Dorit Nuzman wrote: >> Is there documentation on how to get individual optimizations without >> invoking all of -ffast-math for gcc 4.3? In particular, vectorization >> of sum reduction, and abrupt underflow? > > w.r.t vectorization it's briefly mentioned here - > http://gcc.gnu.org/projects/tree-ssa/vectorization.html#news. > Thanks, I'll test -freassociative-math However, I would like to observe parentheses in accordance with language standards, as applications I work with lose accuracy or performance when they are ignored.