This is the mail archive of the fortran@gcc.gnu.org mailing list for the GNU Fortran project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: RFC: Reorganization of gfortran.texi


Jerry DeLisle wrote:
This looks good. Under invocations please add some cautionary notes about using -ffast-math if we do not have that already. This is in particular related to complex math.

Thanks! We don't have any mention of it at all, currently. I've made a note to add something.


What, precisely, should the warning say? Something along the lines of this?: "Enables code optimizations which occasionally produce incorrect answers to mathematical operations, but will produce them faster than reliably correct answers can be generated. This option should be avoided unless the compiled program can be thoroughly tested for correctness. Operations on complex numbers with -ffast-math are particularly likely to obtain incorrect results."

Any other particular things, other than complex numbers, that should be warned about with it?

- Brooks


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]