muks: libvde: fedora just released a 1.5.0 X server RPM and a working radeonhd stopped working past that
muks: gdm tries to start under X several times, and it drops to text mode with a flicker on the screen
muks: this was with the DRI option
muks: when i turned DRI off,
airlied: muks: can you pastebin the xorg log?
airlied: I wonder what changed...
muks: and booted into runlevel 3, startx worked and started with radeonhd, but quitting X will bring back flicker to text mode
muks: airlied: i'm running radeon now.. 2 mins.. will replace config and record log
airlied: it sounds like X segfaults but I don't know what changed that much between pre-release and 1.5 servers
muks: X didn't start during "graphical boot"
muks: waiting for runlevel 5 now
muks: oh that started X this time
muks: but the flicker is still there in text mode
muks: http://radeonhd.pastebin.com/m71bd12dd
muks: this is a radeon X1400
muks: on a thinkpad t60
airlied: wierd.. actually it might just have been gdm segfaulting either :)
muks: but that doesn't explain the screen flicker
airlied: oh that I'm not sure about, sounds more driver specific alright.
Nightwulf|work: hi all
Zajec: how should I use rhd_dump for writing values?
Zajec: rhd_dump -w 0x74FC 0x2DB42200 1:00.0
Zajec: or
Zajec: rhd_dump -w 0x74FC: 0x2DB42200 1:00.0
libvde: muks: the screen flicker is a well known issue on the X1400
libvde: muks: i keep on pestering a manager to retire his laptop so we can finally spend loads of time on fixing that one :)
muks: libvde: wuhoo so it's known :)
libvde: muks: yeah, known for way too long sadly... but this manager has been telling us all the time that he was looking for a new system anyway, but he's procrastinating it :)
libvde: i know how that goes, i still carry my tatty old unichrome based laptop around everywhere, as this is working well enough still and has everything i need installed :)
libvde: but every chance i get i remind him of it :)
libvde: it's turned into a running gag actually
libvde: so at one point in the future, we will be able to fix the issue properly :)
ndim: libvde: If we can squeeze in a hack session at FOSDem 2009, I'll lend you my T60 for testing.
libvde: ndim: ok :) but i do hope this guy gets a new laptop by then :)
muks: libvde: X seems to be ok, so i'll keep using radeonhd
libv: muks: cool :)
NeatFires: hi
NeatFires: what's the status of xv for r600?
libv: NeatFires: non-existant, only got the first triangles out of this hw so far
Honk: so you did manage to draw triangles now? cheers! :)
Obscene_CNN: I believe they did but I came into this conversation late.
MostAwesomeDude: Triangles, yes.
MostAwesomeDude: Anything else, dunno.
MostAwesomeDude: *really* wants docs
Obscene_CNN: notes that two triangles of equal shape and size form a rhombus so they may have drawn that two ;)
MostAwesomeDude: Actually, drawing a quad is roughly the same amount of work, really.
Obscene_CNN: but drawing two triangles allows code reuse :D
dvandyk: libv: so what was the problem?
dvandyk: libv: or rather: what did you initially forget to do?
libv: dvandyk: with what?
dvandyk: libv: drawing triangles on r600
agd5f: dvandyk: figuring out what low level 3D engine init was missing
agd5f: and some fixes to the command streams
dvandyk: i see
dvandyk: agd5f: ah, while you're here... what are your plans for fusion regarding userspace interface?
dvandyk: agd5f: will there be something like spufs for the gpu part in the cpu?
agd5f: dvandyk: you mean for stuff like GPGPU?
dvandyk: yes
agd5f: drm multi-master
MostAwesomeDude: dvandyk: Three easy steps.
MostAwesomeDude: 1) Write CUDA/BTM/Brook compilers for Gallium.
MostAwesomeDude: 2) ???
MostAwesomeDude: 3) GPGPU!
dvandyk: why oh why does everybody want to use gallium?
