mjg59: Or run Fedora
DanaG: I can't stand the package manager.
DanaG: And I don't know my way around where their config files are, and such.
mharris: /etc/X11
mharris: Not sure where other distributions put the X config files, but Fedora puts them in /etc/X11 as per the X.Org documtentation. :)
DanaG: That's the least of the config files I deal with.
DanaG: I also like to use things like dnsmasq and such.
pkt: yes, I find debian much easier for day-to-day stuff as well
pkt: but for testing / helping with the bleeding edge it is much easier to just use rawhide
pkt: you don't need the 'guru level' admin skills you have with debian to just test a few features / apply some patches etc
blindraven: enouf, I'm not following :P
blindraven: enouf, Linux didn't work out for ya?
DanaG: ooooooooh, the fixed drm does work with suspend.
DanaG: drm + suspend means more to me than kms right now.
DanaG: oh, and thanks for fixing it!
pkt: It seems radeon works with suspend here too. Some other module is breaking the suspend :/
enouf: blindraven: don't follow ; lead
blindraven: enouf, With?
enouf: blindraven: your heart!
blindraven: enouf, I don't think you're the same guy I used to talk to :) nvm
enouf: then stfu please
dileX: airlied: there are problems in radeon-gem-cs3 (git20090313.e1ce90a) with "drmModeModeInfo", see
DanaG: oh yeah, another odd thing with kms on r600: fbset doesn't let you change modes.
MostAwesomeDude: fbset doesn't do mode-changing on my r300 either.
pkt: yep
pkt: it seems it was non-obvious how to make fbset work in e.g., dual-head scenarios
pkt: perhaps a /sys interface would be great though
pkt: echo "1920x1200" > /sys/class/dri/card0/crtc1 or something
DanaG: I have just one head, though.
DanaG: er, my laptop, also.
DanaG: =þ
b0le: dileX: what version of libdrm do you have?
DanaG: If I had two heads, I'd be freaky. =þ
b0le: dileX: regardless, either getting latest master, or applying this patch http://cgit.freedesktop.org/mesa/drm/commit/?id=eb78c53aa1a980e60c0dd1f2d0d2f04cb9cb2622 to whatever branch you are using should fix it.
twoerner: airlied: ping
airlied: twoerner: pong
airlied: dileX: it should in xf86drmMode.h
airlied: I mgiht need to chererpick something into modesetting-gem
twoerner: airlied: it happened when i closed mplayer
twoerner: airlied: so it might be cleanup code
airlied: twoerner: btw I've fixed a few things that might have cause memory corruption in the las tcouple of days
airlied: I haven't tested mplayer too much though
airlied: I'm away from r600 hw until monday now
twoerner: airlied: it happened after a day
airlied: dileX: fixed that in modesetting-gem
airlied: twoerner: you know if it is Fedora package or git dose it as well?
twoerner: airlied: i did not have had the problem with the GIT package
twoerner: the GIT version
airlied: I'll have another look at my merge patch again.
twoerner: i was using it for days before rebooting
airlied: there mgiht be another mis-merge
airlied: if you can try latest -ati it might help I built another one this afternoon
twoerner: ok, but i have to install a fresh F-11 before that again
airlied: gotta go get dinner &
dileX: airlied b0le: thanks
pkt: Are there any docs for the radeon_workaround_list in radeonfb?
pkt: I wonder which ones should I use for my laptop
lx2nt: I'm unsure if that's on topic: how hard it is to add ARB_vertex_shader+ARB_fragment_shader suppoort to mesa/r300?
mharris: lx2nt: kokobolo
lx2nt: mharris: ?
mharris: lx2nt: sorry... cocobolo
scarabeus: heya, i have a bit problem, when i fire up xdm i got in log (EE) No devices detected.
scarabeus: no other error
scarabeus: this is what i synced few minutes ago
scarabeus: last sync was 2 days ago and everything worked just peachy :]
lx2nt: mharris: thanks, thought as much. Can you outline major steps in getting SM2 support so that I can dig further myself?
mharris: lx2nt: Might want to bounce it off airlied, not 100% sure but I think that's something he's either working on or might be in the future.
mharris: if not, he'd know who might be
lx2nt: mharris: thank you
nanonyme: mharris: Hrm, related to FBO's?
mharris: nanonyme: hrm?
nanonyme: As in, which way is it actually. :) Can you do fragment shaders using fbo's or fbo's using fragment shaders. :D I'm quite certain there's a link, just don't know/remember/something else vague which way.
