re PR target/63173 (performance problem with simd intrinsics vld2_dup_* on aarch64...
authorFelix Yang <felix.yang@huawei.com>
Fri, 24 Oct 2014 10:53:08 +0000 (10:53 +0000)
committerFei Yang <fyang@gcc.gnu.org>
Fri, 24 Oct 2014 10:53:08 +0000 (10:53 +0000)
commit77efea31205aab4e4fa0d2760d2fa6108d4a56f3
tree5eb609950b2a805788039bc39a7f1679236df936
parente7d8c7020c4e0e392c5d64af86d1135ab02d5902
re PR target/63173 (performance problem with simd intrinsics vld2_dup_* on aarch64-none-elf)

        PR target/63173
        * config/aarch64/arm_neon.h (__LD2R_FUNC): Remove macro.
        (__LD3R_FUNC): Ditto.
        (__LD4R_FUNC): Ditto.
        (vld2_dup_s8, vld2_dup_s16, vld2_dup_s32, vld2_dup_f32, vld2_dup_f64,
         vld2_dup_u8, vld2_dup_u16, vld2_dup_u32, vld2_dup_p8, vld2_dup_p16
         vld2_dup_s64, vld2_dup_u64, vld2q_dup_s8, vld2q_dup_p8,
         vld2q_dup_s16, vld2q_dup_p16, vld2q_dup_s32, vld2q_dup_s64,
         vld2q_dup_u8, vld2q_dup_u16, vld2q_dup_u32, vld2q_dup_u64
         vld2q_dup_f32, vld2q_dup_f64): Rewrite using builtin functions.
        (vld3_dup_s64, vld3_dup_u64, vld3_dup_f64, vld3_dup_s8
         vld3_dup_p8, vld3_dup_s16, vld3_dup_p16, vld3_dup_s32
         vld3_dup_u8, vld3_dup_u16, vld3_dup_u32, vld3_dup_f32
         vld3q_dup_s8, vld3q_dup_p8, vld3q_dup_s16, vld3q_dup_p16
         vld3q_dup_s32, vld3q_dup_s64, vld3q_dup_u8, vld3q_dup_u16
         vld3q_dup_u32, vld3q_dup_u64, vld3q_dup_f32, vld3q_dup_f64): Likewise.
        (vld4_dup_s64, vld4_dup_u64, vld4_dup_f64, vld4_dup_s8
         vld4_dup_p8, vld4_dup_s16, vld4_dup_p16, vld4_dup_s32
         vld4_dup_u8, vld4_dup_u16, vld4_dup_u32, vld4_dup_f32
         vld4q_dup_s8, vld4q_dup_p8, vld4q_dup_s16, vld4q_dup_p16
         vld4q_dup_s32, vld4q_dup_s64, vld4q_dup_u8, vld4q_dup_u16
         vld4q_dup_u32, vld4q_dup_u64, vld4q_dup_f32, vld4q_dup_f64): Likewise.
        * config/aarch64/aarch64.md (define_c_enum "unspec"): Add
        UNSPEC_LD2_DUP, UNSPEC_LD3_DUP, UNSPEC_LD4_DUP.
        * config/aarch64/aarch64-simd-builtins.def (ld2r, ld3r, ld4r): New
        builtins.
        * config/aarch64/aarch64-simd.md (aarch64_simd_ld2r<mode>): New pattern.
        (aarch64_simd_ld3r<mode>): Likewise.
        (aarch64_simd_ld4r<mode>): Likewise.
        (aarch64_ld2r<mode>): New expand.
        (aarch64_ld3r<mode>): Likewise.
        (aarch64_ld4r<mode>): Likewise.

Co-Authored-By: Jiji Jiang <jiangjiji@huawei.com>
From-SVN: r216630
gcc/ChangeLog
gcc/config/aarch64/aarch64-simd-builtins.def
gcc/config/aarch64/aarch64-simd.md
gcc/config/aarch64/aarch64.md
gcc/config/aarch64/arm_neon.h