perlvar does mention the issue, but it also says this has been fully fixed since v5.20.

Edit: so this would mean that you might still get the same relative positions for the different versions on older version of perls, because although $& would be significantly worse than the other solutions on their own, it would actually lower the performances of all other versions when used in the benchmark.

Edit: I thought I had a better version but no. I ran the same benchmark again and the results were not the same at all (actually the three solutions had very similar performances). Something went wrong with my first attempt

I'm actually consistantly getting result that are worse without backreferences which which I don't understand...

my $string = "AAATTTAGTTCTTAAGGCTGACATCGGTTTACGTCAGCGTTACCCCCCAAGTTATT
+GGGGACTTT";
my %expect = qw( CCCCCC 1 GGGG 1 AAA 1 TTT 3 AA 2 GG 2 TT 5 );

my $n = shift // 1;
if ($n > 1) {
    $string = $string x $n;
    $_ *= $n for values %expect;
    }

use Test::More;
use Benchmark qw(cmpthese);

my %subs;

my @v = map { "v$_" } 1 .. 8;

my %f; @f{@v} = (
  sub { %subs = ();
    $subs{$_}++ for grep { length >= 2 } split m/,/ => ($string =~ s/(
+[ACGT])\K(?!\1)/,/gr);
    }, # v1
  sub { %subs = ();
    $subs{$_}++ for grep m/^([ACGT])\1+$/ => split m/,/ => ($string =~
+ s/(\w)\K(?!\1)/,/gr);
    }, # v2
  sub { %subs = ();
    $subs{$_}++ for $string =~ m/(AA+|CC+|GG+|TT+)/g;
    }, # v3
  sub { %subs = ();
    $subs{$1}++ while $string =~ m{(([ACGT])\2+)}g;
    }, # v4
  sub { %subs = ();
    $subs{$&}++ while $string =~ m{([ACGT])\1+}g;
    }, # v5
  sub { %subs = ();
    $subs{$&}++ while $string =~ m{A{2,}|C{2,}|G{2,}|T{2,}}g;
    }, # v6
  sub { %subs = ();
    $subs{$&}++ while $string =~ m{AA+|CC+|GG+|TT+}g;
    }, # v7
  sub { %subs = ();
    $subs{$&}++ while $string =~ m{()AA+|CC+|GG+|TT+}g;
    }, # v8
  );

for (@v) {
    $f{$_}->(); is_deeply (\%subs, \%expect, $_);
    }

printf "%5d %3d %s\n", $subs{$_->[1]}, @$_ for sort { $b->[0] <=> $a->
+[0] || $a->[1] cmp $b->[1] } map {[ length, $_ ]} keys %subs;

cmpthese (-2, { map {( $_ => $f{$_} )} @v });

done_testing;
[download]

$ test.pl 1
ok 1 - v1
ok 2 - v2
ok 3 - v3
ok 4 - v4
ok 5 - v5
ok 6 - v6
ok 7 - v7
ok 8 - v8
    1   6 CCCCCC
    1   4 GGGG
    1   3 AAA
    3   3 TTT
    2   2 AA
    2   2 GG
    5   2 TT
       Rate   v2   v1   v7   v3   v4   v5   v6   v8
v2  41819/s   -- -30% -45% -53% -57% -58% -60% -63%
v1  60150/s  44%   -- -21% -32% -38% -40% -43% -47%
v7  76560/s  83%  27%   -- -13% -22% -23% -28% -32%
v3  88071/s 111%  46%  15%   -- -10% -12% -17% -22%
v4  97745/s 134%  63%  28%  11%   --  -2%  -8% -13%
v5  99555/s 138%  66%  30%  13%   2%   --  -6% -12%
v6 105700/s 153%  76%  38%  20%   8%   6%   --  -6%
v8 112783/s 170%  88%  47%  28%  15%  13%   7%   --
1..8
[download]

$ test.pl 20
ok 1 - v1
ok 2 - v2
ok 3 - v3
ok 4 - v4
ok 5 - v5
ok 6 - v6
ok 7 - v7
ok 8 - v8
   20   6 CCCCCC
   20   4 GGGG
   20   3 AAA
   60   3 TTT
   40   2 AA
   40   2 GG
  100   2 TT
     Rate   v2   v1   v7   v3   v4   v5   v6   v8
v2 2327/s   -- -29% -47% -52% -55% -57% -61% -65%
v1 3284/s  41%   -- -26% -32% -37% -39% -45% -50%
v7 4419/s  90%  35%   --  -9% -15% -17% -26% -33%
v3 4853/s 109%  48%  10%   --  -7%  -9% -18% -27%
v4 5215/s 124%  59%  18%   7%   --  -3% -12% -21%
v5 5351/s 130%  63%  21%  10%   3%   -- -10% -19%
v6 5934/s 155%  81%  34%  22%  14%  11%   -- -10%
v8 6604/s 184% 101%  49%  36%  27%  23%  11%   --
1..8
[download]

$ test.pl 2000
ok 1 - v1
ok 2 - v2
ok 3 - v3
ok 4 - v4
ok 5 - v5
ok 6 - v6
ok 7 - v7
ok 8 - v8
 2000   6 CCCCCC
 2000   4 GGGG
 2000   3 AAA
 6000   3 TTT
 4000   2 AA
 4000   2 GG
10000   2 TT
     Rate   v2   v1   v7   v3   v4   v5   v6   v8
v2 21.3/s   -- -35% -50% -54% -60% -61% -64% -68%
v1 32.7/s  54%   -- -23% -30% -38% -39% -45% -51%
v7 42.6/s 100%  30%   --  -9% -19% -21% -28% -36%
v3 46.6/s 119%  42%   9%   -- -12% -14% -21% -30%
v4 52.7/s 147%  61%  24%  13%   --  -2% -11% -21%
v5 54.0/s 154%  65%  27%  16%   3%   --  -9% -19%
v6 59.2/s 178%  81%  39%  27%  13%  10%   -- -11%
v8 66.3/s 212% 103%  56%  42%  26%  23%  12%   --
1..8
[download]


Enjoy, Have FUN! H.Merijn

you might want to run the different variants thru use re "debug" to see how they are translated into regex primitives. This might give you a clue what is happening.

Cheers Rolf
_{(addicted to the Perl Programming Language :)
Wikisyntax for the Monastery}

Note also that the benchmark results may be different for other input strings. The one in the OP is short and all the same-char substrings are also short, so for instance, results may be different if you use a long string containing long same-char substrings.

Yes indeed. Benchmarking correctly is hard :)