Ian wrote:
> > I don't know. It's not real efficient in the non-DoC case. I think when I
> > did measure it it was 4k.
> hmm ok ...
>
> > Given that the total of linuxbios is 30K or so, what will that buy? I'm
> > just asking. To jump to memory in a PCI slot you need to do a fair amount
> > of work, so why not just boot Linux at that point?
> Because I'm running more than 512k worth of software with my kernel ... mind
> you my current kernel is 300k (I'd like to know how you squeezed it to 30k) ...
> my initrd is just under 2mbytes, and I have a technology partner who's prepared
> to make some modifications to his PCI card, to my specifications ... Basically,
> I can run all of the BIOS specific crap on a small flash in the DIP32 socket, then
> when the universe is resolved I can jump to the common stuff in the PCI bus
You may want to know that the PCI specification, as detailed in "PCI System
Architecture" Third Edition by MindShare, states on page 363 that "The PCI
specification requires that device ROM code never be executed in place"
but must always be shadowed into ram. The ROMs are slow, but it doesn't
say why the spec REQUIRES it. FYI
This is very interesting though, I've thought of using the VGA bios flash and
SCSI, or especially tempting the Netboot rom socket on most ethernet cards.
Together they might eliminate the need for DOC in some cases.
Keep posting what you find out.
>
> somewhere ... should be cheaper than playing with the DoC too ... should be
> able to format off the Award BIOS and use the existing chip ..
>
> --
> http://HumanHeuristic.com/
> "Bringing people together in a world full of computers"