MrCooper: nanonyme: they're completely orthogonal concepts
nanonyme: Hrm.
MrCooper: shader-less cards can support FBOs just fine
nanonyme: MrCooper: Thanks for clarification then...
MrCooper: GLSL support should be easy with Gallium, but for r5xx cards it might not be too hard with the traditional driver either
scarabeus: problem solved it was distro related :]
pkt: fwiw, I got suspend to ram / resume to work reliably on radeonfb (r300/AGP) with pm-suspend --quirk-radeon-off --quirk-vbe-post --quirk-vbestate-restore
pkt: next step, kms
pkt: yep, the same trick worked for KMS too :))
pkt: that will do until suspend is fixed "properly"
buggs: ok, after suspend only some resolutions are broken (they have a mirror image)
buggs: 1920x1200 is broken and 1920x1080 not
buggs: this is without DRI as far as i can tell
nanonyme: Isn't 1920x1200 somewhat unorthodox? :)
pkt: why? it is 16:10
pkt: this is what my laptop uses :)
buggs: it's my native 24" res
buggs: most 24" have that
pkt: my laptop's screen is 15.4'' but it is still ok
nanonyme: buggs: Except the ones that are HDTV compliant? :)
pkt: well, some people don't care about HDTV :)
nanonyme: Why would you want that big a screen then? Coding?
pkt: of course :)
nanonyme: headdesks
pkt: no screen is too large for coding :)
buggs: it's the best display/resultion i ever had
buggs: had doubts before but it's great
nanonyme: You know, there probably is a maximum amount of information your brain can absorb through your eyes.
pkt: yeah, but 1920x1200 ain't it :)
nanonyme: It's not about resolution, it's about the amount of text.
pkt: bring this 2560 on :)
buggs: 1600x1200 20" was much worse
pkt: nanonyme: being able to see lots of text simultaneously does not mean you read it all at once
pkt: it is just much easier to find things
nanonyme: That's what I use interactive search for. :)
pkt: a slight eye movement will always be faster than that :)
nanonyme: Only if you remember by heart approximately where exactly everything is.
nanonyme: Hrm, that sentence sounded almost redundant to me.
pkt: or an oxymoron
nanonyme: Or that.
buggs: i limit the window manager to use a margin of 600px on the right
nanonyme: Ah.
buggs: in that areay i have a conky running or drag some windows over by hand for later use
buggs: or calculation etc.
nanonyme: Well, that is a nice setup.
pkt: there are lots of tricks when you have large screen real estate
buggs: using eclipse i use all off hte screen though
mharris: nanonyme: how is 1920x1200 unorthodox? ;)
nanonyme: mharris: It's not 1080p. :)
mharris: It is if your software draws black on the lines in excess of 1080
mharris: Why *wouldn't* you want a screen this big? ;)
mharris: I've got 2 of them, and I want more space!
pkt: yeah
mharris: I'd gladly take a 3008WFP or two if someone wanted to give them away
mharris: A nice 2560x1600 or 5120x1600 dualhead desktop would be spiffy to say the least.
nanonyme: mharris: You know, if you were inside a display bubble weightless, you could have a muuuch bigger display. ;)
nanonyme: Heck, half of it would always be behind you.
nanonyme: At least.
