Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2008-11-06

Search This Log:

jnoah1984: hmm, I would go to #ubuntu and see if you can get help. I apologize, I'm not sure how to help.
Hilton: bows and thanks all.
Hilton: Still no freeze; I wonder if somewhere in playing around it got fixed unknowingly?
jnoah1984: Hilton, stay in here and idle though, someone else may be able to help.
jnoah1984: Hilton, It is possible. I've done that a few times myself back in my Ubuntu days.
Hilton: Actually I can't, I have to free up this machine for others. :P Good night, all
jnoah1984: Curious though, what have you done to remedy the problem?
Hilton: jnoah1984: I hardly remember. It was a week ago and I don't know what some the commands I was given did ;)
jnoah1984: Lol, ok. Good luck
Hilton: no sooner do I quit, and go to shut down Ubuntu, than it finally freezes
Hilton: now these are some interesting log entries.
Hilton: It would be nice if they put timestamps in these logs...
Hilton: So all these lines start with [mi] rather than (EE) (II) (WW) etc.
Hilton: [mi] EQ overflowing. The server is probably stuckin an infinite loop.
Hilton: there are dozens (hundreds?) of those lines, interspersed with this every 2nd or 3rd line:
Hilton: [mi] mieqEnequeue: out-of-order valuator event; dropping
Hilton: Ah hah! Found my exact problem: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=462590
Hilton: (except on Debian)
jnoah1984: Ubuntu is based off of Debian, well, loosely now.
jnoah1984: So currently no fix I guess...
Hilton: Someone suggested dropping from AGP8x to 4x or 1x to this guy.
jnoah1984: interesting. Worth a try.
spstarr_coding: Hilton: r3xx?
Hilton: There are several bugs listed in Ubuntu relating to this now that I've found the mieqEnequeue thing.
Hilton: spstarr_coding: Yes.
spstarr_coding: Hilton: known
spstarr_coding: I get it too :)
spstarr_coding: Hilton: there are issues with r3xx and composite
spstarr_coding: VT switch with composite on == boom
Hilton: spstarr_coding: Is there a workaround I can use?
spstarr_coding: don't use composite
Hilton: You've lost me, man.
Hilton: *noob*
spstarr_coding: no desktop effects
Hilton: wonders if he should dig up and install his old Diamond Monster Fusion 3D. Mmm, 3dfx Banshee Goodness.
Hilton: 16mb..yeah
spstarr_coding: if I do use them, I canno VT switch
Hilton: Hold on, rebooting Ub.... where do I need to go?
spstarr_coding: if I use kernel mode setting, i get other crashes relating to AGP and issues the radeon devs haven't had time look at yet, bug AMD about :)
spstarr_coding: Hilton: if in GNOME, turn off desktop effects
spstarr_coding: if in KDE, turn off kwin composite
spstarr_coding: turns off brain
Hilton: tries to find where to turn off desktop effects
Hilton: So, why does it all work fine for anywhere from 1-45 minutes then suddenly stop?
Hilton: ah. Visual Effects: None.
jnoah1984: FYI to all 8.542 works with r3xx hardware. https://a248.e.akamai.net/f/674/9206/0/www2.ati.com/drivers/linux/catalyst_810_linux.html
jnoah1984: It's just difficult to find a patch that allows it to work with 2.6.27*
jnoah1984: All that is needed is a correctly configured kernel I guess
z3ro: jnoah1984: the easiest way would be to just use an older kernel.
jnoah1984: That's what I am doing
z3ro: thats my reason (that and I'm too lazy to update the kernel often ;)
rx__: fglrx?
z3ro: yeah, I usually run an older kernel so I can play with revenge.
z3ro: I was going to look at doing some r6xx hacking (probably just play with vert/frag programs) but the current DRM doesn't work.
z3ro: suppose I could still work on a disassembler, but it's more fun to draw tris. :)
rx__: just use fglrx from ubuntu repo :>
z3ro: I tried it, and it broke when trying to run my engine.
rx__: oh
z3ro: I didn't do very much debugging, but it seemed to just give invalid operation on any draw calls.
rx__: suck
z3ro: also I'm pretty sure the new fglrx does some weird optimizations that make dumping harder...
MostAwesomeDude: Hm, think that I've managed to get this to work.
MostAwesomeDude: Manually running ebuilds =/= fun.
z3ro: why did you need to manually run it?
jnoah1984: You have fglrx with xserver 1.5?
jnoah1984: z3ro: he's on debian
z3ro: oh, right.
jnoah1984: he used ebuilds that I patched last night
MostAwesomeDude: z3ro: This mobo has massive issues with anything but Debian+XFCE. Blame Foxconn. :3
rx__: :/
jnoah1984: mobo?
z3ro: hmm thats weird.
MostAwesomeDude: Quite.
jnoah1984: lol, nvm
MostAwesomeDude: What's the best repo to grab revenge from?
rx__: should try a revenge dump on the newest fglrx
rx__: :)
rx__: i second that question ;)
z3ro: MostAwesomeDude: mine is the most up to date one.
z3ro: iirc airlied cloned one a while back, which hasn't been updated.
