perl -MBenchmark=cmpthese -e'cmpthese -1, { cute => sub { [0,1]->[rand
+ 2] }, list => sub { qw(0 1)[rand 2] } }'
Rate cute list
cute 768000/s -- -74%
list 2899719/s 278% --
Yes, the list slice is much faster. But for something that probably runs only once in every 5 minutes, isn't 768000 per second fast enough? Optimizing seems premature here. For things like this, I am against choosing a particular language or syntax for its speed.
I'm not saying that your reply is useless. It's important to know what code does and this information will certainly help some of the readers when they do have to optimize. But the code is written now and not much is gained by changing it, so I'd just let it be. Programmer time is still much more expensive than computer time.