Extra machine modes out of tm.h

[Hans-Peter Nilsson]

On Mon, 10 Jun 2002, Zack Weinberg wrote:
> One reason why the entire bloody compiler needs to include - and
> depend upon - the target configuration headers, is that they *might*
> add extra modes to machmode.def, which *could* change the width of
> enum machine_mode, which would affect every routine that takes an
> enum machine_mode parameter.

So, how much faster does GCC now build or bootstrap with your
change?  (Or to be fair, a proof-of-concept-tweaked Makefile.in
to remove supposedly-eliminable $(TM_P_H) or somesuch.)

I really hope port mechanisms haven't been made more complex
based on *assumptions* of *build-only* speed improvements.

brgds, H-P