MostAwesomeDude: rx__: My list this run: Each of the fogcoord tests in Mesa, and arbocclude.
jnoah1984: http://dri.freedesktop.org/wiki/ReverseEngineering
z3ro: yep, git://people.freedesktop.org/~z3ro/revenge is the git url.
z3ro: btw if anyone has questions/patches/suggestions, I'm happy to answer those.
z3ro: I did try to make revenge pretty friendly to use, but it still has some rough edges
z3ro: I should make some time and add pciid based card detection... that would kill the whole --pci-e/--agp mess.
rx__: z3ro; this should be run from a fresh boot right?
z3ro: rx__: it shouldn't matter really. the more important thing is you're not running compiz or something, and not moving the window all over the place or anything.
z3ro: eg best to just set it going, and wait until it finishes. should just take a few mins.
rx__: brb
MostAwesomeDude: BRB, adventure time. :3
rx__: hm.. it's been a while since i last ran this
rx__: i thought it said stuff like "ok created tarball xyz.tar.bz2" at the end
rx__: or maybe that was one of my requests ;)
rx__: anyone want a tarball?
jnoah1984: tarball of?
rx__: x1600 mobility revenge dump
rx__: on latest fglrx
z3ro: rx__: if you like, email it to me.
z3ro: I might finally write some good analysis tools someday. :P
jnoah1984: sorry, I'm looking for r{3/4/5}xx stuff
rx__: x1600 is r5xx
jnoah1984: lol, ya, just realized that. Sure
MostAwesomeDude: Well, DRI doesn't work. One out of two.
jnoah1984: yo, MostAwesomeDude. rx__ has got r5xx revenge dump from the latest fglrx
MostAwesomeDude: jnoah1984: Awesome.
MostAwesomeDude: should figure out whether revenge already has all the needed tests
jnoah1984: MostAwesomeDude, I had the some problem. Couldn't get dri to work. What kernel/xserver/fglrx are you using?
rx__: sent to z3ro, jnoah1984
jnoah1984: thanks
z3ro: MostAwesomeDude: for fogcoords? it doesn't
z3ro: but you can add them. ;D
MostAwesomeDude: z3ro: Or ARB occlusion.
MostAwesomeDude: But I can add -- yeah. :3
z3ro: also no on ARB occlusion.
rx__: MostAwesomeDude; if you got a patch :)
MostAwesomeDude: z3ro: Hm? How so?
MostAwesomeDude: I know I can't scope the GART memory, but I wanna make sure I'm doing everything else right.
z3ro: MostAwesomeDude: I mean revenge doesn't have a test for ARB occlusion yet.
MostAwesomeDude: z3ro: Oh. Yeah, I know. I shall write patches.
z3ro: and revenge can dump GART memory btw.
MostAwesomeDude: z3ro: Really? That'd be pretty awesome. Assuming, of course, that fglrx uses GART for the OQ.
jnoah1984: OQ?
z3ro: thats the whole mess in revenge_memory.c ;)
z3ro: hehe compare the pci gart code to the agp code. =/
z3ro: jnoah1984: occlusion query
jnoah1984: lol should have figured...
jnoah1984: k, well my turn to try some stuff, brb
jnoah1984: back, ok, now attempting to make the fglrx-8.542 driver that I will be using with xserver 1.4.2
rx__: :)
jnoah1984: 8.542 works with xserver 1.5?
rx__: 8.54.3
jnoah1984: I know, but, ah, nvm... I'm sticking with what I said
jnoah1984: ati-drivers-8.542 is not compatible with RCU Preemption (bug #223281), please disable it
jnoah1984: need to rebuild kernel... :(
jnoah1984: so close yet so far...still
MostAwesomeDude: jnoah1984: http://r300.cvs.sourceforge.net/viewvc/*checkout*/r300/glxtest/pretty_print_command_stream.tcl?revision=1.8
jnoah1984: that is?
MostAwesomeDude: That'll turn dumps into register/value pairs.
jnoah1984: ah, k
jnoah1984: n00b question. How do I use that?
MostAwesomeDude: Cross-reference with the register guides.
MostAwesomeDude: And you'll have a very nice map of what fglrx does.
MostAwesomeDude: It's not that hard. I did it before with texture registers.
z3ro: jnoah1984: and pretty_print_cmd_stream works like: ./pretty_print < revenge_dump_file > human_readable_dump
jnoah1984: ah, k, thanks
MostAwesomeDude: It's actually pretty sweet.
