Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2008-3-28

Search This Log:


edgecase: save->tv_pre_dac_mux_cntl = (RADEON_Y_RED_EN
edgecase: seems like SCART shoud have a different mux setting, i wonder if it even works
airlied_: nggh.. re-wrote some bits of the r300 rs and tcl for rs690, but still no compiz..
airlied_: clearly I need a better test case
MostAwesomeDude: airlied_: Long time no see. Been enjoying break. At any rate... zsnes runs at full speed, although it pegs a core doing so. Wine completely fails in interesting and fglrx-like ways.
airlied_: MostAwesomeDude: hey... was wondering where you disappeared to :)
edgecase: what is the DIN connector type?
agd5f: edgecase: on legacy radeons, we don't explicitly set the panel format as the bios does it for us. it's in LVDS_GEN_CNTL. the information is also in the LVDS table in the bios
agd5f: I've been meaning to add code to the bios parser to grab that data, but the current code works, so I haven't gotten a chance to mess with it
agd5f: edgecase: DIN connectors are usually the ones that you can plug s-video or component dongles into
edgecase: oh what are usually called mini-DIN then?
agd5f: I guess.
edgecase: an obscure feature of ImpacTV2 is SCART output, which i'm gussing is TV sync + RMX + rgb out?
edgecase: i came close to grabbing PCI Rage with video IN yesterday, local university selling G4 Imacs
agd5f: edgecase: the tv encoder can do scart by itself I think
edgecase: ok my question is, is it RGB? SCART devices can do composite, s-video, or rgb
quicksilver: would be nice if was RGB. My native pessimism suggests it will be s-video
edgecase: so, load detection, is that an analog circuit that detects if a given DAC output line is open circuit or 75 ohm terminated... question is, is the logic to deduce output type given multiplexed output pins, in hardware, or software
edgecase: atom_bios_dac_load_detect() coding suggests it's in software
edgecase: quicksilver, what would be the point of having a separate connector type then?
quicksilver: consumer convenience given that lots of people have SCART-SCART leads
quicksilver: rather than needing an s-video-scart lead
quicksilver: but heck, maybe my pessimism is misplaced :)
edgecase: sure, but why driver support if it's just s-video?
edgecase: my guess is, in the mini-DIN connector, there are composite, y, and c pins, and another that's special to scart
edgecase: so based on which pins are loaded/unloaded, driver can deduce what adapter cable is plugged in
edgecase: on my X300 with 7-pin din, I notice it boots up in a different mode depending what RCA jacks are hooked up at power-on
edgecase: if the Y pin is shared between composite, y/c, and "scart" rgb, you make a truth table:
edgecase: y loaded, c unloaded, B unloaded = composite
edgecase: y loaded, c loaded, B unloaded = s-video
edgecase: y (G) loaded, c loaded (R), B loaded = scart RGB mode
edgecase: here's the good stuff RADEONRestoreLVDSRegisters()
agd5f: edgecase: initLVDSregsiters() as well
edgecase: ulData = (RADEON_Y_RED_EN
edgecase: | RADEON_C_GRN_EN
edgecase: | RADEON_CMP_BLU_EN
edgecase: this seems like the part that switches DAS outputs from RGB to Y/C to maybe Component?
agd5f: edgecase: for load detection take a look at radeon_detect_tv()
edgecase: i'm starting to wonder how the hell anyone figured this out... must have had datasheets at some point
agd5f: edgecase: yeah, it's a pain in the butt
edgecase: bodes well for kernel modesetting... most of the stuff is known, and has working code, just need to refactor the code
edgecase: or should I say Just
edgecase: lowercase doesn't do it justice
edgecase: LTU, is that gamma correction?
agd5f: edgecase: LTU = linear transform unit
edgecase: i gathered that, it can be a source for a DAC, what's it transform tho?
