ethana3: oh sorry
ethana3: loud coke
ethana3: Is there any kind of bounty set up for higher radeon performance?
ethana3: I understood that the bottleneck currently
ethana3: wasn't documentation
ethana3: but developers who cared about pushing performance
ethana3: over new features
ethana3: Did I understand incorrectly?
rx__: where did you get that idea?
ethana3: I think someone here said something to that effect
ethana3: are we waiting on more documentation?
rx__: i think you've been misinformed
ethana3: yeah I guess
rx__: the biggest bottleneck happens to be time
ethana3: so a bounty might help
rx__: i have no idea about that
ethana3: oh, ok
rx__: i think it's mostly time and manpower
ethana3: Google should make a linux distro
ethana3: and hire you all ;)
ethana3: Who was the guy here who got hired by AMD?
rx__: i was just going to say
rx__: developers can do one step better and get hired by amd
rx__: like agd5f
rx__: alex deucher
ethana3: so he doesn't really log this channel I take it?
ethana3: I don't see him
rx__: he's usually around
rx__: honestly.. if you're not happy with the speed of development you should step up :)
ethana3: I wish I could code
ethana3: I guess I can test the latest git
rx__: anyone can code
ethana3: If I figured out... how to use git..
ethana3: in C
rx__: anyone can code in C :)
ethana3: as opposed to TI-BASIC
ethana3: Ok then
ethana3: would you help me get started?
rx__: of course
ethana3: I've written C++ code before
ethana3: I should be able to try to familiarize myself with the radeon code base
rx__: yeah i'm trying to figure out how everything fits together as well
rx__: you can always ask questions though
ethana3: ok, making sure I have git installed
ethana3: k-- got it
ethana3: How do I get the source, and where should I put it?
ethana3: Yeah, I don't know what I'm doing
ethana3: I've never messed with my card's bios
rx__: oh no.. nm you had the 9200
ethana3: yes ;)
ethana3: digital memory
ethana3: I love it
rx__: git clone git://anongit.freedesktop.org/git/xorg/driver/xf86-video-ati
ethana3: does it matter what directory I'm in when I do that?
ethana3: I did that kind of thing with svn once
ethana3: when I was in some weird directory
ethana3: and it dumped a bunch of stuff ther
ethana3: same deal?
rx__: it will make a xf86-video ati and put files there
ethana3: I'll make a home/ethan/src
ethana3: uh oh
ethana3: git is already the newest version.
ethana3: bash: git: command not found
ethana3: git is intalled
ethana3: but somehow it managed to not be in my $PATH
rx__: wonders what distro that is
ethana3: Ubuntu Gutsy
rx__: very odd ;)
ethana3: I have to go to bed now
ethana3: if you see ethana2
ethana3: sometimes it increments that
ethana3: freenode can be interesting like that
rx__: so what is this cp microcode and how do i capture commands sent to cp?
ethana3: I'm not seeing my card in this list
ethana3: ATi Radeon 9200SE PCI 128MB
ethana3: I don't see it
OipOS: Hey. Is the x1300 series fully supported?
ethana3: I don't know if I'm looking in the right place for that
ethana3: I don't suppose it wouldn't work just because it's not listed in this file, but...
ethana3: makes me suspicious
ethana3: when I've had to go through hades to get mine working
ethana3: greping x1300...
ethana3: X1300 isn't in this .csv file
OipOS: Otherwise, any good recommendation for a new graphics card?
ethana3: but it probably is still uspported
ethana3: ok, what do you use it for?
OipOS: Though not all too new.
ethana3: then you'd probably want the performance you payed for, right?
OipOS: I can live with this x600 I have now, but it's almost dead.
ethana3: In that case, for Linux, BSD, and OpenSolaris
ethana3: you'd be looking at nVidia, most likely
ethana3: and their proprietary driver
OipOS: That's no fun.
ethana3: unless you're on PPC
OipOS: Nah, x86.
OipOS: I want to keep trying new experimental code. OSS one at that.
ethana3: you're in the right place ;)
OipOS: I know.
ethana3: ok, well, what I just did
ethana3: is git clone the radeon source
ethana3: I'm still trying to figure out what to do with it
ethana3: but that shouldn't be /too/ hard
OipOS: Well, I'm currently running Mesa, Xorg and the OSS ATI driver on git versions.
ethana3: I was using VESA a few days ago
OipOS: About 2 weeks old now.
