[Bug target/64160] msp430 code generation error adding 32-bit integers

[pab at pabigot dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64160

--- Comment #3 from Peter A. Bigot <pab at pabigot dot com> ---
Comment on attachment 34232
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=34232
Proposed patch

I don't trust that the term nonsubreg is being used correctly in that predicate
since the operand does have subregs (there's no documentation of what that
predicate is intended to recognize).  I'm also unconvinced that the test
captures all the ways the operands might overlap: it seems one-sided, and
ignores operand[2], though I'm not going to try to find a counter-example.

I concede it fixes this specific example, but it doesn't leave me confident in
the validity of the split.