Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-6-13

Search This Log:

DanaG: http://www.phoronix.com/scan.php?page=news_item&px=NzMyMA
DanaG: interesting/
DanaG: any chance they could do the same for non-Intel?
airlied: should be possible like anything
agd5f: _Groo_: you'll have to manually add modes other than the native mode of your laptop panel until we port the code that adds scaled modes automatically
MarcOChapeau: Hi there :)
MarcOChapeau: not sure I should report bugs here, but still : blender's menu don't show when using mesa master branch
MarcOChapeau: know issue ?
MostAwesomeDude: I think Blender menus use GL_SELECT, so yeah, known issue.
MarcOChapeau: ok, good
MarcOChapeau: and one more : when activating VBOs in nexuiz, characters and bonuses aren't rendered properly
osiris: airlied: the out of bo problem may be caused also by calling SetupTextures and emitting texoffsets with every render operation (even if textures hadn't changed)
_Groo_: hi/2 all
_Groo_: any dev alive?
_Groo_: glisse: ping
_Groo_: airlied: ping
_Groo_: bridgman_: ping
rhodan: are you aware of some xv-scaling regressions in the newest dri2-enabled git-branch combination?
_Groo_: ping all :P
MostAwesomeDude: _Groo_: What?
_Groo_: hi MostAwesomeDude
_Groo_: MostAwesomeDude: two questions
MostAwesomeDude: You can just ask your questions. You don't have to ping everybody. :3
rhodan: What is your name, what is your quest?
_Groo_: MostAwesomeDude: first.. should i stay following glisses branches as per his blog post, or since some trees are being merged into master, i should follow other path?
MostAwesomeDude: _Groo_: The only branch merged to master so far is radeon-rewrite, from Mesa.
_Groo_: MostAwesomeDude: but if it sync using the radeon-rewrite branch will i still be updated?
MostAwesomeDude: Not anymore.
_Groo_: so, i gust pull from git://anongit.freedesktop.org/git/mesa/mesa exclusively?
MostAwesomeDude: Well, if you have branches set up, just "git checkout master" and then "git pull" will still work.
rhodan: MostAwesomeDude: does xv with dri2 work for you atm?
MostAwesomeDude: rhodan: Haven't tested it in a long time.
MostAwesomeDude: I don't see why DRI2 would affect Xv though.
rhodan: MostAwesomeDude: ok
rhodan: MostAwesomeDude: for me, xv doesn't seem to scale the video correctly
rhodan: i only see a small part of it
rhodan: and the scaling changes as i move a window atop the xv window
_Groo_: rhodan: theres a patch for it
bridgman_: I saw that the first time I tried F11 on rv570
bridgman_: thought it was fixed though
rhodan: patch for what?
_Groo_: rhodan: for the xv problem
rhodan: i mean which software....
_Groo_: rhodan: ??? its a patch for the xv code, xf86video-ati, what are you asking?
rhodan: xf86-video-ati then. it could have been a lot of other things
rhodan: like, the kernel, although unlikely
_Groo_: rhodan: its the video ati code alright. i had this problem (scaling) three times.. and all the three i bugged the devs till they provide a patch :D
_Groo_: MostAwesomeDude: my second question is actually a bug? dri2/kms only gives me one mode setting (fortunatelly its the maximum for my LVDS).. is this normal?
rhodan: the patch doesn't seem to be included in the radeon-gem-cs3-branch
_Groo_: rhodan: its not
rhodan: why not?
_Groo_: rhodan: ask the devs :P
MostAwesomeDude: _Groo_: If your panel's EDID only has one mode, then yes, it's correct.
rhodan: _Groo_: i just did that...
_Groo_: MostAwesomeDude: can you explain a little better? i always had several modes with dri1... 1024x768, 800x600, etc etc... now i only have the max, 1280x800, even if i force it in xorg.conf
MostAwesomeDude: _Groo_: This is a KMS thing, not a DRI2 thing.
_Groo_: MostAwesomeDude: yes, i suspected that, since its now kms which changes modes
_Groo_: MostAwesomeDude: so xrandr only shows me this
_Groo_: MostAwesomeDude: http://pastebin.ca/1459106
MostAwesomeDude: _Groo_: I have no idea.
_Groo_: MostAwesomeDude: maybe glisse or airlied could take a look at this
agd5f: rhodan: http://www.botchco.com/alex/xorg/fix_xv.diff
agd5f: _Groo_: you'll have to manually add modes other than the native mode of your laptop panel until we port the code that adds scaled modes automatically
_Groo_: agd5f: hi alex, how do i do that?
agd5f: _Groo_: xrandr --newmode and --addmode
_Groo_: agd5f: hmm ok let me check
agd5f: _Groo_: http://wiki.debian.org/XStrikeForce/HowToRandR12
agd5f: _Groo_: for example to add 800x600, you'd need xrandr --newmode "800x600" 38.25 800 832 912 1024 600 603 607 624 -hsync +vsync
agd5f: then xrandr --addmode LVDS 800x600
MostAwesomeDude: agd5f: Is the idea to have the scaler automatically turn on for those modes? Auto-letterbox?
agd5f: MostAwesomeDude: the panel is always driven at it's native mode
_Groo_: agd5f: and how do i know how to make a modeline?
agd5f: the scaler can scale however you want
agd5f: centered, full, aspect
agd5f: _Groo_: use cvt
agd5f: cvt 800 600
_Groo_: agd5f: ahh nice :)
agd5f: MostAwesomeDude: in the ddx we just add a bunch of common modes for convenience since users expect non native modes to work
MostAwesomeDude: agd5f: Ah.
MostAwesomeDude: Wouldn't it be possible, with the scaler's assistance, to do *any* mode smaller than the actual panel res?
agd5f: MostAwesomeDude, _Groo_: the function in question is radeon_add_common_modes() in radeon_modes.c
agd5f: MostAwesomeDude: yes
agd5f: you can manually add any mode
MostAwesomeDude: Hm.
