mn10300/AM33: Simplify movqi patterns

[Alexandre Oliva]

On Apr 26, 2000, Jeffrey A Law <law@cygnus.com> wrote:

> Note that you collapsed I->a into the I->d*x case, which looks like
> a harmless simplification.  However, when you look at the cc status
> bits you'll find that you took a "none_0hit" for the I->a case and
> turned it into a "clobber" which could potentially lose some
> optimizations.

Yep, I actually intended to *remove* the I->a case, since `clr a#' is
more costly than `mov 0,a#'.  I had noticed this just after posting
the patch, and I think I've posted a note that I'd fix the patch
before checking it in, but it may have been only to the
Cygnus-internal mailing list.

Here's the patch I'm checking in, and the updated size changes:

v mn10300-elf/am33/libio/libio.a 23772 23763
v mn10300-elf/am33/libio/libiostream.a 64999 64872
v mn10300-elf/am33/libstdc++/libstdc++.a 70837 70710
v mn10300-elf/am33/newlib/libc.a 47573 47533
v mn10300-elf/am33/newlib/libc/libc.a 46127 46087
v mn10300-elf/am33/newlib/libc/stdio/lib.a 16931 16922
v mn10300-elf/am33/newlib/libc/stdlib/lib.a 20333 20302
v mn10300-elf/am33/newlib/libg.a 47573 47533