edgecase: # define RADEON_LVDS_PANEL_TYPE (1 << 2)
edgecase: # define RADEON_LVDS_PANEL_FORMAT (1 << 3)
edgecase: # define RADEON_LVDS_RST_FM (1 << 6)
agd5f: edgecase: part of the overlay
edgecase: ic
edgecase: those constants look like they might be related to 1/2 ch and 18/24 bit LVDS modes
agd5f: RADEON_LVDS_PANEL_FORMAT toggles 18/24 bpp panel
edgecase: ah sweet that's part of the secret sauce i've been looking for
agd5f: RADEON_LVDS_RST_FM resets the frame modulation
agd5f: RADEON_LVDS_PANEL_TYPE single/dual pixel per clock panel
edgecase: ok that's the rest, woot!
edgecase: now i just need some hardware that exposes those pins
edgecase: i guess i could write some code to print those bits, in order to test various laptops and see how they're running
edgecase: a laptop with dual pixel and 24 bit, could then be wired to a 18bit and/or single pixel display, and the registers set accordingly
edgecase: hmm where is it in LVDS bios table? I could download bios upgrades, and scan them, no need for real hardware
agd5f: edgecase: add this function to radeontool: http://www.botchco.com/alex/xorg/lcdbiostable.txt
agd5f: it will print out the lcd into table
agd5f: *info
edgecase: i see 888vs666, but not single/dual pixel per clock
agd5f: edgecase: i might have left that out
agd5f: lemme check
agd5f: edgecase: offset 0x1d-0x1e is a big bitfield
edgecase: I think RADEON_TV_RGB_CNTL will help with SCART RGB tvout
edgecase: looks like TVout can do genlock sync in?
agd5f: edgecase: updated http://www.botchco.com/alex/xorg/lcdbiostable.txt
edgecase: which version of radeontool should i apply that to?
agd5f: edgecase: any version from airlied'd git tree
agd5f: you need to actually call the function too
edgecase: did you just write it?
edgecase: got a git url? i've not used his repos before
edgecase: http://gitweb.freedesktop.org/?p=users/airlied/radeontool.git;a=summary
edgecase: ok maybe call in radeon_rom_tables() for now
agd5f: git clone git://people.freedesktop.org/~airlied/radeontool
edgecase: yeah got it
edgecase: just putting in call to your func
agd5f: I wrote the lcd stuff, I;d send you a patch, but I got a lot of other crap in my radeontool tree at the moment
edgecase: well this'd be nice to merge
edgecase: gimme a minute to try it out
agd5f: edgecase: yeah, I got a bunch of stuff that should be merged
agd5f: edgecase: you'll need a bios to look at
edgecase: running machine?
agd5f: edgecase: you need a file
edgecase: it's trying to mmap
agd5f: cd /sys/bus/pci/devices/
agd5f: echo 1 > rom; cat rom > /tmp/vbios.rom; echo 0 > rom
agd5f: radeontool romtables /tmp/vbios.rom
edgecase: ok it's telling me my Thinkpad T30 is RBG 666, which is consistent
edgecase: color dual TFT <- is that saying dual-pixel per clock? This machine is definitely single-pixel
edgecase: LVDS_GEN_CNTL (02d0) 0x000dffa1
edgecase: PANEL_TYPE and PANEL_FORMAT = 0
edgecase: agd5f, i can't see that it's reporting single/dual status
agd5f: edgecase: yeah, I'm not sure how that bit gets set off hand
edgecase: i'll bug you again when i have a radeon laptop with 24bit and/or dual-pixel panel
agd5f: edgecase: good luck. I don't think I've seen any 24 bit panels
edgecase: well i have one from a viewsonic monitor, not sure about laptops tho
edgecase: what about dual-pixel laptops?
agd5f: edgecase: yeah, I meant LVDS
agd5f: lots of monitors report 24 bit
edgecase: i mean the actual panel inside the monitor is 48bit LVDS
mattmatteh: anyone have a problem with xv blurring greens and blues with RV350 AS [Radeon 9550] ?
