kernel compile: -mno-fpu

["Brian C Miles" <miles@cmu.edu>]

While building 2.4.21 on my SS20 SMP box, I noticed "-mno-fpu" being used
as a CFLAG.  Does anyone know why?  This isn't an indication that fp
emulation is being compiled into the kernel (i.e. though it may very well
be), but a directive to tell GCC to try and use software implementations
of fp math.  Does anyone know why this is enabled by default?  Does this
have any negative performance impact on machines that have hardware FPU? 
Is it safe to compile the kernel without this flag?

There are several dead-end threads about this topic on various newsgroup
archives.  If any one could shed some light on this, it would be much
appreciated.

Thanks,

Brian