[PATCH] Implement __abs?f2 functions in soft-fp

[Roland McGrath]

> Very true.  However, without support for __abs?f2 in soft-fp, glibc's
> implementation of fabs/fabsf/fabsl is probably the same inefficient
> compare against zero, and conditional negation sequence that GCC emits
> inline.

Obviously fabs et al should then be defined in libc with the new code.

I had not considered -ffree-standing and that might be a reason for gcc to
behave differently.  But GCC people and not libc people are the ones who
decide what the libgcc interface is that gcc -ffree-standing emits code for.