I just realized the question is useless.

I'm not sure why you think so, how to take a slice of a hash is a valid question, maybe you could explain what resolution you found for your question?

Anyway, there are several ways to make a (shallow) copy of a part of a hash, here are a few:

1. Using map: %slice = map { $_ => $hash{$_} } @keys;
2. Using Foreach Loops (in this case as a statement modifier): $slice{$_} = $hash{$_} for @keys;
3. Using Slices: @slice{@keys} = @hash{@keys};
4. Using Key/Value Hash Slices, new in Perl 5.20: %slice = %hash{@keys};

my %h is really big and I don't have enough RAM to make a copy of it

This makes the question interesting. It turns out that, at least on my machine with Perl 5.26, from the options above:

1. map takes the most memory, about 180% as much as a regular @ slice,
2. a regular @ slice takes about as much memory as a key/value % slice,
3. a key/value % slice takes about as much memory as a regular @ slice, and
4. the for loop variant takes the least amount of memory, taking only about 88% of a regular @ slice.

use warnings;
use strict;
use feature qw/say/;

my %bighash;
{
    my $sz = 1_000_000;
    my $v = 'aaaaaaaaaaaaaa';
    $bighash{$v++} = int(rand($sz))+1 for 1..$sz;
}
my @keys = keys %bighash;
delete $bighash{aaaaaaaaaaaaab};
my %slice;

say `ps -orss $$`=~s/\s+/ /gr;

#%slice = map { $_ => $bighash{$_} } @keys;
#@slice{@keys} = @bighash{@keys};
#%slice = %bighash{@keys};
$slice{$_} = $bighash{$_} for @keys;

die if defined $slice{aaaaaaaaaaaaab}; # verify the slice
die unless $slice{aaaaaaaaaaaaaa}>0
    && $slice{aaaaaaaaaaaaaa}==$bighash{aaaaaaaaaaaaaa};

say `ps -orss $$`=~s/\s+/ /gr;
[download]

I want to load the slice directly into a hashref

As it turns out, changing my %slice into my $slice and changing the syntaxes into the following doesn't really make a big difference:

1. $slice = { map { $_ => $hash{$_} } @keys }; - about 6% more memory
2. @{$slice}{@keys} = @hash{@keys}; - about the same amount of memory
3. $slice = {%hash{@keys}}; - about the same amount of memory
4. $slice->{$_} = $hash{$_} for @keys; - about the same amount of memory

On the other hand, if for example the values in your hash are huge, then perhaps a different approach is needed - for example, in your second hash you could store references to the values of the first hash. Or, you may not even need to create a second hash in the first place, perhaps simply keeping the list of keys around to select values from the first hash is enough.