swampyankee is correct. Here is some code that shows the principle:

use strict;
use warnings;

use Data::Dumper;
use Digest::MD5 qw(md5);

my %lookup;

my @AoA =
(
  ["a", "b", "c"],
  ["d", "e", "f"],
  ["a", "b", "c"],
  ["g", "h", "i"],
  ["a", "b", "c"],
  ["g", "h", "i"],
);


#Build a lookup hash from AoA, values of the hash will be
#an arrayref to the positions where a given array was found
my $index = 0;

for ( @AoA )
{
    my $k = md5(Dumper($_));
    push @{$lookup{$k}}, $index++;
}

$index = 0;

#test data
my @test =
(
  ["a", "b", "c"],
  ["d", "e", "f"],
  ["g", "h", "i"],
  ["x", "y", "z"],
);

#loop tests
for ( @test )
{
    print "Test ", $index++, ": ", join (", ", @{ref_exists($_)}), "\n
+";
}


#returns the positions or undef when the arrayref is not in AoA
sub ref_exists
{
    return $lookup{md5(Dumper(shift))} || [];
}
[download]

Note that, because I use Data::Dumper for creating the digest, the technique should work for deep structures also, not just for plain arrays of strings.