MostAwesomeDude: (Of course, the AMD guys have big GUI debugger toolkits. :3 )
jnoah1984: Ok, so, tclsh?
z3ro: MostAwesomeDude: yeah, it would be cool to have a nice gui dump reader
jnoah1984: where do I get and what is it?
z3ro: add some diff features, cross ref with pdf docs...
z3ro: although I don't know anything about gui programming (eg gtk etc)
rx__: hm.. nifty pretty print
MostAwesomeDude: jnoah1984: $ sudo aptitude install tcl
rx__: emerge tcl !
jnoah1984: lol ya, emerge
MostAwesomeDude: XD, sorry. Debian's corrupting me.
airlied: not much use in gui readers for that stuff, like really if you are looking at reg dumps you usually know what you are looking for :)
jnoah1984: I eixed for tclsh...oh well
airlied: write a plugin for wireshark :)
airlied: that decodes radeon CP.
jnoah1984: lol
airlied: I've contemplated it before
jnoah1984: Interesting idea...
z3ro: uhh wireshark is a network dumper?
z3ro: although I guess it could work... it's just designed for IP packets.
z3ro: it would be cool to at least have a text-based "smart" diff though. eg there are some regs we can ignore (display base when comparing two dumps etc)
jnoah1984: ya, but it gives you a framework/gui to work with
z3ro: and maybe try and pull register names out of the pdf files.
jnoah1984: or combine revenge with pretty print and slap on a gui
jnoah1984: or just have it feed out pretty HTML and view in a webpage after a bit of gathering...
MostAwesomeDude: z3ro: I don't know TCL, or I'd totally improve the dumper.
MostAwesomeDude: I should go back over the packet format, and write some Python to do it.
z3ro: MostAwesomeDude: well, I have a version of it written in C... but it's kind of ugly
MostAwesomeDude: z3ro, rx__: http://pastebin.ca/1246650 is occlusion query revenge patch.
MostAwesomeDude: You can keep the copyright, I don't think I wanna be associated with such bad C. :3
jnoah1984: MostAwesomeDude, just look at it, it's not that confusing at all. Actually it seems fairly simple
z3ro: MostAwesomeDude: hmm it looks alright.
MostAwesomeDude: z3ro: Compile-tested only; my fglrx doesn't wanna do DRI right now.
MostAwesomeDude: Hm, I should now port some fogcoord stuff.
rx__: k
jnoah1984: brb, restarting with new kernel
rx__: doesn't apply here :/
MostAwesomeDude: rx__: I'm on the git head. Just checked out. :c
rx__: 7c8ec47a?
rx__: hmm..
MostAwesomeDude: Yep.
rx__: confused..
jnoah1984: brb
z3ro: MostAwesomeDude: ok, I added the patch to my local tree, but I'll wait for it to be tested before pushing.
z3ro: I'd test it myself, but I'm on r6xx, so it's a bit hard to interpret the results.
MostAwesomeDude: z3ro: If it runs without crashing Mesa, I'd say it's doing its job. :3
rx__: patch -p1 doesn't like your patch :/
MostAwesomeDude: And if the OGL version is 1.5 or greater, then it should run.
MostAwesomeDude: rx__: Did you get a raw? The syntax coloring that pastebin puts on can confuse patch and diff.
rx__: yea raw
MostAwesomeDude: rx__: Out of curiosity, does git-status/git-diff indicate that your tree's clean? :3
rx__: nm
rx__: figured it
rx__: the file was in DOS mode or some crap
MostAwesomeDude: That's...weird.
z3ro: pastebin might convert it
z3ro: weird though it applied fine for me (got a warning about the newline, though)
jnoah1984: [fglrx:firegl_cail_init] *ERROR* CAIL: AGP acquire failed, error -19
jnoah1984: any ideas? No matter which version of fglrx I use, I keep getting this error! Google is no help either... :(
z3ro: jnoah1984: are you sure you've got the right kernel config options enabled?
jnoah1984: z3ro: Pretty sure, do you have a link or something? I had one, but lost it. I am working off of memory...
z3ro: jnoah1984: just the gentoo ebuild (but if you're buliding on gentoo it will automatically check anyway)
jnoah1984: airlied, any idea?
rx__: there we go
rx__: patched
jnoah1984: z3ro, I used the ebuild, so I guess I have configured it correctly...
rx__: back in a jiffy
z3ro: MostAwesomeDude: small bug with your patch, you have to remember to reset the gl state afterwards. eg glDepthMask (GL_FALSE) has to be reset to true so it doesn't effect subsequent tests. :)
MostAwesomeDude: z3ro: Ooh, I didn't realize I was sharing GL state with other tests.
z3ro: I'll fix it on my commit.
MostAwesomeDude: 'k. Thanks.
z3ro: MostAwesomeDude: yeah, afaik there isn't any easy way to clear the state between each test automatically.
z3ro: I mean, other than bringing down the gl context
MostAwesomeDude: z3ro: I was, for some reason, thinking that you spawned one context per test.
MostAwesomeDude: should read the code he's patching :3
z3ro: MostAwesomeDude: yeah that actually might be a better way, but I think I'd have to ditch SDL in favor of glx.
MostAwesomeDude: z3ro: Yeah, probably.
z3ro: otherwise the window gets destroyed everytime you destroy the context
z3ro: means lots of flickering windows all over the place.
MostAwesomeDude: Right.
MostAwesomeDude: That's what piglit does. :3
z3ro: yeah it's kind of nasty if you want to abort the program for whatever reason.
