NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/sys/sgi/misc/18

Text File | 1992-12-16 | 1.7 KB | 46 lines

Newsgroups: comp.sys.sgi.misc
Path: sparky!uunet!zaphod.mps.ohio-state.edu!uwm.edu!ux1.cso.uiuc.edu!news.cso.uiuc.edu!dweber
From: [e-mail] (Daniel Weber)
Subject: C Compiler & Array subscripting
Message-ID: <[e-mail]>
Originator: [e-mail]
Sender: [e-mail] (Net Noise owner)
Reply-To: [e-mail]
Organization: National Center for Supercomputing Applications
Date: Thu, 17 Dec 1992 03:24:10 GMT
Lines: 33


This is a question that can be directed towards any general C compiler,
but for the moment, it has to deal with SGI's C compiler:

I am working with some code, written by someone else, that contains a *lot*
of math.  In this code, the authors have statements like:

*(R+4) = *(R+3) + h24*(55*(*(RdotS+3)) - 59*(*(RdotS+2)) + 37*(*(RdotS+1)) - 9*(*(RdotS)));

I'm attempting to clean this all up so it is legible (yes, that's a cut &
paste, no indentations here :-( ), and would like to know if using the
following format is less efficient:

   R[4] = R[3] + h24*(55*RdotS[3] - 59*RdotS[2] + 37*RdotS[1] - 9*RdotS[0]);

So, is subscripting like this less efficient?  My guess would be that a
smart compiler would detect that these two statements are identical, and
produce identical object code.  However, is this always true?  Say the
above code were in a loop:

   for (j = 0; j < bignumber; j++) {
      R[j] = ...
         -or-
      *(R+j) = ...

Would a good compiler still interpret the code the same?  Specifically,
would the SGI compiler produce efficient code (with -Ox turned on)?

Thanks for any help.  If you have an answer, post or email, either
way I'm happy!  :-)

-daniel weber
[e-mail]