MostAwesomeDude: Because it's awesome!
dvandyk: rule number one: go easy and simple, then go big and complicated
agd5f: dvandyk: it just provides an API
MostAwesomeDude: dvandyk: Yes, and Gallium is easy and simple. :3
dvandyk: fortran is easier and more simple
agd5f: dvandyk: you could write your own gpgpu drm client and submit your own command buffers kust like mesa would for 3D
dvandyk: i would rather have a (partial) fortran compiler for r6xx/r7xx target
MostAwesomeDude: dvandyk: And I want to be paid, and airlied wants his pony.
agd5f: dvandyk: you could do that too. drm just provides the interface for allocating mem and submitting commands
dvandyk: yeah... i'm still waiting for glisse to show me an r5xx example that sets up one buffer for linear iteration
dvandyk: :-)
dvandyk: MostAwesomeDude: a partial fortran compiler is much more likely than you getting paid or airlied getting his pony ;-)
MostAwesomeDude: dvandyk: It is? Really?
MostAwesomeDude: Who's gonna write it?
dvandyk: always the one that asks
MostAwesomeDude: Well, go ahead, I'm not going to stop you.
Obscene_CNN: stop him please. I don't want to learn fortran ;-)
MostAwesomeDude: Obscene_CNN: There is a world of difference between "I'm going to do it" and "I've done it", which is why certain people get on the team despite having next to no experience. :3
dvandyk: MostAwesomeDude: so you're not going to write it?
dvandyk: MostAwesomeDude: :-)
MostAwesomeDude: dvandyk: I'm going to write a Gallium pipe at some point in the near future.
MostAwesomeDude: But I have no interest in GPGPU, and so I'm not going to write any GPGPU stuff.
dvandyk: great, but the average user of scientific HPC is not a good programmer
dvandyk: ah
dvandyk: ignore me
dvandyk: HPC users do know fortran
MostAwesomeDude: Well, fortunately, somebody *could* always just write CUDA or Brook or BTM into Gallium.
MostAwesomeDude: And please don't complain; I have very little obligation towards new features since I'm unpaid. I'm only wanting to do the Gallium stuff because *somebody* *needs* to do it.
MostAwesomeDude: Gallium support isn't optional for Radeons.
Obscene_CNN: notes prolog lends itself to parallel processing quite easily since we are talking about languages people don't use much anymore
dvandyk: Obscene_CNN: you're refering to fortran as a language that nobody uses anymore?
dvandyk: Obscene_CNN: you're not up-to-date
dvandyk: if you're living in the EU, it's very very likely that your weather reports are based upon the output of fortran programs
MostAwesomeDude: dvandyk: I write in Python, except when I'm working on X.org.
MostAwesomeDude: I don't know Fortran.
MostAwesomeDude: I am *definitely* not qualified to write a Fortran compiler.
dvandyk: MostAwesomeDude: hehe, i don't expect you to do it
Obscene_CNN: I didn't say nobody used fortran, I said its not used much any more.
Bertl: physicists use it all the time ...
Obscene_CNN: more and more they are using c++ with tools like openfoam
Obscene_CNN: fortran code is usually faster though
dvandyk: i'd say "bullshit" towards that
dvandyk: it's a prejudice
dvandyk: however, you can write more optimised code using C++ (think of using SSE with the builtin assembler)
dvandyk: fortran is the workhorse for many scientists, though
Obscene_CNN: Well after having this discussion with someone who developes CFD code, C++ is favored for ease of development and extensibility. Fortran is better for speed.
dvandyk: well, for someone who has developed CFD code: We used C++ code with a custom SSE backend. And it's fast!
Obscene_CNN: well SSE isn'treally C++. Its assembly
dvandyk: and you can't use SSE with either Java or Fortran, so it's a distinct favour on C++/C's side
Blinkiz: I believe I have a HD3300 integrated card on my moderboard. I want to find out if radeonhd supports it. Where should I look? http://wiki.x.org/wiki/radeonhd does not list this card
Honk: Blinkiz: just guessing here, but hd3300 sounds like it's using some r600 chip ;P
Honk: so you'll prolly get a working X but no acceleration whatsoever =)
yangman: Blinkiz: it does, but no acceleration
libv: it is an rs790GX :)
libv: probably the fastest IGP in production ever :)
libv: Blinkiz: good choice
libv: Blinkiz: and it is supported
libv: although we still haven't seen any reports of sideport memory
Zajec2: is this possible that enabling audio in HDMI needs other registers for every resolution?
Blinkiz: libv, Honk, yangman : It's rv620 chipset. So no compiz stuff for me then?
