http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55815
--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> 2013-01-05 14:34:53 UTC --- N.B. I think the only use of _Hash_bytes which users can't already replace is in type_info::hash_code(), which must give a consistent result across ABI-compatible releases. Also worth noting is that neither the murmur2collisions nor murmur3collisions code from https://131002.net/siphash/ produces more than 0.5% collisions for our _Hash_bytes function.
