RE: Subsystem IDs, again

["Kevin D. Davis" <kevind@ti.com>]

I knew what you meant.  I guess you couldn't tell from my reply.

As long as anyone that implements the "pigeon hole" method actually has 
another method, I'm happy.  The problem is that some chip vendors take a 
real OPTION CARD view of the world.  They expect or even demand that you 
implement their function exactly like their reference design or worse yet, 
their retail product.

Kevin

----------
From:  John R Pierce[SMTP:pierce@scruznet.com]
Sent:  Friday, August 02, 1996 2:01 PM
To:  'PCI SIG'; 'Kevin D. Davis'
Subject:  RE: Subsystem IDs, again

Um, actually, what I meant was these chips actually 'pigeon'
hole the subsystem ID configuration reads and reflect them into
a ROM BIOS read of some fixed offset...  I guess I didn't phrase that 
well...

When you read the configuration address for the subsystem/vendor ID info,
the chip actually reads the bios rom itself.  Yes, you are correct, this
won't work on a motherboard solution...  But a motherboard solution
by definition can have the system bios pre-load these registers.  Even a
ACPI system will invoke the motherboard bios on power-back-up...

-jrp

----------
From: 	Kevin D. Davis
Sent: 	Friday, August 02, 1996 7:56 AM
To: 	'PCI SIG'; 'John R Pierce'
Subject: 	RE: Subsystem IDs, again

Please don't count on the BIOS to load up cards.

Once ACPI comes along, there is a good chance that your boot time BIOS
won't get control after a power transistion.  There MIGHT be a chance for
your driver to remember the ids, but that's not entirely clear to me.

As for your chip reading directly from the BIOS, for chips I put in the
notebook, I merge the BIOS into the system BIOS.  Thus, MY core logic chip
controls access to the BIOS, not your chip.  So, if you use this approach,
make sure there is a cheap solution for notebooks.

Kevin

----------
From:  John R Pierce[SMTP:pierce@scruznet.com]
Sent:  Friday, August 02, 1996 1:20 AM
To:  Mailing List Recipients
Subject:  RE: Subsystem IDs, again

Peter wrote...
>A serial ROM is the only solution for expansion cards, but I don't think
it
>violates the spirit or the letter of this requirement to use a write to
the
>user config space to set up the subsystem ID on motherboard devices, and
>this would save motherboard designers the need to have one SROM per PCI
>device.

Actually, there is an alternative for devices with bios roms (mainly
graphics
and scsi adapters)...  The subsystem ID registers can be mapped to fetch
from the bios rom in parallel...

-jrp



ÑTA