https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103750

--- Comment #23 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by hongtao Liu <[email protected]>:

https://gcc.gnu.org/g:594dc80c8e49cb65f77a13c201a0bea9423329ec

commit r16-5369-g594dc80c8e49cb65f77a13c201a0bea9423329ec
Author: liuhongt <[email protected]>
Date:   Thu Nov 13 19:59:08 2025 -0800

    Also handle vptestnm + and15/and3 to just vptestnm.

    r16-1298-gcdfa5fe03512f7 optimizes vpcmp + and15/and3 to vpcmp when VF
    is 2 or 4. vptestnm is a variant of vpcmpeq which accepts
    nonimm_or_0_operand. The patch handles that.

    gcc/ChangeLog:

            PR target/103750
            * config/i386/sse.md (*<avx512>_eq<mode>3_and15): New
            define_insn.
            (*avx512vl_eqv2di_and3): Ditto.
            * config/i386/i386.md (*ior<mode>_ccz_1): Fix the typo in the
            comments above.

    gcc/testsuite/ChangeLog:

            * gcc.target/i386/avx512vl-pr103750-2.c: New test.

Reply via email to