ITYM

@one   = (('a') x 100);
@two   = (('a') x 1000);
@three = (('a') x 10000);
[download]

As it is written each array has one element 100, 1000, or 10000 characters long rather than having 100, 1000, or 10000 elements one character long. With this change (and changing it to run 10,000 times) I get:

Benchmark: timing 10000 iterations of rev 100, rev 1000, rev 10000...
   rev 100:  0 wallclock secs ( 0.54 usr + -0.01 sys =  0.53 CPU) @ 18
+867.92/s (n=10000)
  rev 1000:  4 wallclock secs ( 5.52 usr + -0.01 sys =  5.51 CPU) @ 18
+14.88/s (n=10000)
 rev 10000: 59 wallclock secs (58.33 usr + -0.00 sys = 58.33 CPU) @ 17
+1.44/s (n=10000)
[download]