_Groo_: agd5f: the glisses blog post still applies right? the only thing that changes, is that now i sync directly from mesa master and ignore the radeon-rewrite branch, correct?
agd5f: _Groo_: yes
agd5f: MostAwesomeDude: in the ddx, when the LVDS mode isn't native, we program the crtc with the native mode and then set up the scaler to do the rest
_Groo_: agd5f: ok, conpiling it now :)
MostAwesomeDude: agd5f: Ah. Two steps ahead of me. :3
_Groo_: mesa from master i mean
MostAwesomeDude: agd5f: BTW... can we scanout from any pixmap, under the current CreatePixmap setup?
agd5f: MostAwesomeDude: avivo based radeons (r5xx+) have a much nicer scaler. easier to program and mre flexible
agd5f: MostAwesomeDude: yeah, you can point the crtcs at any address in vram
MostAwesomeDude: agd5f: Woot, that makes my life easier.
agd5f: MostAwesomeDude: You could even point them at gart memory, although I'm not sure how well that would work
MostAwesomeDude: agd5f: I only need to be able to reuse the current CreatePixmap hook.
agd5f: MostAwesomeDude: sometimes for debugging you can use radeontool or avivotool to point the crtc else where if you want to checkout something offscreen
bridgman_: AFAIK you can run our GPUs without video memory, doing everything in system memory
bridgman_: it performs as well as you would expect
MostAwesomeDude: agd5f: Oooh, haxxy.
_Groo_: agd5f: alex btw, any news on that garbled screen bug?
agd5f: _Groo_: nope
_Groo_: agd5f: its the only thing keeping me from using suspend :(
MostAwesomeDude: I'm just thinking about how to have Shatter best set up CRTC pixmaps.
agd5f: _Groo_: just use suspend to disk :)
MostAwesomeDude: Basically, if we can, I'd like to just reuse CreatePixmap for making CRTC root pixmaps.
_Groo_: agd5f: i have an aspire 5102 with ide PATA disk... suspend to disk and real reboot are almost the same time... suspend to sleep was a much better solution
agd5f: MostAwesomeDude: randr has a call to change the crtc offset
agd5f: MostAwesomeDude: you might want to look at the xrandr rotation code
MostAwesomeDude: agd5f: Yeah, I need to dig through more of xrandr.
agd5f: MostAwesomeDude: it creates a shadow framebuffer in vram and points the crtc at that, and then uses composite to do rotated blits from the real front buffer
_Groo_: agd5f: how about if you guys took a look at what vbetool quirk mode was sending to the card? maybe they tripped in something there, since it used to work with dri1 for me
agd5f: using damage to track the changes
agd5f: _Groo_: vbetool executes card specific code from the rom
agd5f: _Groo_: need to dump what regs it's hitting when you call the vbe post call
_Groo_: agd5f: i know, but vbetool fixed the suspend problem... since i dont know squat about this kind of code i cant even compare and see what you doing diferently
_Groo_: agd5f: its already in the bug report
_Groo_: agd5f: and it only worked for dri1 (obviously since kms call the shots now)...
agd5f: _Groo_: need to build vbetool against x86emu so we can dump what regs it's hitting. haven't had time to port it
_Groo_: agd5f: ok, i appreciate all you efforts
agd5f: _Groo_: vbetool doesn't have any machine specific code in it. the code to post the card in on the rom in your laptop, so there's nothing to debug in vbetool itself.
agd5f: just for reference
_Groo_: agd5f: thats what confuses me, why did it worked with vbe but not with the radeon driver itself?
agd5f: _Groo_: the code in the driver and the code in the bios are different
_Groo_: agd5f: so vbetool calls my rom directly and radeon uses its own registry calls, thats the difence?
_Groo_: diference
agd5f: so we have to compare what the bios is doing to fix what's in the driver
agd5f: _Groo_: yes, pretty much
_Groo_: agd5f: theres not even one rs485 card in AMD so you guys can test it 'in loco'?
agd5f: _Groo_: every oem board is different
_Groo_: agd5f: great :P
agd5f: what doesn't work for you might work for others.
_Groo_: agd5f: so my card from hell is evil just for my specific card and board., lol.. lucky me
agd5f: that's why suspend/resume is such a pain in the butt
_Groo_: agd5f: ow i jsu remembered... might be useful.
_Groo_: agd5f: the suspend used to work WITHOUT the vbetool, about a year ago.. maybe ubuntu 6.10 or 7.04 or such... after that it started showing this behaviour.. so some code was introduced at that time that broke my suspend
_Groo_: ok, two years ago
_Groo_: ::D
_Groo_: brb
agd5f: _Groo_: maybe on your laptop, the bios already posts the card or something and posting it again in the driver is what causes the problem
agd5f: it's too bad the acpi spec for suspend to ram is so poorly defined
_Groo_: agd5f: tell me about it :P
_Groo_: agd5f: but that reposting would cause the card to behave like that? it looks like the old crts with bad vertical syncs
agd5f: _Groo_: the change several years ago may be related to how the kernel interprets acpi. we pretend to be windows now, and I think we told acpi we were linux before. that may explain the need or not for vbetool
agd5f: as acpi does different stuff depending on what os it thinks is running
_Groo_: agd5f: can you make a patch emulating the old behavior so i can test and see if that solves the problem
agd5f: _Groo_: I'm not that familiar with the acpi stuff in the kernel, and I suspect it would be pretty invasive. best bet would be to try an older kernel from that time period
agd5f: without kms of course
_Groo_: agd5f: well i know for sure it worked then
agd5f: without vbetool?
