michaellarabel: legume: Should be back up now
legume: michaellarabel: Thanks - it's OK now
mordy^: radeonfb problems - anyone?
glisse: mordy^: fb usage with X is not adviced use pattern
mordy^: glisse - i simply want to disable it - get rid of it - it's forcing itself in middle of boot
glisse: mordy^: i guess you got an initd with some graphical boot stuff
glisse: like ubuntu or fedora does
mordy^: glisse - guess so too - but i don't know how to remove it
mordy^: no, i'm using debian
mordy^: but it's using radeon fb for boot console
glisse: well is there some packet associated to this ?
glisse: maybe remove --purge it will disable fb
MrCooper: mordy^: something like video=radeonfb:off should do
mordy^: the only thing i can see in my init.d is fglrx - i'm wondering how i would remove it
mordy^: but if radeonfb and fglrx are two different, and perhaps even opposed things, i don't see how fglrx would force radeon
mordy^: mrcooper - i've been down that path - i've also tried a different kernel too
mordy^: when i say it's intialized during boot, i mean in middle of boot - not *at* boot
adamk_: mordy^, Is 'radeonfb' listed in /etc/modules ?
mordy^: in other words, i get the regular console video mode for a few seconds - and it initializes some other components, drivers, devices, whatever - and then it initializes radeonfb - and then boom, i get the fb
MrCooper: that does sound like the initrd loads it as a module
mordy^: mr cooper - it doesn't - since i've recompiled intird over and over without the module
mordy^: hmm.. i don't see any radeonfb in /etc/modules
mordy^: if it would be loaded in intird then i would see it immediately - no?
mordy^: i don't know if a find / -iname "*radeonfb*" would help - but here it is http://paste.lisp.org/display/63892#1
mordy^: adamk 00 it's not listed
MrCooper: mordy^: no, that's the point - console is initialized before the initrd runs
mordy^: MrCooper -but it changes while the console has already been initialized
MrCooper: if you see kernel messages in text mode it means it initialized VGA console
MrCooper: which wouldn't happen if radeonfb was built into the kernel
mordy^: indeed MrCooper
mordy^: i'm deleting all the fglrx entries from rc?.d and init.d - i'll see what happens
MrCooper: there's no relation between fglrx and radeonfb...
mordy^: i know, i'm just doing it to troubleshoot - who knows
mordy^: MrCooper - well, indeed that didn't do much. i thought so
glisse: agd5f: i think i need to copy accel in new files and start from their, if i don't want to break anythings
glisse: anyway i will start with simple exa and see how it goes
agd5f: glisse: ok
mordy^: MrCooper - what else is related to radeonfb? so far i've only been searching for radeonfb entries - but perhaps i should be looking for other entires which are related\
MrCooper: mordy^: btw, how have you determined it's radeonfb and not something like (u)vesafb?
mordy^: MrCooper - it's clearly radeonfb - since that's the first line i see after the console enters framebuffer mode
MrCooper: k
mordy^: i wish there was a way to log all boot text
mordy^: MrCooper - nothing in /var/log/boot shows anything about radeonfb - is there a way to log *all* the boot text?
MrCooper: no idea, and this isn't really on topic here...
mordy^: i guess
mordy^: any other suggestions thoug?
ssieb2: mordy^: doesn't dmesg give you all the boot text?
mordy^: hmm...
MrCooper: ssieb2: only the kernel output
mordy^: MrCooper - i found it
mordy^: can i paste two lines? or do i need pastebin for that?
ssieb2: oh, what other boot text is there? all the services starting up?
mordy^: Console: switching to colour frame buffer device 180x56
mordy^: radeonfb (0000:01:00.0): ATI Radeon 1P
mordy^: ok - here is the relevant boot text - in a pastebin - i commented the section where it switches to framebuffer mode http://paste.lisp.org/display/63907
MostAwesomeDude: agd5f: Okay, started work on the bicubic stuff, for real.
agd5f: MostAwesomeDude: cool
MostAwesomeDude: Do you mind if I include the python script I used to generate the bicubic textures?
agd5f: MostAwesomeDude: sure
mordy^: anyone know how to disable radeonfb?
