Unless I'm mistaken, both 'keys' and 'values' will make Perl gather up all the elements of each into temporary lists before even attempting to assign them. If %source is really huge, then that could be a problem?

If I thought memory use was a concern I'd walk the key values instead.

while (my($key, $value) = each %source) { $target{$key} = $value }
[download]

But I imagine Perl is pretty good at handling temporary lists, so I doubt looping through the keys would be as fast.

Update

... and just for fun

Update Again (used the wrong target in the benchmark)

cmpthese( -1, {
   grandfather => sub {
        my %t = %target;
        @t{keys %source} = values %source;
    },
   sleepyjay   => sub {
       my %t = %target;
       %t = ( %t, %source );
    },
   ruzam => sub {
       my %t = %target;
       while ( my($key, $value) = each %source) {
#           $target{$key} = $value   <-- oops!
            $t{$key} = $value 
       }
   },
});

__END__

              Rate   sleepyjay       ruzam grandfather
sleepyjay    905/s          --        -58%        -61%
ruzam       2162/s        139%          --         -8%
grandfather 2349/s        160%          9%          --
[download]