Just addressing the two value case, ignoring the less interesting techniques, but considering inline and sub versions of the ternary technique I get the following:

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

sub ternary {
    $_[0] > $_[1] ? $_[0] : $_[1];
}

sub sauoq {
    ($_[0], $_[1])[$_[0] < $_[1]];
}

sub List {
    max (@_);
}

my @r = map {int rand 1000}(0..1);

cmpthese -1,{
    'ternary' => sub {$r[0] > $r[1] ? $r[0] : $r[1]},
    'ternarySub' => sub {ternary(@r)},
    'sauoq' => sub {($r[0], $r[1])[$r[0] < $r[1]]},
    'sauoqSub' => sub {sauoq(@r)},
    'List' => sub {List(@r)},
};
[download]

Prints:

                Rate       List   sauoqSub ternarySub      sauoq    te
+rnary
List        998323/s         --       -10%       -36%       -75%      
+ -86%
sauoqSub   1106591/s        11%         --       -29%       -73%      
+ -85%
ternarySub 1550115/s        55%        40%         --       -62%      
+ -79%
sauoq      4044842/s       305%       266%       161%         --      
+ -44%
ternary    7270647/s       628%       557%       369%        80%      
+   --
[download]

Update: added sauoq's technique