_Groo_: agd5f: yep
agd5f: if so, track down what kernel you used and see how it announced itself to acpi
_Groo_: agd5f: vbetool was a workaround i devised AFTER you (devs) did something to the radeon code that started giving my card this behaviour
agd5f: _Groo_: could also be that the distro used to use vbetool by default and then changed to not using it
agd5f: _Groo_: if it was a driver change you could try git bisect
_Groo_: agd5f: nope, im pretty sure it was the card alone.. i checked that kinda stuff at the time
_Groo_: agd5f: it was a driver change.. i even at the time recompiled old behaviour and new behaviour.. but that was almost 2 years ago
agd5f: figure out what version of the driver worked at the time and then bisect to find out what change broke it
agd5f: anyway, I gotta run
_Groo_: agd5f: i give up at the time because vbetool solved my problem and airlied told me that the problem would go away when we changed to kms/dri2
_Groo_: unfortunatelly it didnt :P
osiris: bridgman_: are there any limits for vertex buffer size on r300-r500?
bridgman_: I'm not aware of any, other than size fields in PM4 packets etc.
bridgman_: I think we run with some pretty big ones in the proprietary driver
bridgman_: looks like maybe a 2^16 vertex limit in VAP_VF_CNTL
MostAwesomeDude: Hm. For packed verts, 4F, four cols, 8 texcoords, fog, & point size...
bridgman_: good point, has to fit into physical memory ;)
MostAwesomeDude: No, no, that's only 112 MB.
MostAwesomeDude: So the size limits are definitely in the packet size fields. :3
bridgman_: scary
MostAwesomeDude: Heh, "only."
bridgman_: hey, I just noticed VAP_ALT_NUM_VERTICES
bridgman_: Alternate number of vertices to allow >16 bit vertex count
bridgman_: 5xx accel guide v1.3 p241
MostAwesomeDude: That's, um.
MostAwesomeDude: 7GB.
MostAwesomeDude: So let's just say that the limit is 1GB, and leave it at that?
osiris: bridgman_: hmm, but how do I use this reg with PM4 packets?
bridgman_: I think you'd need to use a type 0 (? going from memory) packet to set the alt register
bridgman_: let me check the doc, hold on... I'm cascading too many guesses here ;)
MostAwesomeDude: osiris: There's an alternate pkt3 for it IIRC.
bridgman_: wow, it's amazing how many words have "alt" in them ;)
osiris: MostAwesomeDude: can't find such a packet
bridgman_: I think you need to use type 0 to set the alt register, then set the "use alt" bit in the vap_vf_cntl register slot of the type 3 packet used for draw
bridgman_: eg 3d_draw_vbuf_2 packet has a header plus vap_vf_cntl
bridgman_: I assume you use 3d_load_vbpntr to point to vertex buffers in the first place ?
osiris: bridgman_: will probably stay with INDX_BUFFER, it's almost the same as using this alt_num_vertices - the only difference is that with INDX_BUFFER the max is 8M dwords so 8M uint elts or 16M ushort elts, while with alt_num_vertices you could reach 16M with both types
bridgman_: really ? I thought I saw a limit of 16K indices... hold on
osiris: 16K when you put the indices in packet
bridgman_: oh yeah, 8192 *K*
bridgman_: d'oh !
bridgman_: the packet blurb mentions "further limited by the design of the vertex fetcher/vertex cache"
bridgman_: that sounds ominous ;)
bridgman_: maybe we should go with MostAwesomeDude's 1GB suggestion after all ;)
osiris: the HW limit is 32MB
bridgman_: that's indices though, not vertices; are you talking about drawing with just a vertex buffer or with vertex + index buffers ?
osiris: vertex + index buffer
osiris: for vertex buffer the GART + VRAM size is the limit probably
rhodan: can i has xv-scaling-fixing patch plox? it was mentioned earlier
_Groo_: bridgman_: how do i know i have fbos in dri2/kms? what are the tags in glxinfo?
osiris: _Groo_: EXT_framebuffer_object
rhodan: i have GL_EXT_framebuffer_object, is that a good thing to have?
_Groo_: osiris: ahh let me check
_Groo_: osiris: rhodan: thats what u have also
_Groo_: osiris: im asking this because i cant make wow work with wine and dri2/kms.. it complains of missing fbos :P
rhodan: _Groo_: i know, but is that a good thing?
rhodan: what does it do for me?
_Groo_: rhodan: dont know... fbos are important features for modern graphics but im no developer... osiris or bridgman_ might be able to enlighten you
rhodan: WoW with opensource-ati atm? you best be joking. performance is horrible, even with glxgears
rhodan: _Groo_: can you give me the link to that xv-patch
rhodan: *?
mcgreg: I wonder if wow will ever be plable with radeon
_Groo_: rhodan: unfortunatelly i restart X and lost the irc log.. i can give you the patch itself
osiris: with FBOs you can render to a buffer, that will be later used as a texture for another render operations. this avoids unnecessary data copies from VRAM<->RAM which are very time consuming
_Groo_: mcgreg: ironically i could play it pretty well with old radeon driver.. the new driver works to some extent.. its a matter of waiting for all the pieces joining together, maybe till 09 latest quarter
rhodan: _Groo_: give that patch to me now! ;)
mcgreg: the question is ... what is pretty well?
_Groo_: rhodan: let me dig it
mcgreg: 1680x1050 in high quality at 50fps? :)
bridgman_: I guess that depends a bit on the GPU ;)
rhodan: mcgreg: aliased 800x600 with diagonal tearing and no anisotropic filtering
rhodan: and about 15fps
rhodan: at least for me
_Groo_: rhodan: htto://pastebin.ca/1459183
rhodan: _Groo_: thx
_Groo_: mcgreg: i have a shitty 200m, anything that partially moves is a win :D
mcgreg: bridgman_: I can run it on my radeon 4670 with 1680x1050 in high quality + 2x anti aliasing and stuff at 50fps (in most cases) :)
_Groo_: rhodan: you welcome :)
bridgman_: mcgreg; that sounds about right... a 4670 has roughly 40x the shader power of an Xpress200M
_Groo_: bridgman_: ¬¬
mcgreg: bridgman_: ok, but I am talking abotu windows d3d stuff ;)
rhodan: htto? that's no protocol _i_ ever heard of. they speak hypertext in htto?