mharris: You wouldn't hear me complain, as long as the graphics hardware was adequately supported. ;)
pkt: hehe
mharris: I never understand people that think any display is 'too big' ;) There is no such thing as too big. ;)
mharris: I have a 65" TV and I think it's too small, and I keep the couch about 5 feet away.
nanonyme: mharris: Only if it doesn't fit your desks and bigger desks don't fit in your room.
mharris: 'tis not the size that matters, 'tis the viewing pyramid or field-of-view that matters. I want my eyes filled with screen real estate, preferably like they are when going to the theater. :)
nanonyme: Getting a bigger room just to get a bigger display sounds like an addiction.
mharris: nanonyme: I'd get a bigger room.
mharris: :)
nanonyme: mharris: Ditto.
mharris: I suffered from crappy-displayitus years ago for many years, and then finally 'healed'
nanonyme: mharris: Some people are addicted to higher quality speakers, you are to bigger displays. :)
mharris: I have good quality speakers too... studio monitors ;)
nanonyme: How much did they cost?
mharris: with dual inputs, S/PDIF from my guitar processor, and analog from the 'puter.
mharris: Not too much, around $220 or so. They're mid-end computer studio monitors (Edirol MA-15D)
nanonyme: There's people who spend maybe $100k or so on a speaker set only to replace it later with even more expensive a speaker set.
nanonyme: Better be careful on that road.
mharris: Just got them as a temporary solution, but was happy enough with the quality to keep using them ;)
mharris: Spent the cash I would have spent on a better pair, on my guitar processor
mharris: (Boss GT-Pro)
bryce_: agd5f: the ubuntu kernel guys are about to pull in your patches. Just to doublecheck - are the patches at http://www.botchco.com/alex/xorg/r6xx_drm/ still the latest and greatest they should pull? vs. your drm git tree at fdo?
agd5f: bryce_: should pull in the patches from dave's drm-next tree as there have been a number of fixes there
agd5f: for both old and new chips
agd5f: http://git.kernel.org/?p=linux/kernel/git/airlied/drm-2.6.git;a=summary
agd5f: bryce_: http://git.kernel.org/?p=linux/kernel/git/airlied/drm-2.6.git;a=shortlog;h=drm-next
agd5f: includes my patches as well as a number of fixes since then
agd5f: and rs600 support
agd5f: bryce_: you'll probably want this patch for mesa as well: http://cgit.freedesktop.org/mesa/mesa/commit/?id=69f080cefced8b3395cdf179c107303a1013d196
bryce_: agd5f: thanks, already pulled that one in
bryce_: agd5f: if they want to cherrypick only the stuff required for radeon, would he just pick off the changes that are "11 hours ago" + mention either "radeon" or "r6"?
agd5f: bryce_: yeah
agd5f: bryce_: also grab 87f0da55353e23826a54bff57c457a13b97d18f1
agd5f: drm: add DRM_READ/WRITE64 wrappers around readq/writeq.
agd5f: bryce_: might also want benh's first 3 patches as well
agd5f: those 3 benh patches are probably not necessary, see what your kernel guys think
DanaG: wonders if there's any practical difference between radeon and radeonhd nowadays.
yangman: DanaG: HDMI audio setup, 10-bit precision LUT, tv-out, etc
DanaG: oh, and the dynamic clocks (not that that does as much as voltage changing).
DanaG: s/changing/changing would/
DanaG: has only a 6-bit color LCD. =þ I'd need dithering.
yangman: relevance of dynamic clock is uncertain on hardware that radeonhd is supposed to handle
yangman: IIRC, the EnableDynamicClock AtomBIOS call on rv770 is actually a NOOP
DanaG: ah. Mine's an R600, anyway.
yangman: and playing with the register bit that enables it on M56 did nothing
bryce_: agd5f: they're also asking about 7c1c2871a6a3a114853ec6836e9035ac1c0c7f7a drm: move to kref per-master structures. ?
bryce_: and
bryce_: 8e1004580e0c862cb6bbe2ff8e496f846c54052f drm: Drop unused and broken dri_library_name sysfs attribute.
