[Bug target/107093] AVX512 mask operations not simplified in fully masked loop

crazylht at gmail dot com gcc-bugzilla@gcc.gnu.org
Mon Oct 10 01:50:05 GMT 2022


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107093

--- Comment #2 from Hongtao.liu <crazylht at gmail dot com> ---
For UNSPEC part, we can create a new define_insn with genenral operation and
accept both gpr and mask alternatives just like other logic patterns.
For gpr version, we can split it to xor + not after reload.
For mask version, we can split it to kxnor with UNSPEC after reload.

It should help general simplication for xor and not.


More information about the Gcc-bugs mailing list