_Groo_: rhodan: bad typping hand, bad!
bridgman_: it's what came before http ;)
_Groo_: rhodan: im an old man ;)
bridgman_: probably 6x-8x the performance for simple operations
_Groo_: bridgman_: he got the picture :P ¬¬ no need to humiliate
bridgman_: x200m is 2 pipe with no vertex shaders
bridgman_: 4670 is 8 pipe with vertex shaders
bridgman_: and between 2x-3x the clock speed
_Groo_: bridgman_: 200m is half a pipe and total evil
mcgreg: hmm I like my 4670 a lot. no extra power connectors needed and it doesnt consume too much but still offer a fair 3d speed :)
bridgman_: it was also much faster than any other IGP on the market when it came out IIRC
rhodan: i'd like to have a card that makes my 2d desktop fast
bridgman_: the 200m, not the 4670
_Groo_: curses all ATI devs that devised 200m, burn you bastards burn in the fires of hell
rhodan: kde4.3 is pita atm, even without compositing
bridgman_: Groo, it wasn't the design it was the documentation; there wasn't as much of a paper trail left behind the 200m as there was for other chips
_Groo_: rhodan: what are you talking about???
bridgman_: that's why we gave agd5f a laptop with xpress200m ;)
bridgman_: now we need to get him something else so he can run linux on the laptop
bridgman_: right now it's how he gets into our internal systems
_Groo_: bridgman_: ¬¬ im gonna start killing kitties again
bridgman_: thinks quickly...
_Groo_: bridgman_: give me a better laptop and i send my 200m to agd5f
bridgman_: um... it runs Windows pretty good, doesn't it ?
bridgman_: it's... um... too slow for linux, yeah, that's it
rhodan: _Groo_: my card's 2D performance sucks
_Groo_: rhodan: recompile qt with raster by default.. be happy
_Groo_: rhodan: also use dri2/kms
_Groo_: and new xorg and mesa
_Groo_: be happy again
rhodan: i'll try raster
rhodan: what does raster do? useflag set, recompiling world atm
_Groo_: one kittie thrown into the wall, one kittie thrown... it splashed red all over the wall.. one less kittie to splash on the waaaaaaaallll
bridgman_: I live in the country; feral cats & kittens everywhere
bridgman_: keeps the coyotes fat & happy
_Groo_: rhodan: try it.. open akregator , kmail or other kde app with --graphicssystem raster
_Groo_: bridgman_: i live in brasil...
_Groo_: bridgman_: so i have a lot of stuff to choose when throwing animals at walls
_Groo_: bridgman_: endangered species are very nice to throw at walls
bridgman_: yeah, you still have an ecosystem down there
bridgman_: we just have vermin
_Groo_: bridgman_: not for long they dont ¬¬
rah: *cough*logging*cough*
rah: *cough*no*more*rainforest*cough*
rah: >_>
dileX: _Groo_: you are the kittie-man?
bridgman_: there were some young racoons hanging around by the side of the road a few days ago
bridgman_: I was asking around trying to figure out what they ate so I could drop some food for them
_Groo_: dileX: kittie man? no.. throwing kitties and puppies around is a old sport in #radeon channel
bridgman_: apparently the idea of raccoons eating anything but garbage hadn't occurred to anyone
_Groo_: bridgman_: ehehehe
bridgman_: fortunately they hooked up with their mother again
_Groo_: bridgman_: raccoon city... nice...
bridgman_: yeah, I figure there's some kind of underground hive out there
_Groo_: bridgman_: i would looooove to shoot some zombies... well.. who doesnt like to shoot zombies
bridgman_: just haven't found it yet
bridgman_: Groo; good idea, maybe I'll set up a zombie stage at the next IDPA competition
bridgman_: I've been meaning to get back into setting up the shoots
rhodan: but what does it do?
bridgman_: I'm getting sick of the wild west stuff
_Groo_: and we dont even have to wait for them to become zombies ... just shoot them.. eventually they will buahahahahahha <- evil, cold laughter
rhodan: http://www.psycoreboard.de/component/option,com_puarcade/gid,241/
bridgman_: yeah, why wait until the last minute
_Groo_: dileX: we throw little fluffy things everytime anyone mentions the evil green one in here
osiris: glisse or airlied: the big comment in r300RunRenderPrimitive (r300_render.c) doesn't apply anymore, right?
dileX: _Groo_: speaking of superstition?
_Groo_: rhodan: 1260 meters!
bridgman_: I can't get the kitty over the phone booth
bridgman_: 671 meters
_Groo_: bridgman_: thrown him at the woman
bridgman_: no strategy, it's a weekend
rhodan: throw it at 1/2pi rad
_Groo_: 1298 meters
DTEIT: i'm trying to use radeon driver on a ati hd3300 (790gx chipset)
DTEIT: it worls, but often i get blank screen
DTEIT: is this normal?
phoenix64: 2713m *g*
phoenix64: no, 2961
_Groo_: phoenix64: funf
bridgman_: funf like "four" ?
dileX: 2943 meters
bridgman_: or maybe five, I forget...
DTEIT: funf = five
bridgman_: ein, svei, drei,
bridgman_: thanks ;)
mcgreg: zwei
_Groo_: bridgman_: funf, like i cant make the damn cat go above 1300 meters ¬¬
phoenix64: lol, my scores were like, 100, 2400, 200, -100, 300
phoenix64: no, wait, the third was much more
bridgman_: ok, my German is basically what my father learned in WW2, passed on through a series of anecdotes ;)
phoenix64: fuenf btw *g*
bridgman_: ok, good to know
phoenix64: or, fünf for those with german keyboard
phoenix64: (as in, five)
bridgman_: I knew there had to be an umlaut in there somewhere
DTEIT: uhm...is this a secret coded language or are we really on #radeon? :-P
dileX: fuenf is OK
bridgman_: we were talking about kitten abuse and rhodan posted a link to a flash game
bridgman_: so right now this is more like #kittenthrowing
rhodan: über neuntausend!
DTEIT: ahhh....good to know :-)
dileX: rhodan: lüger!
rhodan: das ist der wahrheit!
_Groo_: bridgman_: so your father learned how to count corpses in german LOL...
dileX: rhodan: I have a screenshot
bridgman_: no, he was actually town major somewhere in Austria
rhodan: my grandfather died in auschwitz
_Groo_: bridgman_: oh... austrian descendent
rhodan: he fell off a watchtower
Wizzup: I have a question. I've recently bought a new laptop, and I've installed gentoo on it, with a custom made kernel. I've said 'Y' to the parts in the kernel that should be enabled for radeon. I however, am not able to find any radeon notification in dmesg, is this normal?