agd5f: 8e1004580e0c862cb6bbe2ff8e496f846c54052f should be fine
DanaG: hmm... theatre_detect_drv.... is that a v4l driver for Rage Theater chips?
agd5f: DanaG: not v4l
DanaG: and theatre_drv
agd5f: bryce_: you'll need 7c1c2871a6a3a114853ec6836e9035ac1c0c7f7a for all the other ones I think
agd5f: DanaG: it's a driver for rage theatre, but it doesn't provide a v4l interface
DanaG: ah.
bryce_: agd5f: ok. the three benh patches are 41c2e75e, f77d390c, and d883f7f1 yes?
agd5f: bryce_: yes
bryce_: agd5f: ok, emailed list of commits for your final review before I send it to kernel
agd5f: cool
DanaG: sticks with radeon for tear-free video.
DanaG: ... reminds me of shampoo commercials.
tlp: lol
nanonyme: Wish I could use the open driver too already. :)
ajax: agd5f: is there a bit in the connector table to tell whether a dvi link really is dual-link?
agd5f: ajax: yeah, but only in the object table so r6xx+
agd5f: ajax: btw, depth 16 seems to be broken on xserver 1.6
agd5f: seems colormap related
ajax: weird.
ajax: haven't tried a 16bpp config in ages, i admit
ajax: but i can't think why that would be true for 1.6 but not 1.5
agd5f: I haven't tested until recently myself. just noticed it today
nanonyme: agd5f: How'd you obtain the object table and how would you interpret that information from it, btw? (And would it make sense to have the info displayed automatically somewhere, eg with xrandr?)
agd5f: nanonyme: it's in the bios
agd5f: we use it to get the connectors and encoders to build the outputs for randr
nanonyme: Right...
ajax: agd5f: pre-avivo dual-link would have been off-chip dvo, i assume. don't know how we'd set about checking on r500 though.
nanonyme: So the second question was if you could expose the information to be easily readable by user.
agd5f: nanonyme: you could print it out. We should save it and use it for mode validation
nanonyme: Right.
arekm: what's "LUT" ?
airlied: look up table
agd5f: yangman: hey, did Dave's shader reload patch fix resume for you?
yangman: agd5f: I tried it this morning and it just gave me a blank screen and no inputs were responsive. but I wasn't sure if I actually had the latest bits running, and haven't had time to try it again
ajax: agd5f: any idea what would be special about hdmi-b? i can make it work for single-link modes but dual-link comes out all trippy
agd5f: not that I know of
ajax: it gets the timings right, and you can see the desktop, it's just mad colors.
ajax: and they don't seem to be stripe-aligned, otherwise i'd suspect that i need to upload the lut to both links or something
agd5f: maybe something in the FMT regs?
agd5f: is this dce3x or previous?
ajax: dce3 i'm pretty sure, let me swap that card back in...
ajax: yeah, rv635
agd5f: ok
buggs: agd5f, using the latest released versions of everything, after suspend i get a shadow picture, maybe 5 milimeter off to the right
agd5f: buggs: what chip?
buggs: but only in some resolutions
buggs: 4850
agd5f: buggs: try git master
buggs: 1920x1200 and 1600x1200 are broken, 1920x1080, 800x600 work
buggs: agd5f, of drm or -ati??
agd5f: ati
buggs: yangman, told it was a drm issue?
agd5f: you should also grab the latest drm bits of you are using accel
buggs: no dri here
buggs: guess that needs to be especially enabled?
agd5f: if you aren't using dri than the drm is not involved
agd5f: buggs: dri/accel support is only available in git, not any released versions yet
agd5f: for your card
buggs: radeonhd crashed on suspend btw
buggs: agd5f, same as before
DanaG: Well, I think I figured out what was up with Plymouth: for some reason, it was trying to draw on ttyS0, and understandably, failing to do so.