OipOS: But x600 is fully supported. So that's not a problem.
ethana3: Well, I haven't been in the market for a card really
ethana3: I want to go from my Radeon 9200SE to a stream coprocessor
ethana3: under windows, it can do Halo at 640x480, 30fps
ethana3: with the really awesome effects turned off
OipOS: Lol, nice.
ethana3: pathetic card
ethana3: but actually
ethana3: under linux, it can do tremulous at like 1280x1024
ethana3: at 30fps
ethana3: until the poly counts get high
OipOS: Yeh, but that's Linux quality :P
ethana3: which intrigues me
OipOS: Though it's weird that Linux and Windows speed veries.
ethana3: I think so too
ethana3: but it might just be the distribution of GPU power
ethana3: but that would mean that if radeon were to be more flexible...
ethana3: meh, I don't know what I'm talking about yet
ethana3: I should keep browsing this source I just cloned
OipOS: I wanted to mess with it too, but I don't understand much of it.
ethana3: Well, I figure I'll just keep reading...
ethana3: maybe something'll click ;)
OipOS: It's a good thing that my linux is entirely self-compiled. It saves a lot of trouble with missing packages and stuff.
OipOS: Or headers.
ethana3: oh wow
ethana3: this README.ati file is very informative
OipOS: Really? Must read.
ethana3: this is going to be some good reading
ethana3: git clone git://anongit.freedesktop.org/git/xorg/driver/xf86-video-ati
ethana3: but first
ethana3: make a directory where you want to put your sources
ethana3: you might already have one in /, I guess
OipOS: It's somewhere in my home actually.
ethana3: mine is too :)
OipOS: I wonder if the clocks are still limited to 80MHz "on most adapters".
ethana3: You mean 800?
ethana3: or seriously just 80?
OipOS: It states 80 here.
OipOS: Also, I have no idea what the common speed for clocks is.
OipOS: Is it RAMDAC? Memory speed? Or something completely different?
ethana3: around 600MHz
ethana3: draw engine probably
ethana3: memory is usually around 1 GHz
ethana3: weird that way, memory faster than core
OipOS: It's in the README.ati btw, in case you're reading it.
ethana3: I'm not that far yet
OipOS: I think the readme is a bit old.
ethana3: what are you reading it with?
ethana3: this formatting is horrible
ethana3: in gedit and kate anyway
ethana3: Those configuring the driver's generic VGA support for a non-ATI
ethana3: adapter, can skip ahead to the [1m``Clocks for non-ATI adapters'' [22msection
ethana3: below. Those [4mnot[24m trying to configure the driver for a Mach64 adapter,
ethana3: can skip ahead to the [1m``Clocks for fixed clock generators on ATI[0m
ethana3: [1madapters'' [22msection below.
OipOS: kwrite for KDE4. But it still has those [22m things
ethana3: asian characters?
OipOS: Doubt it.
ethana3: gotta go to lunch
ethana3: be back later
jrmuizel: how difficult is it to add support for resizing the framebuffer to the xf86-video-ati driver?
rx__: jrmuizel; the code you want to look at is radeonfb in the kernel
jrmuizel: rx__: really?
rx__: i think so
rx__: what are you trying to accomplish?
jrmuizel: eliminate the need for having to specify a large virtual size
rx__: oh wait.. err.. i was thinking of framebuffer in VT
rx__: ignore me :)
airlied: jrmuizel: ttm.
jrmuizel: airlied: that's the type of answer I was looking for
jrmuizel: how far along is the radeon ttm implementation?
umtc: airlied, thanks for adding back zaphod mode
airlied: jrmuizel: not far enough :)
jrmuizel: airlied: is there a todo list or something someplace, if I wanted to take a look at helping out?
jrmuizel: or what's preventing it from being farther along?
jrmuizel: just available time?
airlied: jrmuizel: well I mostly did the kernel component, and glisse did some more with it..
airlied: jrmuizel: then the DDX sorta needed conversion, but the big blocker is the 3x mesa drivers..
jrmuizel: because all three would need to be migrated
airlied: pretty much.
airlied: jrmuizel: its much more likely to all get re-written around new frameworks..
airlied: jrmuizel: but that sorta slows down migration plans etc..
airlied: jrmuizel: like it would be nice to have a modesetting/ttm/ kernel driver,
airlied: jrmuizel: a port of the DDX that requires said driver
airlied: jrmuizel: then gallium based mesa drivers..
jrmuizel: what are the chances of a gallium based r100 driver?
