Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site mit-eddie.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!smh
From: smh@mit-eddie.UUCP (Steven M. Haflich)
Newsgroups: net.lang.c
Subject: Re: reversing a mask
Message-ID: <3178@mit-eddie.UUCP>
Date: Sun, 25-Nov-84 08:42:43 EST
Article-I.D.: mit-eddi.3178
Posted: Sun Nov 25 08:42:43 1984
Date-Received: Mon, 26-Nov-84 02:04:37 EST
References: <1@imd.UUCP> <469@ncoast.UUCP> <691@gloria.UUCP> <4082@elsie.UUCP>
Reply-To: smh@mit-eddie.UUCP (Steven M. Haflich)
Organization: MIT, Cambridge, MA
Lines: 14

Much more space-efficient and only slightly slower is the following:

  char revbytes[256] = { 0x00, 0x80, 0x40, 0xc0, ... , 0x7f, 0xff };
  #define reverse(v) ((revbytes((v)&0xff)<<8)|revbytes(((v)>>8)&0xff))

That's the general idea.  This is not perfectly portable, but then,
neither was the original problem.  Depending on particular C
implementation, various dodges may be necessary to prevent the infamous
char sign extension problem: either by declaring revbytes to be
"unsigned char", or else by anding each mention of revbytes with 0xff.
Also, note that the argument gets evaluated twice, but this could be
corrected if necessary.

Steve Haflich