DanaG: There doesn't seem to be any config file to tell Plymouth which tty to draw on.
spstarr: new ddx -10
airlied: r600 mostly
spstarr: oh
spstarr: airlied: have you confirmed all those crazy things i experienced?
spstarr: so those are legit bugs?
airlied: nope I've been fixing r600
spstarr: ok
airlied: which is why the ddx has r600 fixes in it
spstarr: i have the video/png files when you return to r3xx stuff
airlied: damn remoted rebooted r600 box and it died, not on network power strip either :(
spstarr: ugh
rnoland_: agd5f: did the xv sync issue get fixed in 6.12?
rnoland_: Xv vsync support on r6xx/r7xx cards.
rnoland_: was that it?
yangman: no go on r770 resume. wait for idle failed with 0xA0003028 0x00000002
airlied: yangman: wierd I wonder whats different about rv770
airlied: got an xorg log file?
yangman: airlied: nothing out of the ordinary. there's no new messages after suspending
airlied: does it suspend/resume if you VT switch first?
airlied: just wanted to make sure it was allocating the right sizes for the pci gart table
airlied: just wondering if its the VT switch back to X that is broken or the resume.
yangman: same, I think. pm-utils VT switches for me. the last one I explicitly sent 'mem' to /sys/power to get around pm-utils
mjg59: Even echo mem forces a VT switch
mjg59: But if you're using pm-utils, you need to make sure that it doesn't run any quirks
airlied: I normally do an explict VT switch, pm-suspend, and then explicit VT switch
airlied: well I think we want post quirks as we don't have kms
yangman: mjg59: so it does do that eh? I wasn't sure
mjg59: The kernel vt switches
mjg59: airlied: Oh, right, fair enough
airlied: agd5f: btw not sure I like setting the color bases all to 0 :)
mjg59: airlied: Though the quirks run after kernel resume functions, so may damage state
airlied: I wonder could that cause DMA by accident.
yangman: I'll give explicit VT switch a go when I have time
airlied: mjg59: yeah the driver should redo all that on VT switch
mjg59: The pm-utils vt switch is so that quirks can be run after kernel resume but before re-entering X
mjg59: airlied: Fair enough
airlied: hmm I wonder does krh realise the kernel VT switches
mjg59: We could do with a way of disabling that
yangman: http://yangman.ca/Xorg.0.log_radeonhd.resume http://yangman.ca/dmesg_radeonhd.resume
yangman: that's from suspend to mem. card didn't seem to be posted properly coming out. I'll have to give hibernate a try later
mjg59: Hibernate's much more likely to work
mjg59: Since the BIOS posts the card for you
yangman: yeah.
airlied: yangman: ah you might not have pm-quirks enabled, also radeonhd, I've no idea
yangman: I'll try and run through hibernate quickly
yangman: that was without pm-utils
cxo: I thought i'd ask here since you guys are so cleva! I need to set a bit in a register, like PIN2OUT |= BIT3; is there a way of putting that into a macro, like #define PORT_I_NEED_TO_SET(x) (x) ? PIN2OUT|=BIT3 : PIN2OUT &= ~BIT3 Does that make sense?
airlied: yangman: when I debugged I did vt switch, pm-suspend, resume broguht back text mode and then I vt swtich back
yangman: with post quirk set? I don't think I have it set on my system
airlied: yes I think Fedora set the post quirk for me
yangman: ok. I'll try that after hibernate test
yangman: it seems like hibernate will even VT switch away from VT1. that's just weird
yangman: hm... nope
mjg59: yangman: Yes, the kernel does that
yangman: the last time I've dealt with suspend without tool aid was back in 2.6.1x. it's all so different now :p
yangman: I'll run the quirks test later, then give radeon a try
yangman: there's people expecting me at a party
dmb: yangman, go get drunk!
cxo: Nooooooo
cxo: yangman has to write drivers!
airlied: hmm write dirvers or lie in hammock, I think hammock wins today.
cxo: any suggestions for my macro question by the way?
airlied: cxo: generally we just define something like SetBit and TestBit or explicitly do it
airlied: you are probably wrapping something and making it uglier
cxo: ugly, i know!
airlied: esp if someone wants to set multiple bits at once
DanaG: Oh yeah, as I noted earlier, I figured out a bit about what was up with Plymouth: it seemed to be trying to draw... on ttyS0.
