuff from Mesa, sometimes created myself from reading the specs
nha: that's why I started piglit, after all
nha: when I wrote the tests, I check them against software rendering Mesa; that's usually a good indicator whether I actually understood the spec correctly
nha: then I set about implementing things
MostAwesomeDude: nha: Well, point_params is in nouveau, intel, and r200, so I'm just borrowing from them for this.
MostAwesomeDude: Fog coords, though, are kind of a thing that requires the entire pipeline to work differently.
nha: woah... shadowtex misrenders the shadow texture in an interesting way on R500
MostAwesomeDude: Hmm, is it correct for non-AA points to be drawn as squares on Radeon?
stefanb: wonders what 1D textures are. A 1-pixel texture? :-)
nha: MostAwesomeDude: I believe so
nha: stefanb: just what the name says: one-dimensional textures
Zajec2: z3ro: ping
nha: i.e. instead of two relevant texture coordinates, there's only one (the S coordinate)
Zajec2: z3ro: i need valgrind-mmt, ddcprobe is not enought
nha: MostAwesomeDude: OpenGL 2.1 specificationl, page 97
nha: is computed from the vertex’s xw and yw , and a square grid of the odd width cen-
nha: tered at (x, y) defines the centers of the rasterized fragments (recall that fragment
nha: that's for the non-AA case
nha: MostAwesomeDude: did you get my reply wrt non-AA points?
MostAwesomeDude: And I checked against the SW-rendered version; it works as expected.
MostAwesomeDude: nha: Yes.
spreeuw: MostAwesomeDude: I recall that it's some workaround
spreeuw: r200 has that too
spreeuw: gl points used to be the wrong size without it
MostAwesomeDude: Ooh, much faster.
MostAwesomeDude: spreeuw: Yeah, it's basically just another way to set point size right now.
MostAwesomeDude: But it DOES work.
MostAwesomeDude: Stuff like attenuation is cool, but doesn't work well/at all without the AA.
MostAwesomeDude: And the smoothing.
xorITor: airlied, last week you mentioned in here you were going to push an update for xorg through f9... would that by any way include updates to support the x1950pro? I don't want to push you to releasing just wondering if i should get my hopes up or not ;-)
MostAwesomeDude: xorITor: I'
MostAwesomeDude: xorITor: I'm pretty sure he pushed Mesa 7.1rc1 and DRM 1.29, so you should have support.
xorITor: MostAwesomeDude, thats most awesome
xorITor: sorry couldn't resist
xorITor: :-D
MostAwesomeDude: xorITor: They never can.
xorITor: hangs his head
MostAwesomeDude: And then they learn my name is Corbin, and the next two hours are "Corbeen Daalas MOULTIPAAAS"
xorITor: oh no!
xorITor: i won't do that one i promise
MostAwesomeDude: 'k.
xorITor: i just got an ati x1950pro so thats why i was wondering
xorITor: just like 20 min ago
MostAwesomeDude: xorITor: Ooh, nice choice.
spreeuw: xorITor: cool, does it work?
MostAwesomeDude: I've got an X1950 XT chilling here. Works very nicely.
spreeuw: I saw an x1650 with passive cooling
xorITor: i was going for an x1950xtx but it was sold by the time i got there
xorITor: no 3d so far
spreeuw: on phoronix they recommend the x1650 now
spreeuw: it should have 3d and 2d
spreeuw: but it's a bit unclear whether at the same time
xorITor: spreeuw, the 1650?
spreeuw: and wether xv is supported
xorITor: or the 1950?
spreeuw: sounds a bit like it's in r300's state 2 years ago
xorITor: searches
nha: MostAwesomeDude: btw, you don't happen to be related to the guy who draws Ozy&Millie?
MostAwesomeDude: nha: Wow, that's the first time I've ever gotten that.
MostAwesomeDude: D.C. Simpson, right?
mcgreg: very stzrange, xorg is uisng 69% of my cpu , doing nothing :/
xorITor: heh
MostAwesomeDude: He's up in Washington. I've wanted to meet him at some point, but never had the chance.
nha: MostAwesomeDude: yep. I randomly noticed that yesterday
xorITor: mcgreg, have you been watching flash video?
spreeuw: lol mcgreg dump all useless crap
xorITor: look for a *.bin process
xorITor: sometimes they "hang around" after
spreeuw: jikes
nha: I've updated the Piglit results page: http://people.freedesktop.org/~nh/piglit/results/all/problems.html
mcgreg: omg, it is at >80% now
nha: the r500 has some catching up to do compared to r300 ;)
nha: though r300 fails texrect-many because of the inefficient texture indirection handling
xorITor: expected as the docs have not been out near as long though right?
nha: actually, R500 docs have been out longer, but R300 had been reverse engineered for a much longer time
MostAwesomeDude: And there's not a lot that the R500 series does differently.
xorITor: oh yea
MostAwesomeDude: It's just the little things.
nha: yep, I think all the difference is in fragment programs
nha: except possibly the blendFunc fail
MostAwesomeDude: nha: And RS setup. And flow control registers.
