The following benchmark concurs with the idea that "Assign" is faster than "Push" for this example.

use Benchmark qw(:all) ;

my $count = 1000;
my $foobar = 'foobar' x 1000;

timethese($count, {
    'Assigned' => sub { my @assigned = $foobar =~ /(foo)/g;},
    'Pushed' => sub { my @pushed; push @pushed, $1 while $foobar =~ /(
+foo)/g; },
});
[download]

Benchmark Results:

Benchmark: timing 1000 iterations of Assigned, Pushed...
  Assigned:  1 wallclock secs ( 0.80 usr +  0.00 sys =  0.80 CPU) @ 12
+50.00/s (n=1000)
    Pushed:  1 wallclock secs ( 0.93 usr +  0.00 sys =  0.93 CPU) @ 10
+75.27/s (n=1000)
[download]