MostAwesomeDude: mordy^: On most distros, just blacklist it, or give the kernel an explicit vga= or video= boot option.
mordy^: MostAwesomeDude - blacklisting doesn't do it- i'm thinking it may be something in init.d
mordy^: MostAwesomeDude - here is the pastebin of dmesg - i only posted the lines of the radeonfb initialization http://paste.lisp.org/display/63907
MostAwesomeDude: mordy^: You may have to give the kernel an explicit vga/video line, then.
MrCooper: mordy^: that doesn't show how it got loaded/initialized
mordy^: MostAwesomeDude- that's been tried as well
rizzo: ssieb2: hey again
rizzo: fyi I ended up filing a bug for my radeon hd 2400 pro issues
MrCooper: mordy^: is it listed in the lsmod output?
MostAwesomeDude: mordy^: Well, just rename radeonfb.ko to radeonfb.blacklisted.ko
mordy^: hmm...
MostAwesomeDude: Unless it's built into the kernel, in which case you'll have to recompile, at which point you might as well just compile it out.
mordy^: MostAwesomeDude - it's not built into the kernel - thanks for the rename suggestion - it's refreshing to hear :D
mordy^: MrCooper, yes, it's in lsmod output
MrCooper: mordy^: okay, so if there's no radeon.ko in the initrd, MostAwesomeDude's suggestion should serve as a workaround
mordy^: MostAwesomeDude - all radeonfb entries? - or just the .ko?
MrCooper: what other 'radeonfb entries' are there?
mordy^: MrCooper - take a look http://paste.lisp.org/display/63907#1
rizzo: ugh guess I missed this one as well
rizzo: https://bugzilla.redhat.com/show_bug.cgi?id=450432
rizzo: same issue as mine
rizzo: they say it only happens when you have 4gb RAM (which I do)
MrCooper: mordy^: just the .ko, that's the kernel module binary
MostAwesomeDude: mordy^: Whatever shows up from: find /lib/modules -name radeonfb.ko
mordy^: let's see if this works
mordy^: \crosses fingers
mordy^: gah - but how can it be in the kernel? i just got a new kernel for this purpose - to get rid of the radeonfb
rizzo: oh airlied is assigned to that bug
mordy^: hmm... let me try a different kernel anyway\
ssieb2: rizzo: whatever do you need that much RAM for anyways? :-P
rizzo: ssieb2: because it is there
rizzo: but seriously I'm an Oracle DBA and test some things out
MostAwesomeDude: ssieb2: Because he's following Moore's Law and helping us take back the world from Microsoft. :3
rizzo: I could probably get by with < 2gb, but I was ordering a new machine so why not
ssieb2: was just kidding
rizzo: I'm hoping 4gb is enough in 5 years
rizzo: doesn't get new machines very often
ssieb2: is currently compiling the linux kernel on an 8-core machine with 8GB of RAM ;-)
rizzo: hot
mordy^: i'm trying another kernel - :-s
mordy^: ssieb - how long is it taking?
ssieb2: not long :-)
ssieb2: hasn't actually measured it
mordy^: gah, i misread that for 8MB. doh
bobbens: takes 4 minutes on my dual core, so probably like a minute :P
ssieb2: heh, now that would be funny, it's got more than that in cache!
rizzo: hm the original bug poster on the redhat bug says enabled memory remapping makes the problem go away
mordy^: 4 minutes :( it takes 1:30 hours on my p4 630
bobbens: mordy^: only takes me 50 minutes on my p3 850 MHz laptop :)
ssieb2: a hot compile (files in cache) took 48 seconds
mordy^: hmm..maybe i've got a 'special' kernel
ssieb2: although I'm trying to limit how many modules I'm compiling...
