How is floating point overflow handled in aeabi library functions such as __aeabi_fadd?
Does the function return a pre-defined value? Or does it update any registers?