Doing a quick benchmark seems to show that it is a little bit more efficient, probably because you are only having to test once. It is certainly more readable.

use strict;
use warnings;

use Benchmark q{cmpthese};

my @states;
push @states, int rand 2 for 1 .. 10000;

cmpthese(-5,
   {
      tirwhan    => sub
         {
             foreach my $mf2 ( @states )
             {
                 my $lt = $mf2 ? q{processed time} : q{n/a};
             }
         },
      upallnight => sub
         {
             foreach my $mf2 ( @states )
             {
                 my $lt;
                 $lt = q{processed time} if $mf2;
                 $lt = q{n/a} unless $mf2;
             }
         },
   });
[download]

Here's the output

             Rate upallnight    tirwhan
upallnight 22.6/s         --       -22%
tirwhan    29.1/s        29%         --
[download]

(Usually when I post benchmarks someone points out that I've got it completely wrong so I'll probably get shot down in flames for this ;-)

Cheers,

JohnGG