hklh: anyone here?
nsna: I was curious about the radeonhd vs fglrx setup
nsna: which one is currently better? and if possible can you patch fglrx to run smoother?
yangman: nsna: depends on what you need. and it's not possible to patch fglrx since no source code is provided
nsna: ahh... well i was curious on getting 3d acceleration up and going and such.. i know fglrx kinda sucks in comparsion with the windows driver
nsna: and ati doesn't really port to linux very well
yangman: radeonhd doesn't support 3D acceleration yet, so I guess that answeres your question
nsna: where can one find the info on helping out with this radeonhd project
yangman: fglrx is a lot of YMVV
yangman: points at topic
nsna: ymvv?
yangman: your milage may vary
yangman: experience usually varies from person to person
nsna: does it also vary from install method to install method?
yangman: I actually don't have a lot of issues with the newer fglrx. I use radeonhd because it does xrandr stuff better, and I don't actually need 3d acceleration at the moment
nsna: what's xrandr?
yangman: intallation shouldn't matter, as long as the distro package handles it properly
yangman: it's the component that drives things like dynamic multi-screen configuration, rotate, etc
nsna: cool stuff
nsna: i suppose you can't setup to where ya use one driver for certain apps and such and another for another taking advantage of the best of both worlds eh?
yangman: no
nsna: figured that much
yangman: well, not automatically
yangman: you can always restart X with a different configuration
nsna: i gotcha by configuring xorg.conf
nsna: so it won't be possible to setup certain workspace to run off of certain drivers either
yangman: well, the driver drives hardware
nsna: can't have a driver reload during selection of certaiin workspace?
yangman: it's inherently a bad idea to swtich controllers without properly reinitializing everything
yangman: that requires restarting X
yangman: so that things can reinitialize
nsna: gotcha
nsna: what distro you use?
yangman: gentoo
nsna: how do you like it? and have you used other distros?
yangman: it does what I need. I wouldn't recommend it to most people
yangman: I've tried Debian briefly, and understand how Ubuntu works
yangman: and I've dealt with RedHat and Fedora at work. wouldn't recommend those at all
nsna: why not fedora? and what do you do if i may ask?
yangman: Arch comes recommended from a friend, for people that want/need to be able to easily add packages compiled from source, but don't quite want to use gentoo
yangman: basically, RPMs suck
yangman: it just gets really messy when you start having a lot of dependencies
yangman: the particular version of RedHat I had to use was also so old that I couldn't easily install libraries I needed for a particular project
nsna-: only thing i hate is the lack of support by ati for linux machines
nsna-: guess i got d/c
yangman: I'm currently a student, actually. work was an summer-intern kind of thing, I guess, although more like a short-term fulltime employment
yangman: well, they're trying
yangman: it's not easy
nsna-: what did ya do?
yangman: miscellaneous programming
yangman: bits of everything
nsna-: i do however like how ati is approaching open sourcing their driver with this radeonhd project, shows guts
yangman: but for linux platforms
nsna-: what company was it under? or was it school based?
yangman: EQO, if you've heard of it
yangman: these days, they mainly do VoIP service for cell phones
nsna-: nice
nsna-: i can see where linux will give them an advantage
nsna-: how was it?
yangman: good
yangman: long commute, but good place to work
yangman: I'll most likely be going back part time in 2 months time
nsna-: that's cool, so how did it feel to be a programmer?
yangman: heh
nsna-: i'm currently studing to become one now, gotta wait till i'm out o military first
yangman: nothing special, really
yangman: I've been coding for years
nsna-: you don't feel a sense of satisfaction knowing that you are now a professional coder rather than someone who codes in spare time?
yangman: the only difference is that you're being paid
yangman: depending on what you end up doing, it's often more satisfying to work on side projects
nsna-: i wanna code games and get some of my ideas into the works
nsna-: i have lots of ideas on how to revolutionize the mmorpg world
yangman: you can only really judge a programmer by the quality of code produces
yangman: frankly, too many people in the field are terrible coders
nsna-: most people don't understand logic very well
yangman: logic is only the first step
yangman: you have to have a willingness to learn and try new things
nsna-: scavaging can be a great tool also
nsna-: why reinvent the wheel you know
yangman: logical programming, functional programming, SQL, CPU architecture, etc
nsna-: of course if you scavage you must also attempt to make it more efficient than it previously was
yangman: these are all things that are useful to know
nsna-: true
nsna-: so do you code for this project?
