https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128836
--- Comment #3 from Drea Pinski <pinskia at gcc dot gnu.org> ---
```
(define_split
[(set (match_operand:SWI48 0 "register_operand")
(and:SWI48 (match_dup 0)
(const_int -65536)))
(clobber (reg:CC FLAGS_REG))]
"(TARGET_FAST_PREFIX && !TARGET_PARTIAL_REG_STALL)
|| optimize_function_for_size_p (cfun)"
[(set (strict_low_part (match_dup 1)) (const_int 0))]
"operands[1] = gen_lowpart (HImode, operands[0]);")
```
gen_lowpart here is not always a LHS version of `operands[0]` that is the
misunderstanding.
This bug seems like it was exposed by some other changes, since this split has
been there since r0-97467-gcf2b99ceb73861 before even GCC 5.