Little.Fleas%hi-multics@sri-unix.UUCP (02/23/84)
From: Gary Little <Little.Fleas@hi-multics>
Re: Z80: 16-bit?
Frankly, comparing a Z80 to a 68K is like comparing a Model A Ford to a
Porsche 925 and poopooing the Porsche because it doesn't have a starting
hand crank - or the Model A because it needs a hand crank. If you want
to compare, put the Z80 against the 8080/5 (yeeccchh) or the MC6809
(huummmm!). Also, it seems that to compare a varmint such as the
Z800X/? against the 68K or the 8018x family equates to comparing an
Edsel to a Lincoln or a Cadillac. Expound on the glories of Zilog all
you want, then tell us the number of machines that are using a Z800x,
and how much software is out there. At last count, Commodore was
threatning a Z8000 machine, and some one else was using it to handle IO
for a 68K.
As to the multipy/divide problem, so what? Try adding 16 bits from a
variable to the BC pair on the Z80 or the 8080 in less than FOUR
instructions using the following psuedo-code:
get a 16 bit variable and put it in BC
add the 16 addend to it
Equivalent 6809 code:
LDD V1
ADDD V2
How about doing indexed indirect addressing in less than
FOUR instructions; i.e. what is the equivalent of the following 6809
instruction.
"LDD [100,Y++]"
THAT you can't even do in the marvy 80186!!!
As you can guess, I am thouroughly enamored with the wonders of the MC68000,
and believe that it is the best thing since sliced bread. The biggest
reason for this is the programming model in the assembly language for
the Motorola family of processors. I find myself severly handicapped
and irritated by the models in Intel and Zilog, to the point of purposely
avoiding them at all costs. And if I am constrained as a programmer, I
can imagine that a compiler is virtually locked in concrete in Intel and
Zilog.
Simply stated, I think you can do more with a 68K, in less space and in
less time than you can with Intel, or Zilog. That is an overall
impression. True, in any given chip there are things that shine, but
remember not all programms are looping on strings, or doing multiplies
or divides, but nearly every program can benefit from strong addressing
capabilites such as indirection and indexing with auto-increment and
decrement. Besides, the unique optimization of string manipulation that
you can do with the 80186 is rather difficult to implement in a
compiler, and usually requires a ten toed biped as an optimizer. In
short, which processor being used is important. If you want a lot of
software, then stick with Intel (Zilog isn't even in the market). But
if system response time is important, then stick with Motorola.
Gary