------- Comment #40 from paolo dot carlini at oracle dot com 2009-10-30 16:30 ------- The last attached patch is also not ok wrt move-only types, the special case for __k == 1 that is (trivially replacing copy -> _GLIBCXX_MOVE3 doesn't fix it). Before going ahead this way, I would also like to see benchmark results for __k close to 1 != 1, make sure that 2 is already fine performance-wise.
-- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41351