yangman: no
nsna-: why not?
yangman: no time
yangman: :P
yangman: it's a huge leap to take
yangman: modern GPUs are magnitudes more complex than most people realize
nsna-: how's that?
yangman: well, these days, even the cheapest dedicated graphics card has more logic and processing power than entire PCs of, say, 8 years ago
yangman: you certainly don't need to understand everything, but there is a lot to understand
nsna-: i heard ati released their information on their 3d accelration do you know where to find that?
yangman: http://ati.amd.com/developer/open_gpu_documentation.html
nsna-: sweet thanks man
nsna-: they should have that in the title also
yangman: well, it's not useful for most people
yangman: and certainly overwhelming
yangman: aside from the developers discussing things, most everyone else in here is looking for support or just keeping an eye on things
yangman: report bugs, etc
yangman: it's certainly worth subscribing to the mailling list, if you haven't already
nsna-: when do they developers discuss things? is there a listing of their scheduled times?
yangman: no, nothing planned like that
yangman: they're usually in the channel. just not necessarily paying attention or actively talking
nsna-: all this looks like assemble language
yangman: yup
yangman: you're talking to hardware. it's all opcodes and memory addresses
nsna-: pretty sweet i say
nsna-: so basically i'd have to recompile and recompile to test over and over and over again
yangman: there certainly is a lot of testing
nsna-: and with each recompile possiblity of complete failure
yangman: ideally you only make changes that are reasonably safe ;)
nsna-: true true
nsna-: i don't see how the 3d acceleration happens tho most of it is just registeries and such
yangman: so, this comes down more to the CPU not being good at 3d graphics processing more than anything
yangman: have you ever studied computer architecture?
nsna-: havn't gotten that far into my computer science studies
yangman: okay
nsna-: mostly studied programing languages and suchk
yangman: the basic idea is that a CPU is designed to be fast for the general case
yangman: if..else construction, integer arithmetic, floating points, etc
yangman: the operations most useful for manipulating 3D graphics, and graphics in general, tend to be vector and matrix operations
nsna-: linear algerbra stuff
yangman: these things are slow on a general purpose CPU, because they're just not designed for it
yangman: right, exactly
yangman: so, a GPU is specifically desgined to do these kind of operations very quickly
yangman: so, it can do things like multiply matricies with a single instruction as opposed to iterating through rows/columns
yangman: different problems, different solutions
nsna-: so basically with programing the driver you are trying to figure out what's going into the gpu and manipulate it in such a manner that you get what you want to come out of display
yangman: at the most basic level, yeah
nsna-: so how do you figure out whats going into the driver?
yangman: the driver mostly serves as an abstraction layer between the hardware and some API
yangman: so, an application that wants to display something to a display (X) calls functions in the driver
nsna-: what language is this driver written in?
yangman: the driver then does its magic and makes whatever it is appear
yangman: C
nsna-: makes it easier to read
nsna-: i might have to d/l the open source and see if i can make sense of it
yangman: you won't actually see any assembly in the code
yangman: I'm not an expert, but, these days, usually how these things work is that you ready a set of instructions for execution, hands it as a batch to the graphics card, and then the graphics card executes them
nsna-: rog
yangman: anyways, good luck with whatever you end up doing
yangman: I'm off to bed
nsna-: later man, i am too
sonne: egbert_away, http://bugs.freedesktop.org/show_bug.cgi?id=13470 is not fixed for me...
sonne: at least not for DVI ...
sonne: actually digital output does not work anymore at all...
sonne: but analogous does so far.......
aljaz: hey there! Is Egbert here? I have a question about the mentioned fix in Bug #14690.
egbert: aljaz: i'm here.
egbert: aljaz: please give me 5 minutes.
aljaz: ok
egbert: i need to take care of some urgent business first :)
bardo: hi all!
bardo: i have a radeon hd3650 which i'd like to be supported, but after modifying the sources i have some connector problems
bardo: specifically, it seems to me that rhd_conntest is returning the wrong connector ids
bardo: but i don't have a very good understanding of the code, so i may be wrong
bardo: is anyone able to help?
bardo: (or, may i help you providing the informations you need?)
libv: bardo: rv620/635 is still being worked on
libv: bardo: whenever i get time to fuse in my pixelclock code (between everything that is going on here atm), it might actually happen in the next few days
libv: the analog outputs are working rather well atm
bardo: libv, great! so you already have a tester?
