http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50160
--- Comment #19 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-08-31 00:04:36 UTC --- I'm wondering if processing an unsigned long at a time wouldn't be a step in the right direction. Then, a compiler intrinsics would be the right place for that, would naturally fit in the set of builtins fiddling with bits of integer types. On top of that the library-proper code would become very simple, in particular no ugly reinterpret_cast & co (by the way, Ch. 7 of "Hacker's Delight" has some info about such operations). Assuming indeed, we *do* have to revert the order, I'm not 101% sure.
