benchmarks are funny things. have i made a mistake here, or is grep faster than first in this case? the real win is factoring out the call to param():

use strict;
use Benchmark qw(cmpthese);
use List::Util;

## a 'reasonable' worst-case scenario?  first of list of twenty .. 
my @display_files = qw(
    one two three four five six seven eight nine ten eleven twelve thi
+rteen 
    fourteen fifteen sixteen seventeen eighteen nineteen twenty
);
sub param { 'one' }

cmpthese( -2, {
    grep_param  => sub { 
        unless( grep { $_ eq param('select') } @display_files ) { 1; }
+ 
    },
    param_grep  => sub { 
        my $select = param('select');
        unless( grep { $_ eq $select } @display_files ) { 1; } 
    },
    found => sub {
        my $select = param('select');
        my $found;
        foreach (@display_files) {
            if ($_ eq $select) {
                $found = 1;
                last;
            }
        }
        unless ($found) { 1; }
    },
    first => sub {
        my $select = param('select');
        my $found = List::Util::first { $_ eq $select } @display_files
+;
        unless ($found) { 1; }
    }
});
[download]

produces:

               Rate grep_param      first param_grep      found
grep_param  47285/s         --       -69%       -81%       -92%
first      154996/s       228%         --       -36%       -73%
param_grep 244086/s       416%        57%         --       -57%
found      565517/s      1096%       265%       132%         --

interestingly, change sub param to return undef (to get a failing test) and i get:

               Rate grep_param      first      found param_grep
grep_param  81919/s         --       -18%       -61%       -68%
first       99685/s        22%         --       -52%       -61%
found      207714/s       154%       108%         --       -18%
param_grep 254449/s       211%       155%        22%         --

personally, i think grep is the clearest and most familiar idiom, and for a shortish list of @display_files the performance gains of using the for () { ... last; } aren't worth it, especially if the failure case is common.