------- Comment #23 from paolo dot carlini at oracle dot com 2010-02-18 00:45 ------- Now I understand the issue this way: the slow down is unavoidable if we want a correct treatment of signed zero, and the slow down itself is due to the way the middle-end handles mixed arithmetic. However, the front-ends have to be also fixed, otherwise we get only the slow down only without the correctness. If my summary is correct, that seems really, really bad for C++ users, I think we should consider fixing the C++ front-end too in time for 4.5.
And eventually we should anyway tell the users in the release notes about -fno-signed-zeros, if they want to restore the performance of 4.4 and can live with the wrong zeros. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43108