I had to tweak the 5.6 solution a bit:

    ike510 => q[
        ( my $out = $in ) =~ s/.\K/\x40/sg;
        print 'ike510',  unpack 'C20', $out if DEBUG;;
    ],
    ike5_6 => q[
        ( my $out = $in ) =~ s/(.)/$1\x40/sg;
        print 'ike5.6',  unpack 'C20', $out if DEBUG;;
    ],
[download]

But the answer to the quesion is 'not well':

          Rate ike5_6 ike510    buk  shmem  salva   jmac   lanX fletch
ike5_6 0.882/s     --   -30%   -56%   -68%   -73%   -90%   -91%  -100%
ike510  1.26/s    43%     --   -37%   -54%   -61%   -85%   -88%  -100%
buk     1.99/s   126%    58%     --   -28%   -39%   -76%   -81%   -99%
shmem   2.77/s   214%   120%    39%     --   -15%   -67%   -73%   -99%
salva   3.25/s   269%   158%    63%    17%     --   -61%   -69%   -99%
jmac    8.42/s   855%   567%   322%   204%   159%     --   -19%   -97%
lanX    10.4/s  1076%   721%   420%   274%   219%    23%     --   -97%
fletch   309/s 34974% 24395% 15418% 11055%  9407%  3575%  2883%     --
[download]

I suspect that constantly reallocating the results string every 8 insertions does for them.

I had to tweak the 5.6 solution a bit

Why? Your "tweak" made it slower

$ perl -le'$_="abc"; s/(.)/$1\x40/sg; print'
a@b@c@

$ perl -le'$_="abc"; s/(?<=.)/\x40/sg; print'
a@b@c@

$ perl -le'$_="abc"; s/(?!^)/\x40/sg; print'
a@b@c@

$ perl -e'
    use Benchmark qw( cmpthese );
    our $i = chr(1) x 1e6;
    cmpthese(-3, {
        cap   => sub { (my $o=our $i) =~ s/(.)/$1\x40/sg;  },
        peek1 => sub { (my $o=our $i) =~ s/(?<=.)/\x40/sg; },
        peek2 => sub { (my $o=our $i) =~ s/(?!^)/\x40/sg;  },
    });
'
        Rate   cap peek1 peek2
cap   1.39/s    --  -25%  -27%
peek1 1.85/s   33%    --   -3%
peek2 1.91/s   37%    4%    --
[download]

I guess I must have copy&pasted wrong somehow the first time because it didn't work:

ike5_6 1100110011001100110011001100110011001100
            (warning: too few iterations for a reliable count)
[download]

Done correctly this time is does work, but frankly the difference it makes to performance is minimal:

C:\test>825494
            (warning: too few iterations for a reliable count)

            (warning: too few iterations for a reliable count)

          Rate ike5_6 ike510    buk  shmem  salva   jmac   lanX fletch
ike5_6 0.298/s     --   -10%   -72%   -79%   -82%   -93%   -95%  -100%
ike510 0.330/s    11%     --   -68%   -76%   -80%   -92%   -94%  -100%
buk     1.05/s   251%   217%     --   -26%   -37%   -75%   -81%   -99%
shmem   1.41/s   371%   325%    34%     --   -15%   -66%   -74%   -99%
salva   1.66/s   457%   402%    59%    18%     --   -60%   -70%   -99%
jmac    4.19/s  1305%  1167%   300%   198%   152%     --   -24%   -97%
lanX    5.51/s  1746%  1566%   426%   292%   232%    31%     --   -97%
fletch   163/s 54408% 49077% 15428% 11461%  9687%  3780%  2852%     --
[download]

---

Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.

"Science is about questioning the status quo. Questioning authority".

In the absence of evidence, opinion is indistinguishable from prejudice.

"I'd rather go naked than blow up my ass"