ssieb2: 599% CPU ;-)
mordy^: gah, i don't know what to do about this radeonfb - i've confirmed it's not in the kernel, since i just tried a different kernel - as opposed to reinstalling the one i already have - but it still won't do
bobbens: grab the 2.6.26 source, compile it yourself without radeonfb support, problem solved
ssieb2: mordy^: did you rename the module?
mordy^: ssieb2 - yes
ssieb2: O_o
mordy^: bobbens - i just downloaded the ready-made 2.6.2-amd kernel - same problem
mordy^: hence i guess it's not in the kernel
mordy^: where can i find more information on this?
MostAwesomeDude: agd5f: Is your allocation of "size * 2" for the video texture based on a double-buffering setup?
mordy^: what else can be forcing my radeonfb is the question
mordy^: :(
MostAwesomeDude: mordy^: If it's really disabled, then it shouldn't be loading since it doesn't exist.
mordy^: MostAwesomeDude - it should be loading from somewhere else - and something else is loading it - the question is where and what
mordy^: now follow me here - if the problem is during initialization, but not in the kernel - it *must* be something in *.d - no?
MostAwesomeDude: mordy^: In your kernel config, please confirm that Enable Loadable Module Support -> Module versioning support is checked.
MostAwesomeDude: Also confirm that radeonfb is disabled.
MostAwesomeDude: There's only two ways to get radeonfb: Either build it into the kernel, or build it as a module (radeonfb.ko).
MostAwesomeDude: Nothing else.
mordy^: hmm...
mordy^: make menuconfig?
MostAwesomeDude: So if you don't build it into the kernel, and you don't have the module, then it's not possible for it to load.
MostAwesomeDude: Yes.
MostAwesomeDude: BRB, testing texture uploading. :3
mordy^: MostAwesomeDude - it's enabled - but i don't know if my config file is actually being used - since the kernel i'm using is a downloaded binary from apt-get whereas the config file is in the source
mordy^: and the source is not the actual kernel i'm using
mordy^: ok
mordy^: welcome back
mordy^:
MostAwesomeDude: mordy^: Your "real" config is in /proc/config.gz
MostAwesomeDude: Also I thought you said you were building your own kernels?
MostAwesomeDude: agd5f: Tex allocation works, now I have to figure out how to upload floats into a texture. Yay?
mordy^: MostAwesomeDude - i didn;'t build my kernel - rather i downloaded them from the repository - i built my kernel once - but not more
MostAwesomeDude: Ah.
MostAwesomeDude: Well.
mordy^: hehe, i'm just a newbie - this is a learning experience for me though - and it's interesting to see how all the heirarchy works in the system - really cool
MostAwesomeDude: At any rate, do "zgrep FB_RADEON /proc/config.gz"
MostAwesomeDude: Please.
mordy^: 3 lines - should i paste it here?
mordy^: gah, i'll just pastebin it
MostAwesomeDude: Well, =m means module, =y means built in, =n or "not set" means not enabled.
MostAwesomeDude: So.
mordy^: MostAwesomeDude - here http://paste.lisp.org/display/63907#2\
mordy^: i want to =n it?
MostAwesomeDude: mordy^: You can't, really.
mordy^: what worries me is the =y
MostAwesomeDude: That's just a read-only copy of what the currently running kernel was built with.
MostAwesomeDude: And we're only concerned with FB_RADEON, not the others.
MostAwesomeDude: So, did you rename your radeonfb.ko?
mordy^: but if i can't =m it, then it means that no kernel ever let me =m it- which means that the problem resides elsewhere
mordy^: yes
MostAwesomeDude: ?
MostAwesomeDude: Okay.
mordy^: hmm... this is weird i renamed it - but something recreated the file
MostAwesomeDude: No radeonfb.ko -> no radeonfb loading.
mordy^: somethign is forcing the module to 'exist
MostAwesomeDude: Gah, initramfs.