MostAwesomeDude: SDL *is* awfully convenient, though.
z3ro: yeah. thats why I used it initially when I just wanted to get something up and running quickly.
z3ro: but I think it's time to replace it now... clearing the state completely between each test seems a lot cleaner imho.
MostAwesomeDude: Hm. I think I'm just going to port over progs/tests/fog from the Mesa tree, since that one *always* breaks.
MostAwesomeDude: Yeah, well, it does help ensure that we get a clean run on each test.
rx__: alright
rx__: got a dump w/occlusion query
rx__: who wants?
MostAwesomeDude: rx__: Ooh, me!
MostAwesomeDude: Actually, I'm probably the only one that cares.
rx__: haha
rx__: wait.. is there something about a bug?
z3ro: rx__: it won't matter for the occlusion query test.
rx__: alright
z3ro: just it could effect tests ran after the OQ test.
rx__: sent
rx__: crashes for the night
MostAwesomeDude: rx__: Thanks.
jnoah1984: hopefully final time. brb with a fixed up kernel... If not, I will just ask others to dump for me... :( No pun intended...
MostAwesomeDude: Hm. That can't be right. 4f5c is never written...
jnoah1984: wtf, I DID IT!!! 2 days of work...
jnoah1984: kernel-2.6.26-r2 xorg-server-1.4.2 fglrx-8.542
jnoah1984: ok, now for revenge!
MostAwesomeDude: Pun intended. :3
jnoah1984: lol that time, pun intended
z3ro: hehe sorry, I just got the image of jnoah1984 going out and buying a shovel and a shotgun. ;)
jnoah1984: lol
MostAwesomeDude: XFD
MostAwesomeDude: Knowing what jnoah1984 looks like makes that so much funnier. I fell out of my chair. :c
jnoah1984: lol
z3ro: haha :P
z3ro: X Error of failed request: BadMatch (invalid parameter attributes)
z3ro: doh! my glx foo is not strong.
MostAwesomeDude: z3ro: I'm gonna just write the fog stuff on top of the occlusion query patch, so it might need some TLC.
jnoah1984: lol
z3ro: MostAwesomeDude: ok
jnoah1984: wow, didn't realize that you z3ro wrote revenge
jnoah1984: I received a segment fault...It told me to report to you
z3ro: hmm shouldn't segfault unless something went really wrong.
jnoah1984: lol
z3ro: how are you running it: ./revenge --agp ?
jnoah1984: I just ran: sudo ./regenge.sh
jnoah1984: I was hoping it would give me some output on how to use it
jnoah1984: but since you are author, how do I use it?
z3ro: well it's strange it segfaulted... possible something got detected wrong...
z3ro: can you run: ./src/revenge --debug --agp &> debug.txt
z3ro: then pastebin debug.txt
jnoah1984: I tried with --agp and got the same thing
jnoah1984: sh: line 1: 8056 Segmentation fault fglrxinfo -v 2> /dev/null > fglrxinfo.txt
jnoah1984: sh: line 1: 8058 Segmentation fault glxinfo -v 2> /dev/null > glxinfo.txt
MrCooper: glisse: what do you mean by 'exa padd width with 1'?
z3ro: I mean run the binary directly, not revenge.sh
z3ro: ./src/revenge --debug --agp &> debug.txt
jnoah1984: I did
jnoah1984: That's what the debug.txt gave
z3ro: ah ok
z3ro: hmm thats weird, but you can try opening revenge_test.c, and look for dump_device_id ();
z3ro: comment that whole block (eg dump_device_id to dump_lspci)
z3ro: then recompile and see what happens.
z3ro: looks like it's segfaulting on the glxinfo calls for some reason.
jnoah1984: lol I think I might know why... module/kernel mismatch
z3ro: oh, yeah, if the gl driver segfaults then probably glxinfo will too.
z3ro: try running just glxgears by itself and see if it works.
jnoah1984: glxgears does work
z3ro: hmm strange. I'm not sure why it's segfaulting then...
jnoah1984: 27017 frames in 5.0 seconds = 5403.339 FPS
MostAwesomeDude: http://pastebin.ca/1246682 is the fog test.
MostAwesomeDude: Pretty much shamelessly copied from Mesa. Compiles fine, and source should be self-explanatory.
MostAwesomeDude: Also, I made sure to reset fog at the end. :3
z3ro: cool. I'll add it in a bit. I'm just trying to bash glx into shape now. :P
MostAwesomeDude: Icky. Good luck.
jnoah1984: z3ro, I tried the commenting out and recompiling. This time, it segfaulted, but no output to debug.txt
z3ro: jnoah1984: if you've got gdb installed it could tell you whats happening.
jnoah1984: I need to emerge it, but I was going to need it sooner or later. emergeing...
jnoah1984: Looks like it's going to be compiling for a while. I'm going to power nap. Be back in 15 minutes. (It's 0215 for me)
z3ro: yeah gdb is quite big.
jnoah1984: MostAwesomeDude or anyone, you know how long gdb takes compared to gcc?
jnoah1984: faster or longer?
z3ro: jnoah1984: it would depend on your cpu speed a lot... but I don't remember how long it takes in relation to gcc.
z3ro: I guess 15min would be enough.