DanaG: Beats me why.
cxo: cos ttyS0 is only the hottest place to draw on
DanaG: s/cos/sin/ cxo
DanaG: is teasing for the use of cosine to mean 'because'.
DanaG: =þ
cxo: multiplies with tan-1 to get back his cos
DanaG: =þ
DanaG: grr, I hate how people post threads like this:
DanaG: "look how awesome Catalyst 9.4 is! But oh! You can't have it! Neener, neener, neener!"
DanaG: http://www.phoronix.com/forums/showthread.php?t=15855
DanaG: punches that person in the face.
DanaG: ... or rather, drops him down an elevator shaft. No wait, that's a bit too violent.
benh: that would imply somebody actually cares about what's on phoronix
cxo: hmms and hmms again
cxo: puppy dawg eyes his HD4870 sitting on the table, soon!
DanaG: has an HD3650 in his laptop... and currently has crap battery life with radeon. =þ
cxo: phew, now to test this crap, i just ported about 100Kb of low level crap for a microcontroller in a fridge to another controller
cxo: kinda funny how that guy doesnt want to share his alpha driver
DanaG: Yeah.
DanaG: And there's a "fake" torrent of it somewhere... with zero seeders.
cxo: heh, fglrx has become more scandalous than an unreleased Tom Cruise film
DanaG: misses compiz, too.
cxo: sucks to have a laptop i guess, cant swap out video cards
DanaG: With mine, I could... but HP charges like 700 bucks for the Quadro that matches this laptop.
tlp: Sucks to have an old desktop PC that only has AGP :)
DanaG: They offer it with both ATI and nvidia; I chose ATI for lower power usage in Windows, and because radeon ⋙ nv.
cxo: would be nice if there was a standard for this stuff, and all you did was add accelerator cards, the ports would be on the motherboard
DanaG: Here's what I have: http://www.notebookreview.com/default.asp?newsID=4631
cxo: I have a Radeon 9700pro in my laptop, well my X-laptop, i gave it to a friend, i didnt need it anymore
cxo: looks expensive
DanaG: yup.
DanaG: But well worth it. Same target market as the Thinkpads.
cxo: i want one of these bad boys, one day http://www.sonystyle.com/webapp/wcs/stores/servlet/CategoryDisplay?catalogId=10551&storeId=10151&langId=-1&categoryId=8198552921644608896&parentCategoryId=16154
DanaG: http://forum.notebookreview.com/showthread.php?t=362354
DanaG: Looks like laptops are now coming out with R700.
cxo: has inbuilt gps too
DanaG: ugh, no touchpad.
cxo: throw it into the Jag, and go on mad road trip... yup, thats the plan
DanaG: http://www.notebookreview.com/default.asp?newsID=4796
DanaG: 16:9 display... really stupid.
DanaG: It means if you watch a 16:9 video, the only place a seek bar can go is ON TOP OF THE VIDEO.
DanaG: ... instead of below it.
cxo: the Vaio is sexier
DanaG: Can you play Neverball with it?
DanaG: =þ
cxo: probably, it has a intel vid card
DanaG: er, s/with/by tilting/
DanaG: =þ
cxo: oh, probably not, but i dont really want it for that sort of stuff, just to keep music in the car really
DanaG: http://www.bit-tech.net/news/hardware/2009/03/13/nvidia-not-to-rebrand-9800-gt-as-gts-240/1
DanaG: heh, the very fact that they even considered it... really says something about their practices.
cxo: everyone rebrands, even AMD
cxo: All is fair in love and business