At 08:25 +0000 09/27/2003, ML wrote:

>>  In theory, there should not be any difference in compatibility
>>  between Bandit and PPB slots.  However, there is definitely a bug in
>>  the Apple firmware (ROM) with respect to PCI-PCI Bridges.

>Yep. The Millennium II drivers didn't work at all behind the bridge, a
>  Millennium I did.

I forgot to mention that the reason I'm fairly certain that the flaw 
is in the Apple ROM is that I replaced the DEC 21052AB PCI-PCI Bridge 
chip with other pin compatible chips and the problem did not change 
its nature in any fashion.

I tried the Intel 21152AB, 21152BA, TI PCI2250 and the Hint 
HB1-SE33P.  With the Intel chips, I figured that Intel had just 
carried over any design flaws in the DEC design when they bought the 
group and updated the chip.   Finding the same results with the TI 
chip pretty much convinced me that the problem was not in the PPB 
chip, but it occurred to me that design engineers move around, so 
TI's design could have inherited flaws from an engineer who moved 
over from the DEC team.

Finding identical results with the Hint chip convinced me that the 
problem could not be in the PPB.  The Hint PPB is a recent design 
compared to the DEC, Intel and TI PPBs and it is extremely unlikely 
that they would all have the same design flaw.

So the problem must be in Apple's ROM or in the Bandit chip.   The 
behaviour of the problem makes me think it is in the firmware, but I 
don't know for certain.   It could be in Bandit.

Has anyone disassembled the OF code and deciphered it?   If someone 
can do bug fixes to the code, I can burn the results to pin 
compatible flash memory which can be substituted for the Apple ROM 
chips.

Jeff Walther