Assuming $sql and $csv are both pointers to the hashes:

my %scsv;
foreach (keys %$csv) {
    $scsv{substr($_, 4)} = %$csv->{$_};
}
&hash_match($sql, \%scsv, 'sql', 'csv');

sub hash_match {
    my ($p1, $p2, $n1, $n2) = @_;
    my ($a, $b);
    my @k1 = sort keys %$p1;
    my @k2 = sort keys %$p2;
    while ($a <= $#k1 && $b <= $#k2) {
        if (@k1[$a] eq @k2[$b]) {
            print "\$$n1\{'" . @k1[$a] . "'\} = " . $p1->{@k1[$a]} . "
+ ne " . 
                  "\$$n2\{'" . @k2[$b] . "'\} = " . $p2->{@k2[$b]} . "
+\n"
                  if $p1->{@k1[$a]} ne $p2->{@k2[$b]};
            $a++; $b++;
        }
        elsif (@k1[$a] lt @k2[$b]) {
            print "Unique \$$n1\{'" . @k1[$a] . "'\} = " . $p1->{@k1[$
+a++]} . "\n";
        }
        else {
            print "Unique \$$n2\{'" . @k2[$b] . "'\} = " . $p2->{@k2[$
+b++]} . "\n";
        }
    }
    print "Unique \$$n1\{'" . @k1[$a] . "'\} = " . $p1->{@k1[$a++]} . 
+"\n"
        while ($a <= $#k1);
    print "Unique \$$n2\{'" . @k2[$b] . "'\} = " . $p2->{@k2[$b++]} . 
+"\n"
        while ($b <= $#k2);
}
[download]

You might want to change the output format somewhat, and this isn't necessarily the prettiest chunk of code possible, but it works in O(n+m) rather than O(n*m) and does what you need.