[Coco] How much memory

[Aaron Wolfe]

On Sun, Nov 28, 2010 at 9:18 PM, Frank Pittel <fwp at deepthought.com> wrote:
> On Sun, Nov 28, 2010 at 08:51:53PM -0500, Aaron Wolfe wrote:
>> On Sun, Nov 28, 2010 at 9:04 PM, Frank Pittel <fwp at deepthought.com> wrote:
>> > On Sun, Nov 28, 2010 at 05:51:45PM -0500, RJRTTY at aol.com wrote:
>> >> In a message dated 11/28/2010 5:35:23 P.M. Eastern Standard Time,
>> >> cmayeux at msbministries.org writes:
>> >>
>> >> >Perhaps we could keep it at 2 megs to eliminate some  hardware
>> >> >And software "backpedaling"... just a thought.
>> >>
>> >>
>> >> I agree.   2 megs is plenty of memory for an 8 bit 2 mhz
>> >> processor to run around in.....
>> >> and it has a certain symmetry to it since the task registers
>> >> will also be 8 bits in a 2 meg system...
>> >
>> > The coco3fpga can run the "6809" as a 25mhz cpu and doesn't have the limitation
>> > of only being able to directly address 64K or memory!
>>
>> The 6809 can only directly access 64k of memory whether it's in
>> silicon, FPGA, or being emulated accurately.  This is a fundamental
>> aspect of the chip, it is true of all 6809s everywhere in every
>> application.
>>
>> The CoCo allows you to use more than 64k by mapping in little pieces
>> of a larger address space into the single 64k space the 6809 can
>> actually talk to.  Giving it massive amounts of RAM doesn't change the
>> fact that at no time will the 6809 ever seen more than 64k of it.
>> This is not a function of an operating system or programming, it is
>> simply how a 6809 works (and also how many other "8 bit" chips which
>> had 16 bit addressing work).
>
> That's only the case if the fpga "6809" is limited to a 16 bit address bus. :-)
>

True.. but changing the addressing is easier said than done, and
easier done than actually used by any software.  I'd think
improvements to the MMU supporting the 6809 might be easier and more
compatible.  Blocks smaller than 8k would allow more efficient use of
the 64k space, more addressing lines on the mmu would give more room
for things to be pulled into the space, that's been done in the real
coco with the > 512k upgrades.

I suppose the important question is "why would we want more than 64k
directly addressable?".  Wayne Campbell wrote a great explanation of
why he'd like it.  I think a lot of what he'd like could be
accomplished with improvements to the software more easily than the
hardware.  I guess I'd say that any task can be done within almost any
amount of addressable ram, if you want to do it badly enough :)   64k
is actually quite a lot for many tasks.

Something that John Kent mentioned is putting multiple 6809 cores into
a single "mega coco".  There is an interesting way to extend the
address space :)  With four cores you've got 256k directly
addressable, sort of.