jnoah1984: done
z3ro: ok, you can run it like: gdb --args ./src/revenge --agp --debug
z3ro: then type "run" at the gdb prompt
z3ro: once it dies, type "bt"
jnoah1984: ya, thanks, I've used gdb before though.
z3ro: ok, sorry, wasn't sure.
jnoah1984: no worries, appreciate the help either way
jnoah1984: z3ro: http://pastebin.com/m754a9fd6
z3ro: hmm ok it fails to read the agp memory... so either it's detecting the agp base wrong, or the kernel has that memory protection thing enabled (I forgot about that before)
z3ro: do you see anything in dmesg
z3ro: I forget exactly what it says but something about memory protection
jnoah1984: [ 1146.851792] revenge[12762]: segfault at 204 ip b7efc65d sp bf9fe350 error 4 in ld-2.6.1.so[b7ee9000+1b000]
jnoah1984: [ 1508.535631] revenge[13084]: segfault at 204 ip b7f4965d sp bf94baa0 error 4 in ld-2.6.1.so[b7f36000+1b000]
z3ro: nope, thats not it.
z3ro: can you pastebin lspci -v? I'll check the agp base.
jnoah1984: http://pastebin.com/m4ac93647
z3ro: yeah. it detected the agp base wrong.
z3ro: should be 0xe8000000
glisse: MrCooper: this was just accidental
glisse: but the radeon span code is broken for computation of tiled addr
jnoah1984: z3ro: so, what does that mean?
glisse: it just assume one tile size
z3ro: jnoah1984: just as a quick hack to confirm this, add "agp_addr = 0xe8000000;" to the end of detect_agp_aperture, recompile and try again.
z3ro: it means there is a bug in the detection code. :)
jnoah1984: is that in revenge_test.c?
z3ro: revenge_detect.c
jnoah1984: replace void?
jnoah1984: oh lol, nvm
z3ro: yeah, I mean just add it at the end of the function.
jnoah1984: ya
z3ro: hopefully that works. :)
jnoah1984: no, same result
jnoah1984: addr=939589632,
jnoah1984: #1 0x0804d2d8 in memory_read_agp (addr=939589632, size=3012898824) at revenge_memory.c:49
z3ro: oh, I feel stupid now... iirc I disabled the FB memmap as it breaks on some cards...
jnoah1984: ok. so what then?
z3ro: open revenge_main.c, and search for fb_mmap. just change those two #if 0 to #if 1
MostAwesomeDude: Uf. Gonna go get some food and do some studyin'.
z3ro: cya
jnoah1984: http://pastebin.com/m2a83477f
jnoah1984: Later MostAwesomeDude
jnoah1984: I've got 20 min or so and I should head to bed. I have class in a few hours...
MostAwesomeDude: I don't have class for 10 hours. :3
z3ro: yeah, thats alright... I'll be here tomorrow. :)
jnoah1984: So what do you make of the SIGABRT?
z3ro: well a bit earlier... tomorrow I'll probably be in some random club/bar. :P
z3ro: jnoah1984: technically it shouldn't happen. it means there was a NULL dword in the buffer. I'm still not exactly sure why it happens.
z3ro: you can try running it again a couple of times and see if it goes away.
jnoah1984: lol k, need a way to flush then?
jnoah1984: hey it workedish. I it got further...
z3ro: well I'm not sure... maybe fglrx is putting NULL dwords into the buffer. the docs I've read seem to suggest that this isn't valid.
z3ro: but maybe the hardware doesn't care.
jnoah1984: same abort, but slightly different
jnoah1984: http://pastebin.com/d22bcb705
z3ro: hmm try running with --fast
jnoah1984: and it works 2 die right away 1 goes for a bit, then 2 die right away and 1 goes for a bit....
z3ro: --fast disabled the sleep which might help if it's some kind of sync issue.
jnoah1984: well, this time I got output other than dump_packet3: 0x00000000 on some of it
adamk: Morning all. I am sometimes getting a weird mouse pointer distortion with my x1300 and x1950 under FreeBSD if i use HW cursor. Enabling SWcursor, and the problem goes away. Once I get this distortion, the only way to fix it seems to be rebooting. This is what I'm talking about:
adamk: http://picpaste.com/mouse-problem.png
adamk: I see this bug report, but it doesn't quite match the problem I'm having, I don't think: https://bugs.freedesktop.org/show_bug.cgi?id=18397
z3ro: jnoah1984: there may still be AGP bugs. I thought I got it working correctly there, but agp seems to be less reliable than PCI(-E)
jnoah1984: I == tail is where it fails a lot on revenge_dump.c:246
adamk: Any thoughts? Should I open this up as a separate bug report?
jnoah1984: z3ro: ah
z3ro: you could try removing that assert, but it's there to check that the packets were OK - eg they don't have a screwed up count which caused a read past the end of the buffer.
jnoah1984: w/o the assertion it is going much further, but still dies quickly here and there
jnoah1984: with a mem_map[i] fail
jnoah1984: wow it started working!
