https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127393
--- Comment #8 from GCC Commits <cvs-commit at gcc dot gnu.org> --- The releases/gcc-16 branch has been updated by Richard Earnshaw <[email protected]>: https://gcc.gnu.org/g:04d9b422a9dbf55726d2f6c4a7c877663bf53089 commit r16-9789-g04d9b422a9dbf55726d2f6c4a7c877663bf53089 Author: Richard Earnshaw <[email protected]> Date: Mon Oct 5 17:00:08 2026 +0100 aarch64: Fix more cases of gen_highpart on subregs [PR127393] Following on from the fix in r17-4565, there were reports of more cases where we could get ICEs during expansion of TImode patterns when subregs were in use. I've now gone through the machine description file and hopefully identified the remaining cases where this could be a problem. gcc/ChangeLog: PR target/127393 * config/aarch64/aarch64.md (movti): Ensure we don't call gen_highpart on a subreg. (addti3): Use pure_register_operand for the destination. (addvti4): Likewise. (uaddvti4): Likewise. (subti3): Likewise. (subvti4): Likewise. (usubti4): Likewise. (<su_optab>mulditi3): Likewise. (multi3): Require pure_register_operand for all operands. (cherry picked from commit a947f841925f5e65c4c47711c995ae4094a20ea4)