libv: bardo: we've had some preproduction cards since early december
bardo: i have no analog outputs on my rv635... i could try the converter
libv: bardo: but unfortunately the relevant information didn't pop up until much much later
libv: bardo: at a time where i was too swamped myself with various other things (one major thing, fosdem)
libv: bardo: but stick around... we hope to have good news for you in the next few days
bardo: libv, thanks... i hope the support will be added soon... fglrx happily freezes my machine
libv: dacs are working here beautifully, but i have to set the pixelclock with a command line utility still, as this code still isn't in the driver ;)
bardo: yeah, i noticed :P
libv: egbert did all the legwork for the dacs and is also looking into the very complicated mess that is the new digital blocks
Tharmas: hello, i have a question, if somebody knows: Is radeonhd ever going to suport x550 (r370)?
libv: Tharmas: no, never, we only do r500 and up, as this is where the modesetting side changed completely
bardo: Tharmas, isn't that supported in radeon?
libv: Tharmas: everything before r500 is radeon driver
Tharmas: so I guess I'm stuck with radeon :(
egbert: aljaz: somehow my push didn't work. now i've pushed the fix. was that your question?
libv: Tharmas: agd5f should be able to help you with any issues there
libv: Tharmas: in #radeon too ;)
Tharmas: ok, i'll go there for help .
Tharmas: thanks, anyway. i'll try to bear with radeon's softwae fallbacks :(
aljaz: egbert: it works now, though xorg.log has still these messages when user (terminal) switching: "MC not idle" and "D1CRTCDisable: Failed to Unsync CRTC 1". Meaningful or less? :)
egbert: the latter is not a big deal. the former sort of worries me.
egbert: aljaz: if you don't see problems it's probably not a big deal.
aljaz: egbert:no, I don't see problems for now, I just wanted to know, if messages are having any meanings to you. If not, then it's ok. :)
aljaz: egbert: when switching between users, I do not get MC not idle, just when going to a terminal and back
sonne: egbert, any idea why DVI does not work with current git?
egbert: sonne: it does here. can you describe your problem?
egbert: sonne: ?
sonne: egbert, sorry I am at work and well... had to work ;)
sonne: egbert, the problem is it claims dvi is disconnected
sonne: although I connected it...
sonne: egbert, I don't think it is a connector problem
sonne: just replugged everything
egbert: sonne: ok.
sonne: so now what?
egbert: if your monitor isn't detected as connected this is probably hpd.
egbert: can you back out the patch 2b16734b998e657a426af5b9dc332700cba2c9be and see if this makes a difference?
egbert: sonne: could you also provide a verbose log in a pastebin?
egbert: sonne: i assume your analog monitor is connected thru a vga2dvi dongle, right?
sonne: no
sonne: yes
sonne: sorry, but on the vga part of the monitor
sonne: to the vga connector of the monitor I mean
egbert: right, but this goes in a vga2dvi dongle which is plugged into a dvi-i.
egbert: this dongle also triggers hpd.
sonne: egbert, log is here http://nn7.de/debugging/Xorg.0.log
egbert: ok.
sonne: logverbose 7
egbert: sonne: you've got a DELL 2407WFP?
sonne: yup
egbert: ah, ok, connected to analog.
sonne: yes - as analog works
egbert: i also see edid data from the panel.
sonne: there should be a DVI try inbetween
egbert: do you have 2 dvi connectors on the macbook?
sonne: no
sonne: 1 dvi connector
sonne: and I have an adapter to vga for that one
egbert: ok, but you can only do vga or tmds on a dvi. but you said there should be a dvi try in between.
egbert: did you hotplug?
sonne: ?
sonne: I did xrandr --auto
sonne: i.e. I disconnected the dvi-analog adapter, then connected the dvi cable
sonne: on the tft-display side both cables are plugged in
egbert: right.
egbert: sonne: this looks indeed funny:
egbert: (II) RADEONHD(0): rhdRROutputDetect: Output DVI-I_1/analog
egbert: (II) RADEONHD(0): rhdRROutputDetect: Output DVI-I_1/digital
egbert: (II) RADEONHD(0): rhdRROutputDetect: Output PANEL
egbert: sonne: it supports the notion that hpd isn't working with your digital monitor.
sonne: what is hdp?
egbert: hot plug detection.
