NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/sys/amiga/programmer/650

Internet Message Format | 1996-08-05 | 3.1 KB

Path: cs.uwa.edu.au!jasonb
From: [e-mail] (Jason S Birch)
Newsgroups: comp.sys.amiga.programmer
Subject: Re: PPC compilers
Date: 10 Jan 96 04:24:30 GMT
Organization: The University of Western Australia
Message-ID: <[e-mail]>
References: <[e-mail]> <MQAQx*[e-mail]> <OWhVx*[e-mail]> <4cuhng$[e-mail]>
NNTP-Posting-Host: decadence.cs.uwa.oz.au
X-Newsreader: NN version 6.5.0 #3 (NOV)

[e-mail] (Stephan Schaem) writes:
>Arno Eigenwillig ([e-mail]) wrote:
>: Why don't you read what he wrote? A reasonable example for that would
>: be operations on clock_t.

> The claim was general: you can forget your variable type when you use
> them in C... I'm just saying thats its not wise.

What we've been (repeatedly) trying to say, is that's *not* what he
said. He said you can forget about the *implementation* of your
variable type on a particular machine (if the data is encapsulated
correctly). The whole point being that in assembler, you cannot forget
the implementation, because pratically every time you do *use* a
"variable" of that type, you have to specify how big it is.

>: Why don't you read what he wrote? You use clock_t and treat it as a
>: 'black box'. As you never name the underlying type, you are fully
>: portable.

> What if clock_t is on other system a pointer to a bcpl string?

Unfortunately that will not work, because clock_t isn't sufficiently
encapsulated - ie. to find the difference between two clock_t values,
you must use "-", and "-" acts as expected only on ordinal types. With
C++, however, "-" can be overloaded to work with *any* definition of
clock_t, including BCPL strings.

For a better example, look at time_t. It comes with a full suite of
support functions that effectively hide its implementation from you.
So it could well be implemented as a BCPL string right now without
causing any problems (at least for programmers who don't see it's eg.
a long and try to use "-" to find differences in time instead of
difftime()).

>: Name an existing assembler that allows such things.

> You can write yourself a simple preprocessor that will do a search replace
> using a mytype.typedef file... The point I wanted to make is this:
> [Store    variable of type A, into variable of type A]
> This is what I think when I write
> chipreg->bltcon0 = con0;
> [Store con0 of type CHIPREG, into chipreg.bltcon0 of type CHIPREG]
> &
> move.w    (GLOBAL_con0_w,a4),(CHIPREG_bltcon0_w,a1)
> [Store con0 of type WORD, into chipreg.bltcon0 of type WORD]
> here I just happen to know WORD is 2 byte.. but so what.

Again, the "so what" is that in assembler, you're using the knowledge
that CHIPREG is a WORD - as soon as you do that, you've tied yourself
to only one implementation of CHIPREG.

> Stephan

--
Jason S Birch                        ,-_|\ email: [e-mail]
Department of Computer Science      /     \ Tel (work): [phone]
The University of Western Australia *_.-._/ Fax (work): [phone]
Nedlands  W. Australia  6907             v  Tel (home): [phone]