Re: 65816 coprocessor options..

[mjmahon@aol.com (Michael J. Mahon)]

Pim wrote:

>In article <385DEE13.4F190C34@dcnet2000.com>, phoenyx
><quazar@dcnet2000.com> wrote:
>
>> I would have liked to see the Apple2 with the ability to map in
>> memory blocks into any address, similar to the PC. Perhaps
>> using 4k blocks swappable to any address from $800 up, no
>> matter where it actually resides in the ram pool.
>
>Processors didn't do that in those days, not even the 8086!
>
>But what is the USE of it? I mean, I know later processors can, but I
>never could envision the benefits of having two blocks of memory in two
>different locations instead of one large block.
>Can you give me an example of a situation where being able to remap a
>piece of RAM from, say $001000 to $401000 would give you a distinct
>advantage over not being able to do that?

Virtual memory comes to mind...  ;-)

At various periods in computer evolution, the amount
of physical memory considerably exceeded the addressing
range of processors.  The transition from 16-bit addresses
to 32-bit addresses comes to mind, as does the current
transition from 32-bit to 64-bit (>32-bit) addressing.  In
each case, systems can be configured which have several
times as much physical memory as any single process
can directly (virtually) address.  Fast remapping of scarce
virtual addresses to plentiful physical addresses is used
to permit larger working sets without I/O or data copying.

Appleworks uses a software "virtual" memory management
technique, but it has to physically move each block of
">64KB" memory that it accesses.  Not having to move
it would be faster.  (Of course, the actual "moving" is
done through a specialized address remapping.)

-michael

 Email:  mjmahon@aol.com
 Home page:  http://members.aol.com/MJMahon/