SuperH: -m4-nofpu flag

[NIIBE Yutaka]

I'm thinking about adding -m4-nofpu for sh-linux target as sh-elf
does.

In sh-elf, when -m4-nofpu is specified, __sh3__ (and __SH4_NOFPU__)
are defined.  It looks weird for me to define __sh3__ for SH-4.

I'm thinking use of -m4-nofpu for the kernel compilation for SH-4 to
avoid use of FPU in kernel.  In this case, not defining __sh3__ and
defining __SH4__ is natural.

Is there any reason why __sh3__ is defined with -m4-nofpu in sh-elf?
--