Be careful with benchmark, your test is not a real test because your $str is lexical ... it is not defined in the 'pack' and 'sprintf' subroutines. You have to declare the lexical variable inside the routines. There can be significant differences between your test and a real test (although the pack and sprintf are still similar in performance):

my $str = 'hello';
timethese( 1000000, {
    'pack'      =>  q{$str = pack("A10",$str);       },
    'sprintf'   =>  q{$str = sprintf("%-10s",$str);   },
});
[download]

Results:

Benchmark: timing 1000000 iterations of pack, sprintf...
      pack:  3 wallclock secs ( 3.84 usr +  0.02 sys =  3.86 CPU)
   sprintf:  4 wallclock secs ( 4.24 usr +  0.00 sys =  4.24 CPU)
[download]

The correct way:

timethese( 1000000, {
    'pack'      =>  q{my $str = 'hello'; $str = pack("A10",$str);     
+},
    'sprintf'   =>  q{my $str = 'hello'; $str = sprintf("%-10s",$str);
+},
});
[download]

Results:

Benchmark: timing 1000000 iterations of pack, sprintf...
      pack:  6 wallclock secs ( 5.82 usr +  0.03 sys =  5.85 CPU)
   sprintf:  7 wallclock secs ( 6.80 usr +  0.01 sys =  6.81 CPU)
[download]