MostAwesomeDude: And AA.
xorITor: is there any way i maybe able to help with testing or something?
MostAwesomeDude: We think.
MostAwesomeDude: xorITor: Sure. Pull the F9 updates, try stuff.
nha: plus random minor stuff ;)
xorITor: MostAwesomeDude, try that daily ;-)
MostAwesomeDude: xorITor: http://wiki.x.org/wiki/RadeonProgram
mcgreg: hmm anybody has an idea what "[ 4883.119769] Clocksource tsc unstable (delta = 62504656 ns)" is?
xorITor: cool
MostAwesomeDude: Love it, use it, update it if it's wrong.
MostAwesomeDude: mcgreg: Not something to worry about. Your RTC is probably being emulated by onboard HPET.
MostAwesomeDude: It just means that the kernel does some different stuff to handle it. Just informative.
mcgreg: I thought it has somehting to do with cpu
nha: I love how the RadeonFeature matrix has "Faster 2D (EXA) .... SLOW (for R500 RHD)"
spreeuw: hey put enemy territory to platinum
spreeuw: for r300 and 400
spreeuw: and also quake3 really
spreeuw: also for r200
spreeuw: I cant edit those pages
xorITor: cool i have an hd3650 at home and now the x1950pro here and i just pulled an x1050 out
xorITor: odd that my x1050 showed up as an x300 though
agd5f: xorITor: it's an r3xx core
spreeuw: nice
spreeuw: is it faster than an X850?
spreeuw: seems X850 is the card to buy still
agd5f: spreeuw: no, x850 is faster
spreeuw: but it's hard to get a hold of
agd5f: spreeuw: x1950 is probably the fastest card available with full 3D support
spreeuw: oh
xorITor: only i have no 3d support on mine with the latest f9 stuff
xorITor: wait
xorITor: unless it is in "testing"
spreeuw: probably is
agd5f: xorITor: it's in testing
xorITor: aaah
xorITor: bonks his forhead
spreeuw: rpm distros are horrible for testing stuff
xorITor: s/or/ore/
spreeuw: judging my brief centos experience
z3ro: Zajec2: pong
xorITor: heh
xorITor: i didnt pull down the testing stuff
z3ro: Zajec2: http://pastebin.ca/raw/1059306
spreeuw: xorITor: lemme know how fast it is in practise
stefanb: MostAwesomeDude: Thanks for pointing out the connection to one of my favorite movies (The Fifth Element) :-)
z3ro: Zajec2: note it's untested, but it should compile at lest.
spreeuw: what games do you have for testing?
xorITor: this 1950?
spreeuw: yes
xorITor: sure thing
xorITor: so far it seems good and i am using exa
xorITor: but have not gotten 3d working yet
xorITor: all i should need is the xorg-x11-drv-ati right?
spreeuw: no you need recent mesa
spreeuw: mesa has the 3d drivers
spreeuw: and the 3d libs
xorITor: aah and the glx-utils or whatever its called too?
xorITor: heh
xorITor: so "most" of it ;-)
stefanb: MostAwesomeDude: Lost your Multipass? :-)
spreeuw: glx utils sounds complementary
xorITor: MostAwesomeDude, sorry i made you deal with that
stefanb: xorITor: Are you trying this on F9?
xorITor: MostAwesomeDude, if your ever in austin texas ill buy you a beer to make up for it
xorITor: stefanb, yep
xorITor: im rsyncing over the testing stuff now
stefanb: xorITor: My recommendation: download libdrm, mesa, xorg-server SRPMS. Unpack libdrm, update source archive to git master, build, install. Build radeon & drm kernel module from libdrm source, install. CLEANUP. Unpack mesa, update source archive to git master, build, install. CLEANUP. Rebuild xorg-server SRPM against new libdrm-devel, mesa-libGL-devel, etc., Install. Build xf86-video-ati git master, install
stefanb: xorITor: Works like a charm for me.
spreeuw: I run handbuild last xorg server
stefanb: xorITor: After that you only need to update libdrm, mesa or driver if you pull changes with git
xorITor: stefanb, im giong to save those directions off incase the testing stuff does not work out of the box
spreeuw: with git ati mesa and drm
spreeuw: ontop
stefanb: xorITor: Unfortunately libdrm, mesa & xorg-server SPECS & patches need small updates as the source code structure has been changed or patches have already been upstreamed
xorITor: stefanb, you mean from whats in f9 testing?
