Quick thought to reduce capturing: m|()A|gs?

Unfortunately it seems the extra capture group actually slows things down - no matter how much I fiddle with the parameters at the top, nocapt is always about 2x as fast:

#!/usr/bin/env perl
use warnings;
use strict;
use Benchmark qw/cmpthese/;

my $count = 1000;
my $match = 'ABC'x20;
my $space = 'xAz'x100;

my $str = ( $space.$match x $count ) . $space;
cmpthese(-3, {
    nocapt => sub {
            die unless $count == ( () = $str =~ m|\Q$match|gs );
        },
    withcapt => sub {
            die unless $count == ( () = $str =~ m|()\Q$match|gs );
        },
});

__END__

           Rate withcapt   nocapt
withcapt 4264/s       --     -50%
nocapt   8596/s     102%       --
[download]