MostAwesomeDude: Did you put "blacklist radeonfb" in a file in /etc/modprobe.d ?
mordy^: yes -
MostAwesomeDude: Hmm.
MostAwesomeDude: This is very confusing.
mordy^: let me reboot once more
MostAwesomeDude: You might want to go consult Debian docs.
jcristau: did you rebuild the initrd after blacklisting the module?
mordy^: yes..
jcristau: ok...
mordy^: ahh.. it's working now
mordy^: after renaming it again, it works
mordy^: i'm jumping for joy
mordy^: but still confused
mordy^: MostAwesomeDude, i am rather indebted to you :D
MostAwesomeDude: mordy^: Aw, shucks. :3
mordy^: hehe, now i can get back to what i was doing before - installing debian on my smaller laptop :D
mordy^: debugging and troubleshooting are always a pain in the ass
mordy: but now we get to the second stage of the problem -why does radeonfb cause problems in the first place? but i guess that's asking for too much
MostAwesomeDude: mordy: We don't talk about radeonfb. :3
MostAwesomeDude: It hasn't been worked on in a long time.
mordy: ROTFL
mordy: :(
MostAwesomeDude: And with kernel modesetting coming up, there's no real interest in it.
mordy: what's that?
MostAwesomeDude: Well, modesetting is hitting registers on a video card in order to get it into a mode of operation.
MostAwesomeDude: And historically, that's what the X drivers are for.
MostAwesomeDude: But people are moving that code into the kernel, which will make things a lot faster, more stable, and generally more awesome.
mordy: as opposed to assuming what the card wants and pre-loading it?
mordy: wishes he knew more about this stuff
MostAwesomeDude: mordy: http://www.x.org/wiki/Development/Documentation/HowVideoCardsWork is a starting point.
MostAwesomeDude: The entire x.org wiki has all kinds of neat stuff.
mordy: yeah, i'll read it - i'm so exhausted - been troubleshooting this for almost a whole day already
ssieb2: how does kernel modesetting replace the FB drivers?
MostAwesomeDude: ssieb2: You could run an FB directly above the KMS.
MostAwesomeDude: I don't know all the details.
agd5f: MostAwesomeDude: use memcpy initially then pipeline it with the RadeonCopyData() and friends in radeon_video.c
MostAwesomeDude: agd5f: Yeah, I see RADEONCopyData.
MostAwesomeDude: And the tex format for four floats per pixel is 32f-32f-32f-32f, yes?
MostAwesomeDude: No munging needed?
agd5f: you shouldn't need munging unless you need to re-arrange the order
MostAwesomeDude: Nope, just 128 pixels that need to get uploaded.
MostAwesomeDude: Well, 512 floats.
MostAwesomeDude: ...Sorry for all the questions. I must get annoying. :3
agd5f: MostAwesomeDude: no worries :)
MostAwesomeDude: agd5f: That's my line. :)
MostAwesomeDude: BRB, tex upload and free.
MostAwesomeDude: Hmm, looks like it works. Now to get the RS and FP stuff down.
MostAwesomeDude: But first, lunch.
MostAwesomeDude: Woot.
MostAwesomeDude: bicubic_offset: 0x8faf040
MostAwesomeDude: bicubic_memory: 0x0x2fa12d0
MostAwesomeDude: bicubic_enabled: 0x1
MostAwesomeDude: The hard part's done.
mordy: hehe, maybe i'll write graphics drivers when i grow uip
MostAwesomeDude: mordy: How old are you?
MostAwesomeDude: I'm only 20.
mordy: same here
mordy: lol
MostAwesomeDude: There's only two requirements for doing drivers.
MostAwesomeDude: 1) Gotta learn how to program.
MostAwesomeDude: 2) Gotta learn how to read code.