stefanb: I don't use F9 testing. Here are my updated SRPMs: libdrm-2.4.0-0.13.0.fc9.src.rpm, mesa-7.1-0.33.0.fc9.src.rpm, xorg-x11-server-1.4.99.902-3.20080612.0.fc9.src.rpm
stefanb: xorITor: I added the ".0" to show the difference between the original SRPM and mine
stefanb: xorITor: Oh I forgot, I also updated libpciaccess-0.10-2.0.fc9.src.rpm. Don't remember the exact reason anymore, but it was needed
xorITor: hmm
xorITor: thats interesting, thanks for that
stefanb: xorITor: basically I have what is in F9 updates but with the git master source code inside libdrm, mesa & libpciaccess
stefanb: and thanks to the awesome work of Corbin, Alex & Dave, compiz now rock&rolls smoothly on my Lenovo T60 with RV515
nha: MostAwesomeDude: from a quick glance, tunneling fogcoord in secondary color seems okay
nha: MostAwesomeDude: however, calling r300SetFogState from the fragment program compiler is definitely incorrect
nha: I don't understand all the fog issues, but I believe it needs to be called basically where fragment programs are configured in r300_state.c
nha: man, the way state changes are handled with fog is pretty different from how everything else is done
nha: weee, point parameters :)
Zajec2: z3ro: thanks, i'l' try it
Zajec2: z3ro: it's for 3.3.1, right?
z3ro: Zajec2: yeah
mcgreg: hmm I wish xserver gut would work here again so I could test new mesa , damn it
mcgreg: *git
lymeca: I am correct in my understanding that my R600-level Radeon Mobility 2600 HD will have *NO* 3D support with this driver?
lymeca: Is it possible to at least code 3D support for R600 cards now that AMD has released the GPU docs for R600? Or is the 3D programming guide NECESSARY? I was under the impression that the guide adds no new information but just makes it easier. Is this correct?
Magnade: lymeca: 2d on the r600 is done via the 3d core
Magnade: lymeca: id assume the 3d guide would be more necessary than say for r500
Magnade: the fglrx drivers should work at least for the short term
airlied: they haven't released the 3d docs for r600 yet.
lymeca: That's what I thought.
lymeca: Someone was trying to convince me that this announcement: http://www.phoronix.com/scan.php?page=news_item&px=NjUyMA
lymeca: meant 3D info as well
airlied: that has some 3D info in it, but not enough to write a driver
lymeca: airlied: Okay that you very much for clearing that up. Any idea when the 3D info for R600 will be released? Sometime in the next month?
airlied: hopefully.
airlied: xorITor: yes updates-testing for F9.
airlied: xorITor: just mesa from that.
xorITor: airlied, hey... thanks a ton!
xorITor: any idea why it would try to pull this "Retrieving http://www.w3.org/TR/html4/loose.dtd>" from "rpm -Fhv mesa-*.rpm"?
xorITor: aah
xorITor: i know
xorITor: never mind
xorITor: something freaked out a download and saved an html document instead of an rpm
xorITor: thanks... time to test it
surfer: sup... any idea why xf86-video-ati immediately puts my monitor into permanent power save? until i reboot?
surfer: radeonhd still works
airlied: surfer: pastebin a logfile?
xorITor: ok well... no joy on the 1950
surfer: http://www.webaccel.net/surfer/xorg.log
airlied: xorITor: what kernel you running?
xorITor: 2.6.25.6-55.fc9.x86_64 do i need a newer one?
xorITor: ie one from testing?
airlied: should be fine..
airlied: xorITor: I have it on my x1950 here.. pastebin you're xorg log file
xorITor: anything else i can give you?
xorITor: k
airlied: maybe try a later -ati from updates testing
xorITor: i have xorg-x11-drv-ati-6.8.0-14.fc9.x86_64
xorITor: could it be the x86_64 not working?
airlied: xorITor: shouldn't be.. granted I didn't give x86-64 a lot of testing.
xorITor: :-D
xorITor: thats why we are here right?
surfer: were u able to check out my log?
airlied: surfer: can't see much wrong in it though.
surfer: me either
surfer: same thing happens with no conf
surfer: too
airlied: xorITor: okay I'm running ati -16, mesa-libGL 7.1.0.33
xorITor: getting the -16 now
xorITor: thx for that
xorITor: (EE) AIGLX error: Calling driver entry point failed(EE) AIGLX: reverting to software rendering
airlied: surfer: is that with latest -git tree?
xorITor: bet that is what is causing that EE
xorITor: ;-)
airlied: xorITor: maybe getting a latest server is worth it :)
xorITor: ok another test
Zajec2: a
surfer: yes
surfer: airlied: everything is from git
airlied: surfer: did it ever work?
surfer: i didn't have this problem when i updated everything 1 week ago
surfer: git 1 week ago + xf86-video-ati = OK
surfer: Blank CRTC 1 success
surfer: Disable CRTC 1 success
surfer: giving up.
