slash-i now seems to be faster:

#!/usr/bin/perl -w

use strict;

use Benchmark;

our $s = "A" x 1024;
timethese(1_000_000, {
    'slash i' => sub { $s =~ /a/i; $s =~ /b/i; },
    'brackets' => sub { $s =~ /[Aa]/; $s =~ /[Bb]/; },
      });
[download]

produces:

Benchmark: timing 1000000 iterations of brackets, slash i...
  brackets: 24 wallclock secs (24.20 usr +  0.00 sys = 24.20 CPU) 
            @ 41322.31/s (n=1000000)
   slash i: 14 wallclock secs (13.47 usr +  0.00 sys = 13.47 CPU) 
            @ 74239.05/s (n=1000000)
[download]