[ltp] ATI Rage mobility dual-head support (was: x20 questions)

[Alex Deucher]

Quick update:

I actually just took another look at the spec and adding dualhead
support to mach64 should be very easy.  the problem is the current
driver, ati_misc, is a complicated maze of code that handles all the
old ati isa and pci chips.  It may take a while for me to sort out the
code.  I'm tempted to write a new driver for mach64 modelled after
r128 and radeon , but I don't know if it's worth the effort for such
an old chip.

Alex

On 7/31/05, Alex Deucher <alexdeucher@gmail.com> wrote:
> On 7/31/05, Rolf Leggewie <no2spam@nospam.arcornews.de> wrote:
> > Alex Deucher wrote:
> >
> > > no, they are all variations on the same chip.  All of the mach64 base=
d
> > > rage mobility support dualhead.
> >
> > You said elsewhere that root access would not be sufficient.  Dual-head
> > support would be an important feature for me so obviously I want to hel=
p
> > out in any way I can.  Call me naive and I hope I am not wasting your
> > time but maybe we can still get something going.  Three things I was
> > considering.
> >
> > 1) Do you happen to know by chance someone in Europe involved in x.org
> >     dual head work?
>=20
> I know there are some Xorg developers in Europe, but whether or not
> they would be interested in writign the driver or have the specs to do
> it, I don't know.
>=20
> > 2) root access coupled with a webcam so you can see what is going on.  =
I
> >     understand and fully appreciate this would be tedious for you.  I'd
> >     be willing to leave this running 24/7 so you can work in your pace.
> > 3) I assume video graphics programming is very low-level, dig-down-deep
> >     kind of stuff.  But maybe there is some kind of way we can work on
> >     this together by you sharing some of your knowledge with me.  I am
> >     afraid this might be tedious for you as well ;-)
>=20
> yeah...I don't think it would be really workable.  The problem is it
> quite easy to hang the chip or laptop in the process and then it would
> need to be rebooted, etc.  I could try and code up some stuff to test
> with and have you test and report back, but that tends to be painful
> on both sides since you have to wait for feedback both ways.
>=20
> >
> > Problem is that I do not even have a grasp about what is involved in
> > getting this supported.  Are there specs to be had?  I guess it is not
> > just a) write support and b) kill out the bugs, is it?
>=20
> I have the databooks from ati, but they are under and NDA so I can't shar=
e them.
> Normally, I find it easied to figure out how to get crtc2 programmed
> in conjunction with crtc1 (like a clone mode).  once I figure that
> out, then I can separate the crtc2 stuff and add support for dualhead.
>  Basically you have two display controllers (usually called crtcs for
> CRT controllers).  the crtcs can be mapped to one or more outputs such
> as a DAC for vga ports or a LVDS contoller for laptop flat panels or a
> TMDS controller for DVI ports.  What happens in Xorg is that the
> driver loads twice for "regular" dualhead.  the first time through it
> sets up the first crtc, the second time through, it sets up the
> second.  since there is only one drawing engine, the server has to
> sync the engine and change the offset between each head use of it.
> this works for 2D accel since it is server side only.  it doesn't work
> for 3d since right now only the direct rendering case is accelerated.
> with direct rendering, the 3d client talks directly to the 3d driver
> via libGL so there's no way to know which head it's on to change the
> 3d engine offset.
>=20
> Mach64 is similar to savage in that the crtc registers are shadowed.
> basically you flip a bit in one register and program one set of crtc
> regs, then flip it back to program the other.  The registers live at
> the same offsets though so your mode setup functions can be pretty
> similar. you just have to make sure you got the right crtc selected.
> the problem is mode handling is usually the quirkiest part of graphics
> programming so it usually the hardest to get working 100%.
>=20
> >
> > BTW, http://www.botchco.com/alex/dualhead/ is your site, isn't it?  Jus=
t
>=20
> yes.
>=20
> > a stupid question, you always talk about laptops on that site.  What is
> > the relation to the normal Mach64?  Are they very similar?
> >
>=20
> It's not just laptop, it lists all chips with dualhead support.  I
> just so happens that dualhead sort originated in laptop chips so that
> they could drive both the LCD and monitor at different refresh rates.
> The rage mobility, M1, and rage LT pro chips are all based on mach64
> core.  As such the 2d and 3d engines and mode controls are all pretty
> similar across the line.  hence they are all handled by the same
> driver.  The mobile mach64 chips just added a flatpanel controller and
> a second crtc.  The M3 and M4 are based on the rage 128 core and m6,
> m7, m9, etc. are mobility radeon chips.  It wasn't until the radeon
> however that ati actually released a dualhead desktop chip.  The basic
> r128 controls are so similar to radeon they really could be handled by
> the same driver.  I've though about combining them, but it'd be a lot
> of effort for not much gain.
>=20
> Alex
>=20
> > regards
> >
> > Rolf
> >
>