FallenWizard: is Austrian
rhodan: and here comes the beard:
rhodan: |||||
rhodan: ||||
_Groo_: rhodan: sorry to hear that :( im portuguese, so i didnt had anything to do with the holocaust
DTEIT: Wizzup: yes if you don't specify which driver to use for the console then is using the vesa one
_Groo_: unless someones beats my record.. im the kitty thrower master for the day :D
bridgman_: actually my father was British Army, but the family is kinda half Irish & half German
_Groo_: Wizzup: try modeset=1 when booting grub
Wizzup: DTEIT: Ok, I use uvesafb for FB. But my X doesn't seem to be able to find the radeon module either. (And I have emerged xf86-video-ati, the radeon driver)
_Groo_: bridgman_: i like irish.. good drinking buddies
rhodan: i was about to flood an ascii-beard, but then my router cut off my internet. wise router
bridgman_: yep; short tempers though ;)
bridgman_: ah, here we go... looks like Hermagor, Austria
DTEIT: Wizzup: look in the /usr/lib/xorg/modules/drivers
bridgman_: and here's the kitten game for all you new people ;)
bridgman_: < rhodan> http://www.psycoreboard.de/component/option,com_puarcade/gid,241/
DTEIT: if you have the radeon driver
DTEIT: (the path is for my debian)
Wizzup: DTEIT: ati_drv.so radeon_drv.so vesa_drv.so
Wizzup: I'll play with it a bit more
hifi: Wizzup: you forgot Direct Rendering Manager completely
mcgreg: rhodan: seriuosly, no offense to you, buit that joke about you father is auschwitz is really lame. it's totally not funny
Wizzup: I got it to work on my old laptop too, so I should be able to get this working too. I'll get back to you
Wizzup: hifi: Radeon has no support for DRI agaik
Wizzup: afaik*
Wizzup: for my card*
hifi: does not?
DTEIT: if you specify the radeon drv in the xorg.conf what does tell you?
Wizzup: ok - radeon is loaded by x
_Groo_: mcgreg: well my conscious tells me its not funny.. but i cant help a little jiggle hhihi
bridgman_: yeah, I kinda laughed and felt guilty at the same time
Wizzup: I'll try turning on DRI. It might solve my problem.
bridgman_: DRI should be turned on automatically with radeon; radeonhd needs it turned on manually for 6xx and up
hifi: I might be full of crap but R500 and up need the firmware stuff in kernel to do any acceleration
bridgman_: hifi; yep, all you get is shadowfb without drm and dri enabled
Wizzup: bridgman_: It indeed told me that it was using shadowfb
Wizzup: Thanks! Will report back
bridgman_: actually you don't need DRI (that's the protocol between a direct rendering driver like Mesa and the X server) but you do need drm
mcgreg: bridgman_: the point is, MANY ppl died and many ppl might feel badly offended about that. telling such a joke in a open chat ro0om isnt really that good.
Wizzup: well, DRM_RADEON and DRM are both on y
_Groo_: mcgreg: exactly, and for the record i WASNT in the watchtower at the time..
hifi: Wizzup: and dmesg shows nothing about drm or radeon?
_Groo_: watchtowers were VERY dangerous back then oO
DTEIT: Wizzup: if your x server starts with radeon driver then should be ok...check the xorg log for any problem
Wizzup: hifi: It inits drm
Wizzup: DTEIT: X is my problem atm ;)
rhodan: mcgreg: i said i was going to flood the channel with a beard, the beard of that joke
DTEIT: Wizzup: it doesn't start?
Wizzup: DTEIT: Well, more like it won't close either.
Wizzup: I have to use kill -9 to kill it
bridgman_: Wizzup, I don't understand "DRM_RADEON and DRM are both on"
DTEIT: what the log says?
Wizzup: bridgman_: kernel settings, so drm is set to yes
DTEIT: check for EE in the log
Wizzup: DTEIT: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed to open the DRM
Wizzup: may I paste three lines?
bridgman_: sounds like drm is too old; which distro are you running, and if you updated kernel which kernel ?
Wizzup: nv, i'll pastebin
Wizzup: bridgman_: kernel .29, Gentoo
bridgman_: 3 lines is probably ok
bridgman_: .29 doesn't have 6xx/7xx support
Wizzup: http://wizzup.pastebin.com/d27453aed
Wizzup: Ah. Could you link me to a page on the naming of the cards? I haven't understood that fully yet
scarabeus: Wizzup: emerge 2.6.30-r1 in there is the support :]
Wizzup: scarabeus: Thanks
scarabeus: np :]
DTEIT: no one getting some blank screen with radeon 3300?
_Groo_: only 2372 meters now :P well bbl.. seeya ppl
stikonas: does any of the developers here know that FBO's are not working with RS480 chip and DRI2?
_Groo_: stikonas: i warned them but they didnt listen!
bridgman_: Wizzup; here's the best list I have found, but it's big :
bridgman_: http://en.wikipedia.org/wiki/Comparison_of_ATI_Graphics_Processing_Units
bridgman_: might be easier to tell us what you have and we'll translate into dev-speak
Wizzup: bridgman_: Thank you, I like big wikipedia pages. ;-)
bridgman_: well then this should make you very happy ;)
Wizzup: It is indeed. :)
Wizzup: bridgman_: My card is a Mobility Radeon HD one, (3650), any other things I should know about it and the radeon driver? (I know 3D isn't supported yet, but anyting else?)
bridgman_: Wizzup; should be pretty much like any 3650; we treat mobiles quite differently with Catalyst drivers because laptop vendors tend to get customized drivers (so it's hard to provide generic fglrx upgrades for them) but the open source drivers treat all the chips pretty much the same
Wizzup: Ok, thanks.
bridgman_: there may be some things that work with Windows drivers obtained through the laptop vendor that either don't work or work differently with the open source drivers, but they tend to be bell & whistly kind of things
Wizzup: Right
bridgman_: on the 3d side we've finished the transition to the radeon-rewrite code base and i think the current problems (not drawing properly after clear etc..) are understood, so hopefully we should be back to a normal rate of progress in a week or so
bridgman_: hybrid stuff is going to be the toughest to bring over because X doesn't really understand the concept of flipping different hardware under a running system
bridgman_: there are also some interesting IP issues since the work was done collaboratively between a bunch of companies, so getting approval to release info on the hardware is kind of non-trivial
bridgman_: everything else is easy by comparison
Wizzup: Very nice
Wizzup: Lookin forward to it :)
bridgman_: we'll get there, but I think it's safe to say 3d will get there first ;)
spstarr: goes for a nap
DanaG: oh yeah, the Win7 drivers 9.5 worked fine for my Mobility "FireGL V5700".
