------- Comment #45 from paolo dot carlini at oracle dot com 2009-11-03 12:37 ------- Created an attachment (id=18954) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18954&action=view) Second draft, regtests fine
Ok, this version works, I fixed it to use _GLIBCXX_MOVE3 in one case and _GLIBCXX_MOVE_BACKWARD3 in the other. My benchmarks are fine for simple types, like ints, I think we should just go ahead, close the PR, and maybe think about a more general solution optimized for k small, not just 1. But there is no hurry, definitely. In my opinion, for 4.5.0 it's more urgent to audit the library vs std::swap, ie, the stable_partition case, but really David, if we come to that make sure to be all set about building / testing mainline, etc. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41351
