In case we search not only in the beginning:

use Benchmark;
use strict;

my $string = 'x' x 100;
my $regex  = qr{(?:abc|xyz)};

timethese 1_000_000 => {
    unanchored => sub { $string =~ /$regex/i },
    anchored => sub { $string =~ /^(?=.*abc)(?=.*xyz)/i },
};

__END__
Benchmark: timing 1000000 iterations of anchored, unanchored...
  anchored:  1 wallclock secs ( 1.55 usr + -0.01 sys =  1.53 CPU) @ 65
+2741.51/s (n=1000000)
unanchored: 19 wallclock secs (19.37 usr +  0.00 sys = 19.37 CPU) @ 51
+615.57/s (n=1000000)
[download]