C Cross compiler for Fairchild 9445

["Walker, Arnold" <WALKEAR at mail dot northgrum dot com>]

Hello,

I've been tasked with providing a C cross compiler for the Fairchild 9445 (a
16 bit microprocessor no longer in production).  I have not found any such
compiler, so I expect to develop a new GCC back end for the F9445.  If you
know of an existing F9445 C compiler, please fill me in; otherwise, would
you please tell me how much work (developer hours) is typically involved in
developing a GNU cross compiler, and what has been the range of effort
(something like "developing a gcc cross compiler for a new target has taken
as few as i TUs and as many as l, but usually has taken between j and k
TUs", where i < j < k < l and TUs (Time Units) are hours, days or whatever).

Any info to help me make a better estimate is much appreciated.  I've
researched my company's effort in developing a Z8000 gcc cross compiler
(which they sent to the gcc team (probably in 1990 or shortly after),
although I don't see it at your site).  Although ideally (please correct me
if I'm wrong about this) only config.gcc and the files in the config
subdirectory should be changed for a new target, from 1990-1994 our
developers changed/added about 1000 lines of code in other gcc source files:
cccp.c, combine.c, cse.c, dbxout.c, expmed.c, expr.c, final.c, fold-const.c,
gcc.c, genemit.c, jump.c, local-alloc.c, regclass.c, symout.c, toplev.c and
varasm.c.  The developers who worked on the Z8000 compiler (based on version
1.37.2) have all departed our company.  Is it typical to make many changes
to these files, and how should that affect my estimate?

Of course, when/if I develop the F9445 C cross compiler (running on a PC
with NT), I will submit the new code as directed, per license.

I appreciate your reply; in its anticipation, Thank You.

Arnie Walker
Northrop Grumman Corporation
(847) 259-9600 x4488