It shouldn't really affect the efficiency much, as you're limitting the split to a maximum of two results returned, and so having a longer string passed to the split won't affect it. Either way, the difference in efficiency will be negligible

use Benchmark;
use strict;

timethese(1000000, { 'One Liner' => \&one_liner, 'Two Liner' => \&two_
+liner });

sub one_liner {

        # Stick all variables in the functions, so we can
        # see if there's any extra overhead for the extra
        # variable in two_liner.

        my $text = q/Here is some test text.
                   I don't really know what I'm typing here, but
                   does it really matter?  I don't think so.  I'm just
                   kind of hitting random keys and stuff/;
        my $next_line;

        ($text, $next_line) = split /\n/, $text.$next_line, 2;
}

sub two_liner {
        my $text = q/Here is some test text.
                   I don't really know what I'm typing here, but
                   does it really matter?  I don't think so.  I'm just
                   kind of hitting random keys and stuff/;
        my($next_line, $pre_next_line);

        ($text, $pre_next_line) = split /\n/, $text, 2;
        $next_line = $pre_next_line . $next_line;
}

Benchmark: timing 1000000 iterations of One Liner, Two Liner...
 One Liner: 21 wallclock secs (19.85 usr +  0.00 sys = 19.85 CPU)
 Two Liner: 22 wallclock secs (22.85 usr +  0.00 sys = 22.85 CPU)
[download]