DanaG: Calls it a 3650, though.
DanaG: I really don't know what differences are left between the Radeon and FireGL in Vista and in Linux, though.
osiris: bridgman_: looks like this VAP_ALT_NUM_VERTICES is also needed for INDX_BUFFER packet. too bad the VAP_ALT_NUM_VERTICES exists on r500 only (the r300 docs don't list this reg)
bridgman_: I think you get different optimizations when running on FireGL hardware
bridgman_: Carmack said it best; paraphrasing, I think he said something like "consumer driver developers do stupid driver tricks to make Quake3 run fast, while workstation driver developers do stupid driver tricks to make CDRS run fast"
bridgman_: or something like that
bridgman_: osiris; give me a minute, if I can get into the office systems I can check to see if the alt register really exists on 3xx
bridgman_: here we go : "The workstation vendors do stupid driver tricks to make CDRS go faster, while consumer vendors do stupid driver tricks to make Q3 go faster."
spstarr: bridgman_: VPN works? :-)
MostAwesomeDude: Personally, I think we should do stupid driver tricks to make *everything* go faster. :3
bridgman_: yeah, but that's really hard
bridgman_: the good news is that the open source drivers still have room for a bunch of generic optimizations
bridgman_: but at some point you get into the "making A go faster makes B go slower" tradeoffs
bridgman_: then everything gets yukky
Emme_NK_: agd5f: ping? Could you analyze the register dumps?
DanaG: I do remember driver-modding my old 9800 Pro to a FireGL X2, and getting massive performance improvement in AutoCAD under XP.
bridgman_: yep; workstation and consumer apps tend to give totally different workloads to the driver, so you can optimize for one or the other
bridgman_: workstation apps tend to have much more vertex processing while consumer apps tend to have much more fragment processing
DanaG: Trying AutoCAD 2006 (the only workstation app I own) under Vista, though, I didn't notice any such difference between FireGL and GRadeon drivers.
DanaG: er, keyboard lag.
bridgman_: the drivers should behave the same; in theory they key off the hardware you have
bridgman_: so you should get workstation code paths with either driver
bridgman_: the FireGL drivers are just ISV certified for a set of workstation apps
bridgman_: so they probably have a small, specific set of bug fixes
bridgman_: and more importantly don`t have a big heap of consumer bug fixes which invalidate the workstation app testing
DanaG: hmm, actually, the consumer performance was basically no different between the two, because they seem to key off the executable file name.
DanaG: Oh, and AutoCAD 2006 uses OpenGL overlays, which don't play nicely in Vista, anyway.
_Groo_: bridgman_: how can i test to check if fbo is working with my rs485?
bridgman_: Groo; ask Osiris ;)
osiris_: _Groo_: search for *fbo* apps in mesa/progs/
hbock: hey all
hbock: having a problem with my hd3850, an output just randomly stopped worked
hbock: working*
osiris_: _Groo_: piglit has some fbo tests too
hbock: now it thinks this DVI-I connector is always disconnected, no matter what monitor i hook up to it :(
osiris_: bridgman_: I tried using VAP_ALT_NUM_VERTICES like you've said, but it ended hanging the GPU
rhodan: raster didn't bring any improvements for me
rhodan: could be i/o related though. sometimes my cursor doesn't move at all when something is loading. i thought that would have been fixed with 2.6.30...
bridgman_: osiris; does it work OK when you don`t have the se alt`bit set in vap_vf_cntl but hangs when you do have it set
bridgman_: yes my system is mucking up punctuation, time for a reboot
[Enrico]: rhodan: only at half. .31 should contain the whole fix, but mouse stopped is........ strange
osiris_: bridgman_: it works ok if I don't set "use alt..." bit
bridgman_: osiris; I confirmed the reg only appears in 5xx, ie it is not in r420
bridgman_: I`ll look around & see if there are any other tricks associated with that register
_Groo_: fbo appears broken with rs485 allright.. i run fbotest1 and 2... garbled window instead of the desired output
_Groo_: fbotexture broken, fbotexture -ds broken
_Groo_: same for -arb
_Groo_: fbo_firecube gives mixed results
_Groo_: broken also
_Groo_: so.. fbo looks pretty bugged in rs485 with dri2/kms
aliverius: goodevening
aliverius: i was wondering if there is a way to read my chip's freq and temp?
soreau: Do you if your hardware already has those features?
aliverius: supposedly yes
aliverius: i remember catalyst showing them
aliverius: i want to see if some flags in my xorg.conf have made any measurable difference in power consumption
aliverius: glxgears go 10x slower now that i activated kms in my kernel...
aliverius: maybe is it mesa? (7.4.x)
stikonas: aliverius: yes, you are using software rendering
stikonas: update to the very latest git version
_Groo_: stikonas: also needs updated libdrm and ddx
airlied: glxgears isn't a benchmark, they wil lgo slower with kms
airlied: kms isn't a performance optimisation
rhodan: airlied: how do i determine if i'm using ttm?
airlied: if you are using kms
rhodan: airlied: then i'm using ttm :)
_Groo_: hi airlied
_Groo_: airlied: did you read the backlog? fbos arent working with rs485
airlied: _Groo_: next week sometime I'll probably take a look
_Groo_: airlied: ok, fbos are part of what tree? mesa, drm, drm-next?
airlied: mesa
_Groo_: airlied: ah ok.. just gonna wait and pull/compile as usual
_Groo_: airlied: at least it explains why wine is complaining about missing fbos :)
Wizzup: Is radeonhd worth a shot if radeon doesn't work?
