Re: Re[2]: PCI I/O Space Consumption Limitation

["John R Pierce" <pierce@scruznet.com>]

----------
> From: Taylor, Mike <mike@buntypost.dundee.ncr.com>
> To: Mailing List Recipients <pci-sig-request@znyx.com>
> Subject: FW: Re[2]: PCI I/O Space Consumption Limitation
> Date: Thursday, August 22, 1996 2:00 AM
> 
> I thought PCI/ISA bridges used subtractive decode and therefore were only   
> active in address spaces not acknowledged by any PCI bus. Aliasing by an   
> ISA card would therefore be prevented in any PCI used I/O space. The 256   
> bytes per I/O BAR restriction is therefore not due to ISA limitations.
> 

Say we had a ISA I/O device like the old Media Vision Pro Audio Spectrum 16...
This used 10 bit decode 0x388 and virtually ALL aliases (0xB88,0x1388, etc etc
etc) for additional registers.  If a PCI card used ANY 4K I/O space, it would
prempt some set of these ports from operation.  This is just a example, and an
extreme one at that, but that was the rational behind restricting PCI I/O
decodes to using addresses that had a bbbb.bb00.xxxx.xxxx binary address.

Basically, any modern PCI device should be memory mapped in PC terminology.  To
a RISC engine, its all the same thing.  PC I/O space exists only for legacy
purposes.

-jrp
i
l[