jrmuizel: having what's nice sounds like an awful lot of work
airlied: jrmuizel: depends on whether marcheus work to make gallium work without shaders...
airlied: jrmuizel: porting the current drivers to TTM is also !trivial
jrmuizel: airlied: yeah, i've gotten that impression
airlied: jrmuizel: like is isn't that impossible, you would just need to move a lot of code around..
jrmuizel: is there much to be gained from having an r200 driver using gallium?
airlied: jrmuizel: other than sharing the winsys layer probably not..
airlied: jrmuizel: I do wonder if we could just build another driver on the common winsys..
airlied: that was old-style mesa but used winsys layer..
jrmuizel: i.e. how realistic a situation is it to have a gallium r300 driver, and have the existing r200 and r100 drivers ported to ttm?
airlied: jrmuizel: thats probably the most realistic..
airlied: jrmuizel: if marcheu does gallium for shaderless cards then r100/r200 gallium drivers are more realistic..
ethana3: the main source file for the driver is one big list of hardware bug workarounds
jrmuizel: well, there's certainly a lot of work that could be done...
ethana3: kinda sad ;)
jrmuizel: i might take a look at porting r200 to ttm
airlied: jrmuizel: tbh r100/r200 should be 90% the same really..
jrmuizel: airlied: a gallium version of r100/r200 you mean?
jrmuizel: i.e. the same as what?
jrmuizel: as each other?
airlied: jrmuizel: each other..
airlied: jrmuizel: talk to marcheu about the shaderless-gallium idea..
airlied: jrmuizel: but playing with ttm driver should be fun enough..
jrmuizel: yes, no doubt
jrmuizel: I recall, long ago, there some patches that worked on merging the r100 and r200 drivers (from marcheu?)
jrmuizel: I don't think they were every applied
jrmuizel: anyone remember any reasons?
jrmuizel: and thanks for the information airlied
airlied: jrmuizel: didn't apply as nobody wanted the re-testing overhead..
jrmuizel: which i suppose is a common problem whenever touching the older drivers
airlied: jrmuizel: yes.. with TTM we have a real reason for doing it though..
airlied: jrmuizel: for the userspace driver, you would try and do an initial port to the bufmgr..
airlied: jrmuizel: like 915..
airlied: the bufmgr tries to abstract the TTM buffers so that the 3D driver can run on non-ttm .
airlied: as well as ttm..
ethana3: if a card isn't in the driver pci database,
ethana3: will it have problems working out of the box?
ethana3: Mine isn't there, and I went through hades just to get it working
airlied: ethana3: what card?
ethana3: with a clean Ubuntu Feisty install
airlied: ethana3: what pci id?
ethana3: ATi Radeon 9200SE PCI 128MB
airlied: ethana3: what pci id?
ethana3: hold on...
ethana3: that right?
airlied: ethana3: lspci -n
airlied: ethana3: search for a line with 1002 in it.
ethana3: two of them
jrmuizel: lspci -n -d 1002:
ethana3: 02:09.0 0300: 1002:5964 (rev 01)
ethana3: 02:09.1 0380: 1002:5d44 (rev 01)
airlied: ethana3: that is in the pci db..
ethana3: oh, ok
ethana3: didn't know what I was looking for
ethana3: wait, how do you tell?
ethana3: the 5d44?
airlied: ethana3: that isn't a card.. that is just a windows placeholder for dualhead
ethana3: and the .1 is the monitor controller
airlied: ethana3: nope.. its just a stupid windows hack..
ethana3: you mean like, the card designer did dumb things because they were making it for Windows?
airlied: ethana3: yes... windows made the card designers add a second pci id placeholder due to a brokenness in the windows driver model..
erudified: didn't someone port radeontool to OSX?
airlied: erudified: I've sorta ported it.. for nouveau..
erudified: is the source available anywhere? I'd like to play with it...
airlied: erudified: http://www.skynet.ie/~airlied/nouveau/AppleSamplePCI.tar.gz
airlied: it needs a kernel driver..
airlied: so I started from apple's sample code..
erudified: IOKit rocks, huh?
airlied: erudified: yes I decided to go back to Linux :?)
airlied: theclaw: git://git.freedesktop.org/git/xorg/driver/xf86-video-ati
airlied: theclaw: atombios-support branch
airlied: theclaw: so cd xf86-video-ati; git checkout -b atombios-support origin/atombios-support
theclaw: thanks for the explanation :)
theclaw: airlied: does this driver support rotation for my ati mobility x1400? :)
theclaw: (i have an external display connected)
airlied: theclaw: not yet.. rotation requires 3d..
