Re: Microcontroller ... which one ??

From: keith (krw_at_att.bizzzz)
Date: 11/29/04 

Date: Sun, 28 Nov 2004 22:17:36 -0500

On Sun, 28 Nov 2004 20:25:52 +0000, Ken Smith wrote:

> In article <pan.2004.11.28.19.29.22.170993@att.bizzzz>,
> keith <krw@att.bizzzz> wrote:
>>
>>Since it's being cross-posted to a new group, I'll trim closely...
>>
>>On Sun, 28 Nov 2004 17:16:07 +0000, Ken Smith wrote:
>>
>>> The discussion is about the portability of C. We are not talking about
>>> just general purpose computers. DSPs and micro-controllers are at issue.
>>
>>Also ignoring Harvard architectures (like the 8051) and special function
>>registers (which are obviously hardware dependant).
>
> Ok, I'll include that limitation to the scope of the argument. Things
> like the transputer are alos off the table. I think they are too weird to
> make a C compiler for. We can limit the discussion to only those machines
> for which a reasonable C compiler can be written.

OTOH, I know people who use C compliers for 8051s. I couldn't figure out
how to make it all work. (again, I'm a hardware type and think in RTL and
assembler)
>
>
>>> In article <pan.2004.11.28.05.03.10.61137@att.bizzzz>, keith
>>> <krw@att.bizzzz> wrote:
>>>>On Sat, 27 Nov 2004 22:07:04 +0000, Ken Smith wrote:
>>>>
>>>>> In article <pan.2004.11.27.19.03.02.32327@att.bizzzz>, keith
>>>>> <krw@att.bizzzz> wrote:
>>>>> [...]
>>>>>>I think if you make that statement over on comp.arch you'll get a
>>>>>>lot of argument. AIUI (I'm just a hardware type, can't spell 'C')
>>>>>>it is possible to write portable 'C', but it's not trivial.
>>>>>
>>>>> For simple programs it may be possible to do but you'd better watch
>>>>> it if you use strings. In some cases it is best to just store one
>>>>> character per integer.
>>>>
>>>>Are you man enough to take it to C.A? Trust me, you'll get chewed up
>>>>and spit out by people who do just this for a living.
>>>
>>> Consider it taken there if I did the headers right.
>>
>>No, not C.A.C. It might work, but I meant C.A. I've seen exactly this
>>argument before in C.A. I've never followed C.A., since I'm a hardware
>>type. A Google NG search might keep from upsetting the natives too. ;-)
>
> If we don't get a reaction, one us will have to change the groups line.
> I did not because it is way too early to expect any reaction.

I added C.A. Though it may be a stale subject...
>
>
>
>>>>> All more complex types such as structures and unions are composed of
>>>>> the simple types. They also can't be passed to subroutines.
>>>>
>>>>Of course not, pointers to structures are. But the basic types are
>>>>defined in the language specification too. Because *you* choose to
>>>>bastardize them, you shoot yourself in the foot. 'C' is really good
>>>>at letting the incompetent do that (which is why I don't do 'C').
>>>
>>> "the basic" types means the same as the "simple types". Or are you
>>> suggesting that some arrays or structures are now part of the language
>>> definition?
>>
>>I'm only reporting what the language experts say. They claim to do
>>portable C reliably, every time.
>
>
> But ... but ... are they porting to microcontrollers etc. In the
> general purpose computer world things can be quite different. Companies
> like Prime and DEC no longer exist so there aren't many non-byte
> oriented general purpose computers any more.

They've been doing it for decades. I've heard of very little x86 (32b)
Linux code that isn't an "automatic" recompile to 64b. Of course 64b
isn't new in the *ix world.
 
>>> I maintain that "complex types" end up being machine dependant because
>>> they are defined from the "simple types". I have also said that "long
>>> int", "short int", "int", "float" and all pointers are machine
>>> dependant because thir meaning changes for non-byte oriented
>>> processors. Keith (the other poster) asserted that "short int" means
>>> 16 bits and that "long int" means 32 bits.
>>
>>I never explicitly stated anything of the kind. I states simply that
>>portable C coding was possible without out polluting the code with
>>massive IFDEFs and such.
>
>
> I thought you did. Maybe I confused you with some other poster. Sorry
> if I did.

's ok. My statement is based on experts' opinions that porting isn't
rocket surgery, but rather attention to detail. Sloppy is bad news and
'C' allows one to be as crude and rude as one is.
>
>>>>Tell it to the real programmers over on comp.arch, if you dare. I'll
>>>>lurk and watch you burn too. I've seen these flame-fests before. The
>>>>incompetents always run away with flaming tails.
>>>
>>> I assume you mean "comp.arch.c"
>>
>>No, I meant "comp.arch", where much of this stuff has been hashed out
>>ad-nausium. I've never followed comp.arch.c, so I know none of the
>>regulars.
>
> Well we will see if this get more hash or just ignored. Fasten your
> seat belt, it may be a bumpy ride.

Ok... ;-) 

-- 
  Keith