Path: utzoo!attcan!uunet!lll-winken!lll-tis!helios.ee.lbl.gov!pasteur!agate!labrea!rutgers!bellcore!u1100a!krohn
From: krohn@u1100a.UUCP (Eric Krohn)
Newsgroups: comp.lang.c
Subject: Re: Efficient coding considered harmful?
Message-ID: <2131@u1100a.UUCP>
Date: 8 Nov 88 20:06:47 GMT
References: <3105@hubcap.UUCP> <34112@XAIT.XEROX.COM> <1700@dataio.Data-IO.COM> <8630@smoke.ARPA> <1704@dataio.Data-IO.COM> <119@twwells.uucp> <7700@bloom-beacon.MIT.EDU> <137@twwells.uucp> <8819@smoke.BRL.MIL> <143@twwells.uucp>
Reply-To: krohn@u1100a.UUCP (Eric Krohn)
Organization: Bell Communications Research, Piscataway, NJ
Lines: 28

In article <143@twwells.uucp> bill@twwells.UUCP (T. William Wells) writes:
] 
] Formulated as a general principle: if an operand in a multiply or the
] divisor in a divide or mod (with the other operand being known to be a
] positive number) is one which in the problem being solved is
] naturally a power of two, it is reasonable to use bitwise operators to
] implement the operation, provided that a #define is used to define
] the number and (when needed) its log base 2.

Instead of #defining the log in terms of the number,
I prefer to #define the number in terms of its log base 2, as in

#define	LOGSIZE	7
#define	SIZE	(1 << (LOGSIZE))

to insure that SIZE really is a power of 2 when I expect it to be.
For the safety conscious, you can add

#if	LOGSIZE <= 0
#include	"LOGSIZE must be non-negative"
#endif
#if	LOGSIZE >= BITSPERWORD
#include	"LOGSIZE is too large"
#endif
-- 
--
Eric J. Krohn
krohn@ctt.ctt.bellcore.com  or  {bcr,bellcore}!u1100a!krohn
Bell Communications Research,	444 Hoes Ln,    Piscataway, NJ 08854