Change bfd_link_hash_entry.type to a bitfield?

[H.J. Lu]

On Wed, Apr 20, 2011 at 6:57 PM, Alan Modra <amodra@gmail.com> wrote:
> How do people feel about poking some flags into bfd_link_hash_entry?
> This can be done without increasing the size of the struct, since
> "type" is an unsigned int but only needs 3 bits.  The negative side is
> that hosts without efficient byte read access will see a linker
> slowdown.  "type" is the most used field of bfd_link_hash_entry.
> On the positive side, this would allow the linker lto plugin code to
> lose a potentially very large symbol hash table.  Also, other fields
> from elf_link_hash_entry and coff_link_hash_entry could move into
> bfd_link_hash_entry, reducing the size of the linker hash table.
>

I like it,

-- 
H.J.