surfer: xinit: Connection refused (errno 111): unable to connect to X server
surfer: xinit: No such process (errno 3): Server error.
airlied: surfer: oh cool... so we broke it since then.. care to git bisect?
surfer: chekc it out
surfer: when i run Xorg by itself
surfer: Blank CRTC 1 success
surfer: Disable CRTC 1 success
surfer: Segmentation fault
surfer: i'll try to bisect
surfer: or maybe i should rebuild the x server with debugging turned on
surfer: and backtrace?
airlied: try bisecting first.. it might just be a driver issue.
surfer: hmm
surfer: i went back 10 days ago
surfer: so far everything has failed
surfer: does this help narrow it down? i had been updating the driver and mesa every day
surfer: the xorg-server from 1 week ago was working
surfer: the current xorg-server does not work with current driver
surfer: i think i should be bisecting the server instead of the driver
surfer: cus with the current xorg server even the 10 day old driver doesn't work
surfer: but the current server works with current radeonhd
airlied: wierd.. no surewhat could have failed on the server.. getting a backtrace would be useful then.
surfer: i'll rebuild it with -O -g
surfer: Program terminated with signal SIGSEGV, Segmentation fault.
surfer: The program no longer exists.
surfer: (gdb) bt
surfer: No stack.
surfer: airlied: i get a weird kernel message
surfer: maybe it's a kernel bug
surfer: airlied: http://www.webaccel.net/surfer/kernel.log
airlied: surfer: what kernel?
airlied: that is wierd..
surfer: the xserver from 6/24 passed the test
surfer: 2.6.26-rc8-zen1
airlied: surfer: it would be nice if the kernel had a useful backtrace
surfer: i dont think i have debugging turned on in the kenrel
surfer: maybe i should turn it on 8)
xorITor: airlied, well the updated stuff did just the trick
xorITor: airlied, great work by the way
xorITor: great work to all of you
xorITor: i will try to throw in some testing in the upcoming days
xorITor: ;-)
xorITor: now its time to go home ... thanks again all and good night
surfer: airlied: i think it has something to do with my kernel and drm
MostAwesomeDude: nha: Backlog. Um, WTF fog coords I found some pretty big issues with what I was doing, so I'm going to rewrite it later.
surfer: current xserver and driver work OK with 2.6.25
airlied: surfer: it might be my fault :), I hope I didn't screw up the upstream kernel too badly :)
surfer: the drm module i tested was built a few weeks ago
surfer: so i'm testing 2.6.25 with current drm
airlied: you mentioned 2.6.26-rc8-zen1..
airlied: whats the zen1?
surfer: zen is a project that adds experimental things to the linux kernel that hasn't yet been approved in linus' tree
surfer: kind of lke mm
surfer: zen-sources.org
surfer: airlied: x will start with 2.6.26-rc8-zen1, but only if the drm module is not present
surfer: when drm is present... crash
surfer: i'm going to bisect the kernel now
surfer: hmm
surfer: i narrowed it down
surfer: it's a kernel change between 2.6.26-rc6-zen1 and 2.6.28-rc8-zen1
Gantex: try running 2.6.26-rc8 without the zen1 patchs
Gantex: patches*
spstarr: looks at git
spstarr: there seems to be a problem with EXA's UploadToScreen or DownloadToScreen (?)
spstarr: airlied?
spstarr: i get corrupt background when a window is minimized sometimes
spstarr: but it can correct itself after sometime
spstarr: this only is visible on the root window
spstarr: i have Option "AccelDFS" "true" enabled
spstarr: DownloadFromScreen
agd5f: spstarr: try disabling DFS
spstarr: agd5f: some issues with that?
spstarr: agd5f: anyone noticing GPU wedges after long hours of use randomly?
spstarr: it's happened twice with composite enabled only
agd5f: spstarr: probably unrelated
agd5f: spstarr: there are occasional corruptions with DFS and small pixmaps
spstarr: that might explain it since its small pixmaps
spstarr: agd5f: is the corruption known? or the cause?
agd5f: spstarr some sort of coherency issue with small transfers
spstarr: I see
spstarr: its more of an irritant than anything, as its only on the root
agd5f: AGP can also be problematic with DFS which is why it's disabled by default on AGP
spstarr: I notice my root window has actually 'shifted' sideways
spstarr: part of the wallpaper is wrapped around
agd5f: spstarr: try disabling it
spstarr: but doesnt it speed things up ? :)
agd5f: DFS accelerates pixmap migration out of vram
agd5f: in a well accelerated driver you hopefully won't hit it very often
spstarr: hmm, ok
spstarr: im just going to track git and build periodically now that the .spec i have is tailored for git
spstarr: agd5f: ah, it seems windows that are left without moving for a long time will show this corruption (which explains why my konsole and xchat) windows show this when moving/minimizing them after a long while