https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85680
--- Comment #4 from Marc Glisse <glisse at gcc dot gnu.org> --- All memset come from ldist, so already quite late in the pipeline. Maybe clang/intel, who avoid a comparison between new and the first memset, generate memset directly from the front-end? (clang generates the first memset even at -O0) Even if(n<0)__builtin_unreachable(); doesn't quite solve the problem, because we don't realize that n==0 is the same as (size_t)n==0 :-( (a mix of single_use restrictions, pass ordering, etc) If I cheat and add if(n<=0)__builtin_unreachable(); surprisingly we fail to remove the first memset in n.0_11 = (sizetype) n_10(D); _14 = operator new [] (n.0_11); __builtin_memset (_14, 0, n.0_11); __builtin_memset (_14, 255, n.0_11); return _14; I really thought we had code to do that...