FallenWizard: Why doesn't it work?
Wizzup: Well,
Wizzup: (Sorry for my late reaction)
Wizzup: I've recently compiled kernel 2.6.30-r1, since it is said to have dri support for r6xx and r7xx, and but X keeps telling me it can't find the radeon module
Wizzup: (Using gentoo)
FallenWizard: is it loaded?
FallenWizard: the module
Wizzup: I compiled it in the kernel first
Wizzup: I'm now rebuilding the kernel with radeon as module
Wizzup: Where does X look for the radeon kernel module?
[Enrico]: it doesn't does it?
airlied: just calls modprobe usually if its missing
Wizzup: I want to fix this -
Wizzup: [drm] failed to load kernel module "radeon"
Wizzup: (EE) RADEON(0): [dri] RADEONDRIGetVersion failed to open the DRM
Wizzup: [dri] Disabling DRI.
airlied: so does modprobe radeon show any radeon lines in dmesg
[Enrico]: Wizzup: what kernel and wha card? (btw next time write the error first :D )
airlied: or does dmesg |grep radeno do anything?
Wizzup: I've just rebooted. Hang on
airlied: grep radeon even
Wizzup: dmesg | grep radeon gives nothing. kernel .30-r1, card: 3650 mobilit radeon
Wizzup: I just want 2d support with radeon, really. But I can't get X to start
Wizzup: (gentoo)
[Enrico]: Wizzup: gentoo? like me, i have a 3470 not problem at all
[Enrico]: Wizzup: i use the lastest testing xf86-video-ati and lastest testing mesa too
[Enrico]: Wizzup: and indeed the lastest testing libdrm
Wizzup: I'm not sure if I am using the latest libdrm - what do you mean?
[Enrico]: x11-libs/libdrm-2.4.11 i'm using this
Wizzup: I'll add it to my keywords
[Enrico]: Wizzup: and you need to to get radeon to work (and the lastest xf86-video-ati too) mesa is not needed since no 3d is availble now
Wizzup: It wasn't dragged in by default, btw
Wizzup: (any libdrm)
Wizzup: I take it you do not use overlays?
Wizzup: Just the latest version in portage?
[Enrico]: indeed
Wizzup: Ok.
Wizzup: Hang on - gotta do some compiling. :-)
[Enrico]: sure
Wizzup: [Enrico]: Do you have Radeon DRM as Module in the kernel, or compiled in it?
[Enrico]: Wizzup: module
[Enrico]: Wizzup: i have also fglrx, just in case i need 3d, so radeon must be unloaded (and anyway i think it is a good practice to have the drm module as module :D)
Wizzup: Oh - a little noob question. How do I make sure the module is in the right place?
Wizzup: Wow- the kernel module actually loaded just now
Wizzup: tests
[Enrico]: lol
Wizzup: At least I am getting somewhere now. heh
[Enrico]: :D
[Enrico]: Wizzup: so what's the situation
Wizzup: Is it OK to ask you guys to take a look at my hand written Xorg.conf? I fear I might have made a mistake
[Enrico]: Wizzup: first show us the errors :D
[Enrico]: errors are very usefull
Wizzup: I don't actually have an error now, I just got a black screen
Wizzup: So the module was loaded :)
[Enrico]: Wizzup: look in xorg.0.log
Wizzup: I am, no errors
[Enrico]: O_O
Wizzup: I've been living in that log lately
[Enrico]: Wizzup: can you paste it? (in something similar to nopaste pls :D )
Wizzup: Yes. There is only one error - and that was obviously caused because I ^C'ed it after 10 seconds
Wizzup: http://paste.pocoo.org/show/122965/
[Enrico]: Wizzup: btw what version of xf86-video-ati are you using?
Wizzup: Latest
Wizzup: 12.2, IIRC. I'll check
[Enrico]: 12.2 is fine
[Enrico]: like me
Wizzup: Yes, latest.
Wizzup: Those last repeating errors were only because I didn't shut X down nicely
[Enrico]: may be you should recompile xf86-video-ati against the newer libdrm
Wizzup: I have :)
Wizzup: and xorg-base too
Wizzup: Just not mesa - but I can compile that now too.
[Enrico]: mesa is not needed
[Enrico]: no 3d for r6xx for now :'(
Wizzup: Well, it's better than no X at all. ;)
[Enrico]: Wizzup: btw what's xorg-base?
Wizzup: x11-base/xorg-server * my bad
[Enrico]: oh ok :D
[Enrico]: well ok paste your xorg.conf i will check it with mine
Wizzup: It's quite a small one. http://paste.pocoo.org/show/122966/
[Enrico]: Wizzup: oh btw if you run eselect opengl list what's the output?
Wizzup: Only xorg-x11
[Enrico]: ok
[Enrico]: Wizzup: btw you can enable Option "AccelMethod" "EXA"
[Enrico]: Wizzup: i've tested with EXA on (and even composite on)
[Enrico]: and it worked!
Wizzup: Ok
Wizzup: Any "Load" statement I might be missing?
