Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!zaphod.mps.ohio-state.edu!wuarchive!emory!att!cbnews!cbnewsm!lfd
From: lfd@cbnewsm.att.com (leland.f.derbenwick)
Newsgroups: comp.lang.c
Subject: Re: Source for "integer to ascii" and multiple file search-n-replace
Message-ID: <1990Nov27.224111.24038@cbnewsm.att.com>
Date: 27 Nov 90 22:41:11 GMT
References: <318@cti1.UUCP> <1990Nov20.014110.16982@ux1.cso.uiuc.edu> <1990Nov26.192218.29782@athena.mit.edu>
Organization: AT&T Bell Laboratories
Lines: 32

In article <1990Nov26.192218.29782@athena.mit.edu>, scs@adam.mit.edu (Steve Summit) writes:
> In article <324@cti1.UUCP> mpledger@cti1.UUCP (Mark Pledger) writes:
> >Maybe nobody is reading my question, but everyone keeps mailing me answers 
> >using sprintf() which is actually what I don't want.
...
> Sprintf simply is the recommended function, and it is "fast
> enough" in the vast majority of applications.  It is the rare
> program that can be significantly sped up with a special-purpose
> itoa-type routine, and you gave us no indication of why your
> program might be one of them.
...

I believe very strongly in coding systems to have good performance.
But here's a rule of thumb for optimizing code:

    If you can't say very clearly _why_ you need a faster version
    of a particular routine, then you probably don't need it.

In the cases I can think of (implementation on bare hardware, with no
sprintf available; or absolutely performance-critical and the overall
algorithm can't be improved instead), knowing _why_ you need it would
imply enough knowledge to be able to code it yourself, easily.  (Well,
maybe if it were a homework assignment and the professor didn't allow
you to use sprintf! :-)    Anyway, if it's truly performance-critical,
you probably need a function tuned to the application: fixed-length
strings, or particular distributions of input, or only unsigned, or if
all else fails maybe even assembly language.  So generic code probably
wouldn't do.

 -- Speaking strictly for myself,
 --   Lee Derbenwick, AT&T Bell Laboratories, Warren, NJ
 --   lfd@cbnewsm.ATT.COM  or  <wherever>!att!cbnewsm!lfd