Typically, I'll revert to a dispatch table for this sort of task:

my %dispatch = (
    test   => sub { print "test\n"  },
    hello  => sub { print "hello\n" },
);

my $name = 'test';
$dispatch{$name}->();

$name = 'hello';
$dispatch{$name}->();
[download]

This has the benefit of confining your subroutines to a very small namespace, as opposed to the entire symbol table. You can validate your input (lookup the key before calling it), and thereby provide better diagnostics.

In the event I actually want the symbol table (I can't actually recall wanting this, outside of calling methods...) then I tend to prefer can(). It does have caveats: if you're dealing with a class, it will search up the inheritance tree. However, this is usually something I want.

my $name = 'test';
__PACKAGE__->can("update_$name")->();

$name = 'hello';
__PACKAGE__->can("update_$name")->();

sub update_test  { print "test\n"  }
sub update_hello { print "hello\n" }
[download]

You can similarly check for the function before simply calling it, by checking the return value of can(), which is a subref.