theclaw: i'm just curious, why does rotation require 3d?
airlied: theclaw: to do the rotation..
airlied: theclaw: most cards don't have rotation hardware..
airlied: so you just allocate a new framebuffer and rotate into it.
theclaw: restarting X..
theclaw: hi there, again
theclaw: airlied: it works faster, I think. thanks :)
airlied: theclaw: you can try installing the kernel module as well if you are feeling really adventurous..
theclaw: airlied: what would be the "possible advantages"?
rx__: kernel module means installing a newer libdrm
airlied: rx__: nope.. just kernel module..
airlied: theclaw: it would be less CPU used and slightly faster..
rx__: i tried kernel module only.. complained about some intel stuff
airlied: rx__: no idea why it would..
rx__: er compiling errored on an intel module
airlied: rx__: just do make radeon.o
theclaw: airlied: so using the drm kernel module doesn't need "3d support" for my card?
rx__: alright :)
theclaw: i thought drm is just use for opengl
theclaw: *just used
rx__: it's also used for 2d acceleration
airlied: theclaw: no it also allow access to more features of the card..
airlied: mainly the command processor..
rx__: how do you log the commands sent to the CP?
airlied: rx__: either in the DDX before it sends the buffer of commands..
airlied: rx__: no either..
rx__: there's a few warnings making the module
rx__: map_page_into_agp, flush_agp_mappings, flush_agp_cache, and unmap_page_from_agp undefined
theclaw: airlied: hrm.. I'll might try it sometimes, but first the basic stuff ;-)
theclaw: Xorg.0.log says: "(**) RADEON(0): DPI set to (94, 108)"
theclaw: what do the two numbers mean?
rx__: just copy all *.o and *.ko right?
airlied: rx__: just copy the drm.ko and radeon.ko
airlied: rx__: over the kernel ones.. after backing the kernel ones up
rx__: too late for that ;/
rx__: failed to open DRM
OipOS: Are the modules loaded?
rx__: drm and radeon
airlied: rx__: you may need to rmmod radeon drm and modprobe radeno
airlied: radeon even.
rx__: this from boot
airlied: rx__: have a look in dmesg for a line with radeon in it..
airlied: rx__: did you use the r500-support branch of the drm git?
rx__: nothing with radeon
rx__: yes r500
rx__: nothing in /dev/dri/
airlied: rx__: then it isn't using the right radeon module..
airlied: rx__: try insmodding them by hand maybe..
rx__: i get unknown symbols
airlied: rx__: wierd..
rx__: i guess i should listen to warnings ;)
rx__: map_page_into_agp, flush_agp_mappings, flush_agp_cache, and unmap_page_from_agp unknown symbols
airlied: rx__: do you have a custom kernel?
airlied: rx__: with no agp?
rx__: uh.. probably
rx__: let me check .config
airlied: actually it should still mostly work.
airlied: it should find asm/agp.h no matter what
rx__: yah.. CONFIG_AGP not set
airlied: wierd it should deal with that..
airlied: rx__: ah can you move the include for asm/agp.h in drmP.h
airlied: rx__: outside the CONFIG_AGP
erudified: Did ATI never release M54 specs?
airlied: erudified: they've only released M56 and rv630 modesetting specs..
airlied: erudified: but most chips are pretty similiar..
rx__: great.. no warnings this time..
rx__: tries insmod
erudified: airlied, is the latest atombios devel going on in radeonhd or xf86-video-ati or what?
theclaw: how do I configure the DPI with the radeon driver?
rx__: okay.. looks like that worked
rx__: another silly bug fixed? ;)
airlied: erudified: there isn't really a central atombios devel.. we haven't modified AMDs code that much yet..
erudified: ok, thanks!
erudified: man, why does VESA suck so much? if you could just set arbitrary modes with VESA, life would be easy
erudified: well... for me =)
rx__: what tool is used to trace ddx?
airlied: rx__: trace what in DDX?
rx__: to log the CP?
rx__: i;'m curious what 3d looks like
airlied: rx__: 3D isn't done in the DDX..
airlied: rx__: 3d looks like binary data..
airlied: lots and lots of it..
rx__: oh great
rx__: decides to watch spooks instead
dli: I retreated back radeonhd. system lockup, with block of display messed up
dli: rx__, oh, broken sentence, I retreated back to radeonhd, because I had system lock up with radeon