On 7/26/06, John Culp <[EMAIL PROTECTED]> wrote:

> yeah, right... then you need an XML parser in the driver... somehow I
> don't think the kernel people will think it's a good idea...
>

I am in no way shape or form arguing for XML, but:

Would not the part that sits in the kernel ideally only need to know how
to manage transfers from user space to kernel space and handle errors
gracefully?  If an invalid command for a non-existent functional unit is
invoked by a series of uploaded commands a standard error could be
thrown by the hardware and handled by the kernel.  X could be entrusted
to parse a hardware description and figure out how to best utilize the
hardware available.

The only bits that would have to be standardized between compatible
variants would be the hardware that provides a dumb framebuffer for the
kernel managed console, the command/data upload/download hardware, the
error handling protocol/mechanism, and perhaps the mechanism for
changing video modes.  Compared to the total hardware that can exist on
the card, this is pretty minimal.


I haven't been following the whole conversation, so I may repeat
something, but...

The PROM is 1 megabit, which is 128KiB.  I think.  I don't know how
big the firmware is going to be, but it shouldn't be huge.  Most of
it'll be written in assembly and various sorts of bytecode (OpenBoot,
EFI or whatever that Intel thing is, etc.).  We should be able to fit
in firmware for more than one architecture with room left over.

Use it however you like.

Some of the info you want to know is going to be available via VPD.
Anything else can be placed in some standard location in the PROM so
that software can get at it.

In any case, since this is all just metadata, there's no need to use
more silicon for it.

I've never been a fan of XML, although that could be a mental
deficiency on my part.  Or a background with small memory-limited
processors and hardware design.  But I have done plenty of application
development and web development, so I have had plenty of experience
with having to store data, which means you need a format.  Given the
overhead, I don't see much point in using XML, but you can if you want
to.

But backing up a bit, why do we want this in the first place?  How
many "standard" designs for OGD1 would we really have?  Actually, it
would be great for business if there were lots of them.  Sometimes, it
would make sense to combine designs.  I just don't see it getting so
complex that even a simple list of features wouldn't be overkill.  How
about we use the wiki to define 32-bit identifiers that we put in a
zero-terminated list somewhere in the PROM.  If you don't know what an
identifier is, you can look it up on the wiki.  Moreover, most codes
would be allocated as 4-letter names that, when viewed as ASCII, may
make some kind of sense.  Things like:  NICx, OGA1, VDIN, ADIO, etc.
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to