This trivial one is basically Moritz version implemented with 'split'. But it comes up as #3. For some reason 'index' is twice as fast in the rate limiting step.

Some crazy ideas I won't write code for:
- misusing internal variables, e. g. $RS
- grep (most likely a slow 2 pass to get the list)
- rindex (Moritz backwards)
- unpack (Instead of split)
- printf
- vec
- ikegami_tr with Bit::Vector

sub nijo {
  use bytes;
  my ($s1, $s2) = @_;
  my $idx = 0;
  my $len;

  my @chunks = split /\0/, $s1;

  foreach my $chunk (@chunks) {
    $len = length($chunk);
    substr ($s2, $idx, $len) = $chunk;
    $idx = $idx + 1 + $len;
  }
  return $s2;
}
[download]

            Rate split1 substr1 ikegami_s avar corion avar2 nijo   ike
+gami_tr moritz
split1     3.07/s     --    -84%      -86% -99%   -99%  -99%   -99%   
+    -99%  -100%
substr1    19.4/s   532%      --       -8% -95%   -96%  -96%   -96%   
+    -96%   -98%
ikegami_s  21.2/s   590%      9%        -- -95%   -95%  -95%   -96%   
+    -96%   -98%
avar        410/s 13254%   2013%     1834%   --    -8%   -8%   -20%   
+    -25%   -59%
corion      445/s 14383%   2192%     1998%   8%     --   -0%   -13%   
+    -19%   -56%
avar2       446/s 14428%   2199%     2004%   9%     0%    --   -13%   
+    -19%   -56%
nijo        514/s 16640%   2549%     2324%  25%    16%   15%     --   
+     -6%   -49%
ikegami_tr  549/s 17784%   2730%     2490%  34%    23%   23%     7%   
+      --   -46%
moritz     1009/s 32749%   5098%     4658% 146%   127%  126%    96%   
+     84%     --
1..9
ok 1 - corion gets some value
ok 2 - substr1 gets same value
ok 3 - split1 gets same value
ok 4 - moritz gets same value
ok 5 - nijo   gets same value
ok 6 - ikegami_tr gets same value
ok 7 - ikegami_s gets same value
ok 8 - avar gets same value
ok 9 - avar2 gets same value
[download]