sub hash1 { 
    my %seen;
    $seen{$_}++ for ( split( //, $string ) );
    my $res = join( '', sort keys %seen );
    return $res;
}

sub hash2 { 
    my $res = $string;
    my %seen;
    $res =~ s/(.)/$seen{$1}++ ? "" : $1/gse;
    return $res;
}

sub joinsplit1 {  
    my $res = join '', split /(.)\1+/, join('', sort( split('', $string )));
}   
    
sub hash3 {
    my $res = join "", sort keys %{ { map{ $_ => 1 } split //, $string } };
}

sub join_tr {
    my $res = join("",sort(split("",$string)));
    $res =~ tr/0-9a-zA-Z//s;
    return $res;
}

use Benchmark;
timethese($ARGV[0] || 100, {
    hash1 => \&hash1,
    hash2 => \&hash2,
    hash3 => \&hash3,
    joinsplit1 => \&joinsplit1,
    join_tr => \&join_tr,
});

Benchmark: timing 40000 iterations of hash1, hash2, hash3, join_tr, joinsplit1...
     hash1:  1 wallclock secs ( 1.23 usr +  0.00 sys =  1.23 CPU) @ 32520.33/s (n=40000)
     hash2:  1 wallclock secs ( 1.11 usr +  0.00 sys =  1.11 CPU) @ 36036.04/s (n=40000)
     hash3:  2 wallclock secs ( 1.55 usr +  0.00 sys =  1.55 CPU) @ 25806.45/s (n=40000)
   join_tr:  1 wallclock secs ( 0.73 usr +  0.00 sys =  0.73 CPU) @ 54794.52/s (n=40000)
joinsplit1:  1 wallclock secs ( 0.94 usr +  0.00 sys =  0.94 CPU) @ 42553.19/s (n=40000)