Re: How expensive is eval { }?

A simple benchmark indicates that the overhead isn't very much (unless the benchmark is completely flawed of course). Consider:

use strict;
use warnings;
use Benchmark qw(cmpthese);


cmpthese (
    -1, {
        BareFor => \&dofor,
        ExceptFor => \&doefor,
        ForExcept => \&dofore,
        }
    );

sub dofor {
    my $sum = 0;

    for (1 .. 1000) {
        $sum += $_;
        $sum += $_;
        $sum += $_;
    }
}


sub dofore {
    my $sum = 0;

    for (1 .. 1000) {
        eval {$sum += $_;}
    }
}


sub doefor {
    my $sum = 0;

    eval {
        for (1 .. 1000) {
            $sum += $_;
            $sum += $_;
            $sum += $_;
        }
    }
}
[download]

Prints:

            Rate   BareFor ExceptFor ForExcept
BareFor   1626/s        --       -2%      -11%
ExceptFor 1661/s        2%        --       -9%
ForExcept 1834/s       13%       10%        --
[download]

which indicates that the eval overhead is close to the cost of 2 += operations - not a whole lot in other words!

Perl is environmentally friendly - it saves trees

Comment on Re: How expensive is eval { }? Select or Download Code

Replies are listed 'Best First'.
Re^2: How expensive is eval { }? by moritz (Cardinal) on Apr 17, 2008 at 09:33 UTC
This benchmark result seems to vary strongly with the platform and/or perl version. Debian GNU/Linux i386 (32 bit, SMP): `$ perl5.8.8 foo.pl Rate ForExcept BareFor ExceptFor ForExcept 2511/s -- -34% -34% BareFor 3794/s 51% -- -1% ExceptFor 3829/s 52% 1% -- $ perl5.10.0 foo.pl Rate ForExcept BareFor ExceptFor ForExcept 2393/s -- -29% -30% BareFor 3350/s 40% -- -2% ExceptFor 3429/s 43% 2% --` [download] That's quite different from the 2% .. 13%. Any idea why the differences between the perl versions and platforms are that big?	[reply] [d/l]
Re^3: How expensive is eval { }? by GrandFather (Saint) on Apr 17, 2008 at 10:05 UTC
No real idea, but bear in mind that most likely we are looking at only a few instructions' execution time difference between each case so they could be highly sensitive to optimization or modest differences in instruction timing between different processors (I ran my tests on a Turion⁶⁴ powered laptop btw). Beyond indicating that the eval overhead is small enough to be discounted in most practical contexts, the benchmark really doesn't tell us very much - except that you can't trust benchmarks very much for this sort of fine grain comparison. The important thing to note is that the eval costs much less than 1 micro-second. Perl is environmentally friendly - it saves trees	[reply]