Re: C code .vs. Assembly code for Microcontrollers/DSPs ?

Risto Jokinen <hatrjo@hat-fi.kone.com>
5 Mar 1996 12:26:01 -0500

          From comp.compilers

Related articles
[3 earlier articles]
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? eyal@dsp.co.il (Eyal Ben-Avraham) (1996-03-03)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? jens.hansson@mailbox.swipnet.se (1996-03-03)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? torbenm@diku.dk (1996-03-04)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? andrew@memex.co.uk (Andrew Cruickshank) (1996-03-04)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? fink@post.tau.ac.il (1996-03-05)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? torbenm@diku.dk (1996-03-05)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? hatrjo@hat-fi.kone.com (Risto Jokinen) (1996-03-05)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? max@gac.edu (1996-03-06)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? rhn@sgi.com (Ron Nicholson) (1996-03-06)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? yatesc@csee.usf.edu (1996-03-06)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? cdg@nullstone.com (1996-03-07)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? thinklab@earth.usa.net (1996-03-08)
Re: C code .vs. Assembly code for Microcontrollers/DSPs ? tchannon@black.demon.co.uk (Tim Channon) (1996-03-08)
[40 later articles]
| List of all articles for this month |
From: Risto Jokinen <hatrjo@hat-fi.kone.com>
Newsgroups: comp.compilers,comp.dsp
Date: 5 Mar 1996 12:26:01 -0500
Organization: Telecom Finland News Service
References: 96-03-006
Keywords: code, DSP

ravindra@hal.com (Ravindra Divekar) wrote:
>I've heard people say so many times, that hand-crafted (assembly) code
>is more compact than compiled C code for microcontrollers or DSPs.
>
>So where is the problem?


It is not a problem in the C-compiler of the microcontroller, it is
more problem if thinking. You can write most of code in C, but you
must be more careful with (at least) following items.


(microcontroller == 8/16 bit processor without fp suppor in this case)


- all algorihms must be adopted to use 8 or 16 bit arithmetic
- check support for signed arithmetic, if it is not supported,
    use unsigned arithmetic by default. (BYTE/WORD/BOOL datatypes)
- parametre passing, how many paremeters can be passed in registers?
    where return parameters are? what is correct parametre order?
- switch/case strures
    how case labels are handled, how to order them?
- allow more global data, and also allow goto's if they are really needed.
- bit-field datatypes can be nice, but can cause _very_ slow code.
- You must know how processor works, to check critical parts from code-listing
    files.




>Is it a limitation of existing compiler technology (so much for the
>optimization hype) ?


- I'll say no. compilers can make good quality code, but bad algorithms can
    cause problems. If you need to write code using microcontroller C, you
    should be familiar of assembler/plm/c/forth, if you can handle those languages
    ( at least read them ) you can also write high quality/fast C-code.


>Is it the choice of high level language i.e. C/C++ that matters ?
>
>Is it the nature of application (scientific/real-time/control/float.pt)
>that creates porting problems?


There is one rule, 32bit dsp with fp support is easy for C. You can
verify most of code in pc, linux etc. 8/16 bits variants need more
careful design.


>Or is it just the architectural features of typical microcontrollers that
>make things difficult ?


Yes, they are somehow too small for normal C-implementation, plm was
very good, forth also, and I have heard that some modula-2 compilers
are also good. the problem is that most of microcontrollers can not
handle C-style pointer algorithms that easy. (too few pointer
registers, and also problems with stack)


-- Risto
--


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.