Beefy Boxes and Bandwidth Generously Provided by pair Networks
Perl-Sensitive Sunglasses

Which code executes faster

by alandev (Scribe)
on Jul 31, 2006 at 04:51 UTC ( [id://564677] : perlquestion . print w/replies, xml ) Need Help??

alandev has asked for the wisdom of the Perl Monks concerning the following question:

hi monks

i have a simmple doubt .i wanna know which of the code executes faster

return 1 if $n < 1;


if($n < 1) { return 1; }


Edited by planetscape - added code tags

( keep:0 edit:8 reap:0 )

Replies are listed 'Best First'.
Re: Which code executes faster
by davido (Cardinal) on Jul 31, 2006 at 06:16 UTC

    The difference is so insignificant that it is actually far more important whether the condition being evaluated has an affirmitive outcome (meaning the return 1 gets executed) or not.

    Consider the following code:

    use strict; use warnings; use Benchmark qw/cmpthese/; cmpthese( -1, { Std_pos => sub { my $n = 0; if( $n < 1 ) { return 1 } }, # Std_neg => sub { my $n = 1; if( $n < 1 ) { return 1 } }, Rflx_pos => sub { my $n = 0; return 1 if $n < 1; }, # Rflx_neg => sub { my $n = 1; return 1 if $n < 1; }, And_pos => sub { my $n = 0; $n < 1 and return 1; }, # And_neg => sub { my $n = 1; $n < 1 and return 1; }, } );

    ...and the results...

    Rate And_pos Std_pos Rflx_pos And_pos 4338935/s -- -4% -5% Std_pos 4504851/s 4% -- -2% Rflx_pos 4574559/s 5% 2% --

    Note, in the preceeding code, we're only comparing the cases where the conditional evaluates to true, so the 'return 1' is executed. Notice how the fastest method (reflexive, I call it) is only 5% faster than "and", and only 1% faster than "Standard".

    When you compare only the code where all conditionals evaluate to false (ie, return 1 is not executed), you get the following results, which are slightly more interesting, but not enough so to get worked up over:

    Rate And_neg Rflx_neg Std_neg And_neg 5286632/s -- -2% -14% Rflx_neg 5401907/s 2% -- -12% Std_neg 6149593/s 16% 14% --

    But now uncomment the negs and the pos's, and see how the comparison turns out when you compare the code that evaluates to true versus the code that evaluates to false:

    Rate Std_pos Rflx_pos And_pos Rflx_neg And_neg Std_n +eg Std_pos 4254827/s -- -3% -3% -15% -21% -2 +2% Rflx_pos 4368927/s 3% -- -0% -13% -19% -2 +0% And_pos 4376827/s 3% 0% -- -13% -19% -2 +0% Rflx_neg 5023657/s 18% 15% 15% -- -7% - +8% And_neg 5386758/s 27% 23% 23% 7% -- - +2% Std_neg 5486976/s 29% 26% 25% 9% 2% +--

    Notice how all the negatives are grouped together, as the fastest (because they don't execute the 'return 1'). ...and they're 18% to 29% faster than their corresponding 'positives'. In other words, the return 1 is much more significant than the syntax chosen to control the flow.

    Perhaps the most important thing to notice is that, when comparing all together, the best case executes almost five and a half million times a second, and the worst case over four and a quarter million times a second. To me that says I'm probably looking in the wrong place to optimize. You should really just profile and see if this is where your issue is, and also examine your algorithms to see if they're well thought-out.

    Focus on profiling, and the algorithms at work, and forget about such insignificant syntax differences. If you have to care whether it's faster to use if( ... ) {...} or  ... if ..., you may have to polish up your assembler skills anyway.


Re: Which code executes faster
by Zaxo (Archbishop) on Jul 31, 2006 at 05:01 UTC

    The statement modifier form,

    return 1 if $n < 1;
    should be a bit faster, since no lexical scope is created. Faster yet could be,
    return $n < 1;
    if that's what you want to happen when the test fails.

    In most non-trivial code it's not going to make any noticeable difference.

    After Compline,

      Faster yet could be.
      hmm, like Yoda do you speak. ;)

      seriously though alandev, there is unlikely to be a real world increase in speed. You should go with whatever's going to be easiest for you (and the programmer who comes after you) to read.

      my name's not Keith, and I'm not reasonable.
      thanks zaxo
Re: Which code executes faster
by Aristotle (Chancellor) on Jul 31, 2006 at 14:47 UTC

    Why? Is that the place where your program is spending too much time?

    If you really need to find out, I suggest you familiarise yourself with the Benchmark module. But make sure you don’t miss Zed Shaw’s article Programmers Need To Learn Statistics Or I Will Kill Them All.

    You’ll also want to know what Perl actually makes of your code; B::Deparse will help you with that.

    Makeshifts last the longest.

      thanks monks