Re: programmer optimizations?

"Afsal C. Majeed" <afsal@genius.tisl.soft.net> writes:
> We are having small debate here as to the use of some of the
> code optimizations by the programmer himself.
>
> e.g.
> replacing (n / 4)
> by (n >> 2)
>
> replacing if ((i < 0) || (j < 0) || (k < 0))
> by if ((i | j | k) < 0)
>
> as you can well imagine, the list is endless
>
> Since this produces somewhat non portable code, the view is that
> such optimizations should be discouraged.




Moderator writes:
> [One reason that such optimizations are discouraged is that they're often
> wrong. Turning n/4 into n>>2 is only correct if n is unsigned.


Not only are such optimizations often wrong, they often are not optimal!
For instance, I remember a machine I worked on not too long ago in which
shift was exceptionally slow, whereas integer divide was pretty fair.
Turns out that dividing by small powers of 2 (e.g., 4) was better done by
shifting, but larger powers of 2 were not.


The best advice is to express the operation you actually want performed,
and let the compiler do the optimizing. Unfortunately, some programmers
write "n / 4" when what they actually meant was "shift n right by 2", and
vice versa. This can often result in sub-optimal code because the compiler
must maintain the semantics of the language, regardless of what you
*meant*, and because the compiler has less knowledge about the data than
the programmer (sometimes).


--
Bill Leonard
Harris Computer Systems Corporation
2101 W. Cypress Creek Road
Fort Lauderdale, FL 33309
Bill.Leonard@mail.hcsc.com
--

Return to the comp.compilers page.
Search the comp.compilers archives again.