Hi, On Friday 14 August 2009 12:59:30 Matthias Kretz wrote: > Is this a bug in my code or in gcc? Do I have to add the alignment > restriction to the Vector class itself also? Thinking about this further I think this is not gcc's responsibility since operator new (or malloc) are library functions and those know nothing of the alignment restrictions of the underlying type, right? So the correct thing to do is to overload operator new/new[]/delete/delete[] in my classes an let them return _mm_malloc(size, 16); and respectively call _mm_free(ptr); . Did I get this right, now? Regards, Matthias
