John Fine wrote: > > The Intel version of exp() does not use f2xm1. It uses a much more > complicated method involving far more instructions. But none of those > instructions are anywhere near as slow as f2xm1. > Yeah! that's it! Today I look into assembler code genetated by Intel - amazing! They always use SSE version of exp regardless of any complier options. Code generated with MSVC uses f2xm1 and it is 2 times slower (on my example) -- View this message in context: http://www.nabble.com/GCC--is-7-times-slower-than-Intel--How-to-optimize---Need-help%21%21-tp19817010p19875492.html Sent from the gcc - Help mailing list archive at Nabble.com.