MostAwesomeDude: And the rest is just patience and experience.
mordy: you've aslo gotta be creative or at least planning
mordy: i've programmed before, but never anything serious
MostAwesomeDude: (Having airlied patiently and repeatedly explain the guts of Radeons helps too. :)
MostAwesomeDude: Mm, not a lot of planning needed for most parts of drivers.
mordy: hehe, he helped me once before too
MostAwesomeDude: You have an interface, and you have a piece of HW.
MostAwesomeDude: And you need to make the HW do things according to the interface.\
MostAwesomeDude: Lot less planning than, say, a game, or a word processor.
mordy: true
mordy: but i like creating problems and difficult planning when it's not needed
MostAwesomeDude: Heh.
mordy: why else would i choose linux? for useability?
MostAwesomeDude: Well, I chose it for price, stability, features, and hardware compatibility.
MostAwesomeDude: But sure, usability is a good reason, I suppose.
mordy: i chose it because i jsut wanted something different - and something i guess more powerful and would actually let me check under the hood of the operating system
mordy: or maybe because i realized i'd be dropping out soon and decided to go back to working with computers
MostAwesomeDude: XD
MostAwesomeDude: onestone: Just the guy I was hoping to see.
onestone: runs away ;-)
MostAwesomeDude: onestone: Should the helper tex for bicubic also get the linear filtering?
onestone: no
MostAwesomeDude: Okay.
onestone: only repeat wrap
MostAwesomeDude: Hmm. 'k.
MostAwesomeDude: Wait, repeat wrap? I thought it was stretch nearest?
MostAwesomeDude: Maybe I'm reading wrong?
MostAwesomeDude: Well, okay. You're da boss. :3
MostAwesomeDude: BRB, RS init. :3
rizzo: so where does airlied live?
rizzo: Australia?
rizzo: his livejournal says yes
agd5f: rizzo: yeah
rizzo: I guessed from the fact that he went to eat lunch at like 11pm central US time
MostAwesomeDude: onestone: Um, bit of an ambiguous typo.
MostAwesomeDude: param[0] = {1 / texture[0].width, 0, 0, 0}
MostAwesomeDude: param[1] = {0, 1 / -texture[0].height, 0, 0, 0}
MostAwesomeDude: What should param[1] be?
onestone: ARB program environment variables
MostAwesomeDude: onestone: No, there's five coords there, when there's only four spots...
onestone: MostAwesomeDude: remove the last 0
MostAwesomeDude: 'k.
rizzo: so the radeonhd driver is something completely different from what you guys work on?
MostAwesomeDude: rizzo: Correct.
rizzo: but your driver should support HD cards just fine?
MostAwesomeDude: rizzo: Yep.
rizzo: wonders what the correlation between having 4G of RAM is to the breakage
airlied: rizzo: PCI address space goes > 4GB..
MostAwesomeDude: ...It just occurred to me that KMS -> two radeon drivers enter, one goes upstream
airlied: MostAwesomeDude: yes strange that :)
airlied: MostAwesomeDude: but really only one is entering.
rizzo: airlied: so 4GB shouldn't be a problem?
MostAwesomeDude: airlied: I know. :3
MostAwesomeDude: Not that I dislike the RHD team, but I honestly think a lot of their efforts are on the silly side, especially with the atombios_support.
rizzo: airlied: I assume you've seen that I found the earlier redhat bug on this issue as well
airlied: rizzo: oh its a problem.. some BIOSes do things wrng..
airlied: rizzo: does your BIOS have a remapping option..
airlied: rizzo: it might be worth playing with
rizzo: no clue
rizzo: I'll take a look later tonight
libv: MostAwesomeDude: what is so silly about the atombios_support branch?
MostAwesomeDude: libv: Well, I thought the entire point of RHD was to not rely on atombios for modesetting...
libv: MostAwesomeDude: well... i'm under nda, and i still need to see the first bit of register info for rv770
MostAwesomeDude: Hmm. So the atombios support is just a crutch, to be replaced eventually by non-atom modesetting?
libv: if we ever get the chance to, yes
MostAwesomeDude: Also keep in mind that me calling things silly is like Mr. Bush calling people stupid. :3
GerbilSoft: odd..