edgecase: i wonder if the Gimp could use openGL or AIGLX to upload large images, and do the manipulation with the GPU
agd5f: edgecase: you could certainly implement certain filters and such with the GPU
edgecase: ok i'm trying to figure out the connection from DAC outputs to "ports" or "connectors"
edgecase: would a card with DVI, VGA, and S-Video, perhaps be DVI-D, and so DAC1 could be VGA, and DAC2 could be wired to S-video?
edgecase: would there be external output buffers for each analog output?
agd5f: edgecase: usually most desktop cards are VGA, TV, DVI-I
agd5f: VGA is generally DACA, DVI-I is TMDS and DACB, tv is DACB
edgecase: hmm so how is DACB shared then?
agd5f: edgecase: what do you mean?
edgecase: well DVI-I and TV out using DACB, do are they just wired in parallel, and you tell the user not to connect both at the same time?
edgecase: or does the chip have separate output buffers for each
agd5f: they are both wired up to DACB
agd5f: the mux is controlled via a GPIO
agd5f: the setup varies by card though
edgecase: so analog switch? i wonder if it's on chip or external
edgecase: older impactv2 probably external
mattmatteh: anyone have a problem with xv blurring greens and blues with RV350 AS [Radeon 9550] ?
agd5f: mattmatteh: overlay or textured video? what source? file, tv card?
mattmatteh: agd5f, xf86-video-ati 6.6.3, xorg-server 1.3.0.0
mattmatteh: agd5f, card was posted ^
mattmatteh: agd5f, i think i was trying both overlay and textured
mattmatteh: agd5f, not sure what i should use or what works. i was trying several
mattmatteh: agd5f, i dont know the differences
mattmatteh: agd5f, i would like to use the open source radeon driver. i was also testing the proprietary driver to see if i could get xv working, but that dont work at all
agd5f: mattmatteh: so overlay. 6.6.3 didn't supporte textured video. you might try upgrading 6.63 is really old
mattmatteh: agd5f, i am using gentoo and try to stick to stable or somewhat tested
agd5f: mattmatteh: well, there have been quite a few bug fixes and such since then
agd5f: 6.8.0 is the latest stable release
mattmatteh: ok, ill try that
mattmatteh: agd5f, another question
agd5f: sure
mattmatteh: when i first startx the screen is stretched 1920x1200 to fit in 1600x1200 on my 1920x1200
mattmatteh: do you know where i should look for that problem ?
mattmatteh: i can only fix it by using xfce4 display manager too and select another resolution, click no, and it goes back to what it was, but correct
agd5f: mattmatteh: I'm not following
mattmatteh: i dont know if thats a driver problem or xorg is not reporting it correctly
mattmatteh: thinks
mattmatteh: the height is fine
mattmatteh: the width is not
mattmatteh: it seems to think i have 1600 pixel left-right
agd5f: mattmatteh: do you use gnome?
mattmatteh: so i get the remaining pixels off screen not displayed to the right
edgecase: agd5f, i'm trying to understand where the tv encoding happens, before the DAC (digital) , or after
mattmatteh: but i do infact have 1920, so it gets stretched
agd5f: edgecase: before the dac
mattmatteh: agd5f, xfce4
mattmatteh: agd5f, not sure if they use the same window manager
agd5f: mattmatteh: ok
PSYCHO___: mattmatteh: you pobably have bad xorg.conf modelnie or something like this
mattmatteh: agd5f, tried with window maker too, same problem, but could not find a display manager to test
agd5f: mattmatteh: they may be fixed in 6.8.0 too, I'm not sure what it would be off hand
mattmatteh: PSYCHO___, i was thinking that a mode line could fix, but this is the first time i have had a problem
mattmatteh: i am not sure where to start looking for the problem
PSYCHO___: mattmatteh: bight screw autodetection so modeline is allways good way
PSYCHO___: s/bight/might/
PSYCHO___: mattmatteh: btw xf..ati use from testing
mattmatteh: PSYCHO___, i saved my xorg logs so i should be able to use those mode lines
PSYCHO___: really big pain in ass with stable ones
mattmatteh: PSYCHO___, dindnt understand your last message
PSYCHO___: add keyword ~x86 for xf86-video-ati
mattmatteh: also to compare the open source and closed source, would you (anyone ) know how to get xv working ? just wanted to compare, i would like to uninstall the closed source driver
mattmatteh: PSYCHO___, ill try that now
mattmatteh: agd5f, i am updating xf86-video-ati, you say that it supports overlay and textured ? which is better ? is one faster ?
