On Thu, 2005-05-05 at 15:58 +0400, Nadezhda IvanÐvna Vyukova wrote: > The __builtin_isless, __builtin_islessequal functions are provided as > implementations of standard C99 functions 'isless', 'isgreater'. Please, > explain why gcc for mips implements them via instructions > > c.lt.FMT and > c.le.FMT > > instead of > > c.olt.FMT and > c.ole.FMT.
No reason in particular. The only difference is that the first will signal an exception on QNaN and the second won't, thereby simplifying the programming model. Do you have a good reason for doing it the other way? -eric
