Question demonstrating complete ignorance:

Do any implementations use hardware-assisted copy-on-write to speed vector
copying?  If so, what should sorting impls do to take advantage of this?

David