mattmatteh: i just want to play dvd's
mattmatteh: and have it look the best
agd5f: mattmatteh: each has advantages and disadvantages
mattmatteh: is there one that is better for dvd movies ?
mattmatteh: agd5f, or something that explains it ?
agd5f: overlay can handle planar formats natively and has better scaling. textured works with composite, provides multiple ports and can span crtcs
mattmatteh: just single lcd here, using dvi, widescreen
agd5f: mattmatteh: overlay is probably preferred unless you run compiz or some other compositor
mattmatteh: no, nothing fancy
mattmatteh: might try it one day, but dont see that happening soon
mattmatteh: agd5f, the newer version didnt fix
agd5f: mattmatteh: the mode thing or the overlay?
mattmatteh: agd5f, xf86-video-ati and xv
mattmatteh: i restarted and played the same movie
mattmatteh: still dark blue and dark green blur
agd5f: mattmatteh: ok what play are you using?
agd5f: mythtv?
mattmatteh: also, another problem i didnt mention is that there are random dots all over, like noise, mostly blue. i have gotten them to go away by doing the display manager change
mattmatteh: agd5f, mplayer
agd5f: take a look at the output of xvinfo and see if any of the attributes are way off like XV_HUE
mattmatteh: agd5f, here is xvinfo output info http://rafb.net/p/P7roLc84.html
mattmatteh: agd5f, i do not understand most of the output
mattmatteh: i was going to compare it to the close source driver, but can not get any xv to work on that
agd5f: mattmatteh: you can adjust those attributes to adjust how the video is displayed in the overlay
mattmatteh: agd5f, would one of those being off give a blur on certain colors ?
mattmatteh: not even sure if its certain colors
mattmatteh: hard to decribe
mattmatteh: not sure how to grab a screen shot from a movie
agd5f: mattmatteh: what's teh source again? video file or tvcard?
mattmatteh: agd5f, harry potter 2
agd5f: dvd then
mattmatteh: yup
agd5f: ok
mattmatteh: could try another ?
mattmatteh: someone in #ati suggest to try a different movie
agd5f: mattmatteh: you can try adjusting the xv attributes
agd5f: xvattr -a XV_BRIGHTNESS -v 500
agd5f: for example
mattmatteh: agd5f, there a way to compare that to my other graphics card ? i had nvidia working before in this computer
agd5f: mattmatteh: try it on the other card? how about another player like xine or totem?
mattmatteh: tried xine
mattmatteh: think it was the same
mattmatteh: tries again
agd5f: do other DVD's have the same problem?
mattmatteh: agd5f, havent tried others yet
mattmatteh: just tried xine, no difference at all
mattmatteh: same blue green blur, and same blue dot noise
agd5f: mattmatteh: I have the same card and it works fine here (on various video files), but no dvd drive in that computer
mattmatteh: agd5f, ill see if i can test some other video's
agd5f: mattmatteh: sounds good. let me know how it goes
mattmatteh: agd5f, how long are you going to be around ?
agd5f: mattmatteh: not much longer, but I can read the log later tonight or tomorrow, if I don't happen to be around
mattmatteh: agd5f, ok. i should be on late tonight and tomarrow. thanks for your help
scobby: hi all
scobby: i have problems with my radeon 9600, it seems to be video acceleration isnt on. i use the free radeon driver and compiz. for example youtube videos work but cpu usage goes up to 100 %.
scobby: and other videos in fullscreen dont work correct
scobby: with compiz off it even dont work fast enought
roh: hm.. re
CRXLPY: I have a radeon7200 Radeon AIW I have not been able to access the tv-in. I am running linux. anyone here able to help with that?