This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


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

Re: What is acceptable for -ffast-math? (Was: associative law in combine)


by the way, I find that flush to zero is a perfectly reasonable semantics
for a C compiler, especially on machines that do this naturally. For example,
on an Alpha, I would do this by default, and only do real denormals in -mieee
mode, because otherwise the cost is indeed high, same think for the R10K.
There is certainly nothing in the C or C++ standard requiring denormals.


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