CRIS port, prerequisite patches

[Hans-Peter Nilsson <hans-peter dot nilsson at axis dot com>]

I'd like to submit the CRIS port, a combined ELF and a.out port
for the CRIS CPU core in Axis Etrax chip series.  Testsuite
failures have been on par with those for the i686-pc-linux-gnu
host for a few days now.

Random note: I had to add expanders/splitters for movdf, to
filter out failures from what seems like a bug in SSA.  I'll try
and see if I can do something with getting libstdc++ v3 working
too.

The port depends on previously submitted patches.  I'd like to
get those sorted out, but will send patches shortly anyway, as
depending on all these previous patches.

This unreviewed patch (or at least the issues solved in some
way) is a prerequisite to compile it at all:
 Build libgcc1.a from C with extra flags; gcc/Makefile.in,
 gcc/mklibgcc.in:
 <URL:http://gcc.gnu.org/ml/gcc-patches/2000-11/msg00240.html>

And these patches are quite important too:
 Support not running "strip" for "ld -s" in collect2.c:
 <URL:http://gcc.gnu.org/ml/gcc-patches/2000-11/msg00732.html>

 Enable EH frame info in data_section (dwarf2out.c mostly):
 <URL:http://gcc.gnu.org/ml/gcc-patches/2000-11/msg00731.html>

But this isn't as important (it's just fixing bugs ;-) :
 Take 2: Adjust testsuite to "linker input file unused *because* ..."
 <URL:http://gcc.gnu.org/ml/gcc-patches/2000-11/msg00330.html>

And this is only tagging along as an slightly related improvement:
 CPP: Move REGISTER_PREFIX out of initializers:
 <URL:http://gcc.gnu.org/ml/gcc-patches/2000-11/msg00450.html>

Then I stupidly went and submitted a configure.in patch
<URL:http://gcc.gnu.org/ml/gcc-patches/2000-11/msg00843.html> so
now I confused myself whether I should submit the two trivial
configuration lines for configure.in or config.gcc :-)

brgds, H-P
PS.  Right, this is a cc0 machine.