[Enrico]: Wizzup: and btw your log says that (==) AIGLX enabled
[Enrico]: so composite is enabled
Wizzup: Weird... Cause I set to "Disabled"
Wizzup: Maybe I should just comment it fully, heh
[Enrico]: Wizzup: no i don't think so, Loaded plugins are almost automatically choosen
Wizzup: Ok
Wizzup: I enabled EXA and it did not matter. Perhaps I should reboot once againb
Wizzup: again*
Wizzup: Since my framebuffer is gone too - even after I've killed X
[Enrico]: no it is not windows
Wizzup: I haven't been able to get my FrameBuffer back in another way though
[Enrico]: well i see that there are no Modelines in your xorg.conf
[Enrico]: Wizzup: try to comment all the screen section but leaves Identifier "My ATI Screen"
Wizzup: ok
[Enrico]: Wizzup: anyway i suggest you to leave EXA enabled, it helps a lot :D
Wizzup: ok
Wizzup: 4496 pts/3 Z< 0:00 [X]
Wizzup: heh
[Enrico]: EXA provide 2d acceleration
[Enrico]: that's not good
[Enrico]: kill any instance of X first
Wizzup: I am, hehe
[Enrico]: :D
Wizzup: Still no luck after commenting the Screen section. :-(
[Enrico]: another black screen i guess
Wizzup: It is *still* black, I don't get my framebuffer back
[Enrico]: Wizzup: let's try a little mad thing. try to move xorg.conf to xorg.conf.bak and try to startx without xorg.conf (i doubt it will work with an ati card but..... we will see) btw how did you start X?
Wizzup: xinit
[Enrico]: xinit?
[Enrico]: not startx?
Wizzup: No, but I've tried startx too
[Enrico]: and btw do you have something in your .xinitrc ?
[Enrico]: anyway i suggest you to use startx
Wizzup: Ok. I will. Let's hope it makes a difference
[Enrico]: well startx is the standard way to start the X server
[Enrico]: and it calls automatically xinit
Wizzup: Starting without a conf gives a black screen
[Enrico]: again???
Wizzup: The only way I've gotten X to work currently is with vesa
[Enrico]: ok that's surprise me
[Enrico]: it should just work without an xorg.conf
[Enrico]: or not work at all
[Enrico]: Wizzup: what you ~/.xinitrc contain?
Wizzup: empty
[Enrico]: mhm
[Enrico]: Wizzup: what's your favourite DE/WM?
[Enrico]: and is it installed?+
Wizzup: I usually use gnome
Wizzup: No, I haven't installed gnome and/or any window manager
[Enrico]: eheheheh
Wizzup: I just wanted to get X to work first. ;)
[Enrico]: without any WM ?
[Enrico]: how?
[Enrico]: X need a wm
[Enrico]: needs*
Wizzup: I think X has one
[Enrico]: i don't remeber :'(
Wizzup: It has some default at least
Wizzup: Doesn't look great - but if that one works, so will the others
[Enrico]: did you have xterm installed?
Wizzup: Yes
Wizzup: With Driver "vesa" it opens xterms
[Enrico]: then put exec xterm in ~/.xinitrc
[Enrico]: and then try to startx again
[Enrico]: and let's see
Wizzup: Hm, I can try. I thought it started an xterm by default though?
[Enrico]: eheheh again i don't know, never tested
[Enrico]: i've installed kde and X at the same time
Wizzup: It opens one if I use the vesa driver - so it should
Wizzup: Very brave. :P
[Enrico]: should is not an option, i want to be sure, and if you put exec xterm in ~/.xinitrc we are sure :D
Wizzup: ok
[Enrico]: (btw if with vesa it starts xterm there is no def wm :D )
Wizzup: Nothing happens
[Enrico]: same blackscreen ?
Wizzup: The screen stays black. (Still without a config btw)
[Enrico]: this is odd
Wizzup: That's what I thought
[Enrico]: i'm missing something but i don't get what
[Enrico]: here works perfectly
Wizzup: I think I'll call it a night - I've made progress today, but it's 4 AM now... I'd better get some sleep
Wizzup: Thanks for the great help
[Enrico]: even composite works (ok xrender is slow, but it works)
[Enrico]: Wizzup: yeah better to sleep a little then try again (it's 4 am here too :D )
Wizzup: :-) Bye
[Enrico]: may be with some dev around :D
[Enrico]: Wizzup: good night
Wizzup: night :)
[Enrico]: Wizzup: btw i'm sure a dev can solve your problem quickly
Wizzup: I have confidence and patience.
[Enrico]: Wizzup: wise D:
[Enrico]: :D
Wizzup: Night
[Enrico]: noght
[Enrico]: night*
TCW: it's impressive how good radeon works compared to fglrx. I may have some issues which prevent me from using 3D fully, but at least 2D performs extremely well.
patapouf: TCW: what is you card ?
TCW: patapouf, rv530
patapouf: TCW: good it's work fine !
TCW: 04:00.0 VGA compatible controller: ATI Technologies Inc RV530 [Radeon X1600]
TCW: patapouf, huh?
patapouf: Well, on my side nothing work fine. But it's doesn't seems to be related to the driver ..
TCW: patapouf, what issues do you have?
patapouf: It's metacity + compositing that are in problem : http://yfrog.com/18screenshot1gp
patapouf: TCW: Doesn't look very nice hein ?
TCW: patapouf, looks interesting... whyt chip do you have?
TCW: what even
patapouf: TCW: it's a M22 (support by r300)
TCW: I have the impression this chip schould be supported very well
TCW: patapouf, lspci?
patapouf: TCW: 01:00.0 VGA compatible controller: ATI Technologies Inc M22 [Mobility Radeon X300]
patapouf: TCW: withing Xorg.0.log I got (II) AIGLX: Loaded and initialized /usr/lib/dri/r300_dri.so
TCW: patapouf, me too... r300_dri.so has support for r300-r500 chips
patapouf: TCW: ha .. interesting ...
TCW: is it?
patapouf: TCW: No idea
patapouf: TCW: We should ask one of the expert
TCW: it may be confusing to a certain degree, but it isn't :)
TCW: patapouf, I got that one from one of the experts
patapouf: TCW: Wll, it seams there is no r400.so or r500.so :S lol
TCW: right
patapouf: good to know
spstarr: yes
spstarr: r3xx = r3/4/5 in DRI
TCW: spstarr, ah... hi there :)
spstarr: im just a user ;p
TCW: spstarr, me too, so what? :)
TCW: spstarr, but at least you are a more knwoing user? :)
spstarr: sometimes
TCW: knowing even
TCW: spstarr, I mean related to radeon stuff ;)
TCW: no one can bet my knowledge in other areas! ;)
soreau: Hi TCW
TCW: cheers soreau
soreau: Cheers indeed =)
TCW: :)