z3ro: yeah. I'm pretty sure it's some kind of memory sync issue, but I'd have to setup some agp gear to debug it.
jnoah1984: I have a triangle, but I ^C it cause I didn't want to lose it
z3ro: yeah thats the tricky thing with agp, I remember having some of these problems, but I could never reliably reproduce them.
z3ro: which made it very hard to debug. :(
jnoah1984: ya, so I can't reproduce it working...
jnoah1984: lol
z3ro: pci gart cards seem much better behaved.
z3ro: I can see what I can do about debugging it... I do have some agp cards and a agp mobo somewhere.
z3ro: although probably not right away (waiting on a case so I can rebuild this computer and fit some more hdd's :P)
jnoah1984: Alright, well, I can work on debugging it as well. For now though it's 3 in the morning and like I said, I have class in a few hours. I'm headed to bed. I'll be back in around 15 or so hours, maybe sooner maybe longer, we'll see.
z3ro: ok, bye.
z3ro: I'll be here sometime before 10pm-ish (NZ time) tomorrow.
jnoah1984: I'm PST btw
z3ro: ok, I'm usually awake at these times anyway (midnight here)
jnoah1984: Near Auckland?
z3ro: Christchurch
jnoah1984: lol, you said midnight and I was using my calender applet to add your timezone. My guess is NZDT?
z3ro: yeah, New Zealand time (which is based on Auckland)
jnoah1984: ah, cool. later
z3ro: cya
andrew_r: I have NetBSD 4.0 with Xserver 1.4.2nb3, and rv280/agp as videocard. But after VT switch or server exit my text consoles loss more and more brightness (or contrast?). For first one or two cycles i can compensate it with monitur's buttons. It happen with 6.9.0 and current git, too
hifi: had the same problem on linux, but it didn't get any darker after many switches, it was damn dark from the beginning
hifi: and I don't know any fixes to that problem
hifi: oh wait, maybe it was on openbsd
andrew_r: on Linux i usually have xserver from git master ... and it works without problem with plain VGA console. But on *bsd they used something a bit different (?) as VT driver (or layer, or infrastructure). I think it is good time to add new bug to bugs.fd.org ...
andrew_r: bug 18407
andrew_r: bug 18407 updated wit radeontool info ...
andrew_r: *with
adamk: Has anyone seen this before with HWcursor: http://picpaste.com/mouse-problem.png ?
yangman: adamk: known issue, afaik
adamk: Well I checked bugs.freedesktop.org this morning... The closest I came up with is this: https://bugs.freedesktop.org/show_bug.cgi?id=18397
yangman: adamk: is that on an external monitor connected to a laptop, dual-head on a desktop, or just single monitor?
yangman: adamk: or the laptop monitor with an external monitor connected
adamk: yangman, Sorry for the delay. dualhead on a desktop.
yangman: adamk: does it happen on both or only one monitor?
adamk: Both. And it doesn't go away again till I reboot. Even if I restart the X server, it comes back.
adamk: I have had other distortions, as well, that usually go away when I move from one monitor to another, though. I don't mind those much since they do go away :-)
yangman: hm.. ok
yangman: radeonhd's affected by hw cursor issues as well, which isn't surprising since the cursor handling code is really simple and probably identical in both
yangman: seems to be limited to dual-screen setup, though
adamk: Yes, in fact, I believe I've seen this problem with radeonhd, too.
adamk: This is with two dualhead machines, one with an x1300 and the other an x1950.
adamk: And, unfortunately, I'm not sure how to trigger it, either.
Ferocanis: part
Ferocanis: err
Ferocanis: jeese >_>
rx__: saves patch for later
tlp: eh?
rx__: ?
tlp: what patch? :)
rx__: fog test patch for revenge
rx__: oh screw it..
rx__: sends MostAwesomeDude a present
tlp: ah
Kano: hi agd5f , is it possible to get rid of the extreme flickering when xrandr is called?
Kano: with RV410 on vga port
Kano: gnome does not use xrandr so often, but kde4 does about every 10s, that kills eyes
agd5f: Kano: fix kde to not poll the outputs every 10s
agd5f: Kano: you can also edit the code to disable load detection
Kano: agd5f: but no other driver inkl. fglrx has that problem! only radeon
jcristau: that's not true.
Kano: i tested fglrx and nvidia binary, none flickered with same monitor (kvm)
agd5f: they may not do load detection, and only the recent fglrx got randr 1.2 support
tlp: are there any definite plans to improve r5xx 3D support (extra extensions, performance improvements, etc), or is it mostly a patches free-for-all?
tlp: isn't sure whether the scope was just to match r5xx with r3xx
glisse: tlp: we got plan but actually i think the biggest chunk of work are in the foundation
glisse: ie memory manager & command stream submission
tlp: ah
mcgreg: perhaps I'll ask today again ... how is the current support for Radeon 4670 cards?
mcgreg: does any of you have one?
mcgreg: I'm seriously planning to get one
mcgreg: they're cheap, don't need much power but still strong in performance , compared to my oold card ;)
tlp: you going to be dual-booting it?
