My preference is to do the space trimming in two stages as it seems to be faster than either the capture or alternation methods.

use strict;
use warnings;

use Benchmark q{cmpthese};

my @arr = ( q{   fdsgehw fw wwfe w    } ) x 5000;

cmpthese(
   -5,
   {
       alternation  => sub
          {
              my @new = @arr;
              s{ ^\s* | \s*$ }{}gx for @new;
          },
       capture      => sub
          {
              my @new = @arr;
              s{ ^\s* (\S.*?) \s*$ }{$1}x for @new;
          },
       twoStage     => sub
          {
              my @new = @arr;
              s{ ^\s* }{}x for @new;
              s{ \s*$ }{}x for @new;
          },
   },
   );
[download]

The results.

              Rate     capture alternation    twoStage
capture     8.96/s          --        -26%        -50%
alternation 12.2/s         36%          --        -33%
twoStage    18.1/s        102%         48%          --
[download]

I hope this is of interest.

Cheers,

JohnGG

Update: Fixed code indentation problems caused by TABs