(btw, don't you mean sub blah instead of my blah)

Benchmark it!

use Benchmark qw( cmpthese );

$i=0;

sub blah1 {
   my @list = qw(# list of things);
   foreach my $item (@list) { $i++; }
}

{
   my @list;
   BEGIN {
      @list = qw(# list of things);
   }
   sub blah2 {
      foreach my $item (@list) { $i++; }
   }
}

sub blah3 {
   foreach my $item qw(# list of things) { $i++; }
}

cmpthese(0, {
   blah1 => \&blah1,
   blah2 => \&blah2,
   blah3 => \&blah3,
});

__END__
(partial) output
================
          Rate blah1 blah2 blah3
blah1 118024/s    --  -52%  -53%
blah2 245479/s  108%    --   -1%
blah3 248532/s  111%    1%    --
[download]

2 and 3 are equally fast, so you might as well use the cleaner 3.