glisse: mcgreg: only framebuffer support right now
mcgreg: tlp: well, only for WOW actually
tlp: ah
mcgreg: glisse: any change about that in near future?
tlp: just curious to know why you don't go with a presently-supported (but older) card :)
MostAwesomeDude: tlp: Yes. We are *definitely* wanting to bring r300 all the way up to OGL 2.0. :3
tlp: although WoW causes Xorg to use 99% CPU and kills the machine
tlp: MostAwesomeDude: sweet.
glisse: mcgreg: depends on you definition of near :), i would say few months
tlp: (WoW starts and runs fine until you actually click "Enter World")
mcgreg: tlp: since I have a r500 card.. but it sucks, and I see no point in getting another r500 card. and I dont want linux to run 3d .. not needed since I only use 3d for wow and that fine with windows
tlp: ohh, I see.
tlp: have you tried WoW with your r500 and the Radeon driver?
tlp: curious to know if you have different results.
tlp: (and/or WarCraft 3, which runs too slow to play here)
mcgreg: tlp: sure. it used to work. with gfx errors and VERY slow. like 10x slower than windows
MostAwesomeDude: tlp: WoW's going to hit various fallbacks due to lack of fogcoords.
tlp: MostAwesomeDude: But that wouldn't cause Xorg to grind to a halt, would it?
MostAwesomeDude: tlp: Well, using a software rasterizer for a fullscreen 3D app...
tlp: shouldn't be software.
tlp: unless I misunderstand you
mcgreg: glisse: you mean, even for 2d acceleration and textured video or are you talking about 3d?
glisse: mcgreg: everythings is 3d starting from hd2xxx (not exactly true)
MostAwesomeDude: tlp: Fallbacks for bad fragment programs have to be done in SW.
mcgreg: oh I forgotz that, sry
tlp: ah
tlp: that makes sense
tlp: and fogcoords is what you're working on, eh? :]
Erektium: which is better on r430 chip: textured video or video overlay?
esthar: Hi I've got a ATI Technologies Inc R580 [Radeon X1900]
esthar: And I am running 2.6.26 with the 6.9.0-1 radeon driver in Debian
esthar: glxinfo says direct rendering is working, which is great
esthar: But I can't use KDE4
esthar: I have compositing enabled in Kwin 4
esthar: Also projectM does not work via libvisual
esthar: projectM requires OpenGL acceleration
esthar: I have Mesa 7,2
adamk: esthar, What's the output of 'glxinfo | grep -i render' ?
esthar: OpenGL renderer string: Mesa DRI R300 20060815 TCL
esthar: It says direct rendering is on as well
adamk: Hmmm..
esthar: Why does it say R300?
esthar: I have an R500 chip, do I not?
adamk: Well, having direct rendering on doesn't mean much these days... You can have DRI but fall back to the software raserizer.
adamk: However, that doesn't seem to be the case now.
adamk: esthar, The r500 GPUs use the r300 driver.
esthar: Interesting
esthar: Will radeon be able to play projectM visualiser?
adamk: So in what way can you not use KDE4?
adamk: I'm not at all familiar with projectM.
esthar: It shows the splash screen
adamk: But KDE4 should work fine with compositing.
esthar: then turns white
adamk: Sounds like AIGLX might not be enabled.
esthar: I don't think it is
esthar: Do I have to specifically enable it in xorg.conf?
adamk: How about the output of 'LIBGL_ALWAYS_INDIRECT=1 glxinfo | grep -i render' ?
adamk: esthar, Depends on how X was compiled
esthar: direct rendering: No (LIBGL_ALWAYS_INDIRECT set)
esthar: OpenGL renderer string: Mesa GLX Indirect
adamk: Hmmm.
MostAwesomeDude: Xserver's too old. :3
adamk: Yeah.
chithead: yes, debian uses only xserver 1.4
adamk: esthar, Do you know the X server version? For Mesa 7.2, you really want Xserver 1.5.*
esthar: Errrrr okay yeah I have 1.4.x
adamk: Yeah, that's why AIGLX is broken.
esthar: Ah
esthar: So I need to upgrade X.Org you say
esthar: I'm at 7.3
esthar: and xserver 1.4.2
esthar: I run Debian unstable
esthar: with stuff from experimental
chithead: esthar: more specifically, you are mixing unstable xorg-server with experimental mesa
esthar: Yup
esthar: I'm just trying to get Free 3-D working =[
chithead: esthar: you have to use xorg-server from experimental too
esthar: Okay
esthar: chithead: Does this mean moving to 7.4?
jcristau: pretty much
esthar: What;s the best way of making sure it all updates fine?
esthar: I don't remember this being as extreme on the West Coast.
esthar: HAH!
esthar: Someone should log all IRC rooms, ever, and write an alogorithm to parse through the data to find all instances where people say something in the wrong channel
esthar: Super magic algorithm that would be
MostAwesomeDude: Hey, people. New revenge log parser. Sample: http://pastebin.ca/1247336
MostAwesomeDude: Yea? Nay?
rx__: MostAwesomeDude; need any more revenge dumps?