egbert: the card supplies a voltage for ddc. the monitor connects this line to the hpd line so when the voltage appears the card knows that something is connected.
egbert: the above should look something like that:
egbert: (II) RADEONHD(0): rhdRROutputDetect: Output DVI-I_1/digital
egbert: (II) RADEONHD(0): FUNCTION: TMDSASense
egbert: (II) RADEONHD(0): TMDSASense: Disconnected
egbert: so we don't call sense.
egbert: maybe you can try Option "HPD" "off" in the device section.
sonne: aha
egbert: also you could back out the mentioned patch. but i doubt it is related to that.
libv: we have both hpd and hdp in our hardware btw
egbert: don't we love those confusing acronyms?
libv: ;)
libv: i don't even know what hdp stands for... my guess is host data processor... as it is some aperture system where you can swap bytes for endianness
sonne: egbert OK, I try HDP off now
sonne: brb
sonne: egbert, didn't work -- ahh is it HDP or HPD
sonne: ?
sonne: (WW) RADEONHD(0): !!! Option HPD is set !!!
sonne: so it was HPD, but still :(
egbert: sonne: still doesn't work?
sonne: yes - it does not change anything
egbert: sonne: so can you give me another (verbose 7) log file?
sonne: egbert, updated Xorg.0.log is on the webpage
sonne: hah, I was first :)
egbert: ok :)
egbert: sonne: (II) RADEONHD(0): FUNCTION: DACBSenseCRT
egbert: (II) RADEONHD(0): DACBSense: DAC: 0x00
egbert: (II) RADEONHD(0): rhdRROutputDetect: Output DVI-I_1/digital
egbert: (II) RADEONHD(0): FUNCTION: TMDSASense
egbert: (II) RADEONHD(0): TMDSASense: Disconnected
egbert: this doesn't look right.
egbert: sonne: what if you start the server with the digital display connected?
sonne: I am not in the position to judge this
sonne: OK
sonne: brb
sonne: egbert, nope, updated the log on the webpage
udovdh: hello
udovdh: FWIW: https://bugzilla.redhat.com/show_bug.cgi?id=434695
udovdh: an assignee...
egbert: sonne: do you get text mode on your digital display?
sonne: egbert, no
sonne: but I did not reboot...
egbert: sonne: ok, still this is strange. hpd not working is ineed *very* strange.
egbert: sonne: could it be that the dvi cable is loose on the monitor side?
egbert: sonne: i could imagine that something wonky is happening with tmds. but hpd works with your dongle. so it should work with your digital display.
egbert: it is just a wire between two pins.
sonne: egbert, OK, so let me boot OSX and see if that works
sonne: brb
sonne: egbert, new log new magic :/
sonne: the display has like vga/dvi-d/svideo etc outputs
sonne: and when I switch the display to e.g. VGA (while connected via DVI) xrandr does not detect the DVI output
sonne: if I switch the monitor to DVI and then do xrandr it is detected...
sonne: I don't remember that this happened before.. hmmhh
udovdh: sonne
udovdh: I had to do similar to my crad when I switched to a dvi-d cable
udovdh: from the analog cable
udovdh: somethign like xrandr --auto
udovdh: and some otehr command to turn on the dvi-d output
udovdh: also xrandr
udovdh: it then worked like it should when rebooting
udovdh: but without doint the boot of course
spirithands: Hey guys.. i'm going nutzo here. I've got an ATI Radeon HD 2600 PCIe card. When I use the vesa driver, it works fine. I when through the "radeonhd" xorg driver setup procedure @ http://ubuntuforums.org/archive/index.php/t-641247.html and it still kicks back to the failsafe low resolution driver .. The radeonhd driver exists: http://ubuntuforums.org/archive/index.php/t-641247.html
spirithands: hellos?
TGEN: hi
spirithands: any idea why radeonhd isn't workn? :(
TGEN: what does your Xorg.0.log say about radeonhd?
spirithands: Xorg.0.log doesn't say anything about radeonhd because I'm in the failsafe right now
spirithands: Ubuntu's stupid failsafe thing comes up
spirithands: and replaces everything
marcheu: try /var/log/Xorg.0.log.old....
spirithands: nah, I had tried to switch it back to vesa .. so the old one has vesa stuff
spirithands: which still doesn't work
TGEN: try radeonhd again
TGEN: and save the log
spirithands: ok
spirithands: brb