[MPlayer-users] Re: Radeon frambuffer use if Radeon is not primary display?

[Sven Hartge]

Arpi <arpi at thot.banki.hu> wrote:

>> I mean, AGP-cards register themself as a PCI-Device on PCI-Bus 1 (on
>> normal PC-Mainboards with only one two PCI-busses, namely PCI as bus:0
>> and AGP as "bus":1) and are therefor accessable as PCI-Devices:
 
>> 00:08.1 Multimedia controller: Brooktree Corporation Bt878 (rev 11)
>> 00:09.0 Ethernet controller: Accton Technology Corporation SMC2-1211TX (rev 10)
>> 00:0d.0 SCSI storage controller: Advanced Micro Devices [AMD] 53c974 [PCscsi] (rev 10)
>> 01:05.0 VGA compatible controller: Matrox Graphics, Inc. MGA G400 AGP (rev 82)
 
>> So there is (as far as I know) no limitation from the hardware point of
>> view, but there may be software-specific problems.

> Yes, _if_ the bridge to the card is initialized and working.
> on-board agp bridge is usually initialized and working.
> but the way for pci cards with agp chip:
> 1. initialize MB
> 2. initialize card -> enable pci-to-agp bridge on it
> 3. now you can access the device

> it means your driver has to support the bridge chip on the card to
> be able to access the vag chip.

Question: Who initializes the pci-to-agp bridge if such a card is the
primary VGA adapter in a system? Since it has to be operational before
any driver is run, this has to be done by the BIOS, right?

If such a card is a secondery adapter, the pci-to-agp bridge might be
uninitialized and you've got some problems, did I get this point
correct?

If the pci-to-agp bridge is initialized, access to the device should be
transparent, or do I miss something?

>> But: Every AGP-Card has to be VGA-compatible, or it is not allowd to be
>> in the Display-Controller-Class and has to be put into the
>> Multmedia-Controller-Class (that is where for example Voodoo2-Cards
>> reside, since there are not VGA-compatible), so at least the vesafb
>> should run.

> vesa? it only works on the first card...

Right, I was talking about the first card being an AGP-Card yet
unsupported by the software. 

> two card's vesa bios can't be mapped as they use the same address space
> (c000:0000) and register space (3xx)

Yes, is why I like my Millennium II so much, it has a little jumper to
disable the VGA/VESA-Bios, so this card works in boards, which fail to
correctly disable their on-board-graphics if another card is present.

> i don't know too much about the details (but it seems its' still much more
> than you) but i believe what dri gurus said on dri-devel.

I believe you, since everything I know is only what I have read on
Usenet or mailinglists, so call it third-hand-informations.

(If everybody would only buy hardware which is supported by linux, then
no problems of this kind would arise.)

S!

-- 
BOFH excuse #428:

Firmware update in the coffee machine