jnoah1984: hey, I'm back. :)
jnoah1984: z3ro, you present or idling?
rx__: he probably reads scrollback like everyone else :P
jnoah1984: ya, figured.
jnoah1984: z3ro: when are you merging MostAwesomeDudes fogcoord test into revenge's repo?
rx__: did you try the dump w/fogcoord test?
jnoah1984: no
rx__: oh
jnoah1984: MostAwesomeDude borrowed the code from mesa tree and ported it to revenge last night. I was just wondering when z3ro was going to update the tree with the patch
rx__: yeah i just did a revenge dump this morning with the patch
jnoah1984: rx__: what time zone are you in?
rx__: PST
rx__: PDT.. whatever ;)
jnoah1984: Same here
rx__: no wait.. yeah PST
rx__: it's regular time now
jnoah1984: lol ya, I just go By NorthAmerica/Los_Angeles
rx__: what radeon card do you have?
jnoah1984: r350 (Radeon 9800PRO 256MB AGP)
jnoah1984: you?
rx__: x1600 mobility pcie
jnoah1984: nice. I had the same, but my laptop died
jnoah1984: So I'm back to my old Desktop...
rx__: oh ouch :(
jnoah1984: ya... Oh well.
rx__: well now you can get a better one
rx__: haha
jnoah1984: Exactly what I've been thinking. Although I just might get an eeePC and get a better desktop later on. My laptop is for school and if I want to work on developing I can ssh -y into my desktop. :)
jnoah1984: All my laptop will be good for is using ssh and writing papers. lol
tlp: bought a Dell Mini
tlp: my laptop died too
tlp: screen snapped off
jnoah1984: z3ro, I think I have fixed a major (not all) portion of the AGP code. It seems to be working MUCH better now. :D
z3ro: jnoah1984: cool, send me a patch and I'll look over it.
z3ro: I'm going to go see a friend soon. apparently he's been drinking since 11am so this should be interesting. :P
jnoah1984: I will in a bit. I am still working on it and I keep getting it to run longer and longer. :)
jnoah1984: z3ro, that means you'll just be drunk sitting which is no fun...
z3ro: jnoah1984: nah, I'm drinking tonight too. he just started early. :P
jnoah1984: lol, nice... No coding though, unless it's casual drinking.
z3ro: I usually get a lot more productive after a night drinking. of course I wouldn't commit anything. ;)
jnoah1984: lol
z3ro: anyway, I'd better get going, cya later.
jnoah1984: z3ro, you still around?
jnoah1984: z3ro, can you commit the fogcoord pathc?
jnoah1984: fogcoord patch*
jnoah1984: z3ro, revenge is working on agp through gl_alpha_test/GL_NEVER_0.500000 and dies in GL_NEVER_1.000000
jnoah1984: At least now it's a static behavior. It's still a SIGABRT with assertion 'mem_map[i]' failed
jnoah1984: anyone familiar with revenge around?
Pyrotiger: i am running a tv connected to radeon x850 via multimedia center, but i am getting this "no signal" message while i am watching tv i assume it has to do with sound, i want that message turred off i dont care about having sound on, i dont have the audio input hooked up..
jnoah1984: z3ro: #7 0x0804c1fc in dump_packets (head=0, tail=, mem_map=0x810cbc8,
jnoah1984: always the same no matter when it dies.
Pyrotiger: the tv picture runs just fine...
jnoah1984: Pyrotriger, are you on Windows?
jnoah1984: Pyrotiger * are you using windows?
Pyrotiger: yes
jnoah1984: The opensource radeon driver only works on X,org. Sorry, this is not the room for you
jnoah1984: on -> with
Pyrotiger: where do i go to then?
jnoah1984: #ati
jnoah1984: Or call ati
Pyrotiger: topic says for fglrx, which is open source to use #ati
Pyrotiger: ?
jnoah1984: fglrx is not opensource (parts may be)
jnoah1984: Either way, you are not using radeon or fglrx. You are using the Catalyst driver designed for Windows
jnoah1984: So, see if there is a #catalyst room or just call ATi
jnoah1984: The quickest answer will be if you call ATi, calling ATi will should give you the answer you need. ATi
jnoah1984: damn
MostAwesomeDude: Actually, fgrlx *is* used on both Windows and Linux. Same codebase.
MostAwesomeDude: *fglrx, even.
jnoah1984: But the driver is Windows which he was using
MostAwesomeDude: But there's *no* Windows/fglrx support anywhere on Freenode.
jnoah1984: Besides, this is #radeon, not #ati
jnoah1984: MostAwesomeDude, that's why I told Pyrotiger to CaLl ATi
MostAwesomeDude: Well, I dunno about *calling* ATI. There's tons of help and support forums around the Web that deal with ATI's fun little quirks in their drivers. :3
MostAwesomeDude: But, at any rate, this isn't the place.
esthar: I've got 2.6.26 with X.ORg 7.4 and Mesa 7.2 and radeon driver 6.9.0+git20081012
esthar: Debian
esthar: And I can't seem to get Kwin4's compositing to work
esthar: Do I have to add something specifically to xorg.conf?