[Bug target/106704] [12/13 Regression] avx intrinsic not generating vblendvps instruction with -mavx

[cvs-commit at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106704

--- Comment #10 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-12 branch has been updated by hongtao Liu
<liuho...@gcc.gnu.org>:

https://gcc.gnu.org/g:9f78e7eb8e064556adf466444197aae8e52a1eb3

commit r12-8715-g9f78e7eb8e064556adf466444197aae8e52a1eb3
Author: liuhongt <hongtao....@intel.com>
Date:   Mon Aug 22 10:41:16 2022 +0800

    Don't gimple fold ymm-version vblendvpd/vblendvps/vpblendvb w/o TARGET_AVX2

    Since 256-bit vector integer comparison is under TARGET_AVX2,
    and gimple folding for vblendvpd/vblendvps/vpblendvb relies on that.
    Restrict gimple fold condition to TARGET_AVX2.

    gcc/ChangeLog:

            PR target/106704
            * config/i386/i386-builtin.def (BDESC): Add
            CODE_FOR_avx_blendvpd256/CODE_FOR_avx_blendvps256 to
            corresponding builtins.
            * config/i386/i386.cc (ix86_gimple_fold_builtin):
            Don't fold IX86_BUILTIN_PBLENDVB256, IX86_BUILTIN_BLENDVPS256,
            IX86_BUILTIN_BLENDVPD256 w/o TARGET_AVX2.

    gcc/testsuite/ChangeLog:

            * gcc.target/i386/pr106704.c: New test.