This is all said in my capacity as a member of the RiscBSD team and has nothing
to do with Causality.

On Feb 9,  4:57pm, Neil A. Carson wrote:
> Subject: Re: Generic console device ?
>
> The main thing here is I don't want people to expect they can plug in an
> off-the-shelf whizz-bang bells an whistles PCI card. From *Causality's*
> point of view, on our own PCI hardware, we will support, with XFree86
> and probably pcvt, probably a certain S3 graphics card, maybe a high-end
> card such as a Millenium II, and maybe a Cirrus board. That'll be it.

What I am aiming to do is have a 'good-enough' console across the board with
bells and whistles for specific cards.

> Again on our own hardware we will get XFree86 up, so that all of its
> drivers can be used with the appropriate PCI graphics cards, *provided*
> that the equivalent work of the BIOS on the PCI cards has been done. We
> may have a BIOS emulator to do this on our board, though whether or not
> this BIOS emulator will be ported to RiscPC 2 is again completely up in
> the air, and to a fair extent depends on how many people buy it.
>
> Mark's spent ages faffing around just trying to get _one_ video card to
> work properly, and it really shouldn't be underestimated just how
> colossal a job this is, given the complete absence of documentation and
> a PC BIOS.

Initialisation is hard but after you've got a basic VGA text mode you don't
need the BIOS. If Mark gets the BIOS emulator working I don't expect this to be
an issue. If not, this might just take rather a long time...

> Note that running a bitmap console over PCI is absolutely *not*
> efficient, unless a fair amount of effort is put into providing support
> for some kind of hardware accelerated character plotting via the cards.
> OK if this isn't done, it means sticking with an 80x25 text console, but
> at least that ensures most compatibility with XFree86.

I don't expect it to be worse than the RiscPC bus for sequential access. You
may be limited to standard SVGA modes and DRAM RiscPC type graphics performance
on anything which doesn't have special support. An 80x25 text console is
actually fairly low resolution so the bus bandwidth isn't horrendous to emulate
this bitmapped. You may ask why anybody would want to do this rather than run
pcvt/XFree86. A potential answer is security.

I am deliberately aiming high because if I don't, it won't happen.

Of course this may all take rather a long time to implement.

If anyone wants accelerated support for a specific card that we have no
intention of doing full support for they can either write the driver themselves
or contract Causality to do it.

The bottom line is that if I don't try I'm guaranteed to fail.

Cheers

Rob