You're only doing half the work now :-)

You can make it work like this though:

++$count{$1},pos($genome)-- while $genome =~ /\G(..)/g
[download]

It's interesting that this is slightly faster than dropping the \G.

This however turns out to be a lot slower (seems to do a complete linear search for the proper string start every time):

$genome =~ /./g;
$count{$1}++ while $genome =~ /(.\G.)/g
[download]

update This however means that the lookahead version can be speeded up by about 5% too:

$count{$1}++ while $genome =~ /\G(?=(..))./g
[download]