|
|
Posted by Tim Streater on 03/27/07 10:01
In article <eu9m2t$af5$1@hercules.cohp1>,
Ivαn Sαnchez Ortega
<ivansanchez-alg@rroba-escomposlinux.-.punto.-.org> wrote:
> mgcclx@gmail.com wrote:
>
> > For loop and while loop. which one is faster?
> > I see many articles fighting over it and different people come up with
> > different results.
>
> Unless you're designing a compiler, don't worry about that. Worry about the
> algorithms instead.
>
> That said, the only possible difference between a for loop and a while loop
> is the jump prediction in your architecture's code instruction set. e.g.
> the processor can suppose that the condition for a while loop is going to
> be true, so it feeds the instructions after the conditional jump (that is,
> it's supposing that the jump condition will be successful) and starts
> running them in the first stages of the instruction pipeline, saving a
> couple of pipeline stages.
>
> Do you have to worry about processor pipelines and jump prediction? Not at
> all, unless you're into assembler, or heavy low-level programming. If you
> haven't understood a word of the last paragraph, don't worry about loop
> speed.
>
> In other words: trust your compiler (or JIT compiler, or interpreter). It
> knows about optimizations of conditional jumps better than you (the same
> applies to memory caching). Focus on the algorithmics if you want to
> shorten processor time.
This is quite right. I never worry about such things - that's the
computer's job.
I use whichever is appropriate - try, in other words, to give a hint to
yourself or whoever is going to look at that code in a year's time. If I
have n items and I want to do something for each of them, I use a for
loop. If I need to loop around until some condition is satisfied (maybe
one of several conditions in different parts of the loop), then I use a
while.
-- tim
Navigation:
[Reply to this message]
|