On Nov 15, 21:40, Iggy Drougge wrote:
> Tony Duell skrev:
> >What 26 bit machine were you thinking of?
>
> The old Acorn Archimedes machines.
> I'm sbscribed to the NetBSD/arm26 list, and I seem to recall that
registers...
> Oh, forget it. I'm a know-nothing when it comes to the Acorn.
>
> >> live condor in the northern hemisphere.
> >> What feature about the older ARM processors is 26 bit?
>
> >In some cases, I believe the address bus was 26 bits wide.
>
> What a nice, round figure.
Yeah. Right :-)
Someone decided that they'd not need more memory in the foreseeable future
than could be addressed in 26 bits (where have we seen this reasoning
before, I wonder) and in fact RISC OS even divided that into two images,
logical and physical memory. The other part of the reasoning ws that you
have to put the processor flags somewhere, and putting six bits in a 32-bit
register is wasteful, and we only have sixteen registers to begin with, and
we don't need all the width of the PC, and .. and...
Anyway, that's how it came about, and RISC OS has always been restricted to
26-bit address ARM chips. Hence the recent discussions on comp.sys.acorn.*
about why it would be too much work to make RISC OS run on the newst
incarnation of StrongARM, which has a 32-bit mode but not 26-bit. I
*think* the original StrongARM has both, but I could be wrong. Earlier
ARMs (Arm2, Arm3, Arm{6,7,8}00 are all 26-bit address (No, I don't know how
the memory management deals with the 256MB possible in a RISC PC; it came
after I left Acorn).
In fact, if you want to be pedantic, the PC itself is only 24-bit, because
the bottom two bits of the register are flags used to indicate the
processor mode (user, interrupt, fast interrupt, supervisor). The bottom
two bits of the address bus are alway forced to zero, and all memory
accesses are word-aligned. So how does it read a byte (or half-word) off
an non-aligned address? Easy, it uses it's barrel shifter to rotate the
word while loading!
--
Pete Peter Turnbull
Dept. of Computer Science
University of York
Received on Wed Nov 15 2000 - 17:30:00 GMT