> Seriously, I of course agree that the smaller PICs are not really well
> suited for a C compiler -- but IMO they are not well suited for assembler
> programming either. 68k assembler OTOH is suited for assembler programming
> -- but you'd most likely find that it's quite well suited for C also :)  So
> maybe there's not that much of a difference.

I wouldn't agree that the smaller cores are that ugly.  For the intended
application space, they really aren't that bad.

But the combination of C and the PIC isn't a little bad ... it is
horrible.  I've seen these compilers emit code that is 10 to 20 times
larger than it needs to be for perfectly harmless looking constructs.  The
experienced developer can EASILY work around these, and for him, a
compiler can be a big productivity benefit.

But for someone who doesn't have a good understanding of the PIC, the
results can be quite bad.

Interestingly, the compilers I've looked at do a quite good job of
handling banking and paging.  In that specific case, I would expect the
compiler would do a better job than the assembly language programmer is
likely to.

I agree that when development cost is an issue, an experienced programmer
probably should use a compiler because of the productivity benefits.  But
even there one needs to be careful.  Unlike a "normal" processor where a
compiler choice is likely to lead to a slightly larger/slower result, on
the PIC, it is pretty easy to lead to a MUCH larger/slower result.  One
needs to understand this and balance the development costs against the 2-n
costs.

For the beginner, though, the development time will likely NOT be lower
with a compiler, and the frustration level will be a lot higher.

Again, understand that I am referring here to modern, nested scope
languages.  I suspect (but haven't actually seen the code) that languages
like Fortran or Basic might be a little more easily implemented on the PIC
and might not be quite so pathological as languages like C. Although I
have to admit I've seen some pretty hideous stuff emitted by C compilers
that didn't seem to be related to the nested scope nature of the language.
 And I am pretty confident (although with no data to support it) that a
purpose-built, domain-specific language could do a very good job.  Of
course, such a thing is hard to market, so we are stuck with miserable
implementations of "standard" languages for the most part.

--McD



-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist