Re: [ql-users] Qliberator & SMSQ - problems problems

[ZN]

On 3/24/01 at 2:13 PM Richard Zidlicky wrote:

>> Well, I was told that code and data accesses should actually ignore
>> the top three bits (29, 30, 31). In other words, any combination of
>> bits 29, 30, 31, should effectively act as if those bits are 000.
>
>depends on the degree of compatibility you are trying to achieve, 
>for some software everything above 16 MB must be wrapped down.
>You can play with UQLX and Minerva to see the effects.
>A programmable MMU helps of course.

OK, any examples of software that needs 16Mb wrapping (or 32, 64, 128,
256)?
The GF memory map is 512Mb, wrapped 8 times within the 4G address range.
It's roughly divided into 128M RAM, 128M I/O, a 128M copy of the RAM (with
cache disabled), and a 128M copy of the IO with cache disabled. Doing a
wrap of everything ino 16M would be rather pointless since the evey key
feature implemented on the GF could not be implemented any more.

Nasta