On 6/12/05, Daniel Berlin <[EMAIL PROTECTED]> wrote: > > I could tell you why for PPC (where we aren't that far behind xlc or icc > on a lot of them, if you use the right options), but no clue for x86.
It would be interesting to see what the difference is with http://gcc.gnu.org/ml/gcc-patches/2005-06/msg00754.html and using fpmath=sse, linking against Intel libimf. At least it would tell us if we're not optimizing as good, or if simply Intels math library is better. Richard.
