NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c-part2/14401

Internet Message Format | 1996-08-05 | 2.8 KB

Path: solon.com!not-for-mail
From: Lawrence Kirby <[e-mail]>
Newsgroups: comp.lang.c,comp.lang.c.moderated,hp.unix,comp.sys.hp.apps,comp.sys.hp.hpux
Subject: Re: C coding problem
Date: 14 Apr 1996 10:53:37 -0500
Organization: none
Sender: [e-mail]
Approved: [e-mail]
Message-ID: <4kr721$[e-mail]>
References: <4ianbf$[e-mail]> <4iemcl$[e-mail]> <4io1io$[e-mail]> <4j06na$[e-mail]> <4jttan$[e-mail]> <4kg8gl$[e-mail]>
Reply-To: [e-mail]
NNTP-Posting-Host: solutions.solon.com
X-NNTP-Posting-Host: genesis.demon.co.uk
X-Newsreader: Demon Internet Simple News v1.27
X-Mail2News-Path: genesis.demon.co.uk

In article <4kg8gl$[e-mail]> [e-mail] "Mike Shannon" writes:

>>Secondly, why not use arrays? In many places they are easier to
>>understand - I for one am not going to compromise the readability
>>of my programs because of some theoretical bull about them
>>being second rate objects. (whatever that is intended to mean)
>>a[i] is converted to *(a + i) by the compiler in any case so
>>whats the advantage with your approach?
>>
>>Niall
>
>Sure, a[i] is the same as *(a+i). Both involve keeping up with
>two variables, and, when i incremented, require a multiplication
>of i times sizeof(*a) to add to a to find the address of the next
>row.

Conceptually that's true. However it is rather unlikely that a current
compiler would generate code that looks like that unless it is efficient.
Strength reduction i.e. converting the loop to one based on pointers is
a very standard optimisation these days. Where there is a difference I
would expect the array based version to generate the more efficient code
more often than not since you are giving the optimiser more information
to work with. The key here is that a (the pointer) be invariant in the
loop.

>But the example above is doing neither. p++ involves only one
>variable, and only addition of sizeof(*p) to find the address of
>the next row. Notice especially that there is no multiplication
>required. Agreed the per event difference is small, but if a
>program does a large number of array sweeps, the small
>advantage adds up. Some optimizers may do this for you.

A compiler that doesn't is way behind the current state of play.

>I agree with simple syntax, though, and expect that something
>like:
>
>    while (p < end_p)
>    {
>         ++(*p);

What's wrong with ++*p?

>         p++;
>    }
>
>would produce the same assembly output as the above example,
>and pose less interpretation risk in the future.

--
-----------------------------------------
Lawrence Kirby | [e-mail]
Wilts, England | [e-mail]
-----------------------------------------