https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121778
--- Comment #5 from Jeffrey A. Law <law at gcc dot gnu.org> ---
Yea, I don't see a great target in the rv64 dumps.
As far as the patch itself. I'd bet we want to change all those SIs to X so
that it works on rv64 on a comparable testcase like:
unsigned long test_011 (unsigned long a, unsigned long b)
{ return a << 1 | ((a >> 63) ^ 1); }
Or something like that. And I suspect it should be a define_split, not a
define_insn_and_split.
