Off the top of my head:

use Benchmark qw/cmpthese/;
my $append = 'B' x 50;
cmpthese (100, {
  dot    => sub { my $c = 'A' x 50; $c = $c . $append                f
+or 1..1e5; },
  dot_eq => sub { my $c = 'A' x 50; $c .= $append                    f
+or 1..1e5; },
  substr => sub { my $c = 'A' x 50; substr $c, length $c, 0, $append f
+or 1..1e5; },
});
__END__
         Rate substr    dot dot_eq
substr 16.5/s     --   -35%   -37%
dot    25.5/s    55%     --    -2%
dot_eq 26.0/s    58%     2%     --
[download]

Update: Erm, had to put the 1e5 inside each sub to reduce the overhead of the my declaration.

Update2: After playing with it for a while, I find interesting what happens when $append has a size that is a power of 2. Size of $c doesn't matter:

use Benchmark qw/cmpthese/;
my $append = 'B' x 256;
cmpthese (100, {
  dot    => sub { my $c = 'A' x 50; $c = $c . $append                f
+or 1..1e5; },
  dot_eq => sub { my $c = 'A' x 50; $c .= $append                    f
+or 1..1e5; },
  substr => sub { my $c = 'A' x 50; substr $c, length $c, 0, $append f
+or 1..1e5; },
});
__END__
         Rate substr dot_eq    dot
substr 6.99/s     --   -35%   -49%
dot_eq 10.7/s    53%     --   -21%
dot    13.6/s    95%    27%     --
[download]

Size of $c doesn't matter. So if you are working with 8 Kb chunks, you better use $data = $data . $more.