Zajec2: Blinkiz: didn't we explain that today to you already?
Zajec: Blinkiz: 3D is for R500 only
Blinkiz: Zajec: what? No
Blinkiz: R500 oly. Okay.
Zajec: missed you with someone else :)
Blinkiz: Zajec: :)
Zajec: Blinkiz: http://www.x.org/wiki/RadeonFeature
libv: Blinkiz: hd3300 is rs790
libv: Blinkiz: hd3400 is rv620
Blinkiz: Anyway, keep up the good work. Looking forward to the day when all AMD/ATI graphic cards have open source drivers directly. Will be wonderful! All thanks to you guys :D
Blinkiz: rs790.. okay
libv: Blinkiz: we are getting quite close to getting some acceleration going on r6xx and up now
Blinkiz: libv: Wikked!
Zajec: libv: do you think this can be done before end of october?
Blinkiz: libv: Can't see that radeonhd supports "rs790". Something I have missed? Its working for me anyway.
Zajec: (basic 3D)
Honk: waits till there's opensource acceleration before upgrading his card :]
libv: Zajec: heh, we would be happy if we had basic 2d emulated on 3d in that timeframe :)
Zajec: nice :)
libv: Blinkiz: rs790 was added only after the latest release, we haven't done a proper wiki or README review since
Blinkiz: libv: I see
ndim: Blinkiz: src/rhd_id.c or man/radeonhd.man in the source code. Or if you happen to have the source code locally, run "man ./man/radeonhd.man".
libv: yeah, ndim wrote up some stuff to keep our id file and our manpage automatically up to date :)
Blinkiz: :-P
libv: it's quite nice... we're just now waiting until he writes up the code to keep our wiki up to date as well :ppp
ndim: Preferably, I'd just link to the gitweb page from the Wiki. Unfortunately, gitweb cannot pipe the source through col(1)
ndim: converting man page to txt on the fly
libv: just kidding of course :)
libv: you actually have ideas on how to solve this?
ndim: Not just by using the x.org wiki and freedesktop.org infrastructure.
libv: ndim: i think the wiki can be left alone, i'm not sure whether we want this to change automatically
libv: what if we kept this info also in the README and moved some things from there to an INSTALL file (as you suggested before)
ndim: running 'man man/radeonhd.man | col -b' for conversion to a plain text file is not difficult.
libv: and link to this file in gitweb from the wiki?
ndim: The install/build stuff should be moved froM README to INSTALL
ndim: no question about that.
ndim: libv: Oh. We could just use basically the same sed stuff which updates man/radeonhd.man to update README?
ndim: libv: Was that your suggestion?
libv: yeah, and we can then link to this file, for the latest information, from the wiki
libv: i fixed the issues with CS now :)
libv: finally.
Zajec: sees commits :)
Zajec: nice
yangman: yay
libv: now i'm off to here http://flickr.com/photos/facing-my-life/437867053/ so i can drown the memory of getting this finally fixed :)
libv: tomorrow the road to getting CS into master is started :)
agd5f: libv: fix the hang?
agd5f: what was it?
libv: agd5f: a second instance of the 3d engine init was sitting around still from a quick test commit... really stupid again
libv: another one that, once seen, is hard to miss
libv: there is one solace... a lot of things did become a lot cleaner in the process...
Honk: lol
libv: and i did the 2d engine setup/reset review as well
libv: and verified that the xaa code doesn't need those idlewaits all over on r5xx
libv: etc, etc
glisse: libv: congrats :)
starkmjolk: stupid question; how do I do svn update with a git repo like the radeonhd one? :)
Honk: you use git instead
Honk: ;P
yangman: if you've checked it out properly, just git pull
starkmjolk: pull was the word, and then I mustve checked out in some weird manner I think... hmm
starkmjolk: ah, no. I was at the wrong dir of course :)
ndim: If your builds break... complain to me.
Necos: now, that was interesting...
Necos: second life killed my computer :(
Necos: after launching it with the radeonhd driver, got some pretty nasty gfx corruption... then when i closed it, lost control of my kb/mouse
Necos: hmmmm... what's a good game to test the new 3d code with?
Necos: damn, guess i'll try again tomorrow... laters folks