You could test this yourself using top(1) or whatever the quivalent is on Windows if you're on that platform. There's also a module that will tell you how much memory your data structures are using, but I forget the name of it.

Simple tests of my own show that when I create a large hash, the program uses about 71M. When I pass it as a hash ref (sub foo { ...}  foo(\%hash);), the program still uses about 71M. When I pass it as a hash (sub foo { ...}  foo(%hash);), the program uses about 117M. When I pass it as a hash to a prototyped sub ((sub foo (\%) { ...}  foo(%hash);), the program uses about 117M. Of course, I could have screwed up so test it yourself! :-)

Here's the program I used while I had top(1) running in another window. I just ran it 3 times with two of the three alternatives commented out each time and while it was sleeping I looked at the top window.

#!/usr/bin/perl

print "creating hash\n";
$k = "aaaaaaaa";
for (0..1000000) {  $hash{$k++} = $_ };
print "sleeping\n";
sleep (10);
print "copying\n";
foo(\%hash);
#bar(%hash);
#baz(%hash);
print "done";
exit;

sub foo { print "sleeping in foo\n"; sleep(10); }
sub bar (\%) { print "sleeping in bar\n"; sleep(10); }
sub baz (%) { print "sleeping in baz\n"; sleep(10); }
[download]

sub foo (\%) { ... }
foo(%$hashref);
[download]