Running on an ancient PII 266 this takes less than a minute to process the standard unix dictionary. It uses a fairly efficient algorithm. Basically we bulid a hash which has the sorted letters of the words as its keys and the corresponding words (as an array ref) as its values. Thus we have a hash like:

...
'aet' => [ ate eat ]
...
[download]

As we have represented all the words as their sorted letters in a hash we make the search quick and easy. We work from the longest keys to the shortest and use recursion to test all the possibilities. We keep a list of 'winners' - any word which is in a winning list need not be tested again when we descend to it's level.

#!usr/bin/perl -w
use strict;

# get words into an array and a hash keyed on sorted letters
open DICT, "c:/windows/desktop/unixdict.txt" or die "Oops $!\n";
my (%words,%winners);
while(<DICT>) {
    chomp;
    my $sort = join '', (sort split'',$_);
    push @{$words{$sort}}, $_; 
}

# test the words by length - longest first, don't repeat known winners
for my $word (sort {length $b <=> length $a} keys %words) {
    defined $winners{$word} ? next : test($word);
} 

sub test {
    my $word = $_[-1];
    if (length $word <=3) { winner(@_); return; }   
    my @letters = sort split '', $word;
    for my $i (0..$#letters) {
        my @temp = @letters;
        $temp[$i] = '';
        my $short = join '', @temp;
        test(@_,$short) if exists $words{$short};
    }
}

sub winner {
    local $, = "  ";
    for (@_) {
        $winners{$_}++;
        print @{$words{$_}}, "\n";
    }
    print "\n";
}

__DATA__
planetarium
manipulate
aluminate
laminate
matinal
animal  manila
milan
main
min

[blah]
[download]

cheers

tachyon

s&&rsenoyhcatreve&&&s&n.+t&"$'$`$\"$\&"&ee&&y&srve&&d&&print