Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!necntc!ima!johnl
From: johnl@ima.ISC.COM (John R. Levine)
Newsgroups: comp.lang.c
Subject: Re: another auto-increment problem
Message-ID: <723@ima.ISC.COM>
Date: Sat, 26-Sep-87 17:12:54 EDT
Article-I.D.: ima.723
Posted: Sat Sep 26 17:12:54 1987
Date-Received: Sun, 27-Sep-87 11:21:38 EDT
References: <3680001@wdl1.UUCP>
Reply-To: johnl@ima.UUCP (John R. Levine)
Organization: Not enough to make any difference
Lines: 26
Summary: famous point of confusion

In article <3680001@wdl1.UUCP> rion@wdl1.UUCP (Rion Cassidy) writes:
> ...
>	float junk[3], *junk_ptr;
> ...
>	junk_ptr = (float *) junk;
>	xfpt(*junk_ptr++, *junk_ptr++, *junk_ptr);
> [isn't the same as]
>	xfpt(junk[0], junk[1], junk[2]);

Two points.  First, as has been pointed out elsewhere, no version of C has
ever made any promises about the order of evaluation of arguments nor of
side effects.  It would be entirely legitimate for the compiler in the
first case to push junk[0] three times and then increment the pointer
twice, and on many machines that'd be the best code.

Second, on computers that don't have auto-increment address modes, the
second version is probably faster and smaller than the first, not to
mention making your intentions clearer.  In any event, one or two extra
address calculation instructions are unlikely to make any perceptible
difference if the called routine is going to do a bunch of floating point
multiplication.  Make it right, then make it fast.
-- 
John R. Levine, IECC, PO Box 349, Cambridge MA 02238-0349, +1 617 492 3869
{ ihnp4 | decvax | cbosgd | harvard | yale }!ima!johnl, Levine@YALE.something
The Iran-Contra affair:  None of this would have happened if Ronald Reagan
were still alive.