GerbilSoft: since a recent update of radeon, my CRT's refresh rate is no longer properly detected
GerbilSoft: it's always set to 60 Hz (manually adding a mode works though)
lavish: hi all
lavish: I've noticed a *extreamly slow* rendering - with every browser - of lightbox2 galleries, like http://www.lokeshdhakar.com/projects/lightbox2/ . Could it be related to a wrong configuration of my Radeon 9600 (RV350 chip) using radeon drivers or not?
lavish: I confirm: using vesa it's much faster o_0
MostAwesomeDude: Dangit, and I needed to talk to him.
MostAwesomeDude: onestone: I am having issues with the FP. Do you have any sample code I could use to get a realistic FP dump from?
MostAwesomeDude: Like, is this the same stuff in the bicubic filter in the compiz plugin?
onestone: MostAwesomeDude: dump the shader from the compiz plugin
scobby: are there updated packages for ubuntu to get latest radeon drm driver for my radeon 9600 pro ?
agd5f: scobby: read this thread: http://www.phoronix.com/forums/showthread.php?t=9951
scobby: hmm i just saw i already have the tormavolden packages installed
rizzo: ok going to see if my bios has memory remapping features
rizzo: airlied: I don't see anything like that in my BIOS setup screen
rizzo: says my BIOS Version is fro 03/11/08
rizzo: there is a newer version from 06/20/2008
rizzo: I'll have to find out how to use that .exe to update my BIOS
rizzo: ah I need a bootable DOS USB key
rizzo: the list of Fixes and Enhancements doesn't indicate that anyting about remapping memory was added though
spstarr: newer ati cards can be flashed ?
spstarr: i dont recall seeing any bios updates for radeons (r100, 200, 300)
rizzo: spstarr: no, bios for my mobo
spstarr: oh
airlied: rizzo: it might be hidden in the BIOs, there are some tips in phoronix forums..
airlied: somebody mentioned Ctrl-F1 or something.
rizzo: I'll check again
rizzo: ctrl-P gets me into the Intel Management Engine BIOS Extension
rizzo: but I need a password
airlied: nah it should be just in the normal BIOS I think.
rizzo: none of the F keys are doing anything in the setup screens
rizzo: the security section of the BIOS screens has an Execute Disable Memory Protection Technology on/off setting
rizzo: I don't see much else involving memory
rizzo: would having encrypted swap be an issue?
rizzo: although I've entered the key long before X comes into play
airlied: shouldnt' matter at all
rizzo: let me know if there's something else I should try
rizzo: mabye use -debug packages?
damentz: hey guys
damentz: wicked job with r300 dri support
damentz: i'm trying mesa 7.1 rc3 with warsow
damentz: everything renders beautifully!
damentz: i mean it looks nice, exactly the same as fglrx
damentz: though, not as fast, but simpler code has it's advantages
damentz: but i guess you guys can officially say
damentz: you have the most badass open source drivers
rx__: i'm sure the devs will appreciate that :)
damentz: rx__: hehe, well if you have a r300 card
damentz: try out the latest mesa dri
damentz: it's really nice
damentz: i'm playing it at 1440x900
rx__: r500 here
rx__: still waiting on powerplay :)
damentz: hehe ya
damentz: omfg
damentz: i just enabled low impact fallback
damentz: now it runs even faster
damentz: i'm using hte high profile in warsow
damentz: that's pretty cool
damentz: i'm looking
damentz: the fps floats around 50 to 80 fps as things change on the screen
damentz: ok
damentz: light bloom still seems to be software rendered
damentz: so it's like 1fps
damentz: everything else works perfectly
rbrett: so tempted to swap out my 2600 for my old 9600