>At 02:22 30/10/96 GMT, Andres Djordjalian wrote:
>>And also I guess the 4K-word ROM on the 16C73 is equivalent to almost 10K
>>bytes on another micro, and more if it's programmed with a HLL, ...
>
>What makes me wonder with such comparisons: 4k 16bit words already _equals_
>8k with no added efficiency involved, so the 10k _bytes_ on another micro is
>not too far off the 8k _bytes_ which is 4k words... Seems the microchip did
>a good marketing trick to introduce 16bit _words_ as the base of their
>codes: one of the big arguments is that all instructions are 1 word == 16bit
>wide (on the bigger ones) -- but the average instruction in a typical 8031
>code is way smaller than 2 bytes, as there usually are few with more than 2
>bytes, but some with only 1. They just cannot claim that it is _one_ something.
>

Actually, the word size on a 16Cxx PIC is 14 bits.  But I think that you
miss the point about the word size.  Lets consider a 68hc11.  There are
three ways to address something, depending upon how far away that something
is from where you are right now.  Short jumps (+- 127 bytes) take a single
code space, longer jumps take two or three code spaces.  Since the PIC uses
a 13 bit address, these jumps usually (PCLATH problems notwithstanding) take
only ONE code space.

I have found that rewriting some of my old 'hc11 stuff into PIC format
really does result in a 2 to 3 times reduction in code size.  Part of that
is experience (I am a better programmer now than I was 5 years ago, part of
that is the PIC instruction set (it is possible to do certain things in a
much more elegant manner), and part